forked from M-Labs/nac3
1
0
Fork 0

Compare commits

..

409 Commits

Author SHA1 Message Date
lyken 87d2a4ed59 WIP 2024-07-10 17:27:10 +08:00
lyken 9aae290727 core: irrt general numpy broadcasting 2024-07-10 17:05:01 +08:00
lyken d18c769cdc core: irrt general numpy slicing 2024-07-10 14:05:08 +08:00
lyken f41f06aec7 core: more irrt 2024-07-10 11:56:31 +08:00
lyken 1303265785 core: build.rs rewrite regex to capture `= type` 2024-07-10 10:17:45 +08:00
lyken e9cf6ce1e5 core: move irrt c++ sources to /nac3core/irrt 2024-07-10 10:17:45 +08:00
lyken bc91ab9b13 core: IRRT -Werror=return-type 2024-07-10 10:17:43 +08:00
lyken 1e06a3d199 core: add irrt_test 2024-07-10 10:11:07 +08:00
lyken 87511ac749 core: comment out numpy 2024-07-10 10:05:07 +08:00
Sebastien Bourdeauducq d658d9b00e update dependencies, Python 3.12 on Linux 2024-07-09 23:56:12 +08:00
abdul124 eeb474f9e6 core: reduce code duplication in codegen/extern_fns (#453)
Used macros to reduce code duplication in `codegen/extern_fns`

Reviewed-on: M-Labs/nac3#453
Co-authored-by: abdul124 <ar@m-labs.hk>
Co-committed-by: abdul124 <ar@m-labs.hk>
2024-07-09 16:31:08 +08:00
abdul124 88b72af2d1 core/llvm_intrinsic: improve macro name and comments 2024-07-09 16:30:32 +08:00
abdul124 b73f6c4d68 core: reduce code duplication in codegen/llvm_intrinsic 2024-07-09 16:30:32 +08:00
David Mak f47cdec650 standalone: Fix output format of output_range 2024-07-09 13:55:48 +08:00
David Mak d656880e44 standalone: Fix missing implementation for output_range 2024-07-09 13:53:50 +08:00
David Mak a91602915a core: Fix missing fields in range type 2024-07-09 13:53:50 +08:00
David Mak 1c56005a01 core: Reformat and modernize irrt.cpp
- Use anon namespace instead of static
- Use using declaration instead of typedef
- Align pointers to the type instead of the identifier
2024-07-09 13:53:50 +08:00
David Mak bc40a32524 core: Add report_type_error to enable more code reuse 2024-07-09 13:44:47 +08:00
David Mak c820daf5f8 core: Apply cargo format 2024-07-09 13:32:10 +08:00
David Mak 25d2de67f7 standalone: Add output_range and tests 2024-07-09 04:44:40 +08:00
David Mak 2cfb7a7e10 core: Refactor range function into constructor 2024-07-09 04:44:40 +08:00
David Mak 9238a5e86e standalone: Rename output_str to output_strln and add output_str
output_str is for outputting strings without newline, and the newly
introduced output_strln now has the old behavior of ending with a
newline.
2024-07-09 04:44:40 +08:00
lyken 76defac462 meta: use clang -x c++ instead of clang++ 2024-07-07 20:03:34 +08:00
lyken 650f354b74 core: use C++ for irrt source 2024-07-07 14:36:10 +08:00
abdul124 f062ef5f59 core/llvm_intrinsic: replace roundeven with rint 2024-07-07 14:24:18 +08:00
lyken f52086b706 core: improve binop and cmpop error messages 2024-07-05 16:27:24 +08:00
lyken 0a732691c9 core: refactor typecheck/magic_methods.rs operators & add op symbol name 2024-07-05 16:27:20 +08:00
lyken cbff356d50 core: workaround inkwell on `llvm.stackrestore` 2024-07-05 13:56:12 +08:00
lyken 24ac3820b2 core: check int32 obj_id directly in fold_numpy_function_call_shape_argument 2024-07-05 10:36:47 +08:00
David Mak ba32fab374 standalone: Add demos for list arithmetic operators 2024-07-04 16:01:15 +08:00
David Mak c4052b6342 core: Implement multi-operand __eq__ and __ne__ for lists 2024-07-04 16:01:15 +08:00
David Mak 66c205275f core: Implement list::__add__ 2024-07-04 16:01:11 +08:00
David Mak c85e412206 core: Implement list::__mul__ 2024-07-04 15:53:50 +08:00
David Mak 075536d7bd core: Add BreakContinueHooks for gen_for_callback 2024-07-04 15:32:18 +08:00
David Mak 13beeaa2bf core: Implement handling for zero-length lists 2024-07-04 15:32:18 +08:00
David Mak 2194dbddd5 core/type_annotation: Refactor List type to TObj
In preparation for operators on lists.
2024-07-04 15:32:18 +08:00
David Mak 94a1d547d6 meta: Update dependencies 2024-07-04 15:32:18 +08:00
lyken d6565feed3 core: ndarray_from_ndlist_impl cast size_of to usize 2024-07-04 12:24:52 +08:00
abdul124 83154ef8e1 core/llvm_intrinsics: remove llvm.roundeven call from call_float_roundeven 2024-07-03 14:17:47 +08:00
lyken 0744b938b8 core: fix __nac3_ndarray_calc_size crash due to incorrect typing 2024-07-03 13:03:14 +08:00
lyken 56fa2b6803 core: fix crash on iterating over non-iterables
a
2024-06-28 15:45:53 +08:00
lyken d06c13f936 core: fix crash on invalid subscripting 2024-06-27 16:58:48 +08:00
lyken 9808923258 core: improve comments in type_inferencer/mod.rs 2024-06-27 14:46:48 +08:00
lyken 5b11a1dbdd core: support tuple and int32 input for np_empty, np_ones, and more 2024-06-27 14:30:17 +08:00
lyken b21df53e0d core: fix comment typo in unify_call() 2024-06-27 14:06:39 +08:00
lyken 0ec967a468 core: improve function call errors 2024-06-27 14:06:39 +08:00
lyken ca8459dc7b standalone: prettify TopLevelComposer error reporting 2024-06-27 10:15:14 +08:00
abdul124 b0b804051a nac3artiq: allow class attribute access without init function 2024-06-25 16:06:33 +08:00
abdul124 134af79fd6 core: add support for class attributes 2024-06-25 16:06:33 +08:00
abdul124 7fe2c3496c core: add attribute field to class definition 2024-06-25 16:06:33 +08:00
lyken 144a3fc426 core: more derive Debug in typedef 2024-06-25 15:02:50 +08:00
lyken 74096eb9f6 core: name codegen worker threads 2024-06-25 12:36:37 +08:00
lyken 06e9d90d57 apply clippy changes 2024-06-21 14:14:01 +08:00
lyken d89146aa02 core: use no_run on builtin_fns docs 2024-06-20 13:53:25 +08:00
David Mak 5bade81ddb standalone: Add test for multidim array index with one index 2024-06-20 12:50:30 +08:00
David Mak 0452e6de78 core: Fix codegen for tuple-index into ndarray 2024-06-20 12:50:30 +08:00
David Mak 635c944c90 core: Fix type inference for tuple-index into ndarray
Fixes #420.
2024-06-20 12:50:30 +08:00
lyken e36af3b0a3 core: reduce code duplication in codegen/builtin_fns (#422)
Used macros to generate some unary math functions.

Reviewed-on: M-Labs/nac3#422
Reviewed-by: David Mak <chmakac@connect.ust.hk>
Co-authored-by: lyken <lyken@m-labs.hk>
Co-committed-by: lyken <lyken@m-labs.hk>
2024-06-20 12:48:44 +08:00
Sebastien Bourdeauducq 5b1aa812ed update dependencies 2024-06-20 10:43:55 +08:00
David Mak d3cd2a8d99 artiq: Add support for generating RPC tag for ndarray 2024-06-19 18:56:16 +08:00
David Mak 202a63274d artiq: Implement pyty-to-ty conversion 2024-06-19 18:56:15 +08:00
David Mak 76dd5191f5 artiq: Implement Python-to-LLVM conversion of ndarray 2024-06-19 18:56:15 +08:00
David Mak 8d9df0a615 artiq: Fix ndarray class ID
We want the class ID of the ndarray class, not its corresponding typing
class.
2024-06-19 18:56:15 +08:00
David Mak 07adfb2a18 standalone: Add *.ll to Gitignore list 2024-06-19 18:56:15 +08:00
Sébastien Bourdeauducq f00e458f60 add test for class without __init__ 2024-06-19 18:16:54 +08:00
David Mak 1bc95a7ba6 Add handling for np.bool_ and np.str_ 2024-06-19 15:10:47 +08:00
lyken e85f4f9bd2 core: refactor top_level::builtins::get_builtins() 2024-06-18 11:06:25 +08:00
abdul124 ce3e9bf4fe nac3artiq: add support string attributes in classes 2024-06-17 16:53:51 +08:00
David Mak 82091b1be8 meta: Apply clippy changes 2024-06-17 14:10:31 +08:00
David Mak 32919949e2 Run clippy --tests on pre-commit hook 2024-06-17 12:51:25 +08:00
lyken 2abe75d1f4 core: remove code dup with `make_exception_fields` 2024-06-17 12:01:48 +08:00
lyken 676412fe6d apply cargo fmt 2024-06-14 09:46:42 +08:00
lyken 8b9df7252f core: cleanup with Unifier::generate_var_id 2024-06-14 09:42:04 +08:00
lyken 6979843431 core: fix typo in into_var_map 2024-06-13 16:59:10 +08:00
lyken fed1361c6a core: rename to_var_map to into_var_map 2024-06-13 16:59:10 +08:00
lyken aa94e0c8a4 core: remove pub & add From<TypeVarId> for u32 2024-06-13 16:59:10 +08:00
lyken f523e26227 core: fix typo in fmt::Display of TypeVarId 2024-06-13 16:59:10 +08:00
lyken f026b48e2a core: refactor to use `TypeVarId` and `TypeVar` 2024-06-13 16:59:10 +08:00
lyken dc874f2994 core: use `PrimDef` simple names in make_primitives() 2024-06-13 16:58:32 +08:00
lyken 95de0800b4 core/demo: fix typo in .gitignore 2024-06-13 16:05:33 +08:00
lyken 3d71c6a850 core/demo: gitignore to ignore *.bc & *.o 2024-06-13 16:00:23 +08:00
David Mak be55e2ac80 meta: Update README to include info regarding pre-commit hooks 2024-06-12 16:10:57 +08:00
David Mak 79c8b759ad meta: Add pre-commit configuration 2024-06-12 16:10:57 +08:00
David Mak 4798c53a21 flake: Add pre-commit to dev environment 2024-06-12 16:10:57 +08:00
David Mak 23974feae7 meta: Restrict number of allowed lints 2024-06-12 16:10:57 +08:00
David Mak 40a3bded36 meta: Set clippy lints in {main,lib}.rs
So that this does not have to be manually passed to the `cargo clippy`
command-line every single time. Also allows incrementally addressing
these lints by removing and fixing them one-by-one.
2024-06-12 16:10:57 +08:00
lyken c4420e6ab9 core: refactor `get_builtins()` 2024-06-12 15:09:20 +08:00
lyken fd36f78005 core: refactor `PrimitiveDefinitionId` into enum `PrimDef` 2024-06-12 15:01:01 +08:00
lyken 8168692cc3 apply cargo fmt 2024-06-12 14:45:03 +08:00
David Mak 53d44b9595 standalone: Add np_array tests 2024-06-11 16:44:36 +08:00
David Mak 6153f94b05 core/numpy: Implement codegen for np_array 2024-06-11 16:42:11 +08:00
David Mak 4730b595f3 core/builtins: Add np_array function 2024-06-11 16:42:08 +08:00
David Mak c2fdb12397 core/type_inferencer: Add special rule for np_array 2024-06-11 16:40:35 +08:00
David Mak 82bf14785b core: Add multidimensional array helpers 2024-06-11 15:30:06 +08:00
David Mak 2d4329e23c core/stmt: Use BB of last statement in if-else in phi 2024-06-11 15:30:06 +08:00
David Mak 679656f9e1 core/classes: Fix incorrect field locations for lists 2024-06-11 15:30:06 +08:00
David Mak 210d9e2334 core: Add more creator functions for ProxyType 2024-06-11 15:26:37 +08:00
David Mak 181ac3ec1a core/classes: Fix incorrect pointers of range.{stop,step} 2024-06-11 15:13:31 +08:00
David Mak 3acdfb304d meta: Apply clippy suggestions 2024-06-11 14:58:32 +08:00
David Mak 6e24da9cc5 meta: Update dependencies 2024-06-11 14:58:32 +08:00
David Mak f0ab1b858a core: Refactor class abstractions
- Introduce new Type abstractions
- Rearrange some functions
2024-06-06 13:45:51 +08:00
lyken 08129cc635 nac3core: add TopLevelComposer::new builtin check's assertion msg 2024-06-05 15:30:02 +08:00
David Mak ad4832dcf4 core: Refactor to get LLVM intrinsics via Intrinsics::find 2024-06-05 15:29:40 +08:00
lyken 520bbb246b flake: add llvmPackages_14.llvm to devShells linux default (#405)
Co-authored-by: lyken <lyken@m-labs.hk>
Co-committed-by: lyken <lyken@m-labs.hk>
2024-06-05 11:11:56 +08:00
lyken b857f1e403 nac3core: fix typo in gen_for's comment 2024-06-04 17:15:41 +08:00
Sebastien Bourdeauducq fa8af37e84 flake: update nixpkgs 2024-06-03 22:22:04 +08:00
David Mak 23b2fee4e7 standalone: Add test case for ndarray slicing 2024-06-03 16:40:05 +08:00
David Mak ed79d5bb9e core/expr: Add support for multi-dim slicing of NDArrays 2024-06-03 16:40:05 +08:00
David Mak c35ad06949 core/expr: Add support for 1D slicing of NDArrays 2024-06-03 16:40:05 +08:00
David Mak 135ef557f9 core/numpy: Implement ndarray_sliced_{copy,copyto_impl}
Performing copying with optional support for slicing. Also made
copy_impl delegate to sliced_copy, as sliced_copy now performs a
superset of operations that copy_impl can already do.
2024-06-03 16:40:05 +08:00
David Mak a176c3eb70 core/irrt: Change handle_slice_indices to instead take length of object
So that all other array-like datatypes (e.g. ndarray) can also take
advantage of it.
2024-06-03 16:40:05 +08:00
David Mak 2cf79510c2 core/numpy: Add more helper functions 2024-06-03 16:40:05 +08:00
David Mak b6ff75dcaf core/irrt: Add support for calculating partial size of NDArray 2024-06-03 16:40:05 +08:00
David Mak 588c15f80d core/stmt: Add gen_for_range_callback
For generating for loops over range objects or array slices.
2024-06-03 16:40:05 +08:00
David Mak 82cc693b11 meta: Update dependencies 2024-06-03 16:40:02 +08:00
David Mak 520e1adc56 core/builtins: Add np_minimum/np_maximum 2024-05-09 15:01:20 +08:00
David Mak 73e81259f3 core/builtins: Add np_min/np_max 2024-05-09 15:01:20 +08:00
David Mak 7627acea41 core/type_inferencer: Fix error message 2024-05-09 15:01:20 +08:00
David Mak a777099ea8 core/type_inferencer: Fix missing lowering for some builtin TVars 2024-05-09 15:01:20 +08:00
David Mak 876e6ea7b8 meta: Update dependencies 2024-05-08 17:27:38 +08:00
David Mak 30c6cffbad core/builtins: Refactored numpy builtins to accept scalar and ndarrays 2024-05-06 15:38:29 +08:00
David Mak 51671800b6 core/builtins: Extract codegen portion into functions
We will need to reuse them when implementing elementwise function
application for ndarrays.
2024-05-06 13:21:54 +08:00
David Mak 7195476edb core/builtins: Add llvm_intrinsics prefix 2024-05-06 13:21:54 +08:00
David Mak eecba0b71d core: Add GenCall::create_dummy
A simple abstraction for GenCalls that are already handled elsewhere.
2024-05-06 13:21:54 +08:00
David Mak 7b4253ccd8 core/numpy: Add missing lifetime parameters 2024-05-06 13:21:54 +08:00
David Mak f58c3a11f8 core/builtins: Rework handling of PrimitiveStore-Unifier tuples 2024-05-06 13:21:54 +08:00
David Mak d0766a116f core: Remove Box from GenCallCallback type alias
So that references to the function type can be taken.
2024-05-06 13:21:54 +08:00
David Mak 64a3751fc2 core: Remove custom function type definitions for ndarray operators 2024-05-06 13:21:54 +08:00
David Mak 9566047241 standalone: Fix cbrt never tested 2024-05-06 13:21:54 +08:00
David Mak 062e318dd5 core/magic_methods: Fix clippy warnings 2024-05-06 13:21:54 +08:00
David Mak c4dc36ae99 standalone: Add explicit `--` for delimiting run args vs NAC3 args 2024-05-06 13:21:54 +08:00
David Mak baac348ee6 meta: Update dependencies 2024-05-06 13:21:37 +08:00
David Mak 847615fc2f core: Implement numpy.matmul for 2D-2D ndarrays 2024-04-23 10:27:37 +08:00
David Mak 5dfcc63978 core/classes: Take reference of indexes 2024-04-16 17:20:24 +08:00
David Mak 025b3cd02f core/stmt: Remove gen_if_chained*
Turns out it is really difficult to get lifetimes and closures right, so
let's just provide the most rudimentary if-else codegen and we can nest
them if necessary.
2024-04-16 17:16:50 +08:00
David Mak e0f440040c core/expr: Implement negative indices for ndarray 2024-04-15 12:49:42 +08:00
David Mak f0715e2b6d core/stmt: Add gen_if* functions
For generating if-constructs in IR.
2024-04-15 12:20:34 +08:00
David Mak e7fca67786 core/stmt: Do not generate jumps if bb is already terminated
Future-proofs gen_*_callback functions in case other codegen functions
will delegate to it in the future.
2024-04-15 12:20:34 +08:00
David Mak 52c731c312 core: Implement Not/UAdd/USub for booleans
Not sure if this is deliberate or an oversight, but we implement it
anyway for consistency with other Python implementations.
2024-04-12 18:29:58 +08:00
David Mak 00d1b9be9b core: Fix __inv__ for i8-based boolean operands 2024-04-12 15:35:54 +08:00
David Mak 8404d4c4dc meta: Update dependencies 2024-04-12 15:29:09 +08:00
David Mak e614dd4257 core/type_inferencer: Fix location of unary/compare expressions
Codegen uses this location information to determine the CallId, and if
a function call is the operand of a unary expression or left-hand
operand of a compare expression, codegen will use the type of the
operator expression rather than the actual operand type.
2024-04-05 15:42:10 +08:00
David Mak 937a8b9698 core/magic_methods: Fix type of unary ops with primitive types 2024-04-05 13:23:08 +08:00
David Mak 876ad6c59c core/type_inferencer: Include location info if inferencer fails 2024-04-05 13:22:35 +08:00
David Mak a920fe0501 core: Implement elementwise comparison operators 2024-04-03 00:07:33 +08:00
David Mak 727a1886b3 core: Implement elementwise unary operators 2024-04-03 00:07:33 +08:00
David Mak 6af13a8261 core: Implement elementwise binary operators
Including immediate variants of these operators.
2024-04-03 00:07:33 +08:00
David Mak 3540d0ab29 core/magic_methods: Add typeof_*op
Used to determine the expected type of the binary operator with
primitive operands.
2024-04-03 00:07:33 +08:00
David Mak 3a6c53d760 core/toplevel/numpy: Split ndarray type var utilities 2024-04-03 00:07:33 +08:00
David Mak 87bc34f7ec core: Implement calculations for broadcasting ndarrays 2024-04-03 00:07:31 +08:00
David Mak f50a5f0345 core/type_inferencer: Allow both int32 and isize when indexing ndarray 2024-04-02 16:49:12 +08:00
David Mak a77fd213e0 core/magic_methods: Allow unknown return types
These types can be later inferred by the type inferencer.
2024-04-02 16:49:12 +08:00
David Mak 8f1497df83 core/helper: Add PrimitiveDefinitionIds::iter 2024-04-02 16:49:12 +08:00
David Mak 5ca2dbeec8 core/typedef: Add Type::obj_id to replace get_obj_id 2024-04-02 16:49:10 +08:00
David Mak 9a98cde595 core: Extract codegen portion of gen_*op_expr
This allows *ops to be generated internally using LLVM values as
input. Required in a future change.
2024-04-01 16:48:25 +08:00
David Mak 5ba8601b39 core: Remove ArrayValue variants of functions
These will be lowered and optimized away later anyways, and we have
ArrayLikeAccessor now.
2024-04-01 16:48:25 +08:00
David Mak 26a01b14d5 core: Use more typed slices in APIs 2024-04-01 16:48:25 +08:00
David Mak d5f4817134 core/builtins: Fix len() on ndarrays 2024-04-01 16:48:24 +08:00
David Mak 789bfb5a26 core: Fix index-based operations not returning i32 2024-04-01 16:46:45 +08:00
David Mak 4bb0e60981 core: Apply clippy suggestions 2024-04-01 16:46:41 +08:00
Sebastien Bourdeauducq 623fcf85af msys2: update 2024-03-25 14:45:36 +08:00
David Mak 13f06f3e29 core: Refactor VarMap to IndexMap
This is the only Map I can find that preserves insertion order while
also deduplicating elements by key.
2024-03-22 15:51:23 +08:00
David Mak f0da9c0283 core: Add ArrayLikeValue
For exposing LLVM values that can be accessed like an array.
2024-03-22 15:51:06 +08:00
David Mak 2c4bf3ce59 core: Allow unsized CodeGenerator to be passed to some codegen functions
Enables codegen_callback to call these codegen functions as well.
2024-03-22 15:07:28 +08:00
David Mak e980f19c93 core: Simplify typed value assertions 2024-03-22 15:07:28 +08:00
David Mak cfbc37c1ed core: Add gen_for_callback_incrementing
Simplifies generation of monotonically increasing for loops.
2024-03-22 15:07:28 +08:00
David Mak 50264e8750 core: Add missing unchecked accessors for NDArrayDimsProxy 2024-03-22 15:07:28 +08:00
David Mak 1b77e62901 core: Split numpy into codegen and toplevel 2024-03-22 15:07:28 +08:00
David Mak fd44ee6887 core: Apply clippy suggestions 2024-03-22 15:07:23 +08:00
David Mak c8866b1534 core/classes: Rename get_* functions to remove prefix
As suggested by Rust API Guidelines.
2024-03-21 15:46:10 +08:00
David Mak 84a888758a core: Rename unsafe functions to unchecked
This is this intended name of the functions.
2024-03-21 15:46:10 +08:00
David Mak 9d550725b7 meta: Update cargo dependencies 2024-03-21 15:45:26 +08:00
David Mak 2edc1de0b6 standalone: Update ndarray.py to output all elements in ndarrays 2024-03-07 14:59:13 +08:00
David Mak c3b122acfc core: Implement `ndarray.copy` 2024-03-07 14:59:13 +08:00
David Mak a94927a11d core: Update __builtin_assume expressions
No dimension size should be 0.
2024-03-07 14:59:13 +08:00
David Mak ebf86cd134 core: Use size_t for accessing array elements 2024-03-07 14:59:13 +08:00
David Mak cccd8f2d00 core: Fix ndarray_eye not preserving signness of offset 2024-03-07 14:59:13 +08:00
David Mak 3292aed099 core: Fix ndarray subscript operator returning the wrong object
Should be returning the newly created object instead of the original
ndarray...
2024-03-07 14:59:13 +08:00
David Mak 96b7f29679 core: Implement `ndarray.fill` 2024-03-07 14:59:13 +08:00
David Mak 3d2abf73c8 core: Replace ndarray_init_dims IRRT impl with IR impl
Implementation of that function in IR allows for more flexibility in
terms of different integer type widths.
2024-03-07 14:59:13 +08:00
David Mak f682e9bf7a core: Match IRRT compile flavor with build profile 2024-03-07 14:59:02 +08:00
David Mak b26cb2b360 core: Express member func def IDs as offsets from class def ID 2024-03-06 12:24:39 +08:00
David Mak 2317516cf6 core: Use tvars from ndarray for class definition 2024-03-04 23:58:02 +08:00
David Mak 77de24ef74 core: Use BTreeMap for type variable mapping
There have been multiple instances where I had the need to iterate over
type variables, only to discover that the traversal order is arbitrary.

This commit fixes that by adding SortedMapping, which utilizes BTreeMap
internally to guarantee a traversal order. All instances of VarMap are
now refactored to use this to ensure that type variables are iterated in
 the order of its variable ID, which should be monotonically incremented
 by the unifier.
2024-03-04 23:56:04 +08:00
David Mak 234a6bde2a core: Use TObj for NDArray 2024-03-01 15:41:55 +08:00
David Mak c3db6297d9 core: Add primitive definition-id list
So that we have a single ground truth for the definition IDs of
primitive types.
2024-03-01 11:29:10 +08:00
David Mak 82fdb02d13 core: Extract LLVM intrinsic functions to their functions 2024-02-23 15:41:06 +08:00
David Mak 4efdd17513 core: Add missing From implementations for LLVM wrapper classes 2024-02-23 15:41:06 +08:00
David Mak 49de81ef1e core: Apply clippy suggestions 2024-02-23 15:41:06 +08:00
David Mak 8492503af2 core: Update cargo dependencies 2024-02-23 15:41:04 +08:00
Sebastien Bourdeauducq e1dbe2526a flake: switch to nixpkgs unstable for newer rustc 2024-02-20 15:46:51 +08:00
Sebastien Bourdeauducq f37de381ce update dependencies 2024-02-20 13:33:20 +08:00
Sebastien Bourdeauducq 4452c8986a update ARTIQ version used for PGO profiling 2024-02-20 13:29:00 +08:00
David Mak 22e831cb76 core: Add test for indexing into ndarray 2024-02-19 17:13:10 +08:00
David Mak cc538d221a core: Implement codegen for indexing into ndarray 2024-02-19 17:13:09 +08:00
David Mak 0d5c53e60c core: Implement type inference for indexing into ndarray 2024-02-19 17:13:09 +08:00
David Mak 976a9512c1 core: Add const variants to NDArray element getters 2024-02-19 16:56:21 +08:00
David Mak 1eacaf9afa core: Fix IRRT argument order to ndarray_flatten_index 2024-02-19 16:37:13 +08:00
David Mak 8c7e44098a core: Fix IRRT implementation of ndarray_flatten_index 2024-02-19 16:37:13 +08:00
David Mak 282a3e1911 core: Fix typo in error message 2024-02-14 16:26:13 +08:00
David Mak 5cecb2bb74 core: Fix Literal use in variable type annotation 2024-02-06 18:16:14 +08:00
David Mak 1963c30744 core: Use Display output for locations 2024-02-06 18:11:51 +08:00
David Mak 27011f385b core: Add location to non-primitive value return error 2024-02-02 12:49:21 +08:00
David Mak d6302b6ec8 core: Allow tuple of primitives to be returned 2024-02-02 12:48:52 +08:00
David Mak fef4b2a5ce standalone: Disable tests requiring return of non-primitive values 2024-01-29 12:49:50 +08:00
David Mak b3736c3e99 core: Disallow returning of non-primitive values
Non-primitive values are represented by an `alloca`-ed value in the
function body, and when the pointer is returned from the function, the
`alloca`-ed object is deallocated on the stack.

Related to #54.
2024-01-29 12:49:24 +08:00
Sebastien Bourdeauducq e328e44c9a update MSYS2 2024-01-26 15:55:45 +08:00
Sebastien Bourdeauducq 9e4e90f8a0 update dependencies 2024-01-26 15:52:48 +08:00
David Mak 8470915809 core: Add NDArrayValue and helper functions 2024-01-25 15:51:39 +08:00
David Mak 148900302e core: Add RangeValue and helper functions 2024-01-25 15:51:39 +08:00
David Mak 5ee08b585f core: Add ListValue and helper functions 2024-01-25 15:51:39 +08:00
David Mak f1581299fc core: Minor changes to IRRT
Add missing documentation, remove redundant lifetime variables, and fix
typos.
2024-01-25 15:50:53 +08:00
David Mak af95ba5012 standalone: Add debug flag to run_demo.sh
Allows running demos using the debug build instead of the (default)
release build.
2024-01-25 15:50:53 +08:00
David Mak 9c9756be33 standalone: Use size_t in demo.c 2024-01-25 15:50:53 +08:00
David Mak 2a922c7480 artiq: Fix source module of NDArray
Should be `numpy.typing` instead of `numpy`.
2024-01-17 10:40:08 +08:00
David Mak e3e2c36ef4 core: Mark TNDArray and TLiteral as unimplemented in tests 2024-01-17 09:58:14 +08:00
David Mak 4f9a0110c4 meta: Update insta snapshots 2024-01-17 09:49:50 +08:00
David Mak 12c0eed0a3 core: Fix compilation of tests 2024-01-17 09:49:49 +08:00
David Mak c679474f5c standalone: Fix redefinition of ndarray consumer functions 2024-01-17 09:38:13 +08:00
Sébastien Bourdeauducq ab3fa05996 demo: use portable format strings 2024-01-10 18:35:35 +08:00
David Mak 140f8f8a08 core: Implement most ndarray-creation functions 2023-12-22 16:29:55 +08:00
David Mak 27fcf8926e core: Implement ndarray constructor and numpy.empty 2023-12-22 16:29:54 +08:00
David Mak afa7d9b100 core: Implement helper for creation of generic ndarray 2023-12-21 15:39:49 +08:00
David Mak c395472094 core: Initial infrastructure for ndarray 2023-12-21 15:39:46 +08:00
David Mak 03870f222d core: Extract special method handling in type inferencer
To prepare for more special handling with methods.
2023-12-21 15:38:26 +08:00
David Mak e435b25756 core: Allow implicit promotions of integral literals
It should not matter, since it is the value of the literal that matters
with respect to the const generic variable.
2023-12-21 15:21:08 +08:00
David Mak bd792904f9 core: Add size_t to primitive store
Used for ndims in ndarray.
2023-12-21 15:20:31 +08:00
David Mak 1c3a823670 core: Do not discard value names for IRRT 2023-12-20 15:16:02 +08:00
David Mak f01d833d48 standalone: Add missing parenthesis 2023-12-20 15:15:47 +08:00
David Mak 9d64e606f4 core: Reject multiple literal bounds
This is currently broken due to how we handle function calls in the
unifier.
2023-12-18 10:04:25 +08:00
David Mak 6dccb343bb Revert "core: Do not keep unification result for function arguments"
This reverts commit f09f3c27a5.
2023-12-18 10:01:23 +08:00
Sebastien Bourdeauducq d47534e2ad interpret_demo: add typing.Literal 2023-12-18 08:50:49 +08:00
David Mak 8886964776 core: Remove redundant argument in type annotation parsing 2023-12-16 18:40:48 +08:00
David Mak f09f3c27a5 core: Do not keep unification result for function arguments
For some reason, when unifying a function call parameter with an
argument, subsequent calls to the same function will only accept the
type of the substituted argument.

This affect snippets like:

```
def make1() -> C[Literal[1]]:
    return ...

def make2() -> C[Literal[2]]:
    return ...

def consume(instance: C[Literal[1, 2]]):
    pass

consume(make1())
consume(make2())
```

The last statement will result in a compiler error, as the parameter of
consume is replaced with C[Literal[1]].

We fix this by getting a snapshot before performing unification, and
restoring the snapshot after unification succeeds.
2023-12-16 18:40:48 +08:00
David Mak 0bbc9ce6f5 core: Deduplicate values in `Literal`
Matches the behavior with `typing.Literal`.
2023-12-16 18:40:48 +08:00
David Mak 457d3b6cd7 core: Refactor generic constants to `Literal`
Better matches the syntax of `typing.Literal`.
2023-12-16 18:40:48 +08:00
David Mak 5f692debd8 core: Add PrimitiveStore into Unifier
This will be used during unification between a const generic variable
and a `Literal`.
2023-12-16 18:40:48 +08:00
David Mak c7735d935b standalone: Output id of undefined identifier 2023-12-16 18:40:48 +08:00
David Mak b47ac1b89b core: Minor formatting cleanup 2023-12-15 17:46:44 +08:00
David Mak a19f1065e3 meta: Refactor to use more let-else bindings 2023-12-12 16:31:14 +08:00
Sebastien Bourdeauducq 5bf05c6a69 update ARTIQ version used for PGO profiling 2023-12-12 15:57:48 +08:00
David Mak 32746c37be core: Refactor to return errors by HashSet 2023-12-12 15:41:59 +08:00
David Mak 1d6291b9ba ast: Add Ord implementation to Location 2023-12-12 15:41:59 +08:00
David Mak 16655959f2 meta: Update cargo dependencies 2023-12-12 15:41:59 +08:00
David Mak beee3e1f7e artiq: Pass artiq builtins to NAC3 constructor 2023-12-12 11:28:03 +08:00
David Mak d4c109b6ef core: Add missing generic constant concrete type 2023-12-12 11:28:01 +08:00
David Mak 5ffd06dd61 core: Remove debugging statement 2023-12-12 11:23:51 +08:00
David Mak 95d0c3c93c artiq: Rename const_generic_dummy to const_generic_marker 2023-12-12 11:23:51 +08:00
David Mak bd3d67f3d6 artiq: Apply clippy pedantic changes 2023-12-11 15:16:23 +08:00
David Mak ddfb532b80 standalone: Apply clippy pedantic changes 2023-12-11 15:16:23 +08:00
David Mak 02933753ca core: Apply clippy pedantic changes 2023-12-11 15:16:23 +08:00
David Mak a1f244834f meta: Bringup some documentation 2023-12-11 15:16:23 +08:00
David Mak d304afd333 meta: Apply clippy suggested changes 2023-12-11 15:16:23 +08:00
David Mak ef04696b02 meta: Lift return out of conditional statement 2023-12-11 15:16:23 +08:00
David Mak 4dc5dbb856 meta: Replace equality assertion with assert_eq
Emits a more useful assertion message.
2023-12-11 15:16:23 +08:00
David Mak fd9f66b8d9 meta: Remove redundant casts and brackets 2023-12-11 15:16:23 +08:00
David Mak 5182453bd9 meta: Remove redundant path prefixes 2023-12-11 15:16:23 +08:00
Sebastien Bourdeauducq 68556da5fd update ARTIQ version used for PGO profiling 2023-12-11 09:37:03 +08:00
David Mak 983f080ea7 artiq: Implement handling for const generic variables 2023-12-08 18:02:14 +08:00
David Mak 031e660f18 core: Initial implementation for const generics 2023-12-08 18:02:11 +08:00
David Mak b6dfcfcc38 core: Move some SymbolValue functions to symbol_resolver.rs 2023-12-08 18:00:51 +08:00
David Mak c93ad152d7 core: Codegen for ellipsis expression as NotImplemented
A lot of refactoring was performed, specifically with relaxing
expression codegen to return Option in case where ellipsis are used
within a subexpression.
2023-12-08 18:00:51 +08:00
David Mak 68b97347b1 core: Infer builtins name list using builtin declaration list 2023-12-08 17:29:34 +08:00
David Mak 875d534de4 ast: Use `{filename}:{row}:{col}` for location output 2023-12-08 15:48:54 +08:00
Sebastien Bourdeauducq adadf56e2b nac3standalone: generate PIC 2023-12-04 19:09:50 +08:00
Sebastien Bourdeauducq 9f610745b7 cargo: update dependencies 2023-12-04 18:51:06 +08:00
Sebastien Bourdeauducq 98199768e3 demo: fix 64-bit format strings 2023-12-04 18:51:06 +08:00
Sebastien Bourdeauducq bfa9ceaae3 switch to new nixpkgs release 2023-12-03 10:31:05 +08:00
Sebastien Bourdeauducq 120f8da5c7 fix compilation warnings 2023-11-26 09:09:24 +08:00
Sebastien Bourdeauducq cee62aa6c5 pin down LLVM used for IRRT 2023-11-25 20:15:29 +08:00
Sebastien Bourdeauducq fcda360ad6 flake: update dependencies 2023-11-24 18:11:25 +08:00
Sebastien Bourdeauducq 87c20ada48 windows: switch to CLANG64 MSYS2
For compatibility with MSVC (Anaconda and others).
2023-11-24 18:10:00 +08:00
Sebastien Bourdeauducq 38e968cff6 gitignore: fix msys2 path 2023-11-24 17:18:17 +08:00
David Mak 5c5620692f core: Add np_{round,floor,ceil}
These functions are NumPy variants of round/floor/ceil, which returns
floats instead of ints.
2023-11-23 13:45:07 +08:00
David Mak 0af1e37e99 core: Prefix all NumPy/SciPy functions with np_/sp_spec 2023-11-23 13:35:23 +08:00
David Mak 854e33ed48 meta: Update cargo dependencies 2023-11-23 13:31:24 +08:00
Sebastien Bourdeauducq f020d61cbb update ARTIQ version used for PGO profiling 2023-11-11 11:10:58 +08:00
David Mak 10538b5296 core: Update insta snapshots 2023-11-09 13:00:27 +08:00
David Mak d322c91697 core: Change bitshift operators to accept int32/uint32 for RHS operand 2023-11-09 12:16:20 +08:00
David Mak 3231eb0d78 core: Add compile-time error and runtime assertion for negative shifts 2023-11-09 12:16:20 +08:00
Sebastien Bourdeauducq 1ca4de99b9 update ARTIQ version used for PGO profiling 2023-11-08 17:29:29 +08:00
Sebastien Bourdeauducq bf4b1aae47 update dependencies 2023-11-08 17:23:49 +08:00
David Mak 08a5050f9a core: Implement non-trivial builtin functions using IRRT 2023-11-06 12:57:23 +08:00
David Mak c2ab6b58ff artiq: Implement `with legacy_parallel` block 2023-11-04 13:42:44 +08:00
David Mak 0a84f7ac31 Add CodeGenerator::gen_block and refactor to use it 2023-11-04 13:42:44 +08:00
David Mak fd787ca3f5 core: Remove trunc
The behavior of trunc is already implemented by casts and is therefore
redundant.
2023-11-04 13:35:53 +08:00
David Mak 4dbe07a0c0 core: Revert breaking changes to round-family functions
These functions should return ints as the math.* functions do instead of
following the convention of numpy.* functions.
2023-11-04 13:35:53 +08:00
David Mak 2e055e8ab1 core: Replace rint implementation with LLVM intrinsic 2023-11-04 13:35:53 +08:00
David Mak 9d737743c1 standalone: Add regression test for numeric primitive operations 2023-11-03 16:24:26 +08:00
David Mak c6b9aefe00 core: Fix int32-to-uint64 conversion
This conversion should be sign-extended.
2023-11-03 16:24:26 +08:00
David Mak 8ad09748d0 core: Fix conversion from float to unsigned types
These conversions also need to wraparound.
2023-11-03 16:24:26 +08:00
David Mak 7a5a2db842 core: Fix handling of float-to-int32 casts
Out-of-bound conversions should be wrapped around.
2023-11-03 16:24:26 +08:00
David Mak 447eb9c387 standalone: Fix output format string for output_uint* 2023-11-03 16:24:26 +08:00
David Mak 92d6f0a5d3 core: Implement bitwise not for unsigned ints and fix implementation 2023-11-03 16:24:26 +08:00
David Mak 7e4dab15ae standalone: Add math tests for non-number arguments 2023-11-01 18:03:29 +08:00
David Mak ff1fed112c core: Rework gamma/gammaln to match SciPy behavior
Matches behavior for infinities and NaNs.
2023-11-01 18:03:29 +08:00
David Mak 36a6a7b8cd core: Replace TopLevelDef comments with documentation 2023-11-01 18:03:29 +08:00
David Mak 2b635a0b97 core: Implement numpy and scipy functions 2023-11-01 18:03:29 +08:00
David Mak 60ad100fbb core: Implement and expose {isinf,isnan} 2023-11-01 18:03:29 +08:00
David Mak 316f0824d8 flake: Add scipy 2023-11-01 18:03:29 +08:00
David Mak 7cf7634985 core: Add create_fn_by_* functions
Used for abstracting the creation of function from different sources.
2023-11-01 18:03:29 +08:00
David Mak 068f0d9faf core: Do not cast floor/ceil result to int
NumPy explicitly states that the return type of the floor/ceil is float.
2023-11-01 18:03:29 +08:00
David Mak 95810d4229 core: Remove {ceil64,floor64,round,round64}
These are not present in NumPy or Artiq.
2023-11-01 18:03:29 +08:00
David Mak 630897b779 standalone: Do not output sign if float is NaN
Matches behavior in Python.
2023-11-01 18:03:29 +08:00
Sebastien Bourdeauducq e546535df0 flake: update nixpkgs 2023-11-01 15:53:47 +08:00
David Mak 352f70b885 artiq: Update host exception list to match possibly thrown types 2023-11-01 13:28:48 +08:00
David Mak e95586f61e core: Fix IR generation of `for` loop containing break/continue
Fix cases where the body BB would have two terminators because of a
preceding continue/break statement already emitting a terminator.
2023-11-01 13:21:27 +08:00
David Mak bb27e3d400 standalone: Fix indentation of demo.c 2023-11-01 13:20:26 +08:00
David Mak bb5147521f standalone: Fix indentation of test files 2023-11-01 13:20:26 +08:00
David Mak 9518d3fe14 artiq: Fix timeline not resetting upon exiting sequential block 2023-10-30 14:04:53 +08:00
David Mak cbd333ab10 artiq: Extract `parallel` block timeline utilities 2023-10-30 14:04:53 +08:00
David Mak 65d6104d00 artiq: Improve IR value naming and add documentation 2023-10-30 14:04:53 +08:00
David Mak 8373a6cb0f artiq: Use gen_block when generating "with sequential" 2023-10-30 14:04:53 +08:00
David Mak f75ae78677 cargo: Update dependencies 2023-10-30 14:04:53 +08:00
Sebastien Bourdeauducq ea2ab0ef7c update nixpkgs, python 3.11 2023-10-25 21:09:22 +08:00
David Mak e49b760e34 ld: Support multiple CFIs with different encoding in .eh_frame
We now parse each CFI to read its encoding as opposed to assuming that
all CFIs within the same EH_Frame uses the same encoding. FDEs are now
iterated in a per-CFI manner.
2023-10-20 18:15:03 +08:00
David Mak aa92778363 ld: Fix remapping of FDEs with multiple CFIs 2023-10-20 18:14:27 +08:00
David Mak e1487ed335 cargo: Update dependencies 2023-10-20 18:11:45 +08:00
David Mak 73500c9081 core: Remove lazy_static from dependencies 2023-10-16 15:55:10 +08:00
David Mak 9ca34c714e flake: Enable thread-safe mode for LLVM
This is required as we use the LLVM APIs from multiple threads.
2023-10-16 15:55:10 +08:00
David Mak 7fc2a30c14 Force single-threaded compilation if LLVM is not thread-safe 2023-10-16 15:55:10 +08:00
David Mak 950f431483 standalone: Update help text for `--emit-llvm` 2023-10-16 15:52:51 +08:00
David Mak a50c690428 standalone: Fix run_demo script
- Link main and module*.bc together if using multiple threads
- Fix temporary files not being deleted
2023-10-16 15:52:48 +08:00
David Mak 48eb64403f standalone: Treat -T0 as using all available threads 2023-10-13 14:57:16 +08:00
David Mak 2c44b58bb8 standalone: Require use of `-T` for specifying thread count 2023-10-13 14:36:34 +08:00
David Mak 50230e61f3 core: Simplify loop condition check for list comprehension 2023-10-06 12:24:03 +08:00
David Mak 0205161e35 core: Simplify list creation for comprehension 2023-10-06 12:22:38 +08:00
David Mak a2fce49b26 core: Allocate exceptions at the beginning of function
Only one instance of exception is necessary, as exceptions will always
be initialized before being thrown.
2023-10-06 12:13:20 +08:00
David Mak 60a503a791 core: Allocate more stack variables at the beginning of function
All allocas for temporary objects are now placed in the beginning of the
function. Allocas for on-temporary objects are not modified because
these variables may appear in a loop and thus must be uniquely
allocated by different allocas.
2023-10-06 11:42:47 +08:00
David Mak 0c49b30a90 core: Restore debug info before function call is invoked
Previously, the IR which sets up the call to the target function will
have its debug location pointing at the last argument of the function
call instead of the function call itself.
2023-10-06 11:35:23 +08:00
David Mak c7de22287e core: Fix restoration of stack address
All allocas for temporary objects are now placed in the beginning of the
function. Allocas for on-temporary objects are not modified because
these variables may appear in a loop and thus must be uniquely
represented.
2023-10-06 11:34:23 +08:00
David Mak 1a54aaa1c0 core: Restore debug location when generating allocas
Debug location is lost when moving the builder cursor.
2023-10-06 11:11:50 +08:00
David Mak c5629d4eb5 standalone: Remove redundant const in demo library 2023-10-06 10:32:58 +08:00
David Mak a79286113e standalone: Add output_bool in demo library 2023-10-06 10:19:22 +08:00
Sebastien Bourdeauducq 901e921e00 windows: fix build 2023-10-05 18:02:53 +08:00
Sebastien Bourdeauducq 45a323e969 windows: update msys2 packages 2023-10-05 17:52:29 +08:00
Sebastien Bourdeauducq 11759a722f flake: fix pgo build 2023-10-05 17:38:36 +08:00
David Mak 480a4bc0ad core: Implement comparison operators for unsigned types 2023-10-05 17:13:10 +08:00
Sebastien Bourdeauducq a1d3093196 flake: update dependencies 2023-10-05 17:05:57 +08:00
Sebastien Bourdeauducq 85c5f2c044 cargo: update dependencies 2023-10-05 17:03:35 +08:00
David Mak f34c6053d6 standalone: Add flags to control demo output options 2023-10-04 18:11:44 +08:00
David Mak e8a5f0dfef standalone: Fix parsing NAC3 args in check_demo.sh 2023-10-04 18:03:28 +08:00
David Mak 7140901261 standalone: Fix missing libraries when linking
Fixes `undefined reference to 'pow'` for pow.py using -O0.
2023-10-04 18:03:28 +08:00
David Mak 2a775d822e core: Demote dead code into a stdout warning 2023-10-04 18:03:25 +08:00
David Mak 1659c3e724 standalone: Remove temporary logfiles after execution 2023-09-30 09:31:18 +08:00
David Mak f53cb804ec standalone: Add execution of test cases via lli 2023-09-30 09:31:18 +08:00
David Mak 279376a373 standalone: Emit IRRT IR 2023-09-30 09:31:18 +08:00
David Mak b6afd1bfda standalone: Split check_demos into check_demo
Allows individual tests to be executed.
2023-09-30 09:31:18 +08:00
David Mak be3e8f50a2 standalone: Refactor demo library to C
Needed for use by lli.
2023-09-30 09:31:18 +08:00
David Mak 059d3da58b standalone: Add float64 output tests 2023-09-30 09:31:18 +08:00
David Mak 9b28f23d8c flake: Add clang alongside clang-unwrapped 2023-09-30 09:31:18 +08:00
Sebastien Bourdeauducq 119f4d63e9 cargo: update dependencies 2023-09-29 14:46:22 +08:00
Sebastien Bourdeauducq 458fa12788 flake: update dependencies 2023-09-29 14:07:47 +08:00
David Mak 48c6498d1f core: Fix restoration of loop target in try statement
old_loop_target is only assigned if ctx.loop_target is overwritten,
meaning that if ctx.loop_target is never overwritten, ctx.loop_target
will always be overwritten to None.

We fix this by only restoring from old_loop_target if we previously
assigned to old_loop_target.
2023-09-28 20:00:02 +08:00
David Mak 2a38d5160e meta: Respect opt flags when performing whole-module optimization 2023-09-28 19:58:54 +08:00
David Mak b39831b388 standalone: Update demos
- Add `output_str` for printing a string
- Add demo_test.py to test interop
2023-09-28 19:58:53 +08:00
David Mak cb39f61e79 core: Fix passing structure arguments to extern functions
All parameters with a structure type in extern functions are marked as
`byref` instead of `byval`, as most ABIs require the first several
arguments to be passed in registers before spilling into the stack.

`byval` breaks this contract by explicitly requiring all arguments to be
 passed in the stack, breaking interop with libraries written in other
 languages.
2023-09-28 15:02:35 +08:00
David Mak 176f250bdb core: Fix missing conversion to i1 for IfExp 2023-09-28 10:06:40 +08:00
David Mak acdb1de6fe meta: Improve documentation for various modified classes 2023-09-25 15:42:07 +08:00
David Mak 31dcd2dde9 core: Use i8 for boolean variable allocation
In LLVM, i1 represents a 1-byte integer with a single valid bit; The
rest of the 7 upper bits are undefined. This causes problems when
using these variables in memory operations (e.g. memcpy/memmove as
needed by List slicing and assignment).

We fix this by treating all local boolean variables as i8 so that they
are well-defined for memory operations. Function ABIs will continue to
use i1, as memory operations cannot be directly performed on function
arguments or return types, instead they are always converted back into
local boolean variables (which are i8s anyways).

Fixes #315.
2023-09-25 15:42:07 +08:00
David Mak fc93fc2f0e core: Move bitcode verification error message into panic message 2023-09-22 17:16:29 +08:00
David Mak dd42022633 core: Minor refactor allocate_list 2023-09-22 17:16:29 +08:00
David Mak 6dfc43c8b0 core: Add name to build_gep_and_load 2023-09-22 17:16:29 +08:00
David Mak ab2360d7a0 core: Remove emit_llvm from CodeGenLLVMOptions
We instead output an LLVM bitcode file when the option is specified on
the command-line.
2023-09-22 17:16:29 +08:00
David Mak ee1ee4ab3b core: Replace deprecated _ExtInt with _BitInt 2023-09-22 17:16:29 +08:00
David Mak 3e430b9b40 core: Fix missing changes for codegen tests
Apparently the changes were dropped after rebasing.
2023-09-22 17:16:21 +08:00
David Mak 9e57498958 meta: Update dependencies 2023-09-21 09:38:38 +08:00
David Mak 769fd01df8 meta: Allow specifying compiler arguments for check_demos 2023-09-18 11:35:20 +08:00
David Mak 411837cacd artiq: Specify target CPU when creating LLVM target options
We can try to optimize for the host and Cortex-A9 chips; The RISC-V
ISAs do not target specific chips, so we will fallback to using the
generic CPU.
2023-09-18 11:35:20 +08:00
David Mak f59d45805f standalone: Add command line flags for target properties
For testing codegen for different platforms on the host system.
2023-09-18 11:35:20 +08:00
David Mak 048fcb0a69 core: Switch to LLVM New Pass Manager 2023-09-18 11:35:15 +08:00
David Mak 676d07657a core: Add target field to CodeGenLLVMOptions
For specifying the target machine options when optimizing and linking.

This field is currently unused but will be required in a future
commit.
2023-09-18 09:46:24 +08:00
David Mak 2482a1ef9b core: Add CodeGenTargetMachineOptions
Needed in a future commit.
2023-09-18 09:41:49 +08:00
David Mak eb63f2ad48 meta: Update to Rust Edition 2021 2023-09-15 10:25:50 +08:00
Sebastien Bourdeauducq ff27e22ee6 flake: switch back to nixpkgs unstable
Too many issues with python-updates branch for now.
2023-09-13 19:15:47 +08:00
Sebastien Bourdeauducq d672ef094b msys2: update packages, Python 3.11 2023-09-13 09:50:33 +08:00
Sebastien Bourdeauducq d25921230e switch to Python 3.11 2023-09-13 09:44:08 +08:00
Sebastien Bourdeauducq 66f07b5bf4 flake: switch to nixos-unstable 2023-09-12 18:14:39 +08:00
David Mak 008d50995c meta: Update run_demo.sh
- Allow more than one argument to nac3standalone executable
2023-09-12 16:20:50 +08:00
David Mak 474f9050ce standalone: Expose flags in command-line 2023-09-12 16:20:49 +08:00
David Mak 3993a5cf3f core: Add LLVM options to WorkerRegistry 2023-09-12 10:57:05 +08:00
David Mak 39724de598 core: Add CodeGenLLVMOptions
For specifying LLVM options during code generation.
2023-09-12 10:57:04 +08:00
David Mak e4940247f3 standalone: Implement command-line parser using clap
In preparation for adding more command-line options.
2023-09-12 10:08:34 +08:00
David Mak 4481d48709 core: Use C-style for loop logic for iterables
Index increment is now performed at the end of the loop body.
2023-09-06 20:09:38 +08:00
David Mak b4983526bd core: Remove redundant for.cond BB for iterable loops
Simplifies logic for creating basic blocks.
2023-09-06 20:09:37 +08:00
David Mak b4a9616648 core: Add assertion for when `range` has step of 0
Aligns with the behavior in Python.
2023-09-06 20:09:36 +08:00
David Mak e0de82993f core: Preserve value of variable shadowed by for loop
Previously, the final value of the target expression would be one after
the last element of the loop, which does not match Python's behavior.

This commit fixes this problem while also preserving the last assigned
value of the loop beyond the loop, matching Python's behavior.
2023-09-06 20:09:36 +08:00
David Mak 6805253515 core: Use AST var name for IR name
Aids debugging IR.
2023-09-06 20:09:36 +08:00
David Mak 19915bac79 core: Prepend statement type to basic block label names
Aids debugging IR.
2023-09-06 20:09:36 +08:00
David Mak 17b4686260 standalone: Adapt loop example to output loop variable 2023-09-06 18:56:45 +08:00
David Mak 6de0884dc1 core: Use anonymous name for variables if unspecified
The current default prefix is only derived from the instruction type,
which is not helpful during the comprehension of the IR. Changing to
anonymous names (e.g. %1) helps understand that the variable is only
needed as part of a larger (possibly named) expression.
2023-09-06 14:02:15 +08:00
David Mak f1b0e05b3d core: Rename IR variables
Because it is unclear which variables are expressions and
subexpressions, all variables which are previously anonymous are named
using (1) the control flow statement if available, (2) the possible name
of the variable as inferred from the variable name in Rust, and (3) the
"addr" prefix to indicate that the values are pointers. These three
strings are joint together using '.', forming "for.i.addr" for instance.
2023-09-06 14:02:15 +08:00
David Mak ff23968544 core: Add name parameter to gen_{var_alloc,store_target}
This allows variables in the IR to be assigned a custom name as opposed
to names with a default prefix.
2023-09-06 11:00:02 +08:00
Sebastien Bourdeauducq 049908044a flake: update dependencies 2023-09-04 11:00:15 +08:00
David Mak d37287a33d Cargo: Update dependencies 2023-09-04 10:43:57 +08:00
Sebastien Bourdeauducq 283bd7c69a cargo: update dependencies 2023-07-14 10:57:21 +08:00
Sebastien Bourdeauducq 3d73f5c129 flake: update dependencies 2023-07-10 13:46:00 +08:00
Sebastien Bourdeauducq d824c5d8b5 flake: cleanup dev shells 2023-05-30 16:28:46 +08:00
Sebastien Bourdeauducq b8d637f5c4 cargo: update dependencies 2023-05-27 18:56:21 +08:00
Sebastien Bourdeauducq 3af287d1c4 flake: nixpkgs 23.05 2023-05-27 18:14:55 +08:00
Sebastien Bourdeauducq 5b53be0311 update dependencies 2023-04-30 17:11:47 +08:00
Sebastien Bourdeauducq aead36f0fd update dependencies 2023-03-08 15:19:09 +08:00
Sebastien Bourdeauducq c269444c0b msys2: update packages 2023-01-14 16:09:21 +08:00
Sebastien Bourdeauducq 52cec3c12f msys2: nix store doesn't like tildes 2023-01-14 16:09:00 +08:00
Sebastien Bourdeauducq 2927f2a1d0 msys2: adapt to recent pacman 2023-01-14 16:08:39 +08:00
Sebastien Bourdeauducq c1c45373a6 update dependencies 2023-01-12 19:31:03 +08:00
Sebastien Bourdeauducq 946ea155b8 flake: switch to NixOS release 2022-11-30 11:37:48 +08:00
Sebastien Bourdeauducq 085c6ee738 update dependencies 2022-11-18 16:15:46 +08:00
Sebastien Bourdeauducq cfa67c418a update MSYS2 URL 2022-11-03 19:00:44 +08:00
120 changed files with 27931 additions and 9245 deletions

1
.clippy.toml Normal file
View File

@ -0,0 +1 @@
doc-valid-idents = ["CPython", "NumPy", ".."]

2
.gitignore vendored
View File

@ -1,3 +1,3 @@
__pycache__
/target
windows/msys2
nix/windows/msys2

24
.pre-commit-config.yaml Normal file
View File

@ -0,0 +1,24 @@
# See https://pre-commit.com for more information
# See https://pre-commit.com/hooks.html for more hooks
default_stages: [commit]
repos:
- repo: local
hooks:
- id: nac3-cargo-fmt
name: nac3 cargo format
entry: cargo
language: system
types: [file, rust]
pass_filenames: false
description: Runs cargo fmt on the codebase.
args: [fmt]
- id: nac3-cargo-clippy
name: nac3 cargo clippy
entry: cargo
language: system
types: [file, rust]
pass_filenames: false
description: Runs cargo clippy on the codebase.
args: [clippy, --tests]

863
Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@ -8,6 +8,7 @@ members = [
"nac3artiq",
"runkernel",
]
resolver = "2"
[profile.release]
debug = true

View File

@ -23,21 +23,19 @@ After setting up Nix as above, use ``nix shell git+https://github.com/m-labs/art
### Windows
Install [MSYS2](https://www.msys2.org/), and open "MSYS2 MinGW x64". Edit ``/etc/pacman.conf`` to add:
Install [MSYS2](https://www.msys2.org/), and open "MSYS2 CLANG64". Edit ``/etc/pacman.conf`` to add:
```
[artiq]
SigLevel = Optional TrustAll
Server = https://lab.m-labs.hk/msys2
Server = https://msys2.m-labs.hk/artiq-nac3
```
Then run the following commands:
```
pacman -Syu
pacman -S mingw-w64-x86_64-artiq
pacman -S mingw-w64-clang-x86_64-artiq
```
Note: This build of NAC3 cannot be used with Anaconda Python nor the python.org binaries for Windows. Those Python versions are compiled with Visual Studio (MSVC) and their ABI is incompatible with the GNU ABI used in this build. We have no plans to support Visual Studio nor the MSVC ABI. If you need a MSVC build, please install the requisite bloated spyware from Microsoft and compile NAC3 yourself.
## For developers
This repository contains:
@ -53,3 +51,12 @@ Use ``nix develop`` in this repository to enter a development shell.
If you are using a different shell than bash you can use e.g. ``nix develop --command fish``.
Build NAC3 with ``cargo build --release``. See the demonstrations in ``nac3artiq`` and ``nac3standalone``.
### Pre-Commit Hooks
You are strongly recommended to use the provided pre-commit hooks to automatically reformat files and check for non-optimal Rust practices using Clippy. Run `pre-commit install` to install the hook and `pre-commit` will automatically run `cargo fmt` and `cargo clippy` for you.
Several things to note:
- If `cargo fmt` or `cargo clippy` returns an error, the pre-commit hook will fail. You should fix all errors before trying to commit again.
- If `cargo fmt` reformats some files, the pre-commit hook will also fail. You should review the changes and, if satisfied, try to commit again.

View File

@ -2,16 +2,16 @@
"nodes": {
"nixpkgs": {
"locked": {
"lastModified": 1659689094,
"narHash": "sha256-cXrWxpPYpV1PeEhtpQf9W++8aCgwzxpx2PzfszPofJE=",
"lastModified": 1720418205,
"narHash": "sha256-cPJoFPXU44GlhWg4pUk9oUPqurPlCFZ11ZQPk21GTPU=",
"owner": "NixOS",
"repo": "nixpkgs",
"rev": "697fc6ae98d077f6448cada3ecd63465c48c6af5",
"rev": "655a58a72a6601292512670343087c2d75d859c1",
"type": "github"
},
"original": {
"owner": "NixOS",
"ref": "master",
"ref": "nixos-unstable",
"repo": "nixpkgs",
"type": "github"
}

View File

@ -1,7 +1,7 @@
{
description = "The third-generation ARTIQ compiler";
inputs.nixpkgs.url = github:NixOS/nixpkgs/master;
inputs.nixpkgs.url = github:NixOS/nixpkgs/nixos-unstable;
outputs = { self, nixpkgs }:
let
@ -9,6 +9,13 @@
in rec {
packages.x86_64-linux = rec {
llvm-nac3 = pkgs.callPackage ./nix/llvm {};
llvm-tools-irrt = pkgs.runCommandNoCC "llvm-tools-irrt" {}
''
mkdir -p $out/bin
ln -s ${pkgs.llvmPackages_14.clang-unwrapped}/bin/clang $out/bin/clang-irrt
ln -s ${pkgs.llvmPackages_14.clang}/bin/clang $out/bin/clang-irrt-test
ln -s ${pkgs.llvmPackages_14.llvm.out}/bin/llvm-as $out/bin/llvm-as-irrt
'';
nac3artiq = pkgs.python3Packages.toPythonModule (
pkgs.rustPlatform.buildRustPackage rec {
name = "nac3artiq";
@ -16,14 +23,12 @@
src = self;
cargoLock = {
lockFile = ./Cargo.lock;
outputHashes = {
"inkwell-0.1.0" = "sha256-+ih3SO0n6YmZ/mcf+rLDwPAy/1MEZ/A+tI4pM1pUhvU=";
};
};
cargoTestFlags = [ "--features" "test" ];
passthru.cargoLock = cargoLock;
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_14.clang-unwrapped pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_14.clang llvm-tools-irrt pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
buildInputs = [ pkgs.python3 llvm-nac3 ];
checkInputs = [ (pkgs.python3.withPackages(ps: [ ps.numpy ])) ];
checkInputs = [ (pkgs.python3.withPackages(ps: [ ps.numpy ps.scipy ])) ];
checkPhase =
''
echo "Checking nac3standalone demos..."
@ -63,7 +68,7 @@
name = "nac3artiq-instrumented";
src = self;
inherit (nac3artiq) cargoLock;
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_14.clang-unwrapped pkgs.llvmPackages_14.llvm.out llvm-nac3-instrumented ];
nativeBuildInputs = [ pkgs.python3 packages.x86_64-linux.llvm-tools-irrt llvm-nac3-instrumented ];
buildInputs = [ pkgs.python3 llvm-nac3-instrumented ];
cargoBuildFlags = [ "--package" "nac3artiq" "--features" "init-llvm-profile" ];
doCheck = false;
@ -91,12 +96,12 @@
(pkgs.fetchFromGitHub {
owner = "m-labs";
repo = "artiq";
rev = "dd57fdc530baf926a5f354dc1c2bd90564affd96";
sha256 = "sha256-hcqVcToYWkc3oDFkKr9wZUF65ydiSYVHdmiGiu2Mc1c=";
rev = "923ca3377d42c815f979983134ec549dc39d3ca0";
sha256 = "sha256-oJoEeNEeNFSUyh6jXG8Tzp6qHVikeHS0CzfE+mODPgw=";
})
];
buildInputs = [
(python3-mimalloc.withPackages(ps: [ ps.numpy ps.jsonschema nac3artiq-instrumented ]))
(python3-mimalloc.withPackages(ps: [ ps.numpy ps.scipy ps.jsonschema ps.lmdb nac3artiq-instrumented ]))
pkgs.llvmPackages_14.llvm.out
];
phases = [ "buildPhase" "installPhase" ];
@ -125,7 +130,7 @@
name = "nac3artiq-pgo";
src = self;
inherit (nac3artiq) cargoLock;
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_14.clang-unwrapped pkgs.llvmPackages_14.llvm.out llvm-nac3-pgo ];
nativeBuildInputs = [ pkgs.python3 packages.x86_64-linux.llvm-tools-irrt llvm-nac3-pgo ];
buildInputs = [ pkgs.python3 llvm-nac3-pgo ];
cargoBuildFlags = [ "--package" "nac3artiq" ];
cargoTestFlags = [ "--package" "nac3ast" "--package" "nac3parser" "--package" "nac3core" "--package" "nac3artiq" ];
@ -141,23 +146,27 @@
packages.x86_64-w64-mingw32 = import ./nix/windows { inherit pkgs; };
devShell.x86_64-linux = pkgs.mkShell {
devShells.x86_64-linux.default = pkgs.mkShell {
name = "nac3-dev-shell";
buildInputs = with pkgs; [
# build dependencies
packages.x86_64-linux.llvm-nac3
llvmPackages_14.clang-unwrapped # IRRT
pkgs.llvmPackages_14.llvm.out # IRRT
llvmPackages_14.clang llvmPackages_14.llvm.out # for running nac3standalone demos
packages.x86_64-linux.llvm-tools-irrt
cargo
rustc
# runtime dependencies
lld_14 # for running kernels on the host
(packages.x86_64-linux.python3-mimalloc.withPackages(ps: [ ps.numpy ]))
(packages.x86_64-linux.python3-mimalloc.withPackages(ps: [ ps.numpy ps.scipy ]))
# development tools
cargo-insta
clippy
pre-commit
rustfmt
rust-analyzer
];
# https://nixos.wiki/wiki/Rust#Shell.nix_example
RUST_SRC_PATH = "${pkgs.rust.packages.stable.rustPlatform.rustLibSrc}";
};
devShells.x86_64-linux.msys2 = pkgs.mkShell {
name = "nac3-dev-shell-msys2";

View File

@ -2,22 +2,23 @@
name = "nac3artiq"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2018"
edition = "2021"
[lib]
name = "nac3artiq"
crate-type = ["cdylib"]
[dependencies]
pyo3 = { version = "0.16", features = ["extension-module"] }
itertools = "0.13"
pyo3 = { version = "0.21", features = ["extension-module", "gil-refs"] }
parking_lot = "0.12"
tempfile = "3"
tempfile = "3.10"
nac3parser = { path = "../nac3parser" }
nac3core = { path = "../nac3core" }
nac3ld = { path = "../nac3ld" }
[dependencies.inkwell]
git = "https://github.com/TheDan64/inkwell.git"
version = "0.4"
default-features = false
features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]

View File

@ -18,6 +18,13 @@ class EmbeddingMap:
"SPIError",
"0:ZeroDivisionError",
"0:IndexError",
"0:ValueError",
"0:RuntimeError",
"0:AssertionError",
"0:KeyError",
"0:NotImplementedError",
"0:OverflowError",
"0:IOError",
"0:UnwrapNoneError"])
def preallocate_runtime_exception_names(self, names):

View File

@ -10,7 +10,7 @@ from embedding_map import EmbeddingMap
__all__ = [
"Kernel", "KernelInvariant", "virtual",
"Kernel", "KernelInvariant", "virtual", "ConstGeneric",
"Option", "Some", "none", "UnwrapNoneError",
"round64", "floor64", "ceil64",
"extern", "kernel", "portable", "nac3",
@ -67,6 +67,12 @@ def Some(v: T) -> Option[T]:
none = Option(None)
class _ConstGenericMarker:
pass
def ConstGeneric(name, constraint):
return TypeVar(name, _ConstGenericMarker, constraint)
def round64(x):
return round(x)
@ -80,7 +86,13 @@ def ceil64(x):
import device_db
core_arguments = device_db.device_db["core"]["arguments"]
compiler = nac3artiq.NAC3(core_arguments["target"])
artiq_builtins = {
"none": none,
"virtual": virtual,
"_ConstGenericMarker": _ConstGenericMarker,
"Option": Option,
}
compiler = nac3artiq.NAC3(core_arguments["target"], artiq_builtins)
allow_registration = True
# Delay NAC3 analysis until all referenced variables are supposed to exist on the CPython side.
registered_functions = set()

View File

@ -0,0 +1,24 @@
from min_artiq import *
from numpy import int32
@nac3
class Demo:
core: KernelInvariant[Core]
attr1: KernelInvariant[str]
attr2: KernelInvariant[int32]
def __init__(self):
self.core = Core()
self.attr2 = 32
self.attr1 = "SAMPLE"
@kernel
def run(self):
print_int32(self.attr2)
self.attr1
if __name__ == "__main__":
Demo().run()

View File

@ -0,0 +1,40 @@
from min_artiq import *
from numpy import int32
@nac3
class Demo:
attr1: KernelInvariant[int32] = 2
attr2: int32 = 4
attr3: Kernel[int32]
@kernel
def __init__(self):
self.attr3 = 8
@nac3
class NAC3Devices:
core: KernelInvariant[Core]
attr4: KernelInvariant[int32] = 16
def __init__(self):
self.core = Core()
@kernel
def run(self):
Demo.attr1 # Supported
# Demo.attr2 # Field not accessible on Kernel
# Demo.attr3 # Only attributes can be accessed in this way
# Demo.attr1 = 2 # Attributes are immutable
self.attr4 # Attributes can be accessed within class
obj = Demo()
obj.attr1 # Attributes can be accessed by class objects
NAC3Devices.attr4 # Attributes accessible for classes without __init__
if __name__ == "__main__":
NAC3Devices().run()

View File

@ -1,12 +1,13 @@
use nac3core::{
codegen::{
expr::gen_call,
llvm_intrinsics::{call_int_smax, call_stackrestore, call_stacksave},
stmt::{gen_block, gen_with},
CodeGenContext, CodeGenerator,
},
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, GenCall},
typecheck::typedef::{FunSignature, FuncArg, Type, TypeEnum}
toplevel::{helper::PrimDef, numpy::unpack_ndarray_var_tys, DefinitionId, GenCall},
typecheck::typedef::{iter_type_vars, FunSignature, FuncArg, Type, TypeEnum, VarMap},
};
use nac3parser::ast::{Expr, ExprKind, Located, Stmt, StmtKind, StrRef};
@ -15,7 +16,10 @@ use inkwell::{
context::Context, module::Linkage, types::IntType, values::BasicValueEnum, AddressSpace,
};
use pyo3::{PyObject, PyResult, Python, types::{PyDict, PyList}};
use pyo3::{
types::{PyDict, PyList},
PyObject, PyResult, Python,
};
use crate::{symbol_resolver::InnerResolver, timeline::TimeFns};
@ -26,13 +30,45 @@ use std::{
sync::Arc,
};
/// The parallelism mode within a block.
#[derive(Copy, Clone, Eq, PartialEq)]
enum ParallelMode {
/// No parallelism is currently registered for this context.
None,
/// Legacy (or shallow) parallelism. Default before NAC3.
///
/// Each statement within the `with` block is treated as statements to be executed in parallel.
Legacy,
/// Deep parallelism. Default since NAC3.
///
/// Each function call within the `with` block (except those within a nested `sequential` block)
/// are treated to be executed in parallel.
Deep,
}
pub struct ArtiqCodeGenerator<'a> {
name: String,
/// The size of a `size_t` variable in bits.
size_t: u32,
/// Monotonic counter for naming `start`/`stop` variables used by `with parallel` blocks.
name_counter: u32,
/// Variable for tracking the start of a `with parallel` block.
start: Option<Expr<Option<Type>>>,
/// Variable for tracking the end of a `with parallel` block.
end: Option<Expr<Option<Type>>>,
timeline: &'a (dyn TimeFns + Sync),
/// The [`ParallelMode`] of the current parallel context.
///
/// The current parallel context refers to the nearest `with parallel` or `with legacy_parallel`
/// statement, which is used to determine when and how the timeline should be updated.
parallel_mode: ParallelMode,
}
impl<'a> ArtiqCodeGenerator<'a> {
@ -42,7 +78,74 @@ impl<'a> ArtiqCodeGenerator<'a> {
timeline: &'a (dyn TimeFns + Sync),
) -> ArtiqCodeGenerator<'a> {
assert!(size_t == 32 || size_t == 64);
ArtiqCodeGenerator { name, size_t, name_counter: 0, start: None, end: None, timeline }
ArtiqCodeGenerator {
name,
size_t,
name_counter: 0,
start: None,
end: None,
timeline,
parallel_mode: ParallelMode::None,
}
}
/// If the generator is currently in a direct-`parallel` block context, emits IR that resets the
/// position of the timeline to the initial timeline position before entering the `parallel`
/// block.
///
/// Direct-`parallel` block context refers to when the generator is generating statements whose
/// closest parent `with` statement is a `with parallel` block.
fn timeline_reset_start(&mut self, ctx: &mut CodeGenContext<'_, '_>) -> Result<(), String> {
if let Some(start) = self.start.clone() {
let start_val = self.gen_expr(ctx, &start)?.unwrap().to_basic_value_enum(
ctx,
self,
start.custom.unwrap(),
)?;
self.timeline.emit_at_mu(ctx, start_val);
}
Ok(())
}
/// If the generator is currently in a `parallel` block context, emits IR that updates the
/// maximum end position of the `parallel` block as specified by the timeline `end` value.
///
/// In general the `end` parameter should be set to `self.end` for updating the maximum end
/// position for the current `parallel` block. Other values can be passed in to update the
/// maximum end position for other `parallel` blocks.
///
/// `parallel`-block context refers to when the generator is generating statements within a
/// (possibly indirect) `parallel` block.
///
/// * `store_name` - The LLVM value name for the pointer to `end`. `.addr` will be appended to
/// the end of the provided value name.
fn timeline_update_end_max(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
end: Option<Expr<Option<Type>>>,
store_name: Option<&str>,
) -> Result<(), String> {
if let Some(end) = end {
let old_end = self.gen_expr(ctx, &end)?.unwrap().to_basic_value_enum(
ctx,
self,
end.custom.unwrap(),
)?;
let now = self.timeline.emit_now_mu(ctx);
let max =
call_int_smax(ctx, old_end.into_int_value(), now.into_int_value(), Some("smax"));
let end_store = self
.gen_store_target(
ctx,
&end,
store_name.map(|name| format!("{name}.addr")).as_deref(),
)?
.unwrap();
ctx.builder.build_store(end_store, max).unwrap();
}
Ok(())
}
}
@ -59,49 +162,62 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
}
}
fn gen_call<'ctx, 'a>(
fn gen_block<'ctx, 'a, 'c, I: Iterator<Item = &'c Stmt<Option<Type>>>>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmts: I,
) -> Result<(), String>
where
Self: Sized,
{
// Legacy parallel emits timeline end-update/timeline-reset after each top-level statement
// in the parallel block
if self.parallel_mode == ParallelMode::Legacy {
for stmt in stmts {
self.gen_stmt(ctx, stmt)?;
if ctx.is_terminated() {
break;
}
self.timeline_update_end_max(ctx, self.end.clone(), Some("end"))?;
self.timeline_reset_start(ctx)?;
}
Ok(())
} else {
gen_block(self, ctx, stmts)
}
}
fn gen_call<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
) -> Result<Option<BasicValueEnum<'ctx>>, String> {
let result = gen_call(self, ctx, obj, fun, params)?;
if let Some(end) = self.end.clone() {
let old_end = self.gen_expr(ctx, &end)?.unwrap().to_basic_value_enum(ctx, self, end.custom.unwrap())?;
let now = self.timeline.emit_now_mu(ctx);
let smax = ctx.module.get_function("llvm.smax.i64").unwrap_or_else(|| {
let i64 = ctx.ctx.i64_type();
ctx.module.add_function(
"llvm.smax.i64",
i64.fn_type(&[i64.into(), i64.into()], false),
None,
)
});
let max = ctx
.builder
.build_call(smax, &[old_end.into(), now.into()], "smax")
.try_as_basic_value()
.left()
.unwrap();
let end_store = self.gen_store_target(ctx, &end)?;
ctx.builder.build_store(end_store, max);
}
if let Some(start) = self.start.clone() {
let start_val = self.gen_expr(ctx, &start)?.unwrap().to_basic_value_enum(ctx, self, start.custom.unwrap())?;
self.timeline.emit_at_mu(ctx, start_val);
// Deep parallel emits timeline end-update/timeline-reset after each function call
if self.parallel_mode == ParallelMode::Deep {
self.timeline_update_end_max(ctx, self.end.clone(), Some("end"))?;
self.timeline_reset_start(ctx)?;
}
Ok(result)
}
fn gen_with<'ctx, 'a>(
fn gen_with(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'_, '_>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String> {
if let StmtKind::With { items, body, .. } = &stmt.node {
let StmtKind::With { items, body, .. } = &stmt.node else { unreachable!() };
if items.len() == 1 && items[0].optional_vars.is_none() {
let item = &items[0];
// Behavior of parallel and sequential:
// Each function call (indirectly, can be inside a sequential block) within a parallel
// block will update the end variable to the maximum now_mu in the block.
@ -116,14 +232,21 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
// - If there is a end variable, it indicates that we are (indirectly) inside a
// parallel block, and we should update the max end value.
if let ExprKind::Name { id, ctx: name_ctx } = &item.context_expr.node {
if id == &"parallel".into() {
if id == &"parallel".into() || id == &"legacy_parallel".into() {
let old_start = self.start.take();
let old_end = self.end.take();
let old_parallel_mode = self.parallel_mode;
let now = if let Some(old_start) = &old_start {
self.gen_expr(ctx, old_start)?.unwrap().to_basic_value_enum(ctx, self, old_start.custom.unwrap())?
self.gen_expr(ctx, old_start)?.unwrap().to_basic_value_enum(
ctx,
self,
old_start.custom.unwrap(),
)?
} else {
self.timeline.emit_now_mu(ctx)
};
// Emulate variable allocation, as we need to use the CodeGenContext
// HashMap to store our variable due to lifetime limitation
// Note: we should be able to store variables directly if generic
@ -137,11 +260,13 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
let start_expr = Located {
// location does not matter at this point
location: stmt.location,
node: ExprKind::Name { id: start, ctx: name_ctx.clone() },
node: ExprKind::Name { id: start, ctx: *name_ctx },
custom: Some(ctx.primitives.int64),
};
let start = self.gen_store_target(ctx, &start_expr)?;
ctx.builder.build_store(start, now);
let start = self
.gen_store_target(ctx, &start_expr, Some("start.addr"))?
.unwrap();
ctx.builder.build_store(start, now).unwrap();
Ok(Some(start_expr)) as Result<_, String>
},
|v| Ok(Some(v)),
@ -150,15 +275,23 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
let end_expr = Located {
// location does not matter at this point
location: stmt.location,
node: ExprKind::Name { id: end, ctx: name_ctx.clone() },
node: ExprKind::Name { id: end, ctx: *name_ctx },
custom: Some(ctx.primitives.int64),
};
let end = self.gen_store_target(ctx, &end_expr)?;
ctx.builder.build_store(end, now);
let end = self.gen_store_target(ctx, &end_expr, Some("end.addr"))?.unwrap();
ctx.builder.build_store(end, now).unwrap();
self.end = Some(end_expr);
self.name_counter += 1;
gen_block(self, ctx, body.iter())?;
self.parallel_mode = match id.to_string().as_str() {
"parallel" => ParallelMode::Deep,
"legacy_parallel" => ParallelMode::Legacy,
_ => unreachable!(),
};
self.gen_block(ctx, body.iter())?;
let current = ctx.builder.get_insert_block().unwrap();
// if the current block is terminated, move before the terminator
// we want to set the timeline before reaching the terminator
// TODO: This may be unsound if there are multiple exit paths in the
@ -172,70 +305,60 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
} else {
false
};
// set duration
let end_expr = self.end.take().unwrap();
let end_val = self
.gen_expr(ctx, &end_expr)?
.unwrap()
.to_basic_value_enum(ctx, self, end_expr.custom.unwrap())?;
let end_val = self.gen_expr(ctx, &end_expr)?.unwrap().to_basic_value_enum(
ctx,
self,
end_expr.custom.unwrap(),
)?;
// inside a sequential block
if old_start.is_none() {
self.timeline.emit_at_mu(ctx, end_val);
}
// inside a parallel block, should update the outer max now_mu
if let Some(old_end) = &old_end {
let outer_end_val = self
.gen_expr(ctx, old_end)?
.unwrap()
.to_basic_value_enum(ctx, self, old_end.custom.unwrap())?;
let smax =
ctx.module.get_function("llvm.smax.i64").unwrap_or_else(|| {
let i64 = ctx.ctx.i64_type();
ctx.module.add_function(
"llvm.smax.i64",
i64.fn_type(&[i64.into(), i64.into()], false),
None,
)
});
let max = ctx
.builder
.build_call(smax, &[end_val.into(), outer_end_val.into()], "smax")
.try_as_basic_value()
.left()
.unwrap();
let outer_end = self.gen_store_target(ctx, old_end)?;
ctx.builder.build_store(outer_end, max);
}
self.start = old_start;
self.timeline_update_end_max(ctx, old_end.clone(), Some("outer.end"))?;
self.parallel_mode = old_parallel_mode;
self.end = old_end;
self.start = old_start;
if reset_position {
ctx.builder.position_at_end(current);
}
return Ok(());
} else if id == &"sequential".into() {
// For deep parallel, temporarily take away start to avoid function calls in
// the block from resetting the timeline.
// This does not affect legacy parallel, as the timeline will be reset after
// this block finishes execution.
let start = self.start.take();
for stmt in body.iter() {
self.gen_stmt(ctx, stmt)?;
if ctx.is_terminated() {
break;
}
}
self.gen_block(ctx, body.iter())?;
self.start = start;
// Reset the timeline when we are exiting the sequential block
// Legacy parallel does not need this, since it will be reset after codegen
// for this statement is completed
if self.parallel_mode == ParallelMode::Deep {
self.timeline_reset_start(ctx)?;
}
return Ok(());
}
}
}
// not parallel/sequential
gen_with(self, ctx, stmt)
} else {
unreachable!()
}
}
}
fn gen_rpc_tag<'ctx, 'a>(
ctx: &mut CodeGenContext<'ctx, 'a>,
// not parallel/sequential
gen_with(self, ctx, stmt)
}
}
fn gen_rpc_tag(
ctx: &mut CodeGenContext<'_, '_>,
ty: Type,
buffer: &mut Vec<u8>,
) -> Result<(), String> {
@ -270,9 +393,32 @@ fn gen_rpc_tag<'ctx, 'a>(
gen_rpc_tag(ctx, *ty, buffer)?;
}
}
TList { ty } => {
TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
let ty = iter_type_vars(params).next().unwrap().ty;
buffer.push(b'l');
gen_rpc_tag(ctx, *ty, buffer)?;
gen_rpc_tag(ctx, ty, buffer)?;
}
TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
let (ndarray_dtype, ndarray_ndims) = unpack_ndarray_var_tys(&mut ctx.unifier, ty);
let ndarray_ndims = if let TLiteral { values, .. } =
&*ctx.unifier.get_ty_immutable(ndarray_ndims)
{
if values.len() != 1 {
return Err(format!("NDArray types with multiple literal bounds for ndims is not supported: {}", ctx.unifier.stringify(ty)));
}
let value = values[0].clone();
u64::try_from(value.clone())
.map_err(|()| format!("Expected u64 for ndarray.ndims, got {value}"))?
} else {
unreachable!()
};
assert!((0u64..=u64::from(u8::MAX)).contains(&ndarray_ndims));
buffer.push(b'a');
buffer.push((ndarray_ndims & 0xFF) as u8);
gen_rpc_tag(ctx, ndarray_dtype, buffer)?;
}
_ => return Err(format!("Unsupported type: {:?}", ctx.unifier.stringify(ty))),
}
@ -280,14 +426,14 @@ fn gen_rpc_tag<'ctx, 'a>(
Ok(())
}
fn rpc_codegen_callback_fn<'ctx, 'a>(
ctx: &mut CodeGenContext<'ctx, 'a>,
fn rpc_codegen_callback_fn<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
generator: &mut dyn CodeGenerator,
) -> Result<Option<BasicValueEnum<'ctx>>, String> {
let ptr_type = ctx.ctx.i8_type().ptr_type(inkwell::AddressSpace::Generic);
let ptr_type = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
let size_type = generator.get_size_type(ctx.ctx);
let int8 = ctx.ctx.i8_type();
let int32 = ctx.ctx.i32_type();
@ -299,7 +445,7 @@ fn rpc_codegen_callback_fn<'ctx, 'a>(
if obj.is_some() {
tag.push(b'O');
}
for arg in fun.0.args.iter() {
for arg in &fun.0.args {
gen_rpc_tag(ctx, arg.ty, &mut tag)?;
}
tag.push(b':');
@ -319,7 +465,7 @@ fn rpc_codegen_callback_fn<'ctx, 'a>(
format!("tagptr{}", fun.1 .0).as_str(),
);
tag_arr_ptr.set_initializer(&int8.const_array(
&tag.iter().map(|v| int8.const_int(*v as u64, false)).collect::<Vec<_>>(),
&tag.iter().map(|v| int8.const_int(u64::from(*v), false)).collect::<Vec<_>>(),
));
tag_arr_ptr.set_linkage(Linkage::Private);
let tag_ptr = ctx.module.add_global(tag_ptr_type, None, &hash);
@ -335,38 +481,28 @@ fn rpc_codegen_callback_fn<'ctx, 'a>(
})
.as_pointer_value();
let arg_length = args.len() + if obj.is_some() { 1 } else { 0 };
let arg_length = args.len() + usize::from(obj.is_some());
let stacksave = ctx.module.get_function("llvm.stacksave").unwrap_or_else(|| {
ctx.module.add_function("llvm.stacksave", ptr_type.fn_type(&[], false), None)
});
let stackrestore = ctx.module.get_function("llvm.stackrestore").unwrap_or_else(|| {
ctx.module.add_function(
"llvm.stackrestore",
ctx.ctx.void_type().fn_type(&[ptr_type.into()], false),
None,
)
});
let stackptr = ctx.builder.build_call(stacksave, &[], "rpc.stack");
let args_ptr = ctx.builder.build_array_alloca(
let stackptr = call_stacksave(ctx, Some("rpc.stack"));
let args_ptr = ctx
.builder
.build_array_alloca(
ptr_type,
ctx.ctx.i32_type().const_int(arg_length as u64, false),
"argptr",
);
)
.unwrap();
// -- rpc args handling
let mut keys = fun.0.args.clone();
let mut mapping = HashMap::new();
for (key, value) in args.into_iter() {
for (key, value) in args {
mapping.insert(key.unwrap_or_else(|| keys.remove(0).name), value);
}
// default value handling
for k in keys.into_iter() {
mapping.insert(
k.name,
ctx.gen_symbol_val(generator, &k.default_value.unwrap(), k.ty).into()
);
for k in keys {
mapping
.insert(k.name, ctx.gen_symbol_val(generator, &k.default_value.unwrap(), k.ty).into());
}
// reorder the parameters
let mut real_params = fun
@ -385,17 +521,19 @@ fn rpc_codegen_callback_fn<'ctx, 'a>(
}
for (i, arg) in real_params.iter().enumerate() {
let arg_slot = ctx.builder.build_alloca(arg.get_type(), &format!("rpc.arg{}", i));
ctx.builder.build_store(arg_slot, *arg);
let arg_slot = ctx.builder.build_bitcast(arg_slot, ptr_type, "rpc.arg");
let arg_slot =
generator.gen_var_alloc(ctx, arg.get_type(), Some(&format!("rpc.arg{i}"))).unwrap();
ctx.builder.build_store(arg_slot, *arg).unwrap();
let arg_slot = ctx.builder.build_bitcast(arg_slot, ptr_type, "rpc.arg").unwrap();
let arg_ptr = unsafe {
ctx.builder.build_gep(
args_ptr,
&[int32.const_int(i as u64, false)],
&format!("rpc.arg{}", i),
&format!("rpc.arg{i}"),
)
};
ctx.builder.build_store(arg_ptr, arg_slot);
}
.unwrap();
ctx.builder.build_store(arg_ptr, arg_slot).unwrap();
}
// call
@ -405,26 +543,20 @@ fn rpc_codegen_callback_fn<'ctx, 'a>(
ctx.ctx.void_type().fn_type(
&[
int32.into(),
tag_ptr_type.ptr_type(AddressSpace::Generic).into(),
ptr_type.ptr_type(AddressSpace::Generic).into(),
tag_ptr_type.ptr_type(AddressSpace::default()).into(),
ptr_type.ptr_type(AddressSpace::default()).into(),
],
false,
),
None,
)
});
ctx.builder.build_call(
rpc_send,
&[service_id.into(), tag_ptr.into(), args_ptr.into()],
"rpc.send",
);
ctx.builder
.build_call(rpc_send, &[service_id.into(), tag_ptr.into(), args_ptr.into()], "rpc.send")
.unwrap();
// reclaim stack space used by arguments
ctx.builder.build_call(
stackrestore,
&[stackptr.try_as_basic_value().unwrap_left().into()],
"rpc.stackrestore",
);
call_stackrestore(ctx, stackptr);
// -- receive value:
// T result = {
@ -450,86 +582,91 @@ fn rpc_codegen_callback_fn<'ctx, 'a>(
let alloc_bb = ctx.ctx.append_basic_block(current_function, "rpc.continue");
let tail_bb = ctx.ctx.append_basic_block(current_function, "rpc.tail");
let ret_ty = ctx.get_llvm_type(generator, fun.0.ret);
let ret_ty = ctx.get_llvm_abi_type(generator, fun.0.ret);
let need_load = !ret_ty.is_pointer_type();
let slot = ctx.builder.build_alloca(ret_ty, "rpc.ret.slot");
let slotgen = ctx.builder.build_bitcast(slot, ptr_type, "rpc.ret.ptr");
ctx.builder.build_unconditional_branch(head_bb);
let slot = ctx.builder.build_alloca(ret_ty, "rpc.ret.slot").unwrap();
let slotgen = ctx.builder.build_bitcast(slot, ptr_type, "rpc.ret.ptr").unwrap();
ctx.builder.build_unconditional_branch(head_bb).unwrap();
ctx.builder.position_at_end(head_bb);
let phi = ctx.builder.build_phi(ptr_type, "rpc.ptr");
let phi = ctx.builder.build_phi(ptr_type, "rpc.ptr").unwrap();
phi.add_incoming(&[(&slotgen, prehead_bb)]);
let alloc_size = ctx
.build_call_or_invoke(rpc_recv, &[phi.as_basic_value()], "rpc.size.next")
.unwrap()
.into_int_value();
let is_done = ctx.builder.build_int_compare(
inkwell::IntPredicate::EQ,
int32.const_zero(),
alloc_size,
"rpc.done",
);
let is_done = ctx
.builder
.build_int_compare(inkwell::IntPredicate::EQ, int32.const_zero(), alloc_size, "rpc.done")
.unwrap();
ctx.builder.build_conditional_branch(is_done, tail_bb, alloc_bb);
ctx.builder.build_conditional_branch(is_done, tail_bb, alloc_bb).unwrap();
ctx.builder.position_at_end(alloc_bb);
let alloc_ptr = ctx.builder.build_array_alloca(ptr_type, alloc_size, "rpc.alloc");
let alloc_ptr = ctx.builder.build_bitcast(alloc_ptr, ptr_type, "rpc.alloc.ptr");
let alloc_ptr = ctx.builder.build_array_alloca(ptr_type, alloc_size, "rpc.alloc").unwrap();
let alloc_ptr = ctx.builder.build_bitcast(alloc_ptr, ptr_type, "rpc.alloc.ptr").unwrap();
phi.add_incoming(&[(&alloc_ptr, alloc_bb)]);
ctx.builder.build_unconditional_branch(head_bb);
ctx.builder.build_unconditional_branch(head_bb).unwrap();
ctx.builder.position_at_end(tail_bb);
let result = ctx.builder.build_load(slot, "rpc.result");
let result = ctx.builder.build_load(slot, "rpc.result").unwrap();
if need_load {
ctx.builder.build_call(
stackrestore,
&[stackptr.try_as_basic_value().unwrap_left().into()],
"rpc.stackrestore",
);
call_stackrestore(ctx, stackptr);
}
Ok(Some(result))
}
pub fn attributes_writeback<'ctx, 'a>(
ctx: &mut CodeGenContext<'ctx, 'a>,
pub fn attributes_writeback(
ctx: &mut CodeGenContext<'_, '_>,
generator: &mut dyn CodeGenerator,
inner_resolver: &InnerResolver,
host_attributes: PyObject,
host_attributes: &PyObject,
) -> Result<(), String> {
Python::with_gil(|py| -> PyResult<Result<(), String>> {
let host_attributes = host_attributes.cast_as::<PyList>(py)?;
let host_attributes: &PyList = host_attributes.downcast(py)?;
let top_levels = ctx.top_level.definitions.read();
let globals = inner_resolver.global_value_ids.read();
let int32 = ctx.ctx.i32_type();
let zero = int32.const_zero();
let mut values = Vec::new();
let mut scratch_buffer = Vec::new();
for (_, val) in globals.iter() {
for val in (*globals).values() {
let val = val.as_ref(py);
let ty = inner_resolver.get_obj_type(py, val, &mut ctx.unifier, &top_levels, &ctx.primitives)?;
let ty = inner_resolver.get_obj_type(
py,
val,
&mut ctx.unifier,
&top_levels,
&ctx.primitives,
)?;
if let Err(ty) = ty {
return Ok(Err(ty))
return Ok(Err(ty));
}
let ty = ty.unwrap();
match &*ctx.unifier.get_ty(ty) {
TypeEnum::TObj { fields, obj_id, .. }
if *obj_id != ctx.primitives.option.get_obj_id(&ctx.unifier) =>
if *obj_id != ctx.primitives.option.obj_id(&ctx.unifier).unwrap() =>
{
// we only care about primitive attributes
// for non-primitive attributes, they should be in another global
let mut attributes = Vec::new();
let obj = inner_resolver.get_obj_value(py, val, ctx, generator, ty)?.unwrap();
for (name, (field_ty, is_mutable)) in fields.iter() {
for (name, (field_ty, is_mutable)) in fields {
if !is_mutable {
continue
continue;
}
if gen_rpc_tag(ctx, *field_ty, &mut scratch_buffer).is_ok() {
attributes.push(name.to_string());
let index = ctx.get_attr_index(ty, *name);
values.push((*field_ty, ctx.build_gep_and_load(
let (index, _) = ctx.get_attr_index(ty, *name);
values.push((
*field_ty,
ctx.build_gep_and_load(
obj.into_pointer_value(),
&[zero, int32.const_int(index as u64, false)])));
&[zero, int32.const_int(index as u64, false)],
None,
),
));
}
}
if !attributes.is_empty() {
@ -538,33 +675,46 @@ pub fn attributes_writeback<'ctx, 'a>(
pydict.set_item("fields", attributes)?;
host_attributes.append(pydict)?;
}
},
TypeEnum::TList { ty: elem_ty } => {
if gen_rpc_tag(ctx, *elem_ty, &mut scratch_buffer).is_ok() {
}
TypeEnum::TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
let elem_ty = iter_type_vars(params).next().unwrap().ty;
if gen_rpc_tag(ctx, elem_ty, &mut scratch_buffer).is_ok() {
let pydict = PyDict::new(py);
pydict.set_item("obj", val)?;
host_attributes.append(pydict)?;
values.push((ty, inner_resolver.get_obj_value(py, val, ctx, generator, ty)?.unwrap()));
values.push((
ty,
inner_resolver.get_obj_value(py, val, ctx, generator, ty)?.unwrap(),
));
}
}
},
_ => {}
}
}
let fun = FunSignature {
args: values.iter().enumerate().map(|(i, (ty, _))| FuncArg {
args: values
.iter()
.enumerate()
.map(|(i, (ty, _))| FuncArg {
name: i.to_string().into(),
ty: *ty,
default_value: None
}).collect(),
default_value: None,
})
.collect(),
ret: ctx.primitives.none,
vars: Default::default()
vars: VarMap::default(),
};
let args: Vec<_> = values.into_iter().map(|(_, val)| (None, ValueEnum::Dynamic(val))).collect();
if let Err(e) = rpc_codegen_callback_fn(ctx, None, (&fun, DefinitionId(0)), args, generator) {
let args: Vec<_> =
values.into_iter().map(|(_, val)| (None, ValueEnum::Dynamic(val))).collect();
if let Err(e) =
rpc_codegen_callback_fn(ctx, None, (&fun, PrimDef::Int32.id()), args, generator)
{
return Ok(Err(e));
}
Ok(Ok(()))
}).unwrap()?;
})
.unwrap()?;
Ok(())
}

View File

@ -1,3 +1,21 @@
#![deny(
future_incompatible,
let_underscore,
nonstandard_style,
rust_2024_compatibility,
clippy::all
)]
#![warn(clippy::pedantic)]
#![allow(
unsafe_op_in_unsafe_fn,
clippy::cast_possible_truncation,
clippy::cast_sign_loss,
clippy::enum_glob_use,
clippy::similar_names,
clippy::too_many_lines,
clippy::wildcard_imports
)]
use std::collections::{HashMap, HashSet};
use std::fs;
use std::io::Write;
@ -8,20 +26,22 @@ use std::sync::Arc;
use inkwell::{
memory_buffer::MemoryBuffer,
module::{Linkage, Module},
passes::{PassManager, PassManagerBuilder},
passes::PassBuilderOptions,
support::is_multithreaded,
targets::*,
OptimizationLevel,
};
use nac3core::codegen::gen_func_impl;
use itertools::Itertools;
use nac3core::codegen::{gen_func_impl, CodeGenLLVMOptions, CodeGenTargetMachineOptions};
use nac3core::toplevel::builtins::get_exn_constructor;
use nac3core::typecheck::typedef::{TypeEnum, Unifier};
use nac3core::typecheck::typedef::{TypeEnum, Unifier, VarMap};
use nac3parser::{
ast::{self, ExprKind, Stmt, StmtKind, StrRef},
parser::{self, parse_program},
ast::{ExprKind, Stmt, StmtKind, StrRef},
parser::parse_program,
};
use pyo3::create_exception;
use pyo3::prelude::*;
use pyo3::{exceptions, types::PyBytes, types::PyDict, types::PySet};
use pyo3::create_exception;
use parking_lot::{Mutex, RwLock};
@ -44,7 +64,7 @@ use tempfile::{self, TempDir};
use crate::codegen::attributes_writeback;
use crate::{
codegen::{rpc_codegen_callback, ArtiqCodeGenerator},
symbol_resolver::{InnerResolver, PythonHelper, Resolver, DeferredEvaluationStore},
symbol_resolver::{DeferredEvaluationStore, InnerResolver, PythonHelper, Resolver},
};
mod codegen;
@ -61,6 +81,17 @@ enum Isa {
CortexA9,
}
impl Isa {
/// Returns the number of bits in `size_t` for the [`Isa`].
fn get_size_type(self) -> u32 {
if self == Isa::Host {
64u32
} else {
32u32
}
}
}
#[derive(Clone)]
pub struct PrimitivePythonId {
int: u64,
@ -71,9 +102,14 @@ pub struct PrimitivePythonId {
float: u64,
float64: u64,
bool: u64,
np_bool_: u64,
string: u64,
np_str_: u64,
list: u64,
ndarray: u64,
tuple: u64,
typevar: u64,
const_generic_marker: u64,
none: u64,
exception: u64,
generic_alias: (u64, u64),
@ -97,7 +133,9 @@ struct Nac3 {
top_levels: Vec<TopLevelComponent>,
string_store: Arc<RwLock<HashMap<String, i32>>>,
exception_ids: Arc<RwLock<HashMap<usize, usize>>>,
deferred_eval_store: DeferredEvaluationStore
deferred_eval_store: DeferredEvaluationStore,
/// LLVM-related options for code generation.
llvm_options: CodeGenLLVMOptions,
}
create_exception!(nac3artiq, CompileError, exceptions::PyException);
@ -105,7 +143,7 @@ create_exception!(nac3artiq, CompileError, exceptions::PyException);
impl Nac3 {
fn register_module(
&mut self,
module: PyObject,
module: &PyObject,
registered_class_ids: &HashSet<u64>,
) -> PyResult<()> {
let (module_name, source_file) = Python::with_gil(|py| -> PyResult<(String, String)> {
@ -114,18 +152,16 @@ impl Nac3 {
})?;
let source = fs::read_to_string(&source_file).map_err(|e| {
exceptions::PyIOError::new_err(format!("failed to read input file: {}", e))
exceptions::PyIOError::new_err(format!("failed to read input file: {e}"))
})?;
let parser_result = parser::parse_program(&source, source_file.into())
.map_err(|e| exceptions::PySyntaxError::new_err(format!("parse error: {}", e)))?;
let parser_result = parse_program(&source, source_file.into())
.map_err(|e| exceptions::PySyntaxError::new_err(format!("parse error: {e}")))?;
for mut stmt in parser_result.into_iter() {
for mut stmt in parser_result {
let include = match stmt.node {
ast::StmtKind::ClassDef {
ref decorator_list, ref mut body, ref mut bases, ..
} => {
StmtKind::ClassDef { ref decorator_list, ref mut body, ref mut bases, .. } => {
let nac3_class = decorator_list.iter().any(|decorator| {
if let ast::ExprKind::Name { id, .. } = decorator.node {
if let ExprKind::Name { id, .. } = decorator.node {
id.to_string() == "nac3"
} else {
false
@ -139,11 +175,12 @@ impl Nac3 {
Python::with_gil(|py| -> PyResult<bool> {
let id_fn = PyModule::import(py, "builtins")?.getattr("id")?;
match &base.node {
ast::ExprKind::Name { id, .. } => {
ExprKind::Name { id, .. } => {
if *id == "Exception".into() {
Ok(true)
} else {
let base_obj = module.getattr(py, id.to_string())?;
let base_obj =
module.getattr(py, id.to_string().as_str())?;
let base_id = id_fn.call1((base_obj,))?.extract()?;
Ok(registered_class_ids.contains(&base_id))
}
@ -154,9 +191,9 @@ impl Nac3 {
.unwrap()
});
body.retain(|stmt| {
if let ast::StmtKind::FunctionDef { ref decorator_list, .. } = stmt.node {
if let StmtKind::FunctionDef { ref decorator_list, .. } = stmt.node {
decorator_list.iter().any(|decorator| {
if let ast::ExprKind::Name { id, .. } = decorator.node {
if let ExprKind::Name { id, .. } = decorator.node {
id.to_string() == "kernel"
|| id.to_string() == "portable"
|| id.to_string() == "rpc"
@ -170,9 +207,9 @@ impl Nac3 {
});
true
}
ast::StmtKind::FunctionDef { ref decorator_list, .. } => {
StmtKind::FunctionDef { ref decorator_list, .. } => {
decorator_list.iter().any(|decorator| {
if let ast::ExprKind::Name { id, .. } = decorator.node {
if let ExprKind::Name { id, .. } = decorator.node {
let id = id.to_string();
id == "extern" || id == "portable" || id == "kernel" || id == "rpc"
} else {
@ -193,7 +230,7 @@ impl Nac3 {
fn report_modinit(
arg_names: &[String],
method_name: &str,
resolver: Arc<dyn SymbolResolver + Send + Sync>,
resolver: &Arc<dyn SymbolResolver + Send + Sync>,
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
unifier: &mut Unifier,
primitives: &PrimitiveStore,
@ -201,7 +238,7 @@ impl Nac3 {
let base_ty =
match resolver.get_symbol_type(unifier, top_level_defs, primitives, "base".into()) {
Ok(ty) => ty,
Err(e) => return Some(format!("type error inside object launching kernel: {}", e)),
Err(e) => return Some(format!("type error inside object launching kernel: {e}")),
};
let fun_ty = if method_name.is_empty() {
@ -211,8 +248,7 @@ impl Nac3 {
Some(t) => t.0,
None => {
return Some(format!(
"object launching kernel does not have method `{}`",
method_name
"object launching kernel does not have method `{method_name}`"
))
}
}
@ -233,8 +269,7 @@ impl Nac3 {
Some(n) => n,
None if default_value.is_none() => {
return Some(format!(
"argument `{}` not provided when launching kernel function",
name
"argument `{name}` not provided when launching kernel function"
))
}
_ => break,
@ -248,16 +283,14 @@ impl Nac3 {
Ok(t) => t,
Err(e) => {
return Some(format!(
"type error ({}) at parameter #{} when calling kernel function",
e, i
"type error ({e}) at parameter #{i} when calling kernel function"
))
}
};
if let Err(e) = unifier.unify(in_ty, *ty) {
return Some(format!(
"type error ({}) at parameter #{} when calling kernel function",
e.to_display(unifier).to_string(),
i
"type error ({}) at parameter #{i} when calling kernel function",
e.to_display(unifier),
));
}
}
@ -276,9 +309,11 @@ impl Nac3 {
py: Python,
link_fn: &dyn Fn(&Module) -> PyResult<T>,
) -> PyResult<T> {
let size_t = self.isa.get_size_type();
let (mut composer, mut builtins_def, mut builtins_ty) = TopLevelComposer::new(
self.builtins.clone(),
ComposerConfig { kernel_ann: Some("Kernel"), kernel_invariant_ann: "KernelInvariant" },
size_t,
);
let builtins = PyModule::import(py, "builtins")?;
@ -319,15 +354,16 @@ impl Nac3 {
let mut module_to_resolver_cache: HashMap<u64, _> = HashMap::new();
let mut rpc_ids = vec![];
for (stmt, path, module) in self.top_levels.iter() {
for (stmt, path, module) in &self.top_levels {
let py_module: &PyAny = module.extract(py)?;
let module_id: u64 = id_fn.call1((py_module,))?.extract()?;
let helper = helper.clone();
let class_obj;
if let StmtKind::ClassDef { name, .. } = &stmt.node {
let class = py_module.getattr(name.to_string()).unwrap();
if issubclass.call1((class, exn_class)).unwrap().extract().unwrap() &&
class.getattr("artiq_builtin").is_err() {
let class = py_module.getattr(name.to_string().as_str()).unwrap();
if issubclass.call1((class, exn_class)).unwrap().extract().unwrap()
&& class.getattr("artiq_builtin").is_err()
{
class_obj = Some(class);
} else {
class_obj = None;
@ -339,8 +375,8 @@ impl Nac3 {
module_to_resolver_cache.get(&module_id).cloned().unwrap_or_else(|| {
let mut name_to_pyid: HashMap<StrRef, u64> = HashMap::new();
let members: &PyDict =
py_module.getattr("__dict__").unwrap().cast_as().unwrap();
for (key, val) in members.iter() {
py_module.getattr("__dict__").unwrap().downcast().unwrap();
for (key, val) in members {
let key: &str = key.extract().unwrap();
let val = id_fn.call1((val,)).unwrap().extract().unwrap();
name_to_pyid.insert(key.into(), val);
@ -352,12 +388,12 @@ impl Nac3 {
pyid_to_type: pyid_to_type.clone(),
primitive_ids: self.primitive_ids.clone(),
global_value_ids: global_value_ids.clone(),
class_names: Default::default(),
class_names: Mutex::default(),
name_to_pyid: name_to_pyid.clone(),
module: module.clone(),
id_to_pyval: Default::default(),
id_to_primitive: Default::default(),
field_to_val: Default::default(),
id_to_pyval: RwLock::default(),
id_to_primitive: RwLock::default(),
field_to_val: RwLock::default(),
helper,
string_store: self.string_store.clone(),
exception_ids: self.exception_ids.clone(),
@ -371,28 +407,27 @@ impl Nac3 {
});
let (name, def_id, ty) = composer
.register_top_level(stmt.clone(), Some(resolver.clone()), path.clone(), false)
.register_top_level(stmt.clone(), Some(resolver.clone()), path, false)
.map_err(|e| {
CompileError::new_err(format!(
"compilation failed\n----------\n{}",
e
))
CompileError::new_err(format!("compilation failed\n----------\n{e}"))
})?;
if let Some(class_obj) = class_obj {
self.exception_ids.write().insert(def_id.0, store_obj.call1(py, (class_obj, ))?.extract(py)?);
self.exception_ids
.write()
.insert(def_id.0, store_obj.call1(py, (class_obj,))?.extract(py)?);
}
match &stmt.node {
StmtKind::FunctionDef { decorator_list, .. } => {
if decorator_list.iter().any(|decorator| matches!(decorator.node, ExprKind::Name { id, .. } if id == "rpc".into())) {
store_fun.call1(py, (def_id.0.into_py(py), module.getattr(py, name.to_string()).unwrap())).unwrap();
store_fun.call1(py, (def_id.0.into_py(py), module.getattr(py, name.to_string().as_str()).unwrap())).unwrap();
rpc_ids.push((None, def_id));
}
}
StmtKind::ClassDef { name, body, .. } => {
let class_name = name.to_string();
let class_obj = module.getattr(py, &class_name).unwrap();
for stmt in body.iter() {
let class_obj = module.getattr(py, class_name.as_str()).unwrap();
for stmt in body {
if let StmtKind::FunctionDef { name, decorator_list, .. } = &stmt.node {
if decorator_list.iter().any(|decorator| matches!(decorator.node, ExprKind::Name { id, .. } if id == "rpc".into())) {
if name == &"__init__".into() {
@ -426,7 +461,7 @@ impl Nac3 {
name_to_pyid.insert("base".into(), id_fun.call1((obj,))?.extract()?);
let mut arg_names = vec![];
for (i, arg) in args.into_iter().enumerate() {
let name = format!("tmp{}", i);
let name = format!("tmp{i}");
module.add(&name, arg)?;
name_to_pyid.insert(name.clone().into(), id_fun.call1((arg,))?.extract()?);
arg_names.push(name);
@ -445,10 +480,10 @@ impl Nac3 {
pyid_to_type: pyid_to_type.clone(),
primitive_ids: self.primitive_ids.clone(),
global_value_ids: global_value_ids.clone(),
class_names: Default::default(),
id_to_pyval: Default::default(),
id_to_primitive: Default::default(),
field_to_val: Default::default(),
class_names: Mutex::default(),
id_to_pyval: RwLock::default(),
id_to_primitive: RwLock::default(),
field_to_val: RwLock::default(),
name_to_pyid,
module: module.to_object(py),
helper,
@ -456,47 +491,52 @@ impl Nac3 {
exception_ids: self.exception_ids.clone(),
deferred_eval_store: self.deferred_eval_store.clone(),
});
let resolver = Arc::new(Resolver(inner_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
let resolver =
Arc::new(Resolver(inner_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
let (_, def_id, _) = composer
.register_top_level(synthesized.pop().unwrap(), Some(resolver.clone()), "".into(), false)
.register_top_level(synthesized.pop().unwrap(), Some(resolver.clone()), "", false)
.unwrap();
let fun_signature =
FunSignature { args: vec![], ret: self.primitive.none, vars: HashMap::new() };
FunSignature { args: vec![], ret: self.primitive.none, vars: VarMap::new() };
let mut store = ConcreteTypeStore::new();
let mut cache = HashMap::new();
let signature =
store.from_signature(&mut composer.unifier, &self.primitive, &fun_signature, &mut cache);
let signature = store.from_signature(
&mut composer.unifier,
&self.primitive,
&fun_signature,
&mut cache,
);
let signature = store.add_cty(signature);
if let Err(e) = composer.start_analysis(true) {
// report error of __modinit__ separately
if !e.contains("<nac3_synthesized_modinit>") {
return Err(CompileError::new_err(format!(
"compilation failed\n----------\n{}",
e
)));
} else {
return if e.iter().any(|err| err.contains("<nac3_synthesized_modinit>")) {
let msg = Self::report_modinit(
&arg_names,
method_name,
resolver.clone(),
&resolver,
&composer.extract_def_list(),
&mut composer.unifier,
&self.primitive,
);
return Err(CompileError::new_err(format!(
Err(CompileError::new_err(format!(
"compilation failed\n----------\n{}",
msg.unwrap_or(e)
)));
}
msg.unwrap_or(e.iter().sorted().join("\n----------\n"))
)))
} else {
Err(CompileError::new_err(format!(
"compilation failed\n----------\n{}",
e.iter().sorted().join("\n----------\n"),
)))
};
}
let top_level = Arc::new(composer.make_top_level_context());
{
let rpc_codegen = rpc_codegen_callback();
let defs = top_level.definitions.read();
for (class_data, id) in rpc_ids.iter() {
for (class_data, id) in &rpc_ids {
let mut def = defs[id.0].write();
match &mut *def {
TopLevelDef::Function { codegen_callback, .. } => {
@ -504,7 +544,7 @@ impl Nac3 {
}
TopLevelDef::Class { methods, .. } => {
let (class_def, method_name) = class_data.as_ref().unwrap();
for (name, _, id) in methods.iter() {
for (name, _, id) in &*methods {
if name != method_name {
continue;
}
@ -517,7 +557,9 @@ impl Nac3 {
py,
(
id.0.into_py(py),
class_def.getattr(py, name.to_string()).unwrap(),
class_def
.getattr(py, name.to_string().as_str())
.unwrap(),
),
)
.unwrap();
@ -531,18 +573,18 @@ impl Nac3 {
let instance = {
let defs = top_level.definitions.read();
let mut definition = defs[def_id.0].write();
if let TopLevelDef::Function { instance_to_stmt, instance_to_symbol, .. } =
let TopLevelDef::Function { instance_to_stmt, instance_to_symbol, .. } =
&mut *definition
{
instance_to_symbol.insert("".to_string(), "__modinit__".into());
instance_to_stmt[""].clone()
} else {
else {
unreachable!()
}
};
instance_to_symbol.insert(String::new(), "__modinit__".into());
instance_to_stmt[""].clone()
};
let task = CodeGenTask {
subst: Default::default(),
subst: Vec::default(),
symbol_name: "__modinit__".to_string(),
body: instance.body,
signature,
@ -555,22 +597,26 @@ impl Nac3 {
let mut store = ConcreteTypeStore::new();
let mut cache = HashMap::new();
let signature =
store.from_signature(&mut composer.unifier, &self.primitive, &fun_signature, &mut cache);
let signature = store.from_signature(
&mut composer.unifier,
&self.primitive,
&fun_signature,
&mut cache,
);
let signature = store.add_cty(signature);
let attributes_writeback_task = CodeGenTask {
subst: Default::default(),
subst: Vec::default(),
symbol_name: "attributes_writeback".to_string(),
body: Arc::new(Default::default()),
body: Arc::new(Vec::default()),
signature,
resolver,
store,
unifier_index: instance.unifier_id,
calls: Arc::new(Default::default()),
calls: Arc::new(HashMap::default()),
id: 0,
};
let membuffers: Arc<Mutex<Vec<Vec<u8>>>> = Default::default();
let membuffers: Arc<Mutex<Vec<Vec<u8>>>> = Arc::default();
let membuffer = membuffers.clone();
@ -580,7 +626,8 @@ impl Nac3 {
membuffer.lock().push(buffer);
})));
let size_t = if self.isa == Isa::Host { 64 } else { 32 };
let thread_names: Vec<String> = (0..4).map(|_| "main".to_string()).collect();
let num_threads = if is_multithreaded() { 4 } else { 1 };
let thread_names: Vec<String> = (0..num_threads).map(|_| "main".to_string()).collect();
let threads: Vec<_> = thread_names
.iter()
.map(|s| Box::new(ArtiqCodeGenerator::new(s.to_string(), size_t, self.time_fns)))
@ -588,18 +635,28 @@ impl Nac3 {
let membuffer = membuffers.clone();
py.allow_threads(|| {
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level.clone(), f);
let (registry, handles) =
WorkerRegistry::create_workers(threads, top_level.clone(), &self.llvm_options, &f);
registry.add_task(task);
registry.wait_tasks_complete(handles);
let mut generator = ArtiqCodeGenerator::new("attributes_writeback".to_string(), size_t, self.time_fns);
let mut generator =
ArtiqCodeGenerator::new("attributes_writeback".to_string(), size_t, self.time_fns);
let context = inkwell::context::Context::create();
let module = context.create_module("attributes_writeback");
let builder = context.create_builder();
let (_, module, _) = gen_func_impl(&context, &mut generator, &registry, builder, module,
attributes_writeback_task, |generator, ctx| {
attributes_writeback(ctx, generator, inner_resolver.as_ref(), host_attributes)
}).unwrap();
let (_, module, _) = gen_func_impl(
&context,
&mut generator,
&registry,
builder,
module,
attributes_writeback_task,
|generator, ctx| {
attributes_writeback(ctx, generator, inner_resolver.as_ref(), &host_attributes)
},
)
.unwrap();
let buffer = module.write_bitcode_to_memory();
let buffer = buffer.as_slice().into();
membuffer.lock().push(buffer);
@ -615,13 +672,24 @@ impl Nac3 {
.create_module_from_ir(MemoryBuffer::create_from_memory_range(buffer, "main"))
.unwrap();
main.link_in_module(other)
.map_err(|err| CompileError::new_err(err.to_string()))?;
main.link_in_module(other).map_err(|err| CompileError::new_err(err.to_string()))?;
}
let builder = context.create_builder();
let modinit_return = main.get_function("__modinit__").unwrap().get_last_basic_block().unwrap().get_terminator().unwrap();
let modinit_return = main
.get_function("__modinit__")
.unwrap()
.get_last_basic_block()
.unwrap()
.get_terminator()
.unwrap();
builder.position_before(&modinit_return);
builder.build_call(main.get_function("attributes_writeback").unwrap(), &[], "attributes_writeback");
builder
.build_call(
main.get_function("attributes_writeback").unwrap(),
&[],
"attributes_writeback",
)
.unwrap();
main.link_in_module(load_irrt(&context))
.map_err(|err| CompileError::new_err(err.to_string()))?;
@ -629,16 +697,13 @@ impl Nac3 {
let mut function_iter = main.get_first_function();
while let Some(func) = function_iter {
if func.count_basic_blocks() > 0 && func.get_name().to_str().unwrap() != "__modinit__" {
func.set_linkage(inkwell::module::Linkage::Private);
func.set_linkage(Linkage::Private);
}
function_iter = func.get_next_function();
}
// Demote all global variables that will not be referenced in the kernel to private
let preserved_symbols: Vec<&'static [u8]> = vec![
b"typeinfo",
b"now",
];
let preserved_symbols: Vec<&'static [u8]> = vec![b"typeinfo", b"now"];
let mut global_option = main.get_first_global();
while let Some(global) = global_option {
if !preserved_symbols.contains(&(global.get_name().to_bytes())) {
@ -647,52 +712,73 @@ impl Nac3 {
global_option = global.get_next_global();
}
let builder = PassManagerBuilder::create();
builder.set_optimization_level(OptimizationLevel::Aggressive);
let passes = PassManager::create(());
builder.set_inliner_with_threshold(255);
builder.populate_module_pass_manager(&passes);
passes.run_on(&main);
let target_machine = self
.llvm_options
.target
.create_target_machine(self.llvm_options.opt_level)
.expect("couldn't create target machine");
let pass_options = PassBuilderOptions::create();
pass_options.set_merge_functions(true);
let passes = format!("default<O{}>", self.llvm_options.opt_level as u32);
let result = main.run_passes(passes.as_str(), &target_machine, pass_options);
if let Err(err) = result {
panic!("Failed to run optimization for module `main`: {}", err.to_string());
}
link_fn(&main)
}
fn get_llvm_target_machine(
&self,
) -> TargetMachine {
let (triple, features) = match self.isa {
Isa::Host => (
TargetMachine::get_default_triple(),
TargetMachine::get_host_cpu_features().to_string(),
),
Isa::RiscV32G => {
(TargetTriple::create("riscv32-unknown-linux"), "+a,+m,+f,+d".to_string())
/// Returns the [`TargetTriple`] used for compiling to [isa].
fn get_llvm_target_triple(isa: Isa) -> TargetTriple {
match isa {
Isa::Host => TargetMachine::get_default_triple(),
Isa::RiscV32G | Isa::RiscV32IMA => TargetTriple::create("riscv32-unknown-linux"),
Isa::CortexA9 => TargetTriple::create("armv7-unknown-linux-gnueabihf"),
}
Isa::RiscV32IMA => (TargetTriple::create("riscv32-unknown-linux"), "+a,+m".to_string()),
Isa::CortexA9 => (
TargetTriple::create("armv7-unknown-linux-gnueabihf"),
"+dsp,+fp16,+neon,+vfp3,+long-calls".to_string(),
),
};
let target =
Target::from_triple(&triple).expect("couldn't create target from target triple");
target
.create_target_machine(
&triple,
"",
&features,
OptimizationLevel::Default,
RelocMode::PIC,
CodeModel::Default,
)
}
/// Returns the [`String`] representing the target CPU used for compiling to [isa].
fn get_llvm_target_cpu(isa: Isa) -> String {
match isa {
Isa::Host => TargetMachine::get_host_cpu_name().to_string(),
Isa::RiscV32G | Isa::RiscV32IMA => "generic-rv32".to_string(),
Isa::CortexA9 => "cortex-a9".to_string(),
}
}
/// Returns the [`String`] representing the target features used for compiling to [isa].
fn get_llvm_target_features(isa: Isa) -> String {
match isa {
Isa::Host => TargetMachine::get_host_cpu_features().to_string(),
Isa::RiscV32G => "+a,+m,+f,+d".to_string(),
Isa::RiscV32IMA => "+a,+m".to_string(),
Isa::CortexA9 => "+dsp,+fp16,+neon,+vfp3,+long-calls".to_string(),
}
}
/// Returns an instance of [`CodeGenTargetMachineOptions`] representing the target machine
/// options used for compiling to [isa].
fn get_llvm_target_options(isa: Isa) -> CodeGenTargetMachineOptions {
CodeGenTargetMachineOptions {
triple: Nac3::get_llvm_target_triple(isa).as_str().to_string_lossy().into_owned(),
cpu: Nac3::get_llvm_target_cpu(isa),
features: Nac3::get_llvm_target_features(isa),
reloc_mode: RelocMode::PIC,
..CodeGenTargetMachineOptions::from_host()
}
}
/// Returns an instance of [`TargetMachine`] used in compiling and linking of a program to the
/// target [isa].
fn get_llvm_target_machine(&self) -> TargetMachine {
Nac3::get_llvm_target_options(self.isa)
.create_target_machine(self.llvm_options.opt_level)
.expect("couldn't create target machine")
}
}
fn link_with_lld(
elf_filename: String,
obj_filename: String,
) -> PyResult<()>{
fn link_with_lld(elf_filename: String, obj_filename: String) -> PyResult<()> {
let linker_args = vec![
"-shared".to_string(),
"--eh-frame-hdr".to_string(),
@ -711,9 +797,7 @@ fn link_with_lld(
return Err(CompileError::new_err("failed to start linker"));
}
} else {
return Err(CompileError::new_err(
"linker returned non-zero status code",
));
return Err(CompileError::new_err("linker returned non-zero status code"));
}
Ok(())
@ -723,7 +807,7 @@ fn add_exceptions(
composer: &mut TopLevelComposer,
builtin_def: &mut HashMap<StrRef, DefinitionId>,
builtin_ty: &mut HashMap<StrRef, Type>,
error_names: &[&str]
error_names: &[&str],
) -> Vec<Type> {
let mut types = Vec::new();
// note: this is only for builtin exceptions, i.e. the exception name is "0:{exn}"
@ -736,7 +820,7 @@ fn add_exceptions(
// constructor id
def_id + 1,
&mut composer.unifier,
&composer.primitives_ty
&composer.primitives_ty,
);
composer.definition_ast_list.push((Arc::new(RwLock::new(exception_class)), None));
composer.definition_ast_list.push((Arc::new(RwLock::new(exception_fn)), None));
@ -750,7 +834,7 @@ fn add_exceptions(
#[pymethods]
impl Nac3 {
#[new]
fn new(isa: &str, py: Python) -> PyResult<Self> {
fn new(isa: &str, artiq_builtins: &PyDict, py: Python) -> PyResult<Self> {
let isa = match isa {
"host" => Isa::Host,
"rv32g" => Isa::RiscV32G,
@ -759,16 +843,15 @@ impl Nac3 {
_ => return Err(exceptions::PyValueError::new_err("invalid ISA")),
};
let time_fns: &(dyn TimeFns + Sync) = match isa {
Isa::Host => &timeline::EXTERN_TIME_FNS,
Isa::RiscV32G => &timeline::NOW_PINNING_TIME_FNS_64,
Isa::RiscV32IMA => &timeline::NOW_PINNING_TIME_FNS,
Isa::CortexA9 => &timeline::EXTERN_TIME_FNS,
Isa::CortexA9 | Isa::Host => &timeline::EXTERN_TIME_FNS,
};
let primitive: PrimitiveStore = TopLevelComposer::make_primitives().0;
let primitive: PrimitiveStore = TopLevelComposer::make_primitives(isa.get_size_type()).0;
let builtins = vec![
(
"now_mu".into(),
FunSignature { args: vec![], ret: primitive.int64, vars: HashMap::new() },
FunSignature { args: vec![], ret: primitive.int64, vars: VarMap::new() },
Arc::new(GenCall::new(Box::new(move |ctx, _, _, _, _| {
Ok(Some(time_fns.emit_now_mu(ctx)))
}))),
@ -782,11 +865,12 @@ impl Nac3 {
default_value: None,
}],
ret: primitive.none,
vars: HashMap::new(),
vars: VarMap::new(),
},
Arc::new(GenCall::new(Box::new(move |ctx, _, fun, args, generator| {
let arg_ty = fun.0.args[0].ty;
let arg = args[0].1.clone().to_basic_value_enum(ctx, generator, arg_ty).unwrap();
let arg =
args[0].1.clone().to_basic_value_enum(ctx, generator, arg_ty).unwrap();
time_fns.emit_at_mu(ctx, arg);
Ok(None)
}))),
@ -800,11 +884,12 @@ impl Nac3 {
default_value: None,
}],
ret: primitive.none,
vars: HashMap::new(),
vars: VarMap::new(),
},
Arc::new(GenCall::new(Box::new(move |ctx, _, fun, args, generator| {
let arg_ty = fun.0.args[0].ty;
let arg = args[0].1.clone().to_basic_value_enum(ctx, generator, arg_ty).unwrap();
let arg =
args[0].1.clone().to_basic_value_enum(ctx, generator, arg_ty).unwrap();
time_fns.emit_delay_mu(ctx, arg);
Ok(None)
}))),
@ -817,57 +902,37 @@ impl Nac3 {
let typing_mod = PyModule::import(py, "typing").unwrap();
let types_mod = PyModule::import(py, "types").unwrap();
let get_id = |x| id_fn.call1((x,)).unwrap().extract().unwrap();
let get_attr_id = |obj: &PyModule, attr| id_fn.call1((obj.getattr(attr).unwrap(),))
.unwrap().extract().unwrap();
let get_id = |x: &PyAny| id_fn.call1((x,)).and_then(PyAny::extract).unwrap();
let get_attr_id = |obj: &PyModule, attr| {
id_fn.call1((obj.getattr(attr).unwrap(),)).unwrap().extract().unwrap()
};
let primitive_ids = PrimitivePythonId {
virtual_id: get_id(
builtins_mod
.getattr("globals")
.unwrap()
.call0()
.unwrap()
.get_item("virtual")
.unwrap(
)),
virtual_id: get_id(artiq_builtins.get_item("virtual").ok().flatten().unwrap()),
generic_alias: (
get_attr_id(typing_mod, "_GenericAlias"),
get_attr_id(types_mod, "GenericAlias"),
),
none: id_fn
.call1((builtins_mod
.getattr("globals")
.unwrap()
.call0()
.unwrap()
.get_item("none")
.unwrap(),))
.unwrap()
.extract()
.unwrap(),
none: get_id(artiq_builtins.get_item("none").ok().flatten().unwrap()),
typevar: get_attr_id(typing_mod, "TypeVar"),
const_generic_marker: get_id(
artiq_builtins.get_item("_ConstGenericMarker").ok().flatten().unwrap(),
),
int: get_attr_id(builtins_mod, "int"),
int32: get_attr_id(numpy_mod, "int32"),
int64: get_attr_id(numpy_mod, "int64"),
uint32: get_attr_id(numpy_mod, "uint32"),
uint64: get_attr_id(numpy_mod, "uint64"),
bool: get_attr_id(builtins_mod, "bool"),
np_bool_: get_attr_id(numpy_mod, "bool_"),
string: get_attr_id(builtins_mod, "str"),
np_str_: get_attr_id(numpy_mod, "str_"),
float: get_attr_id(builtins_mod, "float"),
float64: get_attr_id(numpy_mod, "float64"),
list: get_attr_id(builtins_mod, "list"),
ndarray: get_attr_id(numpy_mod, "ndarray"),
tuple: get_attr_id(builtins_mod, "tuple"),
exception: get_attr_id(builtins_mod, "Exception"),
option: id_fn
.call1((builtins_mod
.getattr("globals")
.unwrap()
.call0()
.unwrap()
.get_item("Option")
.unwrap(),))
.unwrap()
.extract()
.unwrap(),
option: get_id(artiq_builtins.get_item("Option").ok().flatten().unwrap()),
};
let working_directory = tempfile::Builder::new().prefix("nac3-").tempdir().unwrap();
@ -879,12 +944,16 @@ impl Nac3 {
primitive,
builtins,
primitive_ids,
top_levels: Default::default(),
pyid_to_def: Default::default(),
top_levels: Vec::default(),
pyid_to_def: Arc::default(),
working_directory,
string_store: Default::default(),
exception_ids: Default::default(),
string_store: Arc::default(),
exception_ids: Arc::default(),
deferred_eval_store: DeferredEvaluationStore::new(),
llvm_options: CodeGenLLVMOptions {
opt_level: OptimizationLevel::Default,
target: Nac3::get_llvm_target_options(isa),
},
})
}
@ -897,11 +966,11 @@ impl Nac3 {
let id_fn = PyModule::import(py, "builtins")?.getattr("id")?;
let getmodule_fn = PyModule::import(py, "inspect")?.getattr("getmodule")?;
for function in functions.iter() {
for function in functions {
let module = getmodule_fn.call1((function,))?.extract()?;
modules.insert(id_fn.call1((&module,))?.extract()?, module);
}
for class in classes.iter() {
for class in classes {
let module = getmodule_fn.call1((class,))?.extract()?;
modules.insert(id_fn.call1((&module,))?.extract()?, module);
class_ids.insert(id_fn.call1((class,))?.extract()?);
@ -910,7 +979,7 @@ impl Nac3 {
})?;
for module in modules.into_values() {
self.register_module(module, &class_ids)?;
self.register_module(&module, &class_ids)?;
}
Ok(())
}
@ -934,7 +1003,7 @@ impl Nac3 {
.expect("couldn't write module to file");
link_with_lld(
filename.to_string(),
working_directory.join("module.o").to_string_lossy().to_string()
working_directory.join("module.o").to_string_lossy().to_string(),
)?;
Ok(())
};
@ -975,14 +1044,14 @@ impl Nac3 {
let link_fn = |module: &Module| {
let working_directory = self.working_directory.path().to_owned();
target_machine
.write_to_file(&module, FileType::Object, &working_directory.join("module.o"))
.write_to_file(module, FileType::Object, &working_directory.join("module.o"))
.expect("couldn't write module to file");
let filename_path = self.working_directory.path().join("module.elf");
let filename = filename_path.to_str().unwrap();
link_with_lld(
filename.to_string(),
working_directory.join("module.o").to_string_lossy().to_string()
working_directory.join("module.o").to_string_lossy().to_string(),
)?;
Ok(PyBytes::new(py, &fs::read(filename).unwrap()).into())
@ -992,7 +1061,7 @@ impl Nac3 {
} else {
let link_fn = |module: &Module| {
let object_mem = target_machine
.write_to_memory_buffer(&module, FileType::Object)
.write_to_memory_buffer(module, FileType::Object)
.expect("couldn't write module to object file buffer");
if let Ok(dyn_lib) = Linker::ld(object_mem.as_slice()) {
Ok(PyBytes::new(py, &dyn_lib).into())

File diff suppressed because it is too large Load Diff

View File

@ -1,10 +1,20 @@
use inkwell::{values::BasicValueEnum, AddressSpace, AtomicOrdering};
use inkwell::{
values::{BasicValueEnum, CallSiteValue},
AddressSpace, AtomicOrdering,
};
use itertools::Either;
use nac3core::codegen::CodeGenContext;
/// Functions for manipulating the timeline.
pub trait TimeFns {
fn emit_now_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>) -> BasicValueEnum<'ctx>;
fn emit_at_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, t: BasicValueEnum<'ctx>);
fn emit_delay_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, dt: BasicValueEnum<'ctx>);
/// Emits LLVM IR for `now_mu`.
fn emit_now_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> BasicValueEnum<'ctx>;
/// Emits LLVM IR for `at_mu`.
fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>);
/// Emits LLVM IR for `delay_mu`.
fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>);
}
pub struct NowPinningTimeFns64 {}
@ -12,141 +22,143 @@ pub struct NowPinningTimeFns64 {}
// For FPGA design reasons, on VexRiscv with 64-bit data bus, the "now" CSR is split into two 32-bit
// values that are each padded to 64-bits.
impl TimeFns for NowPinningTimeFns64 {
fn emit_now_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>) -> BasicValueEnum<'ctx> {
fn emit_now_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> BasicValueEnum<'ctx> {
let i64_type = ctx.ctx.i64_type();
let i32_type = ctx.ctx.i32_type();
let now = ctx
.module
.get_global("now")
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr =
ctx.builder.build_bitcast(now, i32_type.ptr_type(AddressSpace::Generic), "now_hiptr");
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_hiptr = ctx
.builder
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now_gep")
};
if let (BasicValueEnum::IntValue(now_hi), BasicValueEnum::IntValue(now_lo)) = (
ctx.builder.build_load(now_hiptr, "now_hi"),
ctx.builder.build_load(now_loptr, "now_lo"),
) {
let zext_hi = ctx.builder.build_int_z_extend(now_hi, i64_type, "now_zext_hi");
let shifted_hi = ctx.builder.build_left_shift(
zext_hi,
i64_type.const_int(32, false),
"now_shifted_zext_hi",
);
let zext_lo = ctx.builder.build_int_z_extend(now_lo, i64_type, "now_zext_lo");
ctx.builder.build_or(shifted_hi, zext_lo, "now_or").into()
} else {
unreachable!();
}
} else {
unreachable!();
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
}
.unwrap();
let now_hi = ctx
.builder
.build_load(now_hiptr, "now.hi")
.map(BasicValueEnum::into_int_value)
.unwrap();
let now_lo = ctx
.builder
.build_load(now_loptr, "now.lo")
.map(BasicValueEnum::into_int_value)
.unwrap();
let zext_hi = ctx.builder.build_int_z_extend(now_hi, i64_type, "").unwrap();
let shifted_hi =
ctx.builder.build_left_shift(zext_hi, i64_type.const_int(32, false), "").unwrap();
let zext_lo = ctx.builder.build_int_z_extend(now_lo, i64_type, "").unwrap();
ctx.builder.build_or(shifted_hi, zext_lo, "now_mu").map(Into::into).unwrap()
}
fn emit_at_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, t: BasicValueEnum<'ctx>) {
fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>) {
let i32_type = ctx.ctx.i32_type();
let i64_type = ctx.ctx.i64_type();
let i64_32 = i64_type.const_int(32, false);
if let BasicValueEnum::IntValue(time) = t {
let time_hi = ctx.builder.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "now_lshr"),
let time = t.into_int_value();
let time_hi = ctx
.builder
.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "time.hi").unwrap(),
i32_type,
"now_trunc",
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
"",
)
.unwrap();
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
let now = ctx
.module
.get_global("now")
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr = ctx.builder.build_bitcast(
now,
i32_type.ptr_type(AddressSpace::Generic),
"now_bitcast",
);
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_hiptr = ctx
.builder
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now_gep")
};
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
}
.unwrap();
ctx.builder
.build_store(now_hiptr, time_hi)
.unwrap()
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
ctx.builder
.build_store(now_loptr, time_lo)
.unwrap()
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
} else {
unreachable!();
}
} else {
unreachable!();
}
}
fn emit_delay_mu<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
dt: BasicValueEnum<'ctx>,
) {
fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>) {
let i64_type = ctx.ctx.i64_type();
let i32_type = ctx.ctx.i32_type();
let now = ctx
.module
.get_global("now")
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr =
ctx.builder.build_bitcast(now, i32_type.ptr_type(AddressSpace::Generic), "now_hiptr");
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now_loptr")
};
if let (
BasicValueEnum::IntValue(now_hi),
BasicValueEnum::IntValue(now_lo),
BasicValueEnum::IntValue(dt),
) = (
ctx.builder.build_load(now_hiptr, "now_hi"),
ctx.builder.build_load(now_loptr, "now_lo"),
dt,
) {
let zext_hi = ctx.builder.build_int_z_extend(now_hi, i64_type, "now_zext_hi");
let shifted_hi = ctx.builder.build_left_shift(
zext_hi,
i64_type.const_int(32, false),
"now_shifted_zext_hi",
);
let zext_lo = ctx.builder.build_int_z_extend(now_lo, i64_type, "now_zext_lo");
let now_val = ctx.builder.build_or(shifted_hi, zext_lo, "now_or");
let now_hiptr = ctx
.builder
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let time = ctx.builder.build_int_add(now_val, dt, "now_add");
let time_hi = ctx.builder.build_int_truncate(
ctx.builder.build_right_shift(
time,
i64_type.const_int(32, false),
false,
"now_lshr",
),
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
}
.unwrap();
let now_hi = ctx
.builder
.build_load(now_hiptr, "now.hi")
.map(BasicValueEnum::into_int_value)
.unwrap();
let now_lo = ctx
.builder
.build_load(now_loptr, "now.lo")
.map(BasicValueEnum::into_int_value)
.unwrap();
let dt = dt.into_int_value();
let zext_hi = ctx.builder.build_int_z_extend(now_hi, i64_type, "").unwrap();
let shifted_hi =
ctx.builder.build_left_shift(zext_hi, i64_type.const_int(32, false), "").unwrap();
let zext_lo = ctx.builder.build_int_z_extend(now_lo, i64_type, "").unwrap();
let now_val = ctx.builder.build_or(shifted_hi, zext_lo, "now").unwrap();
let time = ctx.builder.build_int_add(now_val, dt, "time").unwrap();
let time_hi = ctx
.builder
.build_int_truncate(
ctx.builder
.build_right_shift(time, i64_type.const_int(32, false), false, "")
.unwrap(),
i32_type,
"now_trunc",
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
"time.hi",
)
.unwrap();
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
ctx.builder
.build_store(now_hiptr, time_hi)
.unwrap()
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
ctx.builder
.build_store(now_loptr, time_lo)
.unwrap()
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
} else {
unreachable!();
}
} else {
unreachable!();
};
}
}
@ -155,68 +167,67 @@ pub static NOW_PINNING_TIME_FNS_64: NowPinningTimeFns64 = NowPinningTimeFns64 {}
pub struct NowPinningTimeFns {}
impl TimeFns for NowPinningTimeFns {
fn emit_now_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>) -> BasicValueEnum<'ctx> {
fn emit_now_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> BasicValueEnum<'ctx> {
let i64_type = ctx.ctx.i64_type();
let now = ctx
.module
.get_global("now")
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_raw = ctx.builder.build_load(now.as_pointer_value(), "now");
if let BasicValueEnum::IntValue(now_raw) = now_raw {
let now_raw = ctx
.builder
.build_load(now.as_pointer_value(), "now")
.map(BasicValueEnum::into_int_value)
.unwrap();
let i64_32 = i64_type.const_int(32, false);
let now_lo = ctx.builder.build_left_shift(now_raw, i64_32, "now_shl");
let now_hi = ctx.builder.build_right_shift(now_raw, i64_32, false, "now_lshr");
ctx.builder.build_or(now_lo, now_hi, "now_or").into()
} else {
unreachable!();
}
let now_lo = ctx.builder.build_left_shift(now_raw, i64_32, "now.lo").unwrap();
let now_hi = ctx.builder.build_right_shift(now_raw, i64_32, false, "now.hi").unwrap();
ctx.builder.build_or(now_lo, now_hi, "now_mu").map(Into::into).unwrap()
}
fn emit_at_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, t: BasicValueEnum<'ctx>) {
fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>) {
let i32_type = ctx.ctx.i32_type();
let i64_type = ctx.ctx.i64_type();
let i64_32 = i64_type.const_int(32, false);
if let BasicValueEnum::IntValue(time) = t {
let time_hi = ctx.builder.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "now_lshr"),
let time = t.into_int_value();
let time_hi = ctx
.builder
.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "").unwrap(),
i32_type,
"now_trunc",
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
"time.hi",
)
.unwrap();
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc").unwrap();
let now = ctx
.module
.get_global("now")
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr = ctx.builder.build_bitcast(
now,
i32_type.ptr_type(AddressSpace::Generic),
"now_bitcast",
);
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_hiptr = ctx
.builder
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now_gep")
};
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now.lo.addr")
}
.unwrap();
ctx.builder
.build_store(now_hiptr, time_hi)
.unwrap()
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
ctx.builder
.build_store(now_loptr, time_lo)
.unwrap()
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
} else {
unreachable!();
}
} else {
unreachable!();
}
}
fn emit_delay_mu<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
dt: BasicValueEnum<'ctx>,
) {
fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>) {
let i32_type = ctx.ctx.i32_type();
let i64_type = ctx.ctx.i64_type();
let i64_32 = i64_type.const_int(32, false);
@ -224,41 +235,47 @@ impl TimeFns for NowPinningTimeFns {
.module
.get_global("now")
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_raw = ctx.builder.build_load(now.as_pointer_value(), "now");
if let (BasicValueEnum::IntValue(now_raw), BasicValueEnum::IntValue(dt)) = (now_raw, dt) {
let now_lo = ctx.builder.build_left_shift(now_raw, i64_32, "now_shl");
let now_hi = ctx.builder.build_right_shift(now_raw, i64_32, false, "now_lshr");
let now_val = ctx.builder.build_or(now_lo, now_hi, "now_or");
let time = ctx.builder.build_int_add(now_val, dt, "now_add");
let time_hi = ctx.builder.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "now_lshr"),
let now_raw = ctx
.builder
.build_load(now.as_pointer_value(), "")
.map(BasicValueEnum::into_int_value)
.unwrap();
let dt = dt.into_int_value();
let now_lo = ctx.builder.build_left_shift(now_raw, i64_32, "now.lo").unwrap();
let now_hi = ctx.builder.build_right_shift(now_raw, i64_32, false, "now.hi").unwrap();
let now_val = ctx.builder.build_or(now_lo, now_hi, "now_val").unwrap();
let time = ctx.builder.build_int_add(now_val, dt, "time").unwrap();
let time_hi = ctx
.builder
.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "time.hi").unwrap(),
i32_type,
"now_trunc",
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
let now_hiptr = ctx.builder.build_bitcast(
now,
i32_type.ptr_type(AddressSpace::Generic),
"now_bitcast",
);
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
)
.unwrap();
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
let now_hiptr = ctx
.builder
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now_gep")
};
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now.lo.addr")
}
.unwrap();
ctx.builder
.build_store(now_hiptr, time_hi)
.unwrap()
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
ctx.builder
.build_store(now_loptr, time_lo)
.unwrap()
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
} else {
unreachable!();
}
} else {
unreachable!();
}
}
}
@ -267,14 +284,18 @@ pub static NOW_PINNING_TIME_FNS: NowPinningTimeFns = NowPinningTimeFns {};
pub struct ExternTimeFns {}
impl TimeFns for ExternTimeFns {
fn emit_now_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>) -> BasicValueEnum<'ctx> {
fn emit_now_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> BasicValueEnum<'ctx> {
let now_mu = ctx.module.get_function("now_mu").unwrap_or_else(|| {
ctx.module.add_function("now_mu", ctx.ctx.i64_type().fn_type(&[], false), None)
});
ctx.builder.build_call(now_mu, &[], "now_mu").try_as_basic_value().left().unwrap()
ctx.builder
.build_call(now_mu, &[], "now_mu")
.map(CallSiteValue::try_as_basic_value)
.map(Either::unwrap_left)
.unwrap()
}
fn emit_at_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, t: BasicValueEnum<'ctx>) {
fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>) {
let at_mu = ctx.module.get_function("at_mu").unwrap_or_else(|| {
ctx.module.add_function(
"at_mu",
@ -282,14 +303,10 @@ impl TimeFns for ExternTimeFns {
None,
)
});
ctx.builder.build_call(at_mu, &[t.into()], "at_mu");
ctx.builder.build_call(at_mu, &[t.into()], "at_mu").unwrap();
}
fn emit_delay_mu<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
dt: BasicValueEnum<'ctx>,
) {
fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>) {
let delay_mu = ctx.module.get_function("delay_mu").unwrap_or_else(|| {
ctx.module.add_function(
"delay_mu",
@ -297,7 +314,7 @@ impl TimeFns for ExternTimeFns {
None,
)
});
ctx.builder.build_call(delay_mu, &[dt.into()], "delay_mu");
ctx.builder.build_call(delay_mu, &[dt.into()], "delay_mu").unwrap();
}
}

View File

@ -2,7 +2,7 @@
name = "nac3ast"
version = "0.1.0"
authors = ["RustPython Team", "M-Labs"]
edition = "2018"
edition = "2021"
[features]
default = ["constant-optimization", "fold"]
@ -10,7 +10,7 @@ constant-optimization = ["fold"]
fold = []
[dependencies]
lazy_static = "1.4"
lazy_static = "1.5"
parking_lot = "0.12"
string-interner = "0.14"
string-interner = "0.17"
fxhash = "0.2"

File diff suppressed because it is too large Load Diff

View File

@ -28,12 +28,12 @@ impl From<bool> for Constant {
}
impl From<i32> for Constant {
fn from(i: i32) -> Constant {
Self::Int(i as i128)
Self::Int(i128::from(i))
}
}
impl From<i64> for Constant {
fn from(i: i64) -> Constant {
Self::Int(i as i128)
Self::Int(i128::from(i))
}
}
@ -50,6 +50,7 @@ pub enum ConversionFlag {
}
impl ConversionFlag {
#[must_use]
pub fn try_from_byte(b: u8) -> Option<Self> {
match b {
b's' => Some(Self::Str),
@ -69,6 +70,7 @@ pub struct ConstantOptimizer {
#[cfg(feature = "constant-optimization")]
impl ConstantOptimizer {
#[inline]
#[must_use]
pub fn new() -> Self {
Self { _priv: () }
}
@ -85,14 +87,10 @@ impl<U> crate::fold::Fold<U> for ConstantOptimizer {
fn fold_expr(&mut self, node: crate::Expr<U>) -> Result<crate::Expr<U>, Self::Error> {
match node.node {
crate::ExprKind::Tuple { elts, ctx } => {
let elts = elts
.into_iter()
.map(|x| self.fold_expr(x))
.collect::<Result<Vec<_>, _>>()?;
let expr = if elts
.iter()
.all(|e| matches!(e.node, crate::ExprKind::Constant { .. }))
{
let elts =
elts.into_iter().map(|x| self.fold_expr(x)).collect::<Result<Vec<_>, _>>()?;
let expr =
if elts.iter().all(|e| matches!(e.node, crate::ExprKind::Constant { .. })) {
let tuple = elts
.into_iter()
.map(|e| match e.node {
@ -100,18 +98,11 @@ impl<U> crate::fold::Fold<U> for ConstantOptimizer {
_ => unreachable!(),
})
.collect();
crate::ExprKind::Constant {
value: Constant::Tuple(tuple),
kind: None,
}
crate::ExprKind::Constant { value: Constant::Tuple(tuple), kind: None }
} else {
crate::ExprKind::Tuple { elts, ctx }
};
Ok(crate::Expr {
node: expr,
custom: node.custom,
location: node.location,
})
Ok(crate::Expr { node: expr, custom: node.custom, location: node.location })
}
_ => crate::fold::fold_expr(self, node),
}
@ -127,7 +118,7 @@ mod tests {
use crate::fold::Fold;
use crate::*;
let location = Location::new(0, 0, Default::default());
let location = Location::new(0, 0, FileName::default());
let custom = ();
let ast = Located {
location,
@ -138,18 +129,12 @@ mod tests {
Located {
location,
custom,
node: ExprKind::Constant {
value: 1.into(),
kind: None,
},
node: ExprKind::Constant { value: 1.into(), kind: None },
},
Located {
location,
custom,
node: ExprKind::Constant {
value: 2.into(),
kind: None,
},
node: ExprKind::Constant { value: 2.into(), kind: None },
},
Located {
location,
@ -160,26 +145,17 @@ mod tests {
Located {
location,
custom,
node: ExprKind::Constant {
value: 3.into(),
kind: None,
},
node: ExprKind::Constant { value: 3.into(), kind: None },
},
Located {
location,
custom,
node: ExprKind::Constant {
value: 4.into(),
kind: None,
},
node: ExprKind::Constant { value: 4.into(), kind: None },
},
Located {
location,
custom,
node: ExprKind::Constant {
value: 5.into(),
kind: None,
},
node: ExprKind::Constant { value: 5.into(), kind: None },
},
],
},
@ -187,9 +163,7 @@ mod tests {
],
},
};
let new_ast = ConstantOptimizer::new()
.fold_expr(ast)
.unwrap_or_else(|e| match e {});
let new_ast = ConstantOptimizer::new().fold_expr(ast).unwrap_or_else(|e| match e {});
assert_eq!(
new_ast,
Located {
@ -199,11 +173,7 @@ mod tests {
value: Constant::Tuple(vec![
1.into(),
2.into(),
Constant::Tuple(vec![
3.into(),
4.into(),
5.into(),
])
Constant::Tuple(vec![3.into(), 4.into(), 5.into(),])
]),
kind: None
},

View File

@ -64,11 +64,4 @@ macro_rules! simple_fold {
};
}
simple_fold!(
usize,
String,
bool,
StrRef,
constant::Constant,
constant::ConversionFlag
);
simple_fold!(usize, String, bool, StrRef, constant::Constant, constant::ConversionFlag);

View File

@ -2,6 +2,7 @@ use crate::{Constant, ExprKind};
impl<U> ExprKind<U> {
/// Returns a short name for the node suitable for use in error messages.
#[must_use]
pub fn name(&self) -> &'static str {
match self {
ExprKind::BoolOp { .. } | ExprKind::BinOp { .. } | ExprKind::UnaryOp { .. } => {
@ -34,10 +35,7 @@ impl<U> ExprKind<U> {
ExprKind::Starred { .. } => "starred",
ExprKind::Slice { .. } => "slice",
ExprKind::JoinedStr { values } => {
if values
.iter()
.any(|e| matches!(e.node, ExprKind::JoinedStr { .. }))
{
if values.iter().any(|e| matches!(e.node, ExprKind::JoinedStr { .. })) {
"f-string expression"
} else {
"literal"

View File

@ -1,3 +1,19 @@
#![deny(
future_incompatible,
let_underscore,
nonstandard_style,
rust_2024_compatibility,
clippy::all
)]
#![warn(clippy::pedantic)]
#![allow(
clippy::missing_errors_doc,
clippy::missing_panics_doc,
clippy::module_name_repetitions,
clippy::too_many_lines,
clippy::wildcard_imports
)]
#[macro_use]
extern crate lazy_static;
@ -9,6 +25,6 @@ mod impls;
mod location;
pub use ast_gen::*;
pub use location::{Location, FileName};
pub use location::{FileName, Location};
pub type Suite<U = ()> = Vec<Stmt<U>>;

View File

@ -1,8 +1,9 @@
//! Datatypes to support source location information.
use crate::ast_gen::StrRef;
use std::cmp::Ordering;
use std::fmt;
#[derive(Clone, Copy, Debug, PartialEq)]
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub struct FileName(pub StrRef);
impl Default for FileName {
fn default() -> Self {
@ -17,16 +18,38 @@ impl From<String> for FileName {
}
/// A location somewhere in the sourcecode.
#[derive(Clone, Copy, Debug, Default, PartialEq)]
#[derive(Clone, Copy, Debug, Default, Eq, PartialEq)]
pub struct Location {
pub row: usize,
pub column: usize,
pub file: FileName
pub file: FileName,
}
impl fmt::Display for Location {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "{}: line {} column {}", self.file.0, self.row, self.column)
write!(f, "{}:{}:{}", self.file.0, self.row, self.column)
}
}
impl Ord for Location {
fn cmp(&self, other: &Self) -> Ordering {
let file_cmp = self.file.0.to_string().cmp(&other.file.0.to_string());
if file_cmp != Ordering::Equal {
return file_cmp;
}
let row_cmp = self.row.cmp(&other.row);
if row_cmp != Ordering::Equal {
return row_cmp;
}
self.column.cmp(&other.column)
}
}
impl PartialOrd for Location {
fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
Some(self.cmp(other))
}
}
@ -53,23 +76,22 @@ impl Location {
)
}
}
Visualize {
loc: *self,
line,
desc,
}
Visualize { loc: *self, line, desc }
}
}
impl Location {
#[must_use]
pub fn new(row: usize, column: usize, file: FileName) -> Self {
Location { row, column, file }
}
#[must_use]
pub fn row(&self) -> usize {
self.row
}
#[must_use]
pub fn column(&self) -> usize {
self.column
}

View File

@ -1,26 +1,31 @@
[features]
test = []
[package]
name = "nac3core"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2018"
edition = "2021"
[dependencies]
itertools = "0.10"
itertools = "0.13"
crossbeam = "0.8"
indexmap = "2.2"
parking_lot = "0.12"
rayon = "1.5"
rayon = "1.8"
nac3parser = { path = "../nac3parser" }
lazy_static = "1.4"
strum = "0.26.2"
strum_macros = "0.26.4"
[dependencies.inkwell]
git = "https://github.com/TheDan64/inkwell.git"
version = "0.4"
default-features = false
features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
[dev-dependencies]
test-case = "1.2.0"
indoc = "1.0"
indoc = "2.0"
insta = "=1.11.0"
[build-dependencies]
regex = "1"
regex = "1.10"

View File

@ -7,30 +7,41 @@ use std::{
process::{Command, Stdio},
};
fn main() {
const FILE: &str = "src/codegen/irrt/irrt.c";
println!("cargo:rerun-if-changed={}", FILE);
let out_dir = env::var("OUT_DIR").unwrap();
let out_path = Path::new(&out_dir);
fn compile_irrt(irrt_dir: &Path, out_dir: &Path) {
let irrt_cpp_path = irrt_dir.join("irrt.cpp");
/*
* HACK: Sadly, clang doesn't let us emit generic LLVM bitcode.
* Compiling for WASM32 and filtering the output with regex is the closest we can get.
*/
const FLAG: &[&str] = &[
let flags: &[&str] = &[
"--target=wasm32",
FILE,
"-O3",
irrt_cpp_path.to_str().unwrap(),
"-x",
"c++",
"-fno-discard-value-names",
"-fno-exceptions",
"-fno-rtti",
match env::var("PROFILE").as_deref() {
Ok("debug") => "-O0",
Ok("release") => "-O3",
flavor => panic!("Unknown or missing build flavor {flavor:?}"),
},
"-emit-llvm",
"-S",
"-Wall",
"-Wextra",
"-Werror=return-type",
"-I",
irrt_dir.to_str().unwrap(),
"-o",
"-",
];
let output = Command::new("clang")
.args(FLAG)
println!("cargo:rerun-if-changed={}", out_dir.to_str().unwrap());
let output = Command::new("clang-irrt")
.args(flags)
.output()
.map(|o| {
assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
@ -42,9 +53,13 @@ fn main() {
let output = std::str::from_utf8(&output.stdout).unwrap().replace("\r\n", "\n");
let mut filtered_output = String::with_capacity(output.len());
let regex_filter = regex::Regex::new(r"(?ms:^define.*?\}$)|(?m:^declare.*?$)").unwrap();
// (?ms:^define.*?\}$) to capture `define` blocks
// (?m:^declare.*?$) to capture `declare` blocks
// (?m:^%.+?=\s*type\s*\{.+?\}$) to capture `type` declarations
let regex_filter =
Regex::new(r"(?ms:^define.*?\}$)|(?m:^declare.*?$)|(?m:^%.+?=\s*type\s*\{.+?\}$)").unwrap();
for f in regex_filter.captures_iter(&output) {
assert!(f.len() == 1);
assert_eq!(f.len(), 1);
filtered_output.push_str(&f[0]);
filtered_output.push('\n');
}
@ -55,18 +70,65 @@ fn main() {
println!("cargo:rerun-if-env-changed=DEBUG_DUMP_IRRT");
if env::var("DEBUG_DUMP_IRRT").is_ok() {
let mut file = File::create(out_path.join("irrt.ll")).unwrap();
let mut file = File::create(out_dir.join("irrt.ll")).unwrap();
file.write_all(output.as_bytes()).unwrap();
let mut file = File::create(out_path.join("irrt-filtered.ll")).unwrap();
let mut file = File::create(out_dir.join("irrt-filtered.ll")).unwrap();
file.write_all(filtered_output.as_bytes()).unwrap();
}
let mut llvm_as = Command::new("llvm-as")
let mut llvm_as = Command::new("llvm-as-irrt")
.stdin(Stdio::piped())
.arg("-o")
.arg(out_path.join("irrt.bc"))
.arg(out_dir.join("irrt.bc"))
.spawn()
.unwrap();
llvm_as.stdin.as_mut().unwrap().write_all(filtered_output.as_bytes()).unwrap();
assert!(llvm_as.wait().unwrap().success())
assert!(llvm_as.wait().unwrap().success());
}
fn compile_irrt_test(irrt_dir: &Path, out_dir: &Path) {
let irrt_test_cpp_path = irrt_dir.join("irrt_test.cpp");
let exe_path = out_dir.join("irrt_test.out");
let flags: &[&str] = &[
irrt_test_cpp_path.to_str().unwrap(),
"-x",
"c++",
"-I",
irrt_dir.to_str().unwrap(),
"-g",
"-fno-discard-value-names",
"-O0",
"-Wall",
"-Wextra",
"-Werror=return-type",
"-lm", // for `tgamma()`, `lgamma()`
"-o",
exe_path.to_str().unwrap(),
];
Command::new("clang-irrt-test")
.args(flags)
.output()
.map(|o| {
assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
o
})
.unwrap();
println!("cargo:rerun-if-changed={}", out_dir.to_str().unwrap());
}
fn main() {
let out_dir = env::var("OUT_DIR").unwrap();
let out_dir = Path::new(&out_dir);
let irrt_dir = Path::new("./irrt");
compile_irrt(irrt_dir, out_dir);
// https://github.com/rust-lang/cargo/issues/2549
// `cargo test -F test` to also build `irrt_test.cpp
if cfg!(feature = "test") {
compile_irrt_test(irrt_dir, out_dir);
}
}

5
nac3core/irrt/irrt.cpp Normal file
View File

@ -0,0 +1,5 @@
#include "irrt_everything.hpp"
/*
This file will be read by `clang-irrt` to conveniently produce LLVM IR for `nac3core/codegen`.
*/

437
nac3core/irrt/irrt.hpp Normal file
View File

@ -0,0 +1,437 @@
#ifndef IRRT_DONT_TYPEDEF_INTS
typedef _BitInt(8) int8_t;
typedef unsigned _BitInt(8) uint8_t;
typedef _BitInt(32) int32_t;
typedef unsigned _BitInt(32) uint32_t;
typedef _BitInt(64) int64_t;
typedef unsigned _BitInt(64) uint64_t;
#endif
// NDArray indices are always `uint32_t`.
typedef uint32_t NDIndex;
// The type of an index or a value describing the length of a range/slice is
// always `int32_t`.
typedef int32_t SliceIndex;
template <typename T>
static T max(T a, T b) {
return a > b ? a : b;
}
template <typename T>
static T min(T a, T b) {
return a > b ? b : a;
}
// adapted from GNU Scientific Library: https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
// need to make sure `exp >= 0` before calling this function
template <typename T>
static T __nac3_int_exp_impl(T base, T exp) {
T res = 1;
/* repeated squaring method */
do {
if (exp & 1) {
res *= base; /* for n odd */
}
exp >>= 1;
base *= base;
} while (exp);
return res;
}
template <typename SizeT>
static SizeT __nac3_ndarray_calc_size_impl(
const SizeT *list_data,
SizeT list_len,
SizeT begin_idx,
SizeT end_idx
) {
__builtin_assume(end_idx <= list_len);
SizeT num_elems = 1;
for (SizeT i = begin_idx; i < end_idx; ++i) {
SizeT val = list_data[i];
__builtin_assume(val > 0);
num_elems *= val;
}
return num_elems;
}
template <typename SizeT>
static void __nac3_ndarray_calc_nd_indices_impl(
SizeT index,
const SizeT *dims,
SizeT num_dims,
NDIndex *idxs
) {
SizeT stride = 1;
for (SizeT dim = 0; dim < num_dims; dim++) {
SizeT i = num_dims - dim - 1;
__builtin_assume(dims[i] > 0);
idxs[i] = (index / stride) % dims[i];
stride *= dims[i];
}
}
template <typename SizeT>
static SizeT __nac3_ndarray_flatten_index_impl(
const SizeT *dims,
SizeT num_dims,
const NDIndex *indices,
SizeT num_indices
) {
SizeT idx = 0;
SizeT stride = 1;
for (SizeT i = 0; i < num_dims; ++i) {
SizeT ri = num_dims - i - 1;
if (ri < num_indices) {
idx += stride * indices[ri];
}
__builtin_assume(dims[i] > 0);
stride *= dims[ri];
}
return idx;
}
template <typename SizeT>
static void __nac3_ndarray_calc_broadcast_impl(
const SizeT *lhs_dims,
SizeT lhs_ndims,
const SizeT *rhs_dims,
SizeT rhs_ndims,
SizeT *out_dims
) {
SizeT max_ndims = lhs_ndims > rhs_ndims ? lhs_ndims : rhs_ndims;
for (SizeT i = 0; i < max_ndims; ++i) {
const SizeT *lhs_dim_sz = i < lhs_ndims ? &lhs_dims[lhs_ndims - i - 1] : nullptr;
const SizeT *rhs_dim_sz = i < rhs_ndims ? &rhs_dims[rhs_ndims - i - 1] : nullptr;
SizeT *out_dim = &out_dims[max_ndims - i - 1];
if (lhs_dim_sz == nullptr) {
*out_dim = *rhs_dim_sz;
} else if (rhs_dim_sz == nullptr) {
*out_dim = *lhs_dim_sz;
} else if (*lhs_dim_sz == 1) {
*out_dim = *rhs_dim_sz;
} else if (*rhs_dim_sz == 1) {
*out_dim = *lhs_dim_sz;
} else if (*lhs_dim_sz == *rhs_dim_sz) {
*out_dim = *lhs_dim_sz;
} else {
__builtin_unreachable();
}
}
}
template <typename SizeT>
static void __nac3_ndarray_calc_broadcast_idx_impl(
const SizeT *src_dims,
SizeT src_ndims,
const NDIndex *in_idx,
NDIndex *out_idx
) {
for (SizeT i = 0; i < src_ndims; ++i) {
SizeT src_i = src_ndims - i - 1;
out_idx[src_i] = src_dims[src_i] == 1 ? 0 : in_idx[src_i];
}
}
template<typename SizeT>
static void __nac3_ndarray_strides_from_shape_impl(
SizeT ndims,
SizeT *shape,
SizeT *dst_strides
) {
SizeT stride_product = 1;
for (SizeT i = 0; i < ndims; i++) {
int dim_i = ndims - i - 1;
dst_strides[dim_i] = stride_product;
stride_product *= shape[dim_i];
}
}
extern "C" {
#define DEF_nac3_int_exp_(T) \
T __nac3_int_exp_##T(T base, T exp) {\
return __nac3_int_exp_impl(base, exp);\
}
DEF_nac3_int_exp_(int32_t)
DEF_nac3_int_exp_(int64_t)
DEF_nac3_int_exp_(uint32_t)
DEF_nac3_int_exp_(uint64_t)
SliceIndex __nac3_slice_index_bound(SliceIndex i, const SliceIndex len) {
if (i < 0) {
i = len + i;
}
if (i < 0) {
return 0;
} else if (i > len) {
return len;
}
return i;
}
SliceIndex __nac3_range_slice_len(
const SliceIndex start,
const SliceIndex end,
const SliceIndex step
) {
SliceIndex diff = end - start;
if (diff > 0 && step > 0) {
return ((diff - 1) / step) + 1;
} else if (diff < 0 && step < 0) {
return ((diff + 1) / step) + 1;
} else {
return 0;
}
}
// Handle list assignment and dropping part of the list when
// both dest_step and src_step are +1.
// - All the index must *not* be out-of-bound or negative,
// - The end index is *inclusive*,
// - The length of src and dest slice size should already
// be checked: if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest)
SliceIndex __nac3_list_slice_assign_var_size(
SliceIndex dest_start,
SliceIndex dest_end,
SliceIndex dest_step,
uint8_t *dest_arr,
SliceIndex dest_arr_len,
SliceIndex src_start,
SliceIndex src_end,
SliceIndex src_step,
uint8_t *src_arr,
SliceIndex src_arr_len,
const SliceIndex size
) {
/* if dest_arr_len == 0, do nothing since we do not support extending list */
if (dest_arr_len == 0) return dest_arr_len;
/* if both step is 1, memmove directly, handle the dropping of the list, and shrink size */
if (src_step == dest_step && dest_step == 1) {
const SliceIndex src_len = (src_end >= src_start) ? (src_end - src_start + 1) : 0;
const SliceIndex dest_len = (dest_end >= dest_start) ? (dest_end - dest_start + 1) : 0;
if (src_len > 0) {
__builtin_memmove(
dest_arr + dest_start * size,
src_arr + src_start * size,
src_len * size
);
}
if (dest_len > 0) {
/* dropping */
__builtin_memmove(
dest_arr + (dest_start + src_len) * size,
dest_arr + (dest_end + 1) * size,
(dest_arr_len - dest_end - 1) * size
);
}
/* shrink size */
return dest_arr_len - (dest_len - src_len);
}
/* if two range overlaps, need alloca */
uint8_t need_alloca =
(dest_arr == src_arr)
&& !(
max(dest_start, dest_end) < min(src_start, src_end)
|| max(src_start, src_end) < min(dest_start, dest_end)
);
if (need_alloca) {
uint8_t *tmp = reinterpret_cast<uint8_t *>(__builtin_alloca(src_arr_len * size));
__builtin_memcpy(tmp, src_arr, src_arr_len * size);
src_arr = tmp;
}
SliceIndex src_ind = src_start;
SliceIndex dest_ind = dest_start;
for (;
(src_step > 0) ? (src_ind <= src_end) : (src_ind >= src_end);
src_ind += src_step, dest_ind += dest_step
) {
/* for constant optimization */
if (size == 1) {
__builtin_memcpy(dest_arr + dest_ind, src_arr + src_ind, 1);
} else if (size == 4) {
__builtin_memcpy(dest_arr + dest_ind * 4, src_arr + src_ind * 4, 4);
} else if (size == 8) {
__builtin_memcpy(dest_arr + dest_ind * 8, src_arr + src_ind * 8, 8);
} else {
/* memcpy for var size, cannot overlap after previous alloca */
__builtin_memcpy(dest_arr + dest_ind * size, src_arr + src_ind * size, size);
}
}
/* only dest_step == 1 can we shrink the dest list. */
/* size should be ensured prior to calling this function */
if (dest_step == 1 && dest_end >= dest_start) {
__builtin_memmove(
dest_arr + dest_ind * size,
dest_arr + (dest_end + 1) * size,
(dest_arr_len - dest_end - 1) * size
);
return dest_arr_len - (dest_end - dest_ind) - 1;
}
return dest_arr_len;
}
int32_t __nac3_isinf(double x) {
return __builtin_isinf(x);
}
int32_t __nac3_isnan(double x) {
return __builtin_isnan(x);
}
double tgamma(double arg);
double __nac3_gamma(double z) {
// Handling for denormals
// | x | Python gamma(x) | C tgamma(x) |
// --- | ----------------- | --------------- | ----------- |
// (1) | nan | nan | nan |
// (2) | -inf | -inf | inf |
// (3) | inf | inf | inf |
// (4) | 0.0 | inf | inf |
// (5) | {-1.0, -2.0, ...} | inf | nan |
// (1)-(3)
if (__builtin_isinf(z) || __builtin_isnan(z)) {
return z;
}
double v = tgamma(z);
// (4)-(5)
return __builtin_isinf(v) || __builtin_isnan(v) ? __builtin_inf() : v;
}
double lgamma(double arg);
double __nac3_gammaln(double x) {
// libm's handling of value overflows differs from scipy:
// - scipy: gammaln(-inf) -> -inf
// - libm : lgamma(-inf) -> inf
if (__builtin_isinf(x)) {
return x;
}
return lgamma(x);
}
double j0(double x);
double __nac3_j0(double x) {
// libm's handling of value overflows differs from scipy:
// - scipy: j0(inf) -> nan
// - libm : j0(inf) -> 0.0
if (__builtin_isinf(x)) {
return __builtin_nan("");
}
return j0(x);
}
uint32_t __nac3_ndarray_calc_size(
const uint32_t *list_data,
uint32_t list_len,
uint32_t begin_idx,
uint32_t end_idx
) {
return __nac3_ndarray_calc_size_impl(list_data, list_len, begin_idx, end_idx);
}
uint64_t __nac3_ndarray_calc_size64(
const uint64_t *list_data,
uint64_t list_len,
uint64_t begin_idx,
uint64_t end_idx
) {
return __nac3_ndarray_calc_size_impl(list_data, list_len, begin_idx, end_idx);
}
void __nac3_ndarray_calc_nd_indices(
uint32_t index,
const uint32_t* dims,
uint32_t num_dims,
NDIndex* idxs
) {
__nac3_ndarray_calc_nd_indices_impl(index, dims, num_dims, idxs);
}
void __nac3_ndarray_calc_nd_indices64(
uint64_t index,
const uint64_t* dims,
uint64_t num_dims,
NDIndex* idxs
) {
__nac3_ndarray_calc_nd_indices_impl(index, dims, num_dims, idxs);
}
uint32_t __nac3_ndarray_flatten_index(
const uint32_t* dims,
uint32_t num_dims,
const NDIndex* indices,
uint32_t num_indices
) {
return __nac3_ndarray_flatten_index_impl(dims, num_dims, indices, num_indices);
}
uint64_t __nac3_ndarray_flatten_index64(
const uint64_t* dims,
uint64_t num_dims,
const NDIndex* indices,
uint64_t num_indices
) {
return __nac3_ndarray_flatten_index_impl(dims, num_dims, indices, num_indices);
}
void __nac3_ndarray_calc_broadcast(
const uint32_t *lhs_dims,
uint32_t lhs_ndims,
const uint32_t *rhs_dims,
uint32_t rhs_ndims,
uint32_t *out_dims
) {
return __nac3_ndarray_calc_broadcast_impl(lhs_dims, lhs_ndims, rhs_dims, rhs_ndims, out_dims);
}
void __nac3_ndarray_calc_broadcast64(
const uint64_t *lhs_dims,
uint64_t lhs_ndims,
const uint64_t *rhs_dims,
uint64_t rhs_ndims,
uint64_t *out_dims
) {
return __nac3_ndarray_calc_broadcast_impl(lhs_dims, lhs_ndims, rhs_dims, rhs_ndims, out_dims);
}
void __nac3_ndarray_calc_broadcast_idx(
const uint32_t *src_dims,
uint32_t src_ndims,
const NDIndex *in_idx,
NDIndex *out_idx
) {
__nac3_ndarray_calc_broadcast_idx_impl(src_dims, src_ndims, in_idx, out_idx);
}
void __nac3_ndarray_calc_broadcast_idx64(
const uint64_t *src_dims,
uint64_t src_ndims,
const NDIndex *in_idx,
NDIndex *out_idx
) {
__nac3_ndarray_calc_broadcast_idx_impl(src_dims, src_ndims, in_idx, out_idx);
}
void __nac3_ndarray_strides_from_shape(uint32_t ndims, uint32_t* shape, uint32_t* dst_strides) {
__nac3_ndarray_strides_from_shape_impl(ndims, shape, dst_strides);
}
void __nac3_ndarray_strides_from_shape64(uint64_t ndims, uint64_t* shape, uint64_t* dst_strides) {
__nac3_ndarray_strides_from_shape_impl(ndims, shape, dst_strides);
}
}

View File

@ -0,0 +1,216 @@
#pragma once
#include "irrt_utils.hpp"
#include "irrt_typedefs.hpp"
/*
This header contains IRRT implementations
that do not deserved to be categorized (e.g., into numpy, etc.)
Check out other *.hpp files before including them here!!
*/
// The type of an index or a value describing the length of a range/slice is
// always `int32_t`.
namespace {
// adapted from GNU Scientific Library: https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
// need to make sure `exp >= 0` before calling this function
template <typename T>
T __nac3_int_exp_impl(T base, T exp) {
T res = 1;
/* repeated squaring method */
do {
if (exp & 1) {
res *= base; /* for n odd */
}
exp >>= 1;
base *= base;
} while (exp);
return res;
}
}
extern "C" {
#define DEF_nac3_int_exp_(T) \
T __nac3_int_exp_##T(T base, T exp) {\
return __nac3_int_exp_impl(base, exp);\
}
DEF_nac3_int_exp_(int32_t)
DEF_nac3_int_exp_(int64_t)
DEF_nac3_int_exp_(uint32_t)
DEF_nac3_int_exp_(uint64_t)
SliceIndex __nac3_slice_index_bound(SliceIndex i, const SliceIndex len) {
if (i < 0) {
i = len + i;
}
if (i < 0) {
return 0;
} else if (i > len) {
return len;
}
return i;
}
SliceIndex __nac3_range_slice_len(
const SliceIndex start,
const SliceIndex end,
const SliceIndex step
) {
SliceIndex diff = end - start;
if (diff > 0 && step > 0) {
return ((diff - 1) / step) + 1;
} else if (diff < 0 && step < 0) {
return ((diff + 1) / step) + 1;
} else {
return 0;
}
}
// Handle list assignment and dropping part of the list when
// both dest_step and src_step are +1.
// - All the index must *not* be out-of-bound or negative,
// - The end index is *inclusive*,
// - The length of src and dest slice size should already
// be checked: if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest)
SliceIndex __nac3_list_slice_assign_var_size(
SliceIndex dest_start,
SliceIndex dest_end,
SliceIndex dest_step,
uint8_t *dest_arr,
SliceIndex dest_arr_len,
SliceIndex src_start,
SliceIndex src_end,
SliceIndex src_step,
uint8_t *src_arr,
SliceIndex src_arr_len,
const SliceIndex size
) {
/* if dest_arr_len == 0, do nothing since we do not support extending list */
if (dest_arr_len == 0) return dest_arr_len;
/* if both step is 1, memmove directly, handle the dropping of the list, and shrink size */
if (src_step == dest_step && dest_step == 1) {
const SliceIndex src_len = (src_end >= src_start) ? (src_end - src_start + 1) : 0;
const SliceIndex dest_len = (dest_end >= dest_start) ? (dest_end - dest_start + 1) : 0;
if (src_len > 0) {
__builtin_memmove(
dest_arr + dest_start * size,
src_arr + src_start * size,
src_len * size
);
}
if (dest_len > 0) {
/* dropping */
__builtin_memmove(
dest_arr + (dest_start + src_len) * size,
dest_arr + (dest_end + 1) * size,
(dest_arr_len - dest_end - 1) * size
);
}
/* shrink size */
return dest_arr_len - (dest_len - src_len);
}
/* if two range overlaps, need alloca */
uint8_t need_alloca =
(dest_arr == src_arr)
&& !(
max(dest_start, dest_end) < min(src_start, src_end)
|| max(src_start, src_end) < min(dest_start, dest_end)
);
if (need_alloca) {
uint8_t *tmp = reinterpret_cast<uint8_t *>(__builtin_alloca(src_arr_len * size));
__builtin_memcpy(tmp, src_arr, src_arr_len * size);
src_arr = tmp;
}
SliceIndex src_ind = src_start;
SliceIndex dest_ind = dest_start;
for (;
(src_step > 0) ? (src_ind <= src_end) : (src_ind >= src_end);
src_ind += src_step, dest_ind += dest_step
) {
/* for constant optimization */
if (size == 1) {
__builtin_memcpy(dest_arr + dest_ind, src_arr + src_ind, 1);
} else if (size == 4) {
__builtin_memcpy(dest_arr + dest_ind * 4, src_arr + src_ind * 4, 4);
} else if (size == 8) {
__builtin_memcpy(dest_arr + dest_ind * 8, src_arr + src_ind * 8, 8);
} else {
/* memcpy for var size, cannot overlap after previous alloca */
__builtin_memcpy(dest_arr + dest_ind * size, src_arr + src_ind * size, size);
}
}
/* only dest_step == 1 can we shrink the dest list. */
/* size should be ensured prior to calling this function */
if (dest_step == 1 && dest_end >= dest_start) {
__builtin_memmove(
dest_arr + dest_ind * size,
dest_arr + (dest_end + 1) * size,
(dest_arr_len - dest_end - 1) * size
);
return dest_arr_len - (dest_end - dest_ind) - 1;
}
return dest_arr_len;
}
int32_t __nac3_isinf(double x) {
return __builtin_isinf(x);
}
int32_t __nac3_isnan(double x) {
return __builtin_isnan(x);
}
double tgamma(double arg);
double __nac3_gamma(double z) {
// Handling for denormals
// | x | Python gamma(x) | C tgamma(x) |
// --- | ----------------- | --------------- | ----------- |
// (1) | nan | nan | nan |
// (2) | -inf | -inf | inf |
// (3) | inf | inf | inf |
// (4) | 0.0 | inf | inf |
// (5) | {-1.0, -2.0, ...} | inf | nan |
// (1)-(3)
if (__builtin_isinf(z) || __builtin_isnan(z)) {
return z;
}
double v = tgamma(z);
// (4)-(5)
return __builtin_isinf(v) || __builtin_isnan(v) ? __builtin_inf() : v;
}
double lgamma(double arg);
double __nac3_gammaln(double x) {
// libm's handling of value overflows differs from scipy:
// - scipy: gammaln(-inf) -> -inf
// - libm : lgamma(-inf) -> inf
if (__builtin_isinf(x)) {
return x;
}
return lgamma(x);
}
double j0(double x);
double __nac3_j0(double x) {
// libm's handling of value overflows differs from scipy:
// - scipy: j0(inf) -> nan
// - libm : j0(inf) -> 0.0
if (__builtin_isinf(x)) {
return __builtin_nan("");
}
return j0(x);
}
}

View File

@ -0,0 +1,14 @@
#pragma once
#include "irrt_utils.hpp"
#include "irrt_typedefs.hpp"
#include "irrt_basic.hpp"
#include "irrt_slice.hpp"
#include "irrt_numpy_ndarray.hpp"
/*
All IRRT implementations.
We don't have any pre-compiled objects, so we are writing all implementations in headers and
concatenate them with `#include` into one massive source file that contains all the IRRT stuff.
*/

View File

@ -0,0 +1,466 @@
#pragma once
#include "irrt_utils.hpp"
#include "irrt_typedefs.hpp"
#include "irrt_slice.hpp"
/*
NDArray-related implementations.
`*/
// NDArray indices are always `uint32_t`.
using NDIndex = uint32_t;
namespace {
namespace ndarray_util {
template <typename SizeT>
static void set_indices_by_nth(SizeT ndims, const SizeT* shape, SizeT* indices, SizeT nth) {
for (int32_t i = 0; i < ndims; i++) {
int32_t dim_i = ndims - i - 1;
int32_t dim = shape[dim_i];
indices[dim_i] = nth % dim;
nth /= dim;
}
}
// Compute the strides of an ndarray given an ndarray `shape`
// and assuming that the ndarray is *fully C-contagious*.
//
// You might want to read up on https://ajcr.net/stride-guide-part-1/.
template <typename SizeT>
static void set_strides_by_shape(SizeT itemsize, SizeT ndims, SizeT* dst_strides, const SizeT* shape) {
SizeT stride_product = 1;
for (SizeT i = 0; i < ndims; i++) {
int dim_i = ndims - i - 1;
dst_strides[dim_i] = stride_product * itemsize;
stride_product *= shape[dim_i];
}
}
// Compute the size/# of elements of an ndarray given its shape
template <typename SizeT>
static SizeT calc_size_from_shape(SizeT ndims, const SizeT* shape) {
SizeT size = 1;
for (SizeT dim_i = 0; dim_i < ndims; dim_i++) size *= shape[dim_i];
return size;
}
template <typename SizeT>
static bool can_broadcast_shape_to(
const SizeT target_ndims,
const SizeT *target_shape,
const SizeT src_ndims,
const SizeT *src_shape
) {
/*
// See https://numpy.org/doc/stable/user/basics.broadcasting.html
This function handles this example:
```
Image (3d array): 256 x 256 x 3
Scale (1d array): 3
Result (3d array): 256 x 256 x 3
```
Other interesting examples to consider:
- `can_broadcast_shape_to([3], [1, 1, 1, 1, 3]) == true`
- `can_broadcast_shape_to([3], [3, 1]) == false`
- `can_broadcast_shape_to([256, 256, 3], [256, 1, 3]) == true`
In cases when the shapes contain zero(es):
- `can_broadcast_shape_to([0], [1]) == true`
- `can_broadcast_shape_to([0], [2]) == false`
- `can_broadcast_shape_to([0, 4, 0, 0], [1]) == true`
- `can_broadcast_shape_to([0, 4, 0, 0], [1, 1, 1, 1]) == true`
- `can_broadcast_shape_to([0, 4, 0, 0], [1, 4, 1, 1]) == true`
- `can_broadcast_shape_to([4, 3], [0, 3]) == false`
- `can_broadcast_shape_to([4, 3], [0, 0]) == false`
*/
// This is essentially doing the following in Python:
// `for target_dim, src_dim in itertools.zip_longest(target_shape[::-1], src_shape[::-1], fillvalue=1)`
for (SizeT i = 0; i < max(target_ndims, src_ndims); i++) {
SizeT target_dim_i = target_ndims - i - 1;
SizeT src_dim_i = src_ndims - i - 1;
bool target_dim_exists = target_dim_i >= 0;
bool src_dim_exists = src_dim_i >= 0;
SizeT target_dim = target_dim_exists ? target_shape[target_dim_i] : 1;
SizeT src_dim = src_dim_exists ? src_shape[src_dim_i] : 1;
bool ok = src_dim == 1 || target_dim == src_dim;
if (!ok) return false;
}
return true;
}
}
typedef uint8_t NDSliceType;
extern "C" {
const NDSliceType INPUT_SLICE_TYPE_INDEX = 0;
const NDSliceType INPUT_SLICE_TYPE_SLICE = 1;
}
struct NDSlice {
// A poor-man's `std::variant<int, UserRange>`
NDSliceType type;
/*
if type == INPUT_SLICE_TYPE_INDEX => `slice` points to a single `SizeT`
if type == INPUT_SLICE_TYPE_SLICE => `slice` points to a single `UserRange`
*/
uint8_t *slice;
};
namespace ndarray_util {
template<typename SizeT>
SizeT deduce_ndims_after_slicing(SizeT ndims, SizeT num_slices, const NDSlice *slices) {
irrt_assert(num_slices <= ndims);
SizeT final_ndims = ndims;
for (SizeT i = 0; i < num_slices; i++) {
if (slices[i].type == INPUT_SLICE_TYPE_INDEX) {
final_ndims--; // An integer slice demotes the rank by 1
}
}
return final_ndims;
}
}
template <typename SizeT>
struct NDArrayIndicesIter {
SizeT ndims;
const SizeT *shape;
SizeT *indices;
void set_indices_zero() {
__builtin_memset(indices, 0, sizeof(SizeT) * ndims);
}
void next() {
for (SizeT i = 0; i < ndims; i++) {
SizeT dim_i = ndims - i - 1;
indices[dim_i]++;
if (indices[dim_i] < shape[dim_i]) {
break;
} else {
indices[dim_i] = 0;
}
}
}
};
// The NDArray object. `SizeT` is the *signed* size type of this ndarray.
//
// NOTE: The order of fields is IMPORTANT. DON'T TOUCH IT
//
// Some resources you might find helpful:
// - The official numpy implementations:
// - https://github.com/numpy/numpy/blob/735a477f0bc2b5b84d0e72d92f224bde78d4e069/doc/source/reference/c-api/types-and-structures.rst
// - On strides (about reshaping, slicing, C-contagiousness, etc)
// - https://ajcr.net/stride-guide-part-1/.
// - https://ajcr.net/stride-guide-part-2/.
// - https://ajcr.net/stride-guide-part-3/.
template <typename SizeT>
struct NDArray {
// The underlying data this `ndarray` is pointing to.
//
// NOTE: Formally this should be of type `void *`, but clang
// translates `void *` to `i8 *` when run with `-S -emit-llvm`,
// so we will put `uint8_t *` here for clarity.
uint8_t *data;
// The number of bytes of a single element in `data`.
//
// The `SizeT` is treated as `unsigned`.
SizeT itemsize;
// The number of dimensions of this shape.
//
// The `SizeT` is treated as `unsigned`.
SizeT ndims;
// Array shape, with length equal to `ndims`.
//
// The `SizeT` is treated as `unsigned`.
//
// NOTE: `shape` can contain 0.
// (those appear when the user makes an out of bounds slice into an ndarray, e.g., `np.zeros((3, 3))[400:].shape == (0, 3)`)
SizeT *shape;
// Array strides (stride value is in number of bytes, NOT number of elements), with length equal to `ndims`.
//
// The `SizeT` is treated as `signed`.
//
// NOTE: `strides` can have negative numbers.
// (those appear when there is a slice with a negative step, e.g., `my_array[::-1]`)
SizeT *strides;
// Calculate the size/# of elements of an `ndarray`.
// This function corresponds to `np.size(<ndarray>)` or `ndarray.size`
SizeT size() {
return ndarray_util::calc_size_from_shape(ndims, shape);
}
// Calculate the number of bytes of its content of an `ndarray` *in its view*.
// This function corresponds to `ndarray.nbytes`
SizeT nbytes() {
return this->size() * itemsize;
}
void set_value_at_pelement(uint8_t* pelement, const uint8_t* pvalue) {
__builtin_memcpy(pelement, pvalue, itemsize);
}
uint8_t* get_pelement(const SizeT *indices) {
uint8_t* element = data;
for (SizeT dim_i = 0; dim_i < ndims; dim_i++)
element += indices[dim_i] * strides[dim_i];
return element;
}
uint8_t* get_nth_pelement(SizeT nth) {
irrt_assert(0 <= nth);
irrt_assert(nth < this->size());
SizeT* indices = (SizeT*) __builtin_alloca(sizeof(SizeT) * this->ndims);
ndarray_util::set_indices_by_nth(this->ndims, this->shape, indices, nth);
return get_pelement(indices);
}
// Get pointer to the first element of this ndarray, assuming
// `this->size() > 0`, i.e., not "degenerate" due to zeroes in `this->shape`)
//
// This is particularly useful for when the ndarray is just containing a single scalar.
uint8_t* get_first_pelement() {
irrt_assert(this->size() > 0);
return this->data; // ...It is simply `this->data`
}
// Is the given `indices` valid/in-bounds?
bool in_bounds(const SizeT *indices) {
for (SizeT dim_i = 0; dim_i < ndims; dim_i++) {
bool dim_ok = indices[dim_i] < shape[dim_i];
if (!dim_ok) return false;
}
return true;
}
// Fill the ndarray with a value
void fill_generic(const uint8_t* pvalue) {
NDArrayIndicesIter<SizeT> iter;
iter.ndims = this->ndims;
iter.shape = this->shape;
iter.indices = (SizeT*) __builtin_alloca(sizeof(SizeT) * ndims);
iter.set_indices_zero();
for (SizeT i = 0; i < this->size(); i++, iter.next()) {
uint8_t* pelement = get_pelement(iter.indices);
set_value_at_pelement(pelement, pvalue);
}
}
// Set the strides of the ndarray with `ndarray_util::set_strides_by_shape`
void set_strides_by_shape() {
ndarray_util::set_strides_by_shape(itemsize, ndims, strides, shape);
}
// https://numpy.org/doc/stable/reference/generated/numpy.eye.html
void set_to_eye(SizeT k, const uint8_t* zero_pvalue, const uint8_t* one_pvalue) {
__builtin_assume(ndims == 2);
// TODO: Better implementation
fill_generic(zero_pvalue);
for (SizeT i = 0; i < min(shape[0], shape[1]); i++) {
SizeT row = i;
SizeT col = i + k;
SizeT indices[2] = { row, col };
if (!in_bounds(indices)) continue;
uint8_t* pelement = get_pelement(indices);
set_value_at_pelement(pelement, one_pvalue);
}
}
// To support numpy complex slices (e.g., `my_array[:50:2,4,:2:-1]`)
//
// Things assumed by this function:
// - `dst_ndarray` is allocated by the caller
// - `dst_ndarray.ndims` has the correct value (according to `ndarray_util::deduce_ndims_after_slicing`).
// - ... and `dst_ndarray.shape` and `dst_ndarray.strides` have been allocated by the caller as well
//
// Other notes:
// - `dst_ndarray->data` does not have to be set, it will be derived.
// - `dst_ndarray->itemsize` does not have to be set, it will be set to `this->itemsize`
// - `dst_ndarray->shape` and `dst_ndarray.strides` can contain empty values
void slice(SizeT num_ndslices, NDSlice* ndslices, NDArray<SizeT>* dst_ndarray) {
// REFERENCE CODE (check out `_index_helper` in `__getitem__`):
// https://github.com/wadetb/tinynumpy/blob/0d23d22e07062ffab2afa287374c7b366eebdda1/tinynumpy/tinynumpy.py#L652
irrt_assert(dst_ndarray->ndims == ndarray_util::deduce_ndims_after_slicing(this->ndims, num_ndslices, ndslices));
dst_ndarray->data = this->data;
SizeT this_axis = 0;
SizeT dst_axis = 0;
for (SizeT i = 0; i < num_ndslices; i++) {
NDSlice *ndslice = &ndslices[i];
if (ndslice->type == INPUT_SLICE_TYPE_INDEX) {
// Handle when the ndslice is just a single (possibly negative) integer
// e.g., `my_array[::2, -5, ::-1]`
// ^^------ like this
SizeT index_user = *((SizeT*) ndslice->slice);
SizeT index = resolve_index_in_length(this->shape[this_axis], index_user);
dst_ndarray->data += index * this->strides[this_axis]; // Add offset
// Next
this_axis++;
} else if (ndslice->type == INPUT_SLICE_TYPE_SLICE) {
// Handle when the ndslice is a slice (represented by UserSlice in IRRT)
// e.g., `my_array[::2, -5, ::-1]`
// ^^^------^^^^----- like these
UserSlice<SizeT>* user_slice = (UserSlice<SizeT>*) ndslice->slice;
Slice<SizeT> slice = user_slice->indices(this->shape[this_axis]); // To resolve negative indices and other funny stuff written by the user
// NOTE: There is no need to write special code to handle negative steps/strides.
// This simple implementation meticulously handles both positive and negative steps/strides.
// Check out the tinynumpy and IRRT's test cases if you are not convinced.
dst_ndarray->data += slice.start * this->strides[this_axis]; // Add offset (NOTE: no need to `* itemsize`, strides count in # of bytes)
dst_ndarray->strides[dst_axis] = slice.step * this->strides[this_axis]; // Determine stride
dst_ndarray->shape[dst_axis] = slice.len(); // Determine shape dimension
// Next
dst_axis++;
this_axis++;
} else {
__builtin_unreachable();
}
}
irrt_assert(dst_axis == dst_ndarray->ndims); // Sanity check on the implementation
}
// Similar to `np.broadcast_to(<ndarray>, <target_shape>)`
// Assumptions:
// - `this` has to be fully initialized.
// - `dst_ndarray->ndims` has to be set.
// - `dst_ndarray->shape` has to be set, this determines the shape `this` broadcasts to.
//
// Other notes:
// - `dst_ndarray->data` does not have to be set, it will be set to `this->data`.
// - `dst_ndarray->itemsize` does not have to be set, it will be set to `this->data`.
// - `dst_ndarray->strides` does not have to be set, it will be overwritten.
//
// Cautions:
// ```
// xs = np.zeros((4,))
// ys = np.zero((4, 1))
// ys[:] = xs # ok
//
// xs = np.zeros((1, 4))
// ys = np.zero((4,))
// ys[:] = xs # allowed
// # However `np.broadcast_to(xs, (4,))` would fails, as per numpy's broadcasting rule.
// # and apparently numpy will "deprecate" this? SEE https://github.com/numpy/numpy/issues/21744
// # This implementation will NOT support this assignment.
// ```
void broadcast_to(NDArray<SizeT>* dst_ndarray) {
dst_ndarray->data = this->data;
dst_ndarray->itemsize = this->itemsize;
irrt_assert(
ndarray_util::can_broadcast_shape_to(
dst_ndarray->ndims,
dst_ndarray->shape,
this->ndims,
this->shape
)
);
SizeT stride_product = 1;
for (SizeT i = 0; i < max(this->ndims, dst_ndarray->ndims); i++) {
SizeT this_dim_i = this->ndims - i - 1;
SizeT dst_dim_i = dst_ndarray->ndims - i - 1;
bool this_dim_exists = this_dim_i >= 0;
bool dst_dim_exists = dst_dim_i >= 0;
// TODO: Explain how this works
bool c1 = this_dim_exists && this->shape[this_dim_i] == 1;
bool c2 = dst_dim_exists && dst_ndarray->shape[dst_dim_i] != 1;
if (!this_dim_exists || (c1 && c2)) {
dst_ndarray->strides[dst_dim_i] = 0; // Freeze it in-place
} else {
dst_ndarray->strides[dst_dim_i] = stride_product * this->itemsize;
stride_product *= this->shape[this_dim_i]; // NOTE: this_dim_exist must be true here.
}
}
}
// Simulates `this_ndarray[:] = src_ndarray`, with automatic broadcasting.
// Caution on https://github.com/numpy/numpy/issues/21744
// Also see `NDArray::broadcast_to`
void assign_with(NDArray<SizeT>* src_ndarray) {
irrt_assert(
ndarray_util::can_broadcast_shape_to(
this->ndims,
this->shape,
src_ndarray->ndims,
src_ndarray->shape
)
);
// Broadcast the `src_ndarray` to make the reading process *much* easier
SizeT* broadcasted_src_ndarray_strides = __builtin_alloca(sizeof(SizeT) * this->ndims); // Remember to allocate strides beforehand
NDArray<SizeT> broadcasted_src_ndarray = {
.ndims = this->ndims,
.shape = this->shape,
.strides = broadcasted_src_ndarray_strides
};
src_ndarray->broadcast_to(&broadcasted_src_ndarray);
// Using iter instead of `get_nth_pelement` because it is slightly faster
SizeT* indices = __builtin_alloca(sizeof(SizeT) * this->ndims);
auto iter = NDArrayIndicesIter<SizeT> {
.ndims = this->ndims,
.shape = this->shape,
.indices = indices
};
const SizeT this_size = this->size();
for (SizeT i = 0; i < this_size; i++, iter.next()) {
uint8_t* src_pelement = broadcasted_src_ndarray_strides->get_pelement(indices);
uint8_t* this_pelement = this->get_pelement(indices);
this->set_value_at_pelement(src_pelement, src_pelement);
}
}
};
}
extern "C" {
uint32_t __nac3_ndarray_size(NDArray<int32_t>* ndarray) {
return ndarray->size();
}
uint64_t __nac3_ndarray_size64(NDArray<int64_t>* ndarray) {
return ndarray->size();
}
void __nac3_ndarray_fill_generic(NDArray<int32_t>* ndarray, uint8_t* pvalue) {
ndarray->fill_generic(pvalue);
}
void __nac3_ndarray_fill_generic64(NDArray<int64_t>* ndarray, uint8_t* pvalue) {
ndarray->fill_generic(pvalue);
}
// void __nac3_ndarray_slice(NDArray<int32_t>* ndarray, int32_t num_slices, NDSlice<int32_t> *slices, NDArray<int32_t> *dst_ndarray) {
// // ndarray->slice(num_slices, slices, dst_ndarray);
// }
}

View File

@ -0,0 +1,80 @@
#pragma once
#include "irrt_utils.hpp"
#include "irrt_typedefs.hpp"
namespace {
// A proper slice in IRRT, all negative indices have be resolved to absolute values.
// Even though nac3core's slices are always `int32_t`, we will template slice anyway
// since this struct is used as a general utility.
template <typename T>
struct Slice {
T start;
T stop;
T step;
// The length/The number of elements of the slice if it were a range,
// i.e., the value of `len(range(this->start, this->stop, this->end))`
T len() {
T diff = stop - start;
if (diff > 0 && step > 0) {
return ((diff - 1) / step) + 1;
} else if (diff < 0 && step < 0) {
return ((diff + 1) / step) + 1;
} else {
return 0;
}
}
};
template<typename T>
T resolve_index_in_length(T length, T index) {
irrt_assert(length >= 0);
if (index < 0) {
// Remember that index is negative, so do a plus here
return max(length + index, 0);
} else {
return min(length, index);
}
}
// NOTE: using a bitfield for the `*_defined` is better, at the
// cost of a more annoying implementation in nac3core inkwell
template <typename T>
struct UserSlice {
uint8_t start_defined;
T start;
uint8_t stop_defined;
T stop;
uint8_t step_defined;
T step;
// Like Python's `slice(start, stop, step).indices(length)`
Slice<T> indices(T length) {
// NOTE: This function implements Python's `slice.indices` *FAITHFULLY*.
// SEE: https://github.com/python/cpython/blob/f62161837e68c1c77961435f1b954412dd5c2b65/Objects/sliceobject.c#L546
irrt_assert(length >= 0);
irrt_assert(!step_defined || step != 0); // step_defined -> step != 0; step cannot be zero if specified by user
Slice<T> result;
result.step = step_defined ? step : 1;
bool step_is_negative = result.step < 0;
if (start_defined) {
result.start = resolve_index_in_length(length, start);
} else {
result.start = step_is_negative ? length - 1 : 0;
}
if (stop_defined) {
result.stop = resolve_index_in_length(length, stop);
} else {
result.stop = step_is_negative ? -1 : length;
}
return result;
}
};
}

658
nac3core/irrt/irrt_test.cpp Normal file
View File

@ -0,0 +1,658 @@
// This file will be compiled like a real C++ program,
// and we do have the luxury to use the standard libraries.
// That is if the nix flakes do not have issues... especially on msys2...
#include <cstdint>
#include <cstdio>
#include <cstdlib>
// Set `IRRT_DONT_TYPEDEF_INTS` because `cstdint` defines them
#define IRRT_DONT_TYPEDEF_INTS
#include "irrt_everything.hpp"
void test_fail() {
printf("[!] Test failed\n");
exit(1);
}
void __begin_test(const char* function_name, const char* file, int line) {
printf("######### Running %s @ %s:%d\n", function_name, file, line);
}
#define BEGIN_TEST() __begin_test(__FUNCTION__, __FILE__, __LINE__)
template <typename T>
void debug_print_array(const char* format, int len, T* as) {
printf("[");
for (int i = 0; i < len; i++) {
if (i != 0) printf(", ");
printf(format, as[i]);
}
printf("]");
}
template <typename T>
void assert_arrays_match(const char* label, const char* format, int len, T* expected, T* got) {
if (!arrays_match(len, expected, got)) {
printf(">>>>>>> %s\n", label);
printf(" Expecting = ");
debug_print_array(format, len, expected);
printf("\n");
printf(" Got = ");
debug_print_array(format, len, got);
printf("\n");
test_fail();
}
}
template <typename T>
void assert_values_match(const char* label, const char* format, T expected, T got) {
if (expected != got) {
printf(">>>>>>> %s\n", label);
printf(" Expecting = ");
printf(format, expected);
printf("\n");
printf(" Got = ");
printf(format, got);
printf("\n");
test_fail();
}
}
void print_repeated(const char *str, int count) {
for (int i = 0; i < count; i++) {
printf("%s", str);
}
}
template<typename SizeT, typename ElementT>
void __print_ndarray_aux(const char *format, bool first, bool last, SizeT* cursor, SizeT depth, NDArray<SizeT>* ndarray) {
// A really lazy recursive implementation
// Add left padding unless its the first entry (since there would be "[[[" before it)
if (!first) {
print_repeated(" ", depth);
}
const SizeT dim = ndarray->shape[depth];
if (depth + 1 == ndarray->ndims) {
// Recursed down to last dimension, print the values in a nice list
printf("[");
SizeT* indices = (SizeT*) __builtin_alloca(sizeof(SizeT) * ndarray->ndims);
for (SizeT i = 0; i < dim; i++) {
ndarray_util::set_indices_by_nth(ndarray->ndims, ndarray->shape, indices, *cursor);
ElementT* pelement = (ElementT*) ndarray->get_pelement(indices);
ElementT element = *pelement;
if (i != 0) printf(", "); // List delimiter
printf(format, element);
printf("(@");
debug_print_array("%d", ndarray->ndims, indices);
printf(")");
(*cursor)++;
}
printf("]");
} else {
printf("[");
for (SizeT i = 0; i < ndarray->shape[depth]; i++) {
__print_ndarray_aux<SizeT, ElementT>(
format,
i == 0, // first?
i + 1 == dim, // last?
cursor,
depth + 1,
ndarray
);
}
printf("]");
}
// Add newline unless its the last entry (since there will be "]]]" after it)
if (!last) {
print_repeated("\n", depth);
}
}
template<typename SizeT, typename ElementT>
void print_ndarray(const char *format, NDArray<SizeT>* ndarray) {
if (ndarray->ndims == 0) {
printf("<empty ndarray>");
} else {
SizeT cursor = 0;
__print_ndarray_aux<SizeT, ElementT>(format, true, true, &cursor, 0, ndarray);
}
printf("\n");
}
void test_calc_size_from_shape_normal() {
// Test shapes with normal values
BEGIN_TEST();
int32_t shape[4] = { 2, 3, 5, 7 };
assert_values_match("size", "%d", 210, ndarray_util::calc_size_from_shape<int32_t>(4, shape));
}
void test_calc_size_from_shape_has_zero() {
// Test shapes with 0 in them
BEGIN_TEST();
int32_t shape[4] = { 2, 0, 5, 7 };
assert_values_match("size", "%d", 0, ndarray_util::calc_size_from_shape<int32_t>(4, shape));
}
void test_set_strides_by_shape() {
// Test `set_strides_by_shape()`
BEGIN_TEST();
int32_t shape[4] = { 99, 3, 5, 7 };
int32_t strides[4] = { 0 };
ndarray_util::set_strides_by_shape((int32_t) sizeof(int32_t), 4, strides, shape);
int32_t expected_strides[4] = {
105 * sizeof(int32_t),
35 * sizeof(int32_t),
7 * sizeof(int32_t),
1 * sizeof(int32_t)
};
assert_arrays_match("strides", "%u", 4u, expected_strides, strides);
}
void test_ndarray_indices_iter_normal() {
// Test NDArrayIndicesIter normal behavior
BEGIN_TEST();
int32_t shape[3] = { 1, 2, 3 };
int32_t indices[3] = { 0, 0, 0 };
auto iter = NDArrayIndicesIter<int32_t> {
.ndims = 3,
.shape = shape,
.indices = indices
};
assert_arrays_match("indices #0", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 0 });
iter.next();
assert_arrays_match("indices #1", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 1 });
iter.next();
assert_arrays_match("indices #2", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 2 });
iter.next();
assert_arrays_match("indices #3", "%u", 3u, iter.indices, (int32_t[3]) { 0, 1, 0 });
iter.next();
assert_arrays_match("indices #4", "%u", 3u, iter.indices, (int32_t[3]) { 0, 1, 1 });
iter.next();
assert_arrays_match("indices #5", "%u", 3u, iter.indices, (int32_t[3]) { 0, 1, 2 });
iter.next();
assert_arrays_match("indices #6", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 0 }); // Loops back
iter.next();
assert_arrays_match("indices #7", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 1 });
}
void test_ndarray_fill_generic() {
// Test ndarray fill_generic
BEGIN_TEST();
// Choose a type that's neither int32_t nor uint64_t (candidates of SizeT) to spice it up
// Also make all the octets non-zero, to see if `memcpy` in `fill_generic` is working perfectly.
uint16_t fill_value = 0xFACE;
uint16_t in_data[6] = { 100, 101, 102, 103, 104, 105 }; // Fill `data` with values that != `999`
int32_t in_itemsize = sizeof(uint16_t);
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = { 2, 3 };
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = in_itemsize,
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides,
};
ndarray.set_strides_by_shape();
ndarray.fill_generic((uint8_t*) &fill_value); // `fill_generic` here
uint16_t expected_data[6] = { fill_value, fill_value, fill_value, fill_value, fill_value, fill_value };
assert_arrays_match("data", "0x%hX", 6, expected_data, in_data);
}
void test_ndarray_set_to_eye() {
// Test `set_to_eye` behavior (helper function to implement `np.eye()`)
BEGIN_TEST();
double in_data[9] = { 99.0, 99.0, 99.0, 99.0, 99.0, 99.0, 99.0, 99.0, 99.0 };
int32_t in_itemsize = sizeof(double);
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = { 3, 3 };
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = in_itemsize,
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides,
};
ndarray.set_strides_by_shape();
double zero = 0.0;
double one = 1.0;
ndarray.set_to_eye(1, (uint8_t*) &zero, (uint8_t*) &one);
assert_values_match("in_data[0]", "%f", 0.0, in_data[0]);
assert_values_match("in_data[1]", "%f", 1.0, in_data[1]);
assert_values_match("in_data[2]", "%f", 0.0, in_data[2]);
assert_values_match("in_data[3]", "%f", 0.0, in_data[3]);
assert_values_match("in_data[4]", "%f", 0.0, in_data[4]);
assert_values_match("in_data[5]", "%f", 1.0, in_data[5]);
assert_values_match("in_data[6]", "%f", 0.0, in_data[6]);
assert_values_match("in_data[7]", "%f", 0.0, in_data[7]);
assert_values_match("in_data[8]", "%f", 0.0, in_data[8]);
}
void test_slice_1() {
// Test `slice(5, None, None).indices(100) == slice(5, 100, 1)`
BEGIN_TEST();
UserSlice<int> user_slice = {
.start_defined = 1,
.start = 5,
.stop_defined = 0,
.step_defined = 0,
};
auto slice = user_slice.indices(100);
assert_values_match("start", "%d", 5, slice.start);
assert_values_match("stop", "%d", 100, slice.stop);
assert_values_match("step", "%d", 1, slice.step);
}
void test_slice_2() {
// Test `slice(400, 999, None).indices(100) == slice(100, 100, 1)`
BEGIN_TEST();
UserSlice<int> user_slice = {
.start_defined = 1,
.start = 400,
.stop_defined = 0,
.step_defined = 0,
};
auto slice = user_slice.indices(100);
assert_values_match("start", "%d", 100, slice.start);
assert_values_match("stop", "%d", 100, slice.stop);
assert_values_match("step", "%d", 1, slice.step);
}
void test_slice_3() {
// Test `slice(-10, -5, None).indices(100) == slice(90, 95, 1)`
BEGIN_TEST();
UserSlice<int> user_slice = {
.start_defined = 1,
.start = -10,
.stop_defined = 1,
.stop = -5,
.step_defined = 0,
};
auto slice = user_slice.indices(100);
assert_values_match("start", "%d", 90, slice.start);
assert_values_match("stop", "%d", 95, slice.stop);
assert_values_match("step", "%d", 1, slice.step);
}
void test_slice_4() {
// Test `slice(None, None, -5).indices(100) == (99, -1, -5)`
BEGIN_TEST();
UserSlice<int> user_slice = {
.start_defined = 0,
.stop_defined = 0,
.step_defined = 1,
.step = -5
};
auto slice = user_slice.indices(100);
assert_values_match("start", "%d", 99, slice.start);
assert_values_match("stop", "%d", -1, slice.stop);
assert_values_match("step", "%d", -5, slice.step);
}
void test_ndslice_1() {
/*
Reference Python code:
```python
ndarray = np.arange(12, dtype=np.float64).reshape((3, 4));
# array([[ 0., 1., 2., 3.],
# [ 4., 5., 6., 7.],
# [ 8., 9., 10., 11.]])
dst_ndarray = ndarray[-2:, 1::2]
# array([[ 5., 7.],
# [ 9., 11.]])
assert dst_ndarray.shape == (2, 2)
assert dst_ndarray.strides == (32, 16)
assert dst_ndarray[0, 0] == 5.0
assert dst_ndarray[0, 1] == 7.0
assert dst_ndarray[1, 0] == 9.0
assert dst_ndarray[1, 1] == 11.0
```
*/
BEGIN_TEST();
double in_data[12] = { 0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0 };
int32_t in_itemsize = sizeof(double);
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = { 3, 4 };
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = in_itemsize,
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides
};
ndarray.set_strides_by_shape();
// Destination ndarray
// As documented, ndims and shape & strides must be allocated and determined by the caller.
const int32_t dst_ndims = 2;
int32_t dst_shape[dst_ndims] = {999, 999}; // Empty values
int32_t dst_strides[dst_ndims] = {999, 999}; // Empty values
NDArray<int32_t> dst_ndarray = {
.data = nullptr,
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides
};
// Create the slice in `ndarray[-2::, 1::2]`
UserSlice<int32_t> user_slice_1 = {
.start_defined = 1,
.start = -2,
.stop_defined = 0,
.step_defined = 0
};
UserSlice<int32_t> user_slice_2 = {
.start_defined = 1,
.start = 1,
.stop_defined = 0,
.step_defined = 1,
.step = 2
};
const int32_t num_ndslices = 2;
NDSlice ndslices[num_ndslices] = {
{ .type = INPUT_SLICE_TYPE_SLICE, .slice = (uint8_t*) &user_slice_1 },
{ .type = INPUT_SLICE_TYPE_SLICE, .slice = (uint8_t*) &user_slice_2 }
};
ndarray.slice(num_ndslices, ndslices, &dst_ndarray);
int32_t expected_shape[dst_ndims] = { 2, 2 };
int32_t expected_strides[dst_ndims] = { 32, 16 };
assert_arrays_match("shape", "%d", dst_ndims, expected_shape, dst_ndarray.shape);
assert_arrays_match("strides", "%d", dst_ndims, expected_strides, dst_ndarray.strides);
assert_values_match("dst_ndarray[0, 0]", "%f", 5.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 0, 0 })));
assert_values_match("dst_ndarray[0, 1]", "%f", 7.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 0, 1 })));
assert_values_match("dst_ndarray[1, 0]", "%f", 9.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 1, 0 })));
assert_values_match("dst_ndarray[1, 1]", "%f", 11.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 1, 1 })));
}
void test_ndslice_2() {
/*
```python
ndarray = np.arange(12, dtype=np.float64).reshape((3, 4))
# array([[ 0., 1., 2., 3.],
# [ 4., 5., 6., 7.],
# [ 8., 9., 10., 11.]])
dst_ndarray = ndarray[2, ::-2]
# array([11., 9.])
assert dst_ndarray.shape == (2,)
assert dst_ndarray.strides == (-16,)
assert dst_ndarray[0] == 11.0
assert dst_ndarray[1] == 9.0
dst_ndarray[1, 0] == 99 # If you write to `dst_ndarray`
assert ndarray[1, 3] == 99 # `ndarray` also updates!!
```
*/
BEGIN_TEST();
double in_data[12] = { 0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0 };
int32_t in_itemsize = sizeof(double);
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = { 3, 4 };
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = in_itemsize,
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides
};
ndarray.set_strides_by_shape();
// Destination ndarray
// As documented, ndims and shape & strides must be allocated and determined by the caller.
const int32_t dst_ndims = 1;
int32_t dst_shape[dst_ndims] = {999}; // Empty values
int32_t dst_strides[dst_ndims] = {999}; // Empty values
NDArray<int32_t> dst_ndarray = {
.data = nullptr,
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides
};
// Create the slice in `ndarray[2, ::-2]`
int32_t user_slice_1 = 2;
UserSlice<int32_t> user_slice_2 = {
.start_defined = 0,
.stop_defined = 0,
.step_defined = 1,
.step = -2
};
const int32_t num_ndslices = 2;
NDSlice ndslices[num_ndslices] = {
{ .type = INPUT_SLICE_TYPE_INDEX, .slice = (uint8_t*) &user_slice_1 },
{ .type = INPUT_SLICE_TYPE_SLICE, .slice = (uint8_t*) &user_slice_2 }
};
ndarray.slice(num_ndslices, ndslices, &dst_ndarray);
int32_t expected_shape[dst_ndims] = { 2 };
int32_t expected_strides[dst_ndims] = { -16 };
assert_arrays_match("shape", "%d", dst_ndims, expected_shape, dst_ndarray.shape);
assert_arrays_match("strides", "%d", dst_ndims, expected_strides, dst_ndarray.strides);
// [5.0, 3.0]
assert_values_match("dst_ndarray[0]", "%f", 11.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 0 })));
assert_values_match("dst_ndarray[1]", "%f", 9.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 1 })));
}
void test_can_broadcast_shape() {
BEGIN_TEST();
assert_values_match(
"can_broadcast_shape_to([3], [1, 1, 1, 1, 3]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 3 }, 5, (int32_t[]) { 1, 1, 1, 1, 3 })
);
assert_values_match(
"can_broadcast_shape_to([3], [3, 1]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 3 }, 2, (int32_t[]) { 3, 1 }));
assert_values_match(
"can_broadcast_shape_to([3], [3]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 3 }, 1, (int32_t[]) { 3 }));
assert_values_match(
"can_broadcast_shape_to([1], [3]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 1 }, 1, (int32_t[]) { 3 }));
assert_values_match(
"can_broadcast_shape_to([1], [1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 1 }, 1, (int32_t[]) { 1 }));
assert_values_match(
"can_broadcast_shape_to([256, 256, 3], [256, 1, 3]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(3, (int32_t[]) { 256, 256, 3 }, 3, (int32_t[]) { 256, 1, 3 })
);
assert_values_match(
"can_broadcast_shape_to([256, 256, 3], [3]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(3, (int32_t[]) { 256, 256, 3 }, 1, (int32_t[]) { 3 })
);
assert_values_match(
"can_broadcast_shape_to([256, 256, 3], [2]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(3, (int32_t[]) { 256, 256, 3 }, 1, (int32_t[]) { 2 })
);
assert_values_match(
"can_broadcast_shape_to([256, 256, 3], [1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(3, (int32_t[]) { 256, 256, 3 }, 1, (int32_t[]) { 1 })
);
// In cases when the shapes contain zero(es)
assert_values_match(
"can_broadcast_shape_to([0], [1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 0 }, 1, (int32_t[]) { 1 })
);
assert_values_match(
"can_broadcast_shape_to([0], [2]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 0 }, 1, (int32_t[]) { 2 })
);
assert_values_match(
"can_broadcast_shape_to([0, 4, 0, 0], [1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(4, (int32_t[]) { 0, 4, 0, 0 }, 1, (int32_t[]) { 1 })
);
assert_values_match(
"can_broadcast_shape_to([0, 4, 0, 0], [1, 1, 1, 1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(4, (int32_t[]) { 0, 4, 0, 0 }, 4, (int32_t[]) { 1, 1, 1, 1 })
);
assert_values_match(
"can_broadcast_shape_to([0, 4, 0, 0], [1, 4, 1, 1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(4, (int32_t[]) { 0, 4, 0, 0 }, 4, (int32_t[]) { 1, 4, 1, 1 })
);
assert_values_match(
"can_broadcast_shape_to([4, 3], [0, 3]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(2, (int32_t[]) { 4, 3 }, 2, (int32_t[]) { 0, 3 })
);
assert_values_match(
"can_broadcast_shape_to([4, 3], [0, 0]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(2, (int32_t[]) { 4, 3 }, 2, (int32_t[]) { 0, 0 })
);
}
void test_ndarray_broadcast_1() {
/*
# array = np.array([[19.9, 29.9, 39.9, 49.9]], dtype=np.float64)
# >>> [[19.9 29.9 39.9 49.9]]
#
# array = np.broadcast_to(array, (2, 3, 4))
# >>> [[[19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]]
# >>> [[19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]]]
#
# assery array.strides == (0, 0, 8)
*/
BEGIN_TEST();
double in_data[4] = { 19.9, 29.9, 39.9, 49.9 };
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = {1, 4};
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = sizeof(double),
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides
};
ndarray.set_strides_by_shape();
const int32_t dst_ndims = 3;
int32_t dst_shape[dst_ndims] = {2, 3, 4};
int32_t dst_strides[dst_ndims] = {};
NDArray<int32_t> dst_ndarray = {
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides
};
ndarray.broadcast_to(&dst_ndarray);
assert_arrays_match("dst_ndarray->strides", "%d", dst_ndims, (int32_t[]) { 0, 0, 8 }, dst_ndarray.strides);
assert_values_match("dst_ndarray[0, 0, 0]", "%f", 19.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 0, 0})));
assert_values_match("dst_ndarray[0, 0, 1]", "%f", 29.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 0, 1})));
assert_values_match("dst_ndarray[0, 0, 2]", "%f", 39.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 0, 2})));
assert_values_match("dst_ndarray[0, 0, 3]", "%f", 49.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 0, 3})));
assert_values_match("dst_ndarray[0, 1, 0]", "%f", 19.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 1, 0})));
assert_values_match("dst_ndarray[0, 1, 1]", "%f", 29.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 1, 1})));
assert_values_match("dst_ndarray[0, 1, 2]", "%f", 39.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 1, 2})));
assert_values_match("dst_ndarray[0, 1, 3]", "%f", 49.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 1, 3})));
assert_values_match("dst_ndarray[1, 2, 3]", "%f", 49.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {1, 2, 3})));
}
void test_assign_with() {
/*
```
xs = np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0], [7.0, 8.0, 9.0]], dtype=np.float64)
ys = xs.shape
```
*/
}
int main() {
test_calc_size_from_shape_normal();
test_calc_size_from_shape_has_zero();
test_set_strides_by_shape();
test_ndarray_indices_iter_normal();
test_ndarray_fill_generic();
test_ndarray_set_to_eye();
test_slice_1();
test_slice_2();
test_slice_3();
test_slice_4();
test_ndslice_1();
test_ndslice_2();
test_can_broadcast_shape();
test_ndarray_broadcast_1();
test_assign_with();
return 0;
}

View File

@ -0,0 +1,14 @@
#pragma once
// This is made toggleable since `irrt_test.cpp` itself would include
// headers that define the `int_t` family.
#ifndef IRRT_DONT_TYPEDEF_INTS
typedef _BitInt(8) int8_t;
typedef unsigned _BitInt(8) uint8_t;
typedef _BitInt(32) int32_t;
typedef unsigned _BitInt(32) uint32_t;
typedef _BitInt(64) int64_t;
typedef unsigned _BitInt(64) uint64_t;
#endif
typedef int32_t SliceIndex;

View File

@ -0,0 +1,37 @@
#pragma once
#include "irrt_typedefs.hpp"
namespace {
template <typename T>
T max(T a, T b) {
return a > b ? a : b;
}
template <typename T>
T min(T a, T b) {
return a > b ? b : a;
}
template <typename T>
bool arrays_match(int len, T *as, T *bs) {
for (int i = 0; i < len; i++) {
if (as[i] != bs[i]) return false;
}
return true;
}
void irrt_panic() {
// Crash the program for now.
// TODO: Don't crash the program
// ... or at least produce a good message when doing testing IRRT
uint8_t* death = nullptr;
*death = 0; // TODO: address 0 on hardware might be writable?
}
// TODO: Make this a macro and allow it to be toggled on/off (e.g., debug vs release)
void irrt_assert(bool condition) {
if (!condition) irrt_panic();
}
}

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@ -3,10 +3,13 @@ use crate::{
toplevel::DefinitionId,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
typedef::{
into_var_map, FunSignature, FuncArg, Type, TypeEnum, TypeVar, TypeVarId, Unifier,
},
},
};
use indexmap::IndexMap;
use nac3parser::ast::StrRef;
use std::collections::HashMap;
@ -44,13 +47,10 @@ pub enum ConcreteTypeEnum {
TTuple {
ty: Vec<ConcreteType>,
},
TList {
ty: ConcreteType,
},
TObj {
obj_id: DefinitionId,
fields: HashMap<StrRef, (ConcreteType, bool)>,
params: HashMap<u32, ConcreteType>,
params: IndexMap<TypeVarId, ConcreteType>,
},
TVirtual {
ty: ConcreteType,
@ -58,11 +58,15 @@ pub enum ConcreteTypeEnum {
TFunc {
args: Vec<ConcreteFuncArg>,
ret: ConcreteType,
vars: HashMap<u32, ConcreteType>,
vars: HashMap<TypeVarId, ConcreteType>,
},
TLiteral {
values: Vec<SymbolValue>,
},
}
impl ConcreteTypeStore {
#[must_use]
pub fn new() -> ConcreteTypeStore {
ConcreteTypeStore {
store: vec![
@ -80,6 +84,7 @@ impl ConcreteTypeStore {
}
}
#[must_use]
pub fn get(&self, cty: ConcreteType) -> &ConcreteTypeEnum {
&self.store[cty.0]
}
@ -159,9 +164,6 @@ impl ConcreteTypeStore {
.map(|t| self.from_unifier_type(unifier, primitives, *t, cache))
.collect(),
},
TypeEnum::TList { ty } => ConcreteTypeEnum::TList {
ty: self.from_unifier_type(unifier, primitives, *ty, cache),
},
TypeEnum::TObj { obj_id, fields, params } => ConcreteTypeEnum::TObj {
obj_id: *obj_id,
fields: fields
@ -194,9 +196,12 @@ impl ConcreteTypeStore {
ty: self.from_unifier_type(unifier, primitives, *ty, cache),
},
TypeEnum::TFunc(signature) => {
self.from_signature(unifier, primitives, &*signature, cache)
self.from_signature(unifier, primitives, signature, cache)
}
_ => unreachable!(),
TypeEnum::TLiteral { values, .. } => {
ConcreteTypeEnum::TLiteral { values: values.clone() }
}
_ => unreachable!("{:?}", ty_enum.get_type_name()),
};
let index = if let Some(ConcreteType(index)) = cache.get(&ty).unwrap() {
self.store[*index] = result;
@ -221,7 +226,7 @@ impl ConcreteTypeStore {
return if let Some(ty) = ty {
*ty
} else {
*ty = Some(unifier.get_dummy_var().0);
*ty = Some(unifier.get_dummy_var().ty);
ty.unwrap()
};
}
@ -249,9 +254,6 @@ impl ConcreteTypeStore {
.map(|cty| self.to_unifier_type(unifier, primitives, *cty, cache))
.collect(),
},
ConcreteTypeEnum::TList { ty } => {
TypeEnum::TList { ty: self.to_unifier_type(unifier, primitives, *ty, cache) }
}
ConcreteTypeEnum::TVirtual { ty } => {
TypeEnum::TVirtual { ty: self.to_unifier_type(unifier, primitives, *ty, cache) }
}
@ -263,10 +265,10 @@ impl ConcreteTypeStore {
(*name, (self.to_unifier_type(unifier, primitives, cty.0, cache), cty.1))
})
.collect::<HashMap<_, _>>(),
params: params
.iter()
.map(|(id, cty)| (*id, self.to_unifier_type(unifier, primitives, *cty, cache)))
.collect::<HashMap<_, _>>(),
params: into_var_map(params.iter().map(|(&id, cty)| {
let ty = self.to_unifier_type(unifier, primitives, *cty, cache);
TypeVar { id, ty }
})),
},
ConcreteTypeEnum::TFunc { args, ret, vars } => TypeEnum::TFunc(FunSignature {
args: args
@ -278,11 +280,14 @@ impl ConcreteTypeStore {
})
.collect(),
ret: self.to_unifier_type(unifier, primitives, *ret, cache),
vars: vars
.iter()
.map(|(id, cty)| (*id, self.to_unifier_type(unifier, primitives, *cty, cache)))
.collect::<HashMap<_, _>>(),
vars: into_var_map(vars.iter().map(|(&id, cty)| {
let ty = self.to_unifier_type(unifier, primitives, *cty, cache);
TypeVar { id, ty }
})),
}),
ConcreteTypeEnum::TLiteral { values, .. } => {
TypeEnum::TLiteral { values: values.clone(), loc: None }
}
};
let result = unifier.add_ty(result);
if let Some(ty) = cache.get(&cty).unwrap() {

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,132 @@
use inkwell::attributes::{Attribute, AttributeLoc};
use inkwell::values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue};
use itertools::Either;
use crate::codegen::CodeGenContext;
/// Macro to generate extern function
/// Both function return type and function parameter type are `FloatValue`
///
/// Arguments:
/// * `unary/binary`: Whether the extern function requires one (unary) or two (binary) operands
/// * `$fn_name:ident`: The identifier of the rust function to be generated
/// * `$extern_fn:literal`: Name of underlying extern function
///
/// Optional Arguments:
/// * `$(,$attributes:literal)*)`: Attributes linked with the extern function
/// The default attributes are "mustprogress", "nofree", "nounwind", "willreturn", and "writeonly"
/// These will be used unless other attributes are specified
/// * `$(,$args:ident)*`: Operands of the extern function
/// The data type of these operands will be set to `FloatValue`
///
macro_rules! generate_extern_fn {
("unary", $fn_name:ident, $extern_fn:literal) => {
generate_extern_fn!($fn_name, $extern_fn, arg, "mustprogress", "nofree", "nounwind", "willreturn", "writeonly");
};
("unary", $fn_name:ident, $extern_fn:literal $(,$attributes:literal)*) => {
generate_extern_fn!($fn_name, $extern_fn, arg $(,$attributes)*);
};
("binary", $fn_name:ident, $extern_fn:literal) => {
generate_extern_fn!($fn_name, $extern_fn, arg1, arg2, "mustprogress", "nofree", "nounwind", "willreturn", "writeonly");
};
("binary", $fn_name:ident, $extern_fn:literal $(,$attributes:literal)*) => {
generate_extern_fn!($fn_name, $extern_fn, arg1, arg2 $(,$attributes)*);
};
($fn_name:ident, $extern_fn:literal $(,$args:ident)* $(,$attributes:literal)*) => {
#[doc = concat!("Invokes the [`", stringify!($extern_fn), "`](https://en.cppreference.com/w/c/numeric/math/", stringify!($llvm_name), ") function." )]
pub fn $fn_name<'ctx>(
ctx: &CodeGenContext<'ctx, '_>
$(,$args: FloatValue<'ctx>)*,
name: Option<&str>,
) -> FloatValue<'ctx> {
const FN_NAME: &str = $extern_fn;
let llvm_f64 = ctx.ctx.f64_type();
$(debug_assert_eq!($args.get_type(), llvm_f64);)*
let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
let fn_type = llvm_f64.fn_type(&[$($args.get_type().into()),*], false);
let func = ctx.module.add_function(FN_NAME, fn_type, None);
for attr in [$($attributes),*] {
func.add_attribute(
AttributeLoc::Function,
ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
);
}
func
});
ctx.builder
.build_call(extern_fn, &[$($args.into()),*], name.unwrap_or_default())
.map(CallSiteValue::try_as_basic_value)
.map(|v| v.map_left(BasicValueEnum::into_float_value))
.map(Either::unwrap_left)
.unwrap()
}
};
}
generate_extern_fn!("unary", call_tan, "tan");
generate_extern_fn!("unary", call_asin, "asin");
generate_extern_fn!("unary", call_acos, "acos");
generate_extern_fn!("unary", call_atan, "atan");
generate_extern_fn!("unary", call_sinh, "sinh");
generate_extern_fn!("unary", call_cosh, "cosh");
generate_extern_fn!("unary", call_tanh, "tanh");
generate_extern_fn!("unary", call_asinh, "asinh");
generate_extern_fn!("unary", call_acosh, "acosh");
generate_extern_fn!("unary", call_atanh, "atanh");
generate_extern_fn!("unary", call_expm1, "expm1");
generate_extern_fn!(
"unary",
call_cbrt,
"cbrt",
"mustprogress",
"nofree",
"nosync",
"nounwind",
"readonly",
"willreturn"
);
generate_extern_fn!("unary", call_erf, "erf", "nounwind");
generate_extern_fn!("unary", call_erfc, "erfc", "nounwind");
generate_extern_fn!("unary", call_j1, "j1", "nounwind");
generate_extern_fn!("binary", call_atan2, "atan2");
generate_extern_fn!("binary", call_hypot, "hypot", "nounwind");
generate_extern_fn!("binary", call_nextafter, "nextafter", "nounwind");
/// Invokes the [`ldexp`](https://en.cppreference.com/w/c/numeric/math/ldexp) function.
pub fn call_ldexp<'ctx>(
ctx: &CodeGenContext<'ctx, '_>,
arg: FloatValue<'ctx>,
exp: IntValue<'ctx>,
name: Option<&str>,
) -> FloatValue<'ctx> {
const FN_NAME: &str = "ldexp";
let llvm_f64 = ctx.ctx.f64_type();
let llvm_i32 = ctx.ctx.i32_type();
debug_assert_eq!(arg.get_type(), llvm_f64);
debug_assert_eq!(exp.get_type(), llvm_i32);
let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
let fn_type = llvm_f64.fn_type(&[llvm_f64.into(), llvm_i32.into()], false);
let func = ctx.module.add_function(FN_NAME, fn_type, None);
for attr in ["mustprogress", "nofree", "nounwind", "willreturn"] {
func.add_attribute(
AttributeLoc::Function,
ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
);
}
func
});
ctx.builder
.build_call(extern_fn, &[arg.into(), exp.into()], name.unwrap_or_default())
.map(CallSiteValue::try_as_basic_value)
.map(|v| v.map_left(BasicValueEnum::into_float_value))
.map(Either::unwrap_left)
.unwrap()
}

View File

@ -1,5 +1,5 @@
use crate::{
codegen::{expr::*, stmt::*, CodeGenContext},
codegen::{bool_to_i1, bool_to_i8, classes::ArraySliceValue, expr::*, stmt::*, CodeGenContext},
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, TopLevelDef},
typecheck::typedef::{FunSignature, Type},
@ -7,7 +7,7 @@ use crate::{
use inkwell::{
context::Context,
types::{BasicTypeEnum, IntType},
values::{BasicValueEnum, PointerValue},
values::{BasicValueEnum, IntValue, PointerValue},
};
use nac3parser::ast::{Expr, Stmt, StrRef};
@ -22,9 +22,9 @@ pub trait CodeGenerator {
/// - fun: Function signature and definition ID.
/// - params: Function parameters. Note that this does not include the object even if the
/// function is a class method.
fn gen_call<'ctx, 'a>(
fn gen_call<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
@ -39,9 +39,9 @@ pub trait CodeGenerator {
/// - signature: Function signature of the constructor.
/// - def: Class definition for the constructor class.
/// - params: Function parameters.
fn gen_constructor<'ctx, 'a>(
fn gen_constructor<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
signature: &FunSignature,
def: &TopLevelDef,
params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
@ -59,20 +59,20 @@ pub trait CodeGenerator {
/// function is a class method.
/// Note that this function should check if the function is generated in another thread (due to
/// possible race condition), see the default implementation for an example.
fn gen_func_instance<'ctx, 'a>(
fn gen_func_instance<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, &mut TopLevelDef, String),
id: usize,
) -> Result<String, String> {
gen_func_instance(ctx, obj, fun, id)
gen_func_instance(ctx, &obj, fun, id)
}
/// Generate the code for an expression.
fn gen_expr<'ctx, 'a>(
fn gen_expr<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
expr: &Expr<Option<Type>>,
) -> Result<Option<ValueEnum<'ctx>>, String>
where
@ -83,30 +83,44 @@ pub trait CodeGenerator {
/// Allocate memory for a variable and return a pointer pointing to it.
/// The default implementation places the allocations at the start of the function.
fn gen_var_alloc<'ctx, 'a>(
fn gen_var_alloc<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
ty: BasicTypeEnum<'ctx>,
name: Option<&str>,
) -> Result<PointerValue<'ctx>, String> {
gen_var(ctx, ty)
gen_var(ctx, ty, name)
}
/// Allocate memory for a variable and return a pointer pointing to it.
/// The default implementation places the allocations at the start of the function.
fn gen_array_var_alloc<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
ty: BasicTypeEnum<'ctx>,
size: IntValue<'ctx>,
name: Option<&'ctx str>,
) -> Result<ArraySliceValue<'ctx>, String> {
gen_array_var(ctx, ty, size, name)
}
/// Return a pointer pointing to the target of the expression.
fn gen_store_target<'ctx, 'a>(
fn gen_store_target<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
pattern: &Expr<Option<Type>>,
) -> Result<PointerValue<'ctx>, String>
name: Option<&str>,
) -> Result<Option<PointerValue<'ctx>>, String>
where
Self: Sized,
{
gen_store_target(self, ctx, pattern)
gen_store_target(self, ctx, pattern, name)
}
/// Generate code for an assignment expression.
fn gen_assign<'ctx, 'a>(
fn gen_assign<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
target: &Expr<Option<Type>>,
value: ValueEnum<'ctx>,
) -> Result<(), String>
@ -118,9 +132,9 @@ pub trait CodeGenerator {
/// Generate code for a while expression.
/// Return true if the while loop must early return
fn gen_while<'ctx, 'a>(
fn gen_while(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'_, '_>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -129,11 +143,11 @@ pub trait CodeGenerator {
gen_while(self, ctx, stmt)
}
/// Generate code for a while expression.
/// Return true if the while loop must early return
fn gen_for<'ctx, 'a>(
/// Generate code for a for expression.
/// Return true if the for loop must early return
fn gen_for(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'_, '_>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -144,9 +158,9 @@ pub trait CodeGenerator {
/// Generate code for an if expression.
/// Return true if the statement must early return
fn gen_if<'ctx, 'a>(
fn gen_if(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'_, '_>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -155,9 +169,9 @@ pub trait CodeGenerator {
gen_if(self, ctx, stmt)
}
fn gen_with<'ctx, 'a>(
fn gen_with(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'_, '_>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -167,10 +181,11 @@ pub trait CodeGenerator {
}
/// Generate code for a statement
///
/// Return true if the statement must early return
fn gen_stmt<'ctx, 'a>(
fn gen_stmt(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'_, '_>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -178,6 +193,36 @@ pub trait CodeGenerator {
{
gen_stmt(self, ctx, stmt)
}
/// Generates code for a block statement.
fn gen_block<'a, I: Iterator<Item = &'a Stmt<Option<Type>>>>(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
stmts: I,
) -> Result<(), String>
where
Self: Sized,
{
gen_block(self, ctx, stmts)
}
/// See [`bool_to_i1`].
fn bool_to_i1<'ctx>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
bool_value: IntValue<'ctx>,
) -> IntValue<'ctx> {
bool_to_i1(&ctx.builder, bool_value)
}
/// See [`bool_to_i8`].
fn bool_to_i8<'ctx>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
bool_value: IntValue<'ctx>,
) -> IntValue<'ctx> {
bool_to_i8(&ctx.builder, ctx.ctx, bool_value)
}
}
pub struct DefaultCodeGenerator {
@ -186,17 +231,20 @@ pub struct DefaultCodeGenerator {
}
impl DefaultCodeGenerator {
#[must_use]
pub fn new(name: String, size_t: u32) -> DefaultCodeGenerator {
assert!(size_t == 32 || size_t == 64);
assert!(matches!(size_t, 32 | 64));
DefaultCodeGenerator { name, size_t }
}
}
impl CodeGenerator for DefaultCodeGenerator {
/// Returns the name for this [`CodeGenerator`].
fn get_name(&self) -> &str {
&self.name
}
/// Returns an LLVM integer type representing `size_t`.
fn get_size_type<'ctx>(&self, ctx: &'ctx Context) -> IntType<'ctx> {
// it should be unsigned, but we don't really need unsigned and this could save us from
// having to do a bit cast...

View File

@ -1,140 +0,0 @@
typedef _ExtInt(8) int8_t;
typedef unsigned _ExtInt(8) uint8_t;
typedef _ExtInt(32) int32_t;
typedef unsigned _ExtInt(32) uint32_t;
typedef _ExtInt(64) int64_t;
typedef unsigned _ExtInt(64) uint64_t;
# define MAX(a, b) (a > b ? a : b)
# define MIN(a, b) (a > b ? b : a)
// adapted from GNU Scientific Library: https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
// need to make sure `exp >= 0` before calling this function
#define DEF_INT_EXP(T) T __nac3_int_exp_##T( \
T base, \
T exp \
) { \
T res = (T)1; \
/* repeated squaring method */ \
do { \
if (exp & 1) res *= base; /* for n odd */ \
exp >>= 1; \
base *= base; \
} while (exp); \
return res; \
} \
DEF_INT_EXP(int32_t)
DEF_INT_EXP(int64_t)
DEF_INT_EXP(uint32_t)
DEF_INT_EXP(uint64_t)
int32_t __nac3_slice_index_bound(int32_t i, const int32_t len) {
if (i < 0) {
i = len + i;
}
if (i < 0) {
return 0;
} else if (i > len) {
return len;
}
return i;
}
int32_t __nac3_range_slice_len(const int32_t start, const int32_t end, const int32_t step) {
int32_t diff = end - start;
if (diff > 0 && step > 0) {
return ((diff - 1) / step) + 1;
} else if (diff < 0 && step < 0) {
return ((diff + 1) / step) + 1;
} else {
return 0;
}
}
// Handle list assignment and dropping part of the list when
// both dest_step and src_step are +1.
// - All the index must *not* be out-of-bound or negative,
// - The end index is *inclusive*,
// - The length of src and dest slice size should already
// be checked: if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest)
int32_t __nac3_list_slice_assign_var_size(
int32_t dest_start,
int32_t dest_end,
int32_t dest_step,
uint8_t *dest_arr,
int32_t dest_arr_len,
int32_t src_start,
int32_t src_end,
int32_t src_step,
uint8_t *src_arr,
int32_t src_arr_len,
const int32_t size
) {
/* if dest_arr_len == 0, do nothing since we do not support extending list */
if (dest_arr_len == 0) return dest_arr_len;
/* if both step is 1, memmove directly, handle the dropping of the list, and shrink size */
if (src_step == dest_step && dest_step == 1) {
const int32_t src_len = (src_end >= src_start) ? (src_end - src_start + 1) : 0;
const int32_t dest_len = (dest_end >= dest_start) ? (dest_end - dest_start + 1) : 0;
if (src_len > 0) {
__builtin_memmove(
dest_arr + dest_start * size,
src_arr + src_start * size,
src_len * size
);
}
if (dest_len > 0) {
/* dropping */
__builtin_memmove(
dest_arr + (dest_start + src_len) * size,
dest_arr + (dest_end + 1) * size,
(dest_arr_len - dest_end - 1) * size
);
}
/* shrink size */
return dest_arr_len - (dest_len - src_len);
}
/* if two range overlaps, need alloca */
uint8_t need_alloca =
(dest_arr == src_arr)
&& !(
MAX(dest_start, dest_end) < MIN(src_start, src_end)
|| MAX(src_start, src_end) < MIN(dest_start, dest_end)
);
if (need_alloca) {
uint8_t *tmp = __builtin_alloca(src_arr_len * size);
__builtin_memcpy(tmp, src_arr, src_arr_len * size);
src_arr = tmp;
}
int32_t src_ind = src_start;
int32_t dest_ind = dest_start;
for (;
(src_step > 0) ? (src_ind <= src_end) : (src_ind >= src_end);
src_ind += src_step, dest_ind += dest_step
) {
/* for constant optimization */
if (size == 1) {
__builtin_memcpy(dest_arr + dest_ind, src_arr + src_ind, 1);
} else if (size == 4) {
__builtin_memcpy(dest_arr + dest_ind * 4, src_arr + src_ind * 4, 4);
} else if (size == 8) {
__builtin_memcpy(dest_arr + dest_ind * 8, src_arr + src_ind * 8, 8);
} else {
/* memcpy for var size, cannot overlap after previous alloca */
__builtin_memcpy(dest_arr + dest_ind * size, src_arr + src_ind * size, size);
}
}
/* only dest_step == 1 can we shrink the dest list. */
/* size should be ensured prior to calling this function */
if (dest_step == 1 && dest_end >= dest_start) {
__builtin_memmove(
dest_arr + dest_ind * size,
dest_arr + (dest_end + 1) * size,
(dest_arr_len - dest_end - 1) * size
);
return dest_arr_len - (dest_end - dest_ind) - 1;
}
return dest_arr_len;
}

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,26 @@
#[cfg(test)]
mod tests {
use std::{path::Path, process::Command};
#[test]
fn run_irrt_test() {
assert!(
cfg!(feature = "test"),
"Please do `cargo test -F test` to compile `irrt_test.out` and run test"
);
let irrt_test_out_path = Path::new(concat!(env!("OUT_DIR"), "/irrt_test.out"));
let output = Command::new(irrt_test_out_path.to_str().unwrap()).output().unwrap();
if !output.status.success() {
eprintln!("irrt_test failed with status {}:", output.status);
eprintln!("====== stdout ======");
eprintln!("{}", String::from_utf8(output.stdout).unwrap());
eprintln!("====== stderr ======");
eprintln!("{}", String::from_utf8(output.stderr).unwrap());
eprintln!("====================");
panic!("irrt_test failed");
}
}
}

View File

@ -0,0 +1,308 @@
use crate::codegen::CodeGenContext;
use inkwell::context::Context;
use inkwell::intrinsics::Intrinsic;
use inkwell::types::AnyTypeEnum::IntType;
use inkwell::types::FloatType;
use inkwell::values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue, PointerValue};
use inkwell::AddressSpace;
use itertools::Either;
/// Returns the string representation for the floating-point type `ft` when used in intrinsic
/// functions.
fn get_float_intrinsic_repr(ctx: &Context, ft: FloatType) -> &'static str {
// Standard LLVM floating-point types
if ft == ctx.f16_type() {
return "f16";
}
if ft == ctx.f32_type() {
return "f32";
}
if ft == ctx.f64_type() {
return "f64";
}
if ft == ctx.f128_type() {
return "f128";
}
// Non-standard floating-point types
if ft == ctx.x86_f80_type() {
return "f80";
}
if ft == ctx.ppc_f128_type() {
return "ppcf128";
}
unreachable!()
}
/// Invokes the [`llvm.stacksave`](https://llvm.org/docs/LangRef.html#llvm-stacksave-intrinsic)
/// intrinsic.
pub fn call_stacksave<'ctx>(
ctx: &CodeGenContext<'ctx, '_>,
name: Option<&str>,
) -> PointerValue<'ctx> {
const FN_NAME: &str = "llvm.stacksave";
let intrinsic_fn = Intrinsic::find(FN_NAME)
.and_then(|intrinsic| intrinsic.get_declaration(&ctx.module, &[]))
.unwrap();
ctx.builder
.build_call(intrinsic_fn, &[], name.unwrap_or_default())
.map(CallSiteValue::try_as_basic_value)
.map(|v| v.map_left(BasicValueEnum::into_pointer_value))
.map(Either::unwrap_left)
.unwrap()
}
/// Invokes the
/// [`llvm.stackrestore`](https://llvm.org/docs/LangRef.html#llvm-stackrestore-intrinsic) intrinsic.
///
/// - `ptr`: The pointer storing the address to restore the stack to.
pub fn call_stackrestore<'ctx>(ctx: &CodeGenContext<'ctx, '_>, ptr: PointerValue<'ctx>) {
const FN_NAME: &str = "llvm.stackrestore";
/*
SEE https://github.com/TheDan64/inkwell/issues/496
We want `llvm.stackrestore`, but the following would generate `llvm.stackrestore.p0i8`.
```ignore
let intrinsic_fn = Intrinsic::find(FN_NAME)
.and_then(|intrinsic| intrinsic.get_declaration(&ctx.module, &[llvm_p0i8.into()]))
.unwrap();
```
Temp workaround by manually declaring the intrinsic with the correct function name instead.
*/
let intrinsic_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
let llvm_void = ctx.ctx.void_type();
let llvm_i8 = ctx.ctx.i8_type();
let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
let fn_type = llvm_void.fn_type(&[llvm_p0i8.into()], false);
ctx.module.add_function(FN_NAME, fn_type, None)
});
ctx.builder.build_call(intrinsic_fn, &[ptr.into()], "").unwrap();
}
/// Invokes the [`llvm.memcpy`](https://llvm.org/docs/LangRef.html#llvm-memcpy-intrinsic) intrinsic.
///
/// * `dest` - The pointer to the destination. Must be a pointer to an integer type.
/// * `src` - The pointer to the source. Must be a pointer to an integer type.
/// * `len` - The number of bytes to copy.
/// * `is_volatile` - Whether the `memcpy` operation should be `volatile`.
pub fn call_memcpy<'ctx>(
ctx: &CodeGenContext<'ctx, '_>,
dest: PointerValue<'ctx>,
src: PointerValue<'ctx>,
len: IntValue<'ctx>,
is_volatile: IntValue<'ctx>,
) {
const FN_NAME: &str = "llvm.memcpy";
debug_assert!(dest.get_type().get_element_type().is_int_type());
debug_assert!(src.get_type().get_element_type().is_int_type());
debug_assert_eq!(
dest.get_type().get_element_type().into_int_type().get_bit_width(),
src.get_type().get_element_type().into_int_type().get_bit_width(),
);
debug_assert!(matches!(len.get_type().get_bit_width(), 32 | 64));
debug_assert_eq!(is_volatile.get_type().get_bit_width(), 1);
let llvm_dest_t = dest.get_type();
let llvm_src_t = src.get_type();
let llvm_len_t = len.get_type();
let intrinsic_fn = Intrinsic::find(FN_NAME)
.and_then(|intrinsic| {
intrinsic.get_declaration(
&ctx.module,
&[llvm_dest_t.into(), llvm_src_t.into(), llvm_len_t.into()],
)
})
.unwrap();
ctx.builder
.build_call(intrinsic_fn, &[dest.into(), src.into(), len.into(), is_volatile.into()], "")
.unwrap();
}
/// Invokes the `llvm.memcpy` intrinsic.
///
/// Unlike [`call_memcpy`], this function accepts any type of pointer value. If `dest` or `src` is
/// not a pointer to an integer, the pointer(s) will be cast to `i8*` before invoking `memcpy`.
pub fn call_memcpy_generic<'ctx>(
ctx: &CodeGenContext<'ctx, '_>,
dest: PointerValue<'ctx>,
src: PointerValue<'ctx>,
len: IntValue<'ctx>,
is_volatile: IntValue<'ctx>,
) {
let llvm_i8 = ctx.ctx.i8_type();
let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
let dest_elem_t = dest.get_type().get_element_type();
let src_elem_t = src.get_type().get_element_type();
let dest = if matches!(dest_elem_t, IntType(t) if t.get_bit_width() == 8) {
dest
} else {
ctx.builder
.build_bitcast(dest, llvm_p0i8, "")
.map(BasicValueEnum::into_pointer_value)
.unwrap()
};
let src = if matches!(src_elem_t, IntType(t) if t.get_bit_width() == 8) {
src
} else {
ctx.builder
.build_bitcast(src, llvm_p0i8, "")
.map(BasicValueEnum::into_pointer_value)
.unwrap()
};
call_memcpy(ctx, dest, src, len, is_volatile);
}
/// Macro to find and generate build call for llvm intrinsic (body of llvm intrinsic function)
///
/// Arguments:
/// * `$ctx:ident`: Reference to the current Code Generation Context
/// * `$name:ident`: Optional name to be assigned to the llvm build call (Option<&str>)
/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function
/// * `$map_fn:ident`: Mapping function to be applied on `BasicValue` (`BasicValue` -> Function Return Type)
/// Use `BasicValueEnum::into_int_value` for Integer return type and `BasicValueEnum::into_float_value` for Float return type
/// * `$llvm_ty:ident`: Type of first operand
/// * `,($val:ident)*`: Comma separated list of operands
macro_rules! generate_llvm_intrinsic_fn_body {
($ctx:ident, $name:ident, $llvm_name:literal, $map_fn:expr, $llvm_ty:ident $(,$val:ident)*) => {{
const FN_NAME: &str = concat!("llvm.", $llvm_name);
let intrinsic_fn = Intrinsic::find(FN_NAME).and_then(|intrinsic| intrinsic.get_declaration(&$ctx.module, &[$llvm_ty.into()])).unwrap();
$ctx.builder.build_call(intrinsic_fn, &[$($val.into()),*], $name.unwrap_or_default()).map(CallSiteValue::try_as_basic_value).map(|v| v.map_left($map_fn)).map(Either::unwrap_left).unwrap()
}};
}
/// Macro to generate the llvm intrinsic function using [`generate_llvm_intrinsic_fn_body`].
///
/// Arguments:
/// * `float/int`: Indicates the return and argument type of the function
/// * `$fn_name:ident`: The identifier of the rust function to be generated
/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function
/// Omit "llvm." prefix from the function name i.e. use "ceil" instead of "llvm.ceil"
/// * `$val:ident`: The operand for unary operations
/// * `$val1:ident`, `$val2:ident`: The operands for binary operations
macro_rules! generate_llvm_intrinsic_fn {
("float", $fn_name:ident, $llvm_name:literal, $val:ident) => {
#[doc = concat!("Invokes the [`", stringify!($llvm_name), "`](https://llvm.org/docs/LangRef.html#llvm-", stringify!($llvm_name), "-intrinsic) intrinsic." )]
pub fn $fn_name<'ctx> (
ctx: &CodeGenContext<'ctx, '_>,
$val: FloatValue<'ctx>,
name: Option<&str>,
) -> FloatValue<'ctx> {
let llvm_ty = $val.get_type();
generate_llvm_intrinsic_fn_body!(ctx, name, $llvm_name, BasicValueEnum::into_float_value, llvm_ty, $val)
}
};
("float", $fn_name:ident, $llvm_name:literal, $val1:ident, $val2:ident) => {
#[doc = concat!("Invokes the [`", stringify!($llvm_name), "`](https://llvm.org/docs/LangRef.html#llvm-", stringify!($llvm_name), "-intrinsic) intrinsic." )]
pub fn $fn_name<'ctx> (
ctx: &CodeGenContext<'ctx, '_>,
$val1: FloatValue<'ctx>,
$val2: FloatValue<'ctx>,
name: Option<&str>,
) -> FloatValue<'ctx> {
debug_assert_eq!($val1.get_type(), $val2.get_type());
let llvm_ty = $val1.get_type();
generate_llvm_intrinsic_fn_body!(ctx, name, $llvm_name, BasicValueEnum::into_float_value, llvm_ty, $val1, $val2)
}
};
("int", $fn_name:ident, $llvm_name:literal, $val1:ident, $val2:ident) => {
#[doc = concat!("Invokes the [`", stringify!($llvm_name), "`](https://llvm.org/docs/LangRef.html#llvm-", stringify!($llvm_name), "-intrinsic) intrinsic." )]
pub fn $fn_name<'ctx> (
ctx: &CodeGenContext<'ctx, '_>,
$val1: IntValue<'ctx>,
$val2: IntValue<'ctx>,
name: Option<&str>,
) -> IntValue<'ctx> {
debug_assert_eq!($val1.get_type().get_bit_width(), $val2.get_type().get_bit_width());
let llvm_ty = $val1.get_type();
generate_llvm_intrinsic_fn_body!(ctx, name, $llvm_name, BasicValueEnum::into_int_value, llvm_ty, $val1, $val2)
}
};
}
/// Invokes the [`llvm.abs`](https://llvm.org/docs/LangRef.html#llvm-abs-intrinsic) intrinsic.
///
/// * `src` - The value for which the absolute value is to be returned.
/// * `is_int_min_poison` - Whether `poison` is to be returned if `src` is `INT_MIN`.
pub fn call_int_abs<'ctx>(
ctx: &CodeGenContext<'ctx, '_>,
src: IntValue<'ctx>,
is_int_min_poison: IntValue<'ctx>,
name: Option<&str>,
) -> IntValue<'ctx> {
debug_assert_eq!(is_int_min_poison.get_type().get_bit_width(), 1);
debug_assert!(is_int_min_poison.is_const());
let src_type = src.get_type();
generate_llvm_intrinsic_fn_body!(
ctx,
name,
"abs",
BasicValueEnum::into_int_value,
src_type,
src,
is_int_min_poison
)
}
generate_llvm_intrinsic_fn!("int", call_int_smax, "smax", a, b);
generate_llvm_intrinsic_fn!("int", call_int_smin, "smin", a, b);
generate_llvm_intrinsic_fn!("int", call_int_umax, "umax", a, b);
generate_llvm_intrinsic_fn!("int", call_int_umin, "umin", a, b);
generate_llvm_intrinsic_fn!("int", call_expect, "expect", val, expected_val);
generate_llvm_intrinsic_fn!("float", call_float_sqrt, "sqrt", val);
generate_llvm_intrinsic_fn!("float", call_float_sin, "sin", val);
generate_llvm_intrinsic_fn!("float", call_float_cos, "cos", val);
generate_llvm_intrinsic_fn!("float", call_float_pow, "pow", val, power);
generate_llvm_intrinsic_fn!("float", call_float_exp, "exp", val);
generate_llvm_intrinsic_fn!("float", call_float_exp2, "exp2", val);
generate_llvm_intrinsic_fn!("float", call_float_log, "log", val);
generate_llvm_intrinsic_fn!("float", call_float_log10, "log10", val);
generate_llvm_intrinsic_fn!("float", call_float_log2, "log2", val);
generate_llvm_intrinsic_fn!("float", call_float_fabs, "fabs", src);
generate_llvm_intrinsic_fn!("float", call_float_minnum, "minnum", val, power);
generate_llvm_intrinsic_fn!("float", call_float_maxnum, "maxnum", val, power);
generate_llvm_intrinsic_fn!("float", call_float_copysign, "copysign", mag, sgn);
generate_llvm_intrinsic_fn!("float", call_float_floor, "floor", val);
generate_llvm_intrinsic_fn!("float", call_float_ceil, "ceil", val);
generate_llvm_intrinsic_fn!("float", call_float_round, "round", val);
generate_llvm_intrinsic_fn!("float", call_float_rint, "rint", val);
/// Invokes the [`llvm.powi`](https://llvm.org/docs/LangRef.html#llvm-powi-intrinsic) intrinsic.
pub fn call_float_powi<'ctx>(
ctx: &CodeGenContext<'ctx, '_>,
val: FloatValue<'ctx>,
power: IntValue<'ctx>,
name: Option<&str>,
) -> FloatValue<'ctx> {
const FN_NAME: &str = "llvm.powi";
let llvm_val_t = val.get_type();
let llvm_power_t = power.get_type();
let intrinsic_fn = Intrinsic::find(FN_NAME)
.and_then(|intrinsic| {
intrinsic.get_declaration(&ctx.module, &[llvm_val_t.into(), llvm_power_t.into()])
})
.unwrap();
ctx.builder
.build_call(intrinsic_fn, &[val.into(), power.into()], name.unwrap_or_default())
.map(CallSiteValue::try_as_basic_value)
.map(|v| v.map_left(BasicValueEnum::into_float_value))
.map(Either::unwrap_left)
.unwrap()
}

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@ -1,18 +1,27 @@
use crate::{
codegen::{
concrete_type::ConcreteTypeStore, CodeGenContext, CodeGenTask, DefaultCodeGenerator,
WithCall, WorkerRegistry,
classes::{ListType, NDArrayType, ProxyType, RangeType},
concrete_type::ConcreteTypeStore,
CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask,
CodeGenerator, DefaultCodeGenerator, WithCall, WorkerRegistry,
},
symbol_resolver::{SymbolResolver, ValueEnum},
toplevel::{
composer::TopLevelComposer, DefinitionId, FunInstance, TopLevelContext, TopLevelDef,
composer::{ComposerConfig, TopLevelComposer},
DefinitionId, FunInstance, TopLevelContext, TopLevelDef,
},
typecheck::{
type_inferencer::{FunctionData, Inferencer, PrimitiveStore},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
},
};
use indexmap::IndexMap;
use indoc::indoc;
use inkwell::{
targets::{InitializationConfig, Target},
OptimizationLevel,
};
use nac3parser::ast::FileName;
use nac3parser::{
ast::{fold::Fold, StrRef},
parser::parse_program,
@ -48,23 +57,23 @@ impl SymbolResolver for Resolver {
_: &PrimitiveStore,
str: StrRef,
) -> Result<Type, String> {
self.id_to_type.get(&str).cloned().ok_or_else(|| format!("cannot find symbol `{}`", str))
self.id_to_type.get(&str).copied().ok_or_else(|| format!("cannot find symbol `{str}`"))
}
fn get_symbol_value<'ctx, 'a>(
fn get_symbol_value<'ctx>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, 'a>,
_: &mut CodeGenContext<'ctx, '_>,
) -> Option<ValueEnum<'ctx>> {
unimplemented!()
}
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, String> {
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, HashSet<String>> {
self.id_to_def
.read()
.get(&id)
.cloned()
.ok_or_else(|| format!("cannot find symbol `{}`", id))
.copied()
.ok_or_else(|| HashSet::from([format!("cannot find symbol `{id}`")]))
}
fn get_string_id(&self, _: &str) -> i32 {
@ -83,9 +92,9 @@ fn test_primitives() {
d = a if c == 1 else 0
return d
"};
let statements = parse_program(source, Default::default()).unwrap();
let statements = parse_program(source, FileName::default()).unwrap();
let composer: TopLevelComposer = Default::default();
let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
let mut unifier = composer.unifier.clone();
let primitives = composer.primitives_ty;
let top_level = Arc::new(composer.make_top_level_context());
@ -94,7 +103,7 @@ fn test_primitives() {
let resolver = Arc::new(Resolver {
id_to_type: HashMap::new(),
id_to_def: RwLock::new(HashMap::new()),
class_names: Default::default(),
class_names: HashMap::default(),
}) as Arc<dyn SymbolResolver + Send + Sync>;
let threads = vec![DefaultCodeGenerator::new("test".into(), 32).into()];
@ -104,7 +113,7 @@ fn test_primitives() {
FuncArg { name: "b".into(), ty: primitives.int32, default_value: None },
],
ret: primitives.int32,
vars: HashMap::new(),
vars: VarMap::new(),
};
let mut store = ConcreteTypeStore::new();
@ -119,12 +128,12 @@ fn test_primitives() {
};
let mut virtual_checks = Vec::new();
let mut calls = HashMap::new();
let mut identifiers: HashSet<_> = ["a".into(), "b".into()].iter().cloned().collect();
let mut identifiers: HashSet<_> = ["a".into(), "b".into()].into();
let mut inferencer = Inferencer {
top_level: &top_level,
function_data: &mut function_data,
unifier: &mut unifier,
variable_mapping: Default::default(),
variable_mapping: HashMap::default(),
primitives: &primitives,
virtual_checks: &mut virtual_checks,
calls: &mut calls,
@ -148,7 +157,7 @@ fn test_primitives() {
});
let task = CodeGenTask {
subst: Default::default(),
subst: Vec::default(),
symbol_name: "testing".into(),
body: Arc::new(statements),
unifier_index: 0,
@ -181,23 +190,17 @@ fn test_primitives() {
; ModuleID = 'test'
source_filename = \"test\"
define i32 @testing(i32 %0, i32 %1) !dbg !4 {
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
define i32 @testing(i32 %0, i32 %1) local_unnamed_addr #0 !dbg !4 {
init:
%add = add i32 %0, %1, !dbg !9
%add = add i32 %1, %0, !dbg !9
%cmp = icmp eq i32 %add, 1, !dbg !10
br i1 %cmp, label %then, label %else, !dbg !10
then: ; preds = %init
br label %cont, !dbg !11
else: ; preds = %init
br label %cont, !dbg !12
cont: ; preds = %else, %then
%if_exp_result.0 = phi i32 [ %0, %then ], [ 0, %else ], !dbg !13
ret i32 %if_exp_result.0, !dbg !14
%. = select i1 %cmp, i32 %0, i32 0, !dbg !11
ret i32 %., !dbg !12
}
attributes #0 = { mustprogress nofree norecurse nosync nounwind readnone willreturn }
!llvm.module.flags = !{!0, !1}
!llvm.dbg.cu = !{!2}
@ -212,15 +215,20 @@ fn test_primitives() {
!8 = !{}
!9 = !DILocation(line: 1, column: 9, scope: !4)
!10 = !DILocation(line: 2, column: 15, scope: !4)
!11 = !DILocation(line: 2, column: 5, scope: !4)
!12 = !DILocation(line: 2, column: 22, scope: !4)
!13 = !DILocation(line: 0, scope: !4)
!14 = !DILocation(line: 3, column: 8, scope: !4)
!11 = !DILocation(line: 0, scope: !4)
!12 = !DILocation(line: 3, column: 8, scope: !4)
"}
.trim();
assert_eq!(expected, module.print_to_string().to_str().unwrap().trim());
})));
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, f);
Target::initialize_all(&InitializationConfig::default());
let llvm_options = CodeGenLLVMOptions {
opt_level: OptimizationLevel::Default,
target: CodeGenTargetMachineOptions::from_host_triple(),
};
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, &llvm_options, &f);
registry.add_task(task);
registry.wait_tasks_complete(handles);
}
@ -231,14 +239,14 @@ fn test_simple_call() {
a = foo(a)
return a * 2
"};
let statements_1 = parse_program(source_1, Default::default()).unwrap();
let statements_1 = parse_program(source_1, FileName::default()).unwrap();
let source_2 = indoc! { "
return a + 1
"};
let statements_2 = parse_program(source_2, Default::default()).unwrap();
let statements_2 = parse_program(source_2, FileName::default()).unwrap();
let composer: TopLevelComposer = Default::default();
let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
let mut unifier = composer.unifier.clone();
let primitives = composer.primitives_ty;
let top_level = Arc::new(composer.make_top_level_context());
@ -247,7 +255,7 @@ fn test_simple_call() {
let signature = FunSignature {
args: vec![FuncArg { name: "a".into(), ty: primitives.int32, default_value: None }],
ret: primitives.int32,
vars: HashMap::new(),
vars: VarMap::new(),
};
let fun_ty = unifier.add_ty(TypeEnum::TFunc(signature.clone()));
let mut store = ConcreteTypeStore::new();
@ -271,7 +279,7 @@ fn test_simple_call() {
let resolver = Resolver {
id_to_type: HashMap::new(),
id_to_def: RwLock::new(HashMap::new()),
class_names: Default::default(),
class_names: HashMap::default(),
};
resolver.add_id_def("foo".into(), DefinitionId(foo_id));
let resolver = Arc::new(resolver) as Arc<dyn SymbolResolver + Send + Sync>;
@ -292,12 +300,12 @@ fn test_simple_call() {
};
let mut virtual_checks = Vec::new();
let mut calls = HashMap::new();
let mut identifiers: HashSet<_> = ["a".into(), "foo".into()].iter().cloned().collect();
let mut identifiers: HashSet<_> = ["a".into(), "foo".into()].into();
let mut inferencer = Inferencer {
top_level: &top_level,
function_data: &mut function_data,
unifier: &mut unifier,
variable_mapping: Default::default(),
variable_mapping: HashMap::default(),
primitives: &primitives,
virtual_checks: &mut virtual_checks,
calls: &mut calls,
@ -326,11 +334,11 @@ fn test_simple_call() {
&mut *top_level.definitions.read()[foo_id].write()
{
instance_to_stmt.insert(
"".to_string(),
String::new(),
FunInstance {
body: Arc::new(statements_2),
calls: Arc::new(inferencer.calls.clone()),
subst: Default::default(),
subst: IndexMap::default(),
unifier_id: 0,
},
);
@ -346,7 +354,7 @@ fn test_simple_call() {
});
let task = CodeGenTask {
subst: Default::default(),
subst: Vec::default(),
symbol_name: "testing".to_string(),
body: Arc::new(statements_1),
calls: Arc::new(calls1),
@ -361,19 +369,23 @@ fn test_simple_call() {
; ModuleID = 'test'
source_filename = \"test\"
define i32 @testing(i32 %0) !dbg !5 {
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
define i32 @testing(i32 %0) local_unnamed_addr #0 !dbg !5 {
init:
%call = call i32 @foo.0(i32 %0), !dbg !10
%mul = mul i32 %call, 2, !dbg !11
ret i32 %mul, !dbg !11
%add.i = shl i32 %0, 1, !dbg !10
%mul = add i32 %add.i, 2, !dbg !10
ret i32 %mul, !dbg !10
}
define i32 @foo.0(i32 %0) !dbg !12 {
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
define i32 @foo.0(i32 %0) local_unnamed_addr #0 !dbg !11 {
init:
%add = add i32 %0, 1, !dbg !13
ret i32 %add, !dbg !13
%add = add i32 %0, 1, !dbg !12
ret i32 %add, !dbg !12
}
attributes #0 = { mustprogress nofree norecurse nosync nounwind readnone willreturn }
!llvm.module.flags = !{!0, !1}
!llvm.dbg.cu = !{!2, !4}
@ -387,15 +399,53 @@ fn test_simple_call() {
!7 = !{!8}
!8 = !DIBasicType(name: \"_\", flags: DIFlagPublic)
!9 = !{}
!10 = !DILocation(line: 1, column: 9, scope: !5)
!11 = !DILocation(line: 2, column: 12, scope: !5)
!12 = distinct !DISubprogram(name: \"foo.0\", linkageName: \"foo.0\", scope: null, file: !3, line: 1, type: !6, scopeLine: 1, flags: DIFlagPublic, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !9)
!13 = !DILocation(line: 1, column: 12, scope: !12)
!10 = !DILocation(line: 2, column: 12, scope: !5)
!11 = distinct !DISubprogram(name: \"foo.0\", linkageName: \"foo.0\", scope: null, file: !3, line: 1, type: !6, scopeLine: 1, flags: DIFlagPublic, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !9)
!12 = !DILocation(line: 1, column: 12, scope: !11)
"}
.trim();
assert_eq!(expected, module.print_to_string().to_str().unwrap().trim());
})));
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, f);
Target::initialize_all(&InitializationConfig::default());
let llvm_options = CodeGenLLVMOptions {
opt_level: OptimizationLevel::Default,
target: CodeGenTargetMachineOptions::from_host_triple(),
};
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, &llvm_options, &f);
registry.add_task(task);
registry.wait_tasks_complete(handles);
}
#[test]
fn test_classes_list_type_new() {
let ctx = inkwell::context::Context::create();
let generator = DefaultCodeGenerator::new(String::new(), 64);
let llvm_i32 = ctx.i32_type();
let llvm_usize = generator.get_size_type(&ctx);
let llvm_list = ListType::new(&generator, &ctx, llvm_i32.into());
assert!(ListType::is_type(llvm_list.as_base_type(), llvm_usize).is_ok());
}
#[test]
fn test_classes_range_type_new() {
let ctx = inkwell::context::Context::create();
let llvm_range = RangeType::new(&ctx);
assert!(RangeType::is_type(llvm_range.as_base_type()).is_ok());
}
#[test]
fn test_classes_ndarray_type_new() {
let ctx = inkwell::context::Context::create();
let generator = DefaultCodeGenerator::new(String::new(), 64);
let llvm_i32 = ctx.i32_type();
let llvm_usize = generator.get_size_type(&ctx);
let llvm_ndarray = NDArrayType::new(&generator, &ctx, llvm_i32.into());
assert!(NDArrayType::is_type(llvm_ndarray.as_base_type(), llvm_usize).is_ok());
}

View File

@ -1,7 +1,26 @@
#![warn(clippy::all)]
#![allow(dead_code)]
#![deny(
future_incompatible,
let_underscore,
nonstandard_style,
rust_2024_compatibility,
clippy::all
)]
#![warn(clippy::pedantic)]
#![allow(
dead_code,
clippy::cast_possible_truncation,
clippy::cast_sign_loss,
clippy::enum_glob_use,
clippy::missing_errors_doc,
clippy::missing_panics_doc,
clippy::module_name_repetitions,
clippy::similar_names,
clippy::too_many_lines,
clippy::wildcard_imports
)]
pub mod codegen;
pub mod symbol_resolver;
pub mod toplevel;
pub mod typecheck;
pub mod util;

View File

@ -1,22 +1,19 @@
use std::fmt::Debug;
use std::rc::Rc;
use std::sync::Arc;
use std::{collections::HashMap, fmt::Display};
use std::{collections::HashMap, collections::HashSet, fmt::Display};
use crate::typecheck::typedef::TypeEnum;
use crate::{
codegen::CodeGenContext,
toplevel::{DefinitionId, TopLevelDef},
};
use crate::{
codegen::CodeGenerator,
codegen::{CodeGenContext, CodeGenerator},
toplevel::{type_annotation::TypeAnnotation, DefinitionId, TopLevelDef},
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{Type, Unifier},
typedef::{Type, TypeEnum, Unifier, VarMap},
},
};
use inkwell::values::{BasicValueEnum, FloatValue, IntValue, PointerValue, StructValue};
use itertools::{chain, izip};
use nac3parser::ast::{Expr, Location, StrRef};
use itertools::{chain, izip, Itertools};
use nac3parser::ast::{Constant, Expr, Location, StrRef};
use parking_lot::RwLock;
#[derive(Clone, PartialEq, Debug)]
@ -33,15 +30,192 @@ pub enum SymbolValue {
OptionNone,
}
impl SymbolValue {
/// Creates a [`SymbolValue`] from a [`Constant`].
///
/// * `constant` - The constant to create the value from.
/// * `expected_ty` - The expected type of the [`SymbolValue`].
pub fn from_constant(
constant: &Constant,
expected_ty: Type,
primitives: &PrimitiveStore,
unifier: &mut Unifier,
) -> Result<Self, String> {
match constant {
Constant::None => {
if unifier.unioned(expected_ty, primitives.option) {
Ok(SymbolValue::OptionNone)
} else {
Err(format!("Expected {expected_ty:?}, but got Option"))
}
}
Constant::Bool(b) => {
if unifier.unioned(expected_ty, primitives.bool) {
Ok(SymbolValue::Bool(*b))
} else {
Err(format!("Expected {expected_ty:?}, but got bool"))
}
}
Constant::Str(s) => {
if unifier.unioned(expected_ty, primitives.str) {
Ok(SymbolValue::Str(s.to_string()))
} else {
Err(format!("Expected {expected_ty:?}, but got str"))
}
}
Constant::Int(i) => {
if unifier.unioned(expected_ty, primitives.int32) {
i32::try_from(*i).map(SymbolValue::I32).map_err(|e| e.to_string())
} else if unifier.unioned(expected_ty, primitives.int64) {
i64::try_from(*i).map(SymbolValue::I64).map_err(|e| e.to_string())
} else if unifier.unioned(expected_ty, primitives.uint32) {
u32::try_from(*i).map(SymbolValue::U32).map_err(|e| e.to_string())
} else if unifier.unioned(expected_ty, primitives.uint64) {
u64::try_from(*i).map(SymbolValue::U64).map_err(|e| e.to_string())
} else {
Err(format!("Expected {}, but got int", unifier.stringify(expected_ty)))
}
}
Constant::Tuple(t) => {
let expected_ty = unifier.get_ty(expected_ty);
let TypeEnum::TTuple { ty } = expected_ty.as_ref() else {
return Err(format!(
"Expected {:?}, but got Tuple",
expected_ty.get_type_name()
));
};
assert_eq!(ty.len(), t.len());
let elems = t
.iter()
.zip(ty)
.map(|(constant, ty)| Self::from_constant(constant, *ty, primitives, unifier))
.collect::<Result<Vec<SymbolValue>, _>>()?;
Ok(SymbolValue::Tuple(elems))
}
Constant::Float(f) => {
if unifier.unioned(expected_ty, primitives.float) {
Ok(SymbolValue::Double(*f))
} else {
Err(format!("Expected {expected_ty:?}, but got float"))
}
}
_ => Err(format!("Unsupported value type {constant:?}")),
}
}
/// Creates a [`SymbolValue`] from a [`Constant`], with its type being inferred from the constant value.
///
/// * `constant` - The constant to create the value from.
pub fn from_constant_inferred(constant: &Constant) -> Result<Self, String> {
match constant {
Constant::None => Ok(SymbolValue::OptionNone),
Constant::Bool(b) => Ok(SymbolValue::Bool(*b)),
Constant::Str(s) => Ok(SymbolValue::Str(s.to_string())),
Constant::Int(i) => {
let i = *i;
if i >= 0 {
i32::try_from(i)
.map(SymbolValue::I32)
.or_else(|_| i64::try_from(i).map(SymbolValue::I64))
.map_err(|_| {
format!("Literal cannot be expressed as any integral type: {i}")
})
} else {
u32::try_from(i)
.map(SymbolValue::U32)
.or_else(|_| u64::try_from(i).map(SymbolValue::U64))
.map_err(|_| {
format!("Literal cannot be expressed as any integral type: {i}")
})
}
}
Constant::Tuple(t) => {
let elems = t
.iter()
.map(Self::from_constant_inferred)
.collect::<Result<Vec<SymbolValue>, _>>()?;
Ok(SymbolValue::Tuple(elems))
}
Constant::Float(f) => Ok(SymbolValue::Double(*f)),
_ => Err(format!("Unsupported value type {constant:?}")),
}
}
/// Returns the [`Type`] representing the data type of this value.
pub fn get_type(&self, primitives: &PrimitiveStore, unifier: &mut Unifier) -> Type {
match self {
SymbolValue::I32(_) => primitives.int32,
SymbolValue::I64(_) => primitives.int64,
SymbolValue::U32(_) => primitives.uint32,
SymbolValue::U64(_) => primitives.uint64,
SymbolValue::Str(_) => primitives.str,
SymbolValue::Double(_) => primitives.float,
SymbolValue::Bool(_) => primitives.bool,
SymbolValue::Tuple(vs) => {
let vs_tys = vs.iter().map(|v| v.get_type(primitives, unifier)).collect::<Vec<_>>();
unifier.add_ty(TypeEnum::TTuple { ty: vs_tys })
}
SymbolValue::OptionSome(_) | SymbolValue::OptionNone => primitives.option,
}
}
/// Returns the [`TypeAnnotation`] representing the data type of this value.
pub fn get_type_annotation(
&self,
primitives: &PrimitiveStore,
unifier: &mut Unifier,
) -> TypeAnnotation {
match self {
SymbolValue::Bool(..)
| SymbolValue::Double(..)
| SymbolValue::I32(..)
| SymbolValue::I64(..)
| SymbolValue::U32(..)
| SymbolValue::U64(..)
| SymbolValue::Str(..) => TypeAnnotation::Primitive(self.get_type(primitives, unifier)),
SymbolValue::Tuple(vs) => {
let vs_tys = vs
.iter()
.map(|v| v.get_type_annotation(primitives, unifier))
.collect::<Vec<_>>();
TypeAnnotation::Tuple(vs_tys)
}
SymbolValue::OptionNone => TypeAnnotation::CustomClass {
id: primitives.option.obj_id(unifier).unwrap(),
params: Vec::default(),
},
SymbolValue::OptionSome(v) => {
let ty = v.get_type_annotation(primitives, unifier);
TypeAnnotation::CustomClass {
id: primitives.option.obj_id(unifier).unwrap(),
params: vec![ty],
}
}
}
}
/// Returns the [`TypeEnum`] representing the data type of this value.
pub fn get_type_enum(
&self,
primitives: &PrimitiveStore,
unifier: &mut Unifier,
) -> Rc<TypeEnum> {
let ty = self.get_type(primitives, unifier);
unifier.get_ty(ty)
}
}
impl Display for SymbolValue {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match self {
SymbolValue::I32(i) => write!(f, "{}", i),
SymbolValue::I64(i) => write!(f, "int64({})", i),
SymbolValue::U32(i) => write!(f, "uint32({})", i),
SymbolValue::U64(i) => write!(f, "uint64({})", i),
SymbolValue::Str(s) => write!(f, "\"{}\"", s),
SymbolValue::Double(d) => write!(f, "{}", d),
SymbolValue::I32(i) => write!(f, "{i}"),
SymbolValue::I64(i) => write!(f, "int64({i})"),
SymbolValue::U32(i) => write!(f, "uint32({i})"),
SymbolValue::U64(i) => write!(f, "uint64({i})"),
SymbolValue::Str(s) => write!(f, "\"{s}\""),
SymbolValue::Double(d) => write!(f, "{d}"),
SymbolValue::Bool(b) => {
if *b {
write!(f, "True")
@ -50,42 +224,82 @@ impl Display for SymbolValue {
}
}
SymbolValue::Tuple(t) => {
write!(f, "({})", t.iter().map(|v| format!("{}", v)).collect::<Vec<_>>().join(", "))
write!(f, "({})", t.iter().map(|v| format!("{v}")).collect::<Vec<_>>().join(", "))
}
SymbolValue::OptionSome(v) => write!(f, "Some({})", v),
SymbolValue::OptionSome(v) => write!(f, "Some({v})"),
SymbolValue::OptionNone => write!(f, "none"),
}
}
}
impl TryFrom<SymbolValue> for u64 {
type Error = ();
/// Tries to convert a [`SymbolValue`] into a [`u64`], returning [`Err`] if the value is not
/// numeric or if the value cannot be converted into a `u64` without overflow.
fn try_from(value: SymbolValue) -> Result<Self, Self::Error> {
match value {
SymbolValue::I32(v) => u64::try_from(v).map_err(|_| ()),
SymbolValue::I64(v) => u64::try_from(v).map_err(|_| ()),
SymbolValue::U32(v) => Ok(u64::from(v)),
SymbolValue::U64(v) => Ok(v),
_ => Err(()),
}
}
}
impl TryFrom<SymbolValue> for i128 {
type Error = ();
/// Tries to convert a [`SymbolValue`] into a [`i128`], returning [`Err`] if the value is not
/// numeric.
fn try_from(value: SymbolValue) -> Result<Self, Self::Error> {
match value {
SymbolValue::I32(v) => Ok(i128::from(v)),
SymbolValue::I64(v) => Ok(i128::from(v)),
SymbolValue::U32(v) => Ok(i128::from(v)),
SymbolValue::U64(v) => Ok(i128::from(v)),
_ => Err(()),
}
}
}
pub trait StaticValue {
/// Returns a unique identifier for this value.
fn get_unique_identifier(&self) -> u64;
fn get_const_obj<'ctx, 'a>(
/// Returns the constant object represented by this unique identifier.
fn get_const_obj<'ctx>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
generator: &mut dyn CodeGenerator,
) -> BasicValueEnum<'ctx>;
fn to_basic_value_enum<'ctx, 'a>(
/// Converts this value to a LLVM [`BasicValueEnum`].
fn to_basic_value_enum<'ctx>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
generator: &mut dyn CodeGenerator,
expected_ty: Type,
) -> Result<BasicValueEnum<'ctx>, String>;
fn get_field<'ctx, 'a>(
/// Returns a field within this value.
fn get_field<'ctx>(
&self,
name: StrRef,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Option<ValueEnum<'ctx>>;
/// Returns a single element of this tuple.
fn get_tuple_element<'ctx>(&self, index: u32) -> Option<ValueEnum<'ctx>>;
}
#[derive(Clone)]
pub enum ValueEnum<'ctx> {
/// [`ValueEnum`] representing a static value.
Static(Arc<dyn StaticValue + Send + Sync>),
/// [`ValueEnum`] representing a dynamic value.
Dynamic(BasicValueEnum<'ctx>),
}
@ -120,6 +334,7 @@ impl<'ctx> From<StructValue<'ctx>> for ValueEnum<'ctx> {
}
impl<'ctx> ValueEnum<'ctx> {
/// Converts this [`ValueEnum`] to a [`BasicValueEnum`].
pub fn to_basic_value_enum<'a>(
self,
ctx: &mut CodeGenContext<'ctx, 'a>,
@ -134,7 +349,7 @@ impl<'ctx> ValueEnum<'ctx> {
}
pub trait SymbolResolver {
// get type of type variable identifier or top-level function type
/// Get type of type variable identifier or top-level function type,
fn get_symbol_type(
&self,
unifier: &mut Unifier,
@ -143,16 +358,16 @@ pub trait SymbolResolver {
str: StrRef,
) -> Result<Type, String>;
// get the top-level definition of identifiers
fn get_identifier_def(&self, str: StrRef) -> Result<DefinitionId, String>;
/// Get the top-level definition of identifiers.
fn get_identifier_def(&self, str: StrRef) -> Result<DefinitionId, HashSet<String>>;
fn get_symbol_value<'ctx, 'a>(
fn get_symbol_value<'ctx>(
&self,
str: StrRef,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Option<ValueEnum<'ctx>>;
fn get_default_param_value(&self, expr: &nac3parser::ast::Expr) -> Option<SymbolValue>;
fn get_default_param_value(&self, expr: &Expr) -> Option<SymbolValue>;
fn get_string_id(&self, s: &str) -> i32;
fn get_exception_id(&self, tyid: usize) -> usize;
@ -160,7 +375,7 @@ pub trait SymbolResolver {
&self,
_unifier: &mut Unifier,
_top_level_defs: &[Arc<RwLock<TopLevelDef>>],
_primitives: &PrimitiveStore
_primitives: &PrimitiveStore,
) -> Result<(), String> {
Ok(())
}
@ -173,23 +388,23 @@ thread_local! {
"float".into(),
"bool".into(),
"virtual".into(),
"list".into(),
"tuple".into(),
"str".into(),
"Exception".into(),
"uint32".into(),
"uint64".into(),
"Literal".into(),
];
}
// convert type annotation into type
/// Converts a type annotation into a [Type].
pub fn parse_type_annotation<T>(
resolver: &dyn SymbolResolver,
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
unifier: &mut Unifier,
primitives: &PrimitiveStore,
expr: &Expr<T>,
) -> Result<Type, String> {
) -> Result<Type, HashSet<String>> {
use nac3parser::ast::ExprKind::*;
let ids = IDENTIFIER_ID.with(|ids| *ids);
let int32_id = ids[0];
@ -197,12 +412,12 @@ pub fn parse_type_annotation<T>(
let float_id = ids[2];
let bool_id = ids[3];
let virtual_id = ids[4];
let list_id = ids[5];
let tuple_id = ids[6];
let str_id = ids[7];
let exn_id = ids[8];
let uint32_id = ids[9];
let uint64_id = ids[10];
let tuple_id = ids[5];
let str_id = ids[6];
let exn_id = ids[7];
let uint32_id = ids[8];
let uint64_id = ids[9];
let literal_id = ids[10];
let name_handling = |id: &StrRef, loc: Location, unifier: &mut Unifier| {
if *id == int32_id {
@ -223,39 +438,33 @@ pub fn parse_type_annotation<T>(
Ok(primitives.exception)
} else {
let obj_id = resolver.get_identifier_def(*id);
match obj_id {
Ok(obj_id) => {
if let Ok(obj_id) = obj_id {
let def = top_level_defs[obj_id.0].read();
if let TopLevelDef::Class { fields, methods, type_vars, .. } = &*def {
if !type_vars.is_empty() {
return Err(format!(
return Err(HashSet::from([format!(
"Unexpected number of type parameters: expected {} but got 0",
type_vars.len()
));
)]));
}
let fields = chain(
fields.iter().map(|(k, v, m)| (*k, (*v, *m))),
methods.iter().map(|(k, v, _)| (*k, (*v, false))),
)
.collect();
Ok(unifier.add_ty(TypeEnum::TObj {
obj_id,
fields,
params: Default::default(),
}))
Ok(unifier.add_ty(TypeEnum::TObj { obj_id, fields, params: VarMap::default() }))
} else {
Err(format!("Cannot use function name as type at {}", loc))
Err(HashSet::from([format!("Cannot use function name as type at {loc}")]))
}
}
Err(_) => {
let ty = resolver
.get_symbol_type(unifier, top_level_defs, primitives, *id)
.map_err(|e| format!("Unknown type annotation at {}: {}", loc, e))?;
} else {
let ty =
resolver.get_symbol_type(unifier, top_level_defs, primitives, *id).map_err(
|e| HashSet::from([format!("Unknown type annotation at {loc}: {e}")]),
)?;
if let TypeEnum::TVar { .. } = &*unifier.get_ty(ty) {
Ok(ty)
} else {
Err(format!("Unknown type annotation {} at {}", id, loc))
}
Err(HashSet::from([format!("Unknown type annotation {id} at {loc}")]))
}
}
}
@ -265,9 +474,6 @@ pub fn parse_type_annotation<T>(
if *id == virtual_id {
let ty = parse_type_annotation(resolver, top_level_defs, unifier, primitives, slice)?;
Ok(unifier.add_ty(TypeEnum::TVirtual { ty }))
} else if *id == list_id {
let ty = parse_type_annotation(resolver, top_level_defs, unifier, primitives, slice)?;
Ok(unifier.add_ty(TypeEnum::TList { ty }))
} else if *id == tuple_id {
if let Tuple { elts, .. } = &slice.node {
let ty = elts
@ -278,8 +484,31 @@ pub fn parse_type_annotation<T>(
.collect::<Result<Vec<_>, _>>()?;
Ok(unifier.add_ty(TypeEnum::TTuple { ty }))
} else {
Err("Expected multiple elements for tuple".into())
Err(HashSet::from(["Expected multiple elements for tuple".into()]))
}
} else if *id == literal_id {
let mut parse_literal = |elt: &Expr<T>| {
let ty = parse_type_annotation(resolver, top_level_defs, unifier, primitives, elt)?;
let ty_enum = &*unifier.get_ty_immutable(ty);
match ty_enum {
TypeEnum::TLiteral { values, .. } => Ok(values.clone()),
_ => Err(HashSet::from([format!(
"Expected literal in type argument for Literal at {}",
elt.location
)])),
}
};
let values = if let Tuple { elts, .. } = &slice.node {
elts.iter().map(&mut parse_literal).collect::<Result<Vec<_>, _>>()?
} else {
vec![parse_literal(slice)?]
}
.into_iter()
.flatten()
.collect_vec();
Ok(unifier.get_fresh_literal(values, Some(slice.location)))
} else {
let types = if let Tuple { elts, .. } = &slice.node {
elts.iter()
@ -295,13 +524,13 @@ pub fn parse_type_annotation<T>(
let def = top_level_defs[obj_id.0].read();
if let TopLevelDef::Class { fields, methods, type_vars, .. } = &*def {
if types.len() != type_vars.len() {
return Err(format!(
return Err(HashSet::from([format!(
"Unexpected number of type parameters: expected {} but got {}",
type_vars.len(),
types.len()
));
)]));
}
let mut subst = HashMap::new();
let mut subst = VarMap::new();
for (var, ty) in izip!(type_vars.iter(), types.iter()) {
let id = if let TypeEnum::TVar { id, .. } = &*unifier.get_ty(*var) {
*id
@ -323,7 +552,7 @@ pub fn parse_type_annotation<T>(
}));
Ok(unifier.add_ty(TypeEnum::TObj { obj_id, fields, params: subst }))
} else {
Err("Cannot use function name as type".into())
Err(HashSet::from(["Cannot use function name as type".into()]))
}
}
};
@ -334,10 +563,13 @@ pub fn parse_type_annotation<T>(
if let Name { id, .. } = &value.node {
subscript_name_handle(id, slice, unifier)
} else {
Err(format!("unsupported type expression at {}", expr.location))
Err(HashSet::from([format!("unsupported type expression at {}", expr.location)]))
}
}
_ => Err(format!("unsupported type expression at {}", expr.location)),
Constant { value, .. } => SymbolValue::from_constant_inferred(value)
.map(|v| unifier.get_fresh_literal(vec![v], Some(expr.location)))
.map_err(|err| HashSet::from([err])),
_ => Err(HashSet::from([format!("unsupported type expression at {}", expr.location)])),
}
}
@ -348,7 +580,7 @@ impl dyn SymbolResolver + Send + Sync {
unifier: &mut Unifier,
primitives: &PrimitiveStore,
expr: &Expr<T>,
) -> Result<Type, String> {
) -> Result<Type, HashSet<String>> {
parse_type_annotation(self, top_level_defs, unifier, primitives, expr)
}
@ -361,13 +593,13 @@ impl dyn SymbolResolver + Send + Sync {
unifier.internal_stringify(
ty,
&mut |id| {
if let TopLevelDef::Class { name, .. } = &*top_level_defs[id].read() {
name.to_string()
} else {
let TopLevelDef::Class { name, .. } = &*top_level_defs[id].read() else {
unreachable!("expected class definition")
}
};
name.to_string()
},
&mut |id| format!("typevar{}", id),
&mut |id| format!("typevar{id}"),
&mut None,
)
}

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@ -1,10 +1,298 @@
use std::convert::TryInto;
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::numpy::unpack_ndarray_var_tys;
use crate::typecheck::typedef::{into_var_map, iter_type_vars, Mapping, TypeVarId, VarMap};
use nac3parser::ast::{Constant, Location};
use strum::IntoEnumIterator;
use strum_macros::EnumIter;
use super::*;
/// All primitive types and functions in nac3core.
#[derive(Clone, Copy, Debug, EnumIter, PartialEq, Eq)]
pub enum PrimDef {
// Classes
Int32,
Int64,
Float,
Bool,
None,
Range,
Str,
Exception,
UInt32,
UInt64,
Option,
List,
NDArray,
// Member Functions
OptionIsSome,
OptionIsNone,
OptionUnwrap,
NDArrayCopy,
NDArrayFill,
FunInt32,
FunInt64,
FunUInt32,
FunUInt64,
FunFloat,
FunNpNDArray,
FunNpEmpty,
FunNpZeros,
FunNpOnes,
FunNpFull,
FunNpArray,
FunNpEye,
FunNpIdentity,
FunRound,
FunRound64,
FunNpRound,
FunRangeInit,
FunStr,
FunBool,
FunFloor,
FunFloor64,
FunNpFloor,
FunCeil,
FunCeil64,
FunNpCeil,
FunLen,
FunMin,
FunNpMin,
FunNpMinimum,
FunMax,
FunNpMax,
FunNpMaximum,
FunAbs,
FunNpIsNan,
FunNpIsInf,
FunNpSin,
FunNpCos,
FunNpExp,
FunNpExp2,
FunNpLog,
FunNpLog10,
FunNpLog2,
FunNpFabs,
FunNpSqrt,
FunNpRint,
FunNpTan,
FunNpArcsin,
FunNpArccos,
FunNpArctan,
FunNpSinh,
FunNpCosh,
FunNpTanh,
FunNpArcsinh,
FunNpArccosh,
FunNpArctanh,
FunNpExpm1,
FunNpCbrt,
FunSpSpecErf,
FunSpSpecErfc,
FunSpSpecGamma,
FunSpSpecGammaln,
FunSpSpecJ0,
FunSpSpecJ1,
FunNpArctan2,
FunNpCopysign,
FunNpFmax,
FunNpFmin,
FunNpLdExp,
FunNpHypot,
FunNpNextAfter,
// Top-Level Functions
FunSome,
}
/// Associated details of a [`PrimDef`]
pub enum PrimDefDetails {
PrimFunction { name: &'static str, simple_name: &'static str },
PrimClass { name: &'static str },
}
impl PrimDef {
/// Get the assigned [`DefinitionId`] of this [`PrimDef`].
///
/// The assigned definition ID is defined by the position this [`PrimDef`] enum unit variant is defined at,
/// with the first `PrimDef`'s definition id being `0`.
#[must_use]
pub fn id(&self) -> DefinitionId {
DefinitionId(*self as usize)
}
/// Check if a definition ID is that of a [`PrimDef`].
#[must_use]
pub fn contains_id(id: DefinitionId) -> bool {
Self::iter().any(|prim| prim.id() == id)
}
/// Get the definition "simple name" of this [`PrimDef`].
///
/// If the [`PrimDef`] is a function, this corresponds to [`TopLevelDef::Function::simple_name`].
///
/// If the [`PrimDef`] is a class, this returns [`None`].
#[must_use]
pub fn simple_name(&self) -> &'static str {
match self.details() {
PrimDefDetails::PrimFunction { simple_name, .. } => simple_name,
PrimDefDetails::PrimClass { .. } => {
panic!("PrimDef {self:?} has no simple_name as it is not a function.")
}
}
}
/// Get the definition "name" of this [`PrimDef`].
///
/// If the [`PrimDef`] is a function, this corresponds to [`TopLevelDef::Function::name`].
///
/// If the [`PrimDef`] is a class, this corresponds to [`TopLevelDef::Class::name`].
#[must_use]
pub fn name(&self) -> &'static str {
match self.details() {
PrimDefDetails::PrimFunction { name, .. } | PrimDefDetails::PrimClass { name } => name,
}
}
/// Get the associated details of this [`PrimDef`]
#[must_use]
pub fn details(self) -> PrimDefDetails {
fn class(name: &'static str) -> PrimDefDetails {
PrimDefDetails::PrimClass { name }
}
fn fun(name: &'static str, simple_name: Option<&'static str>) -> PrimDefDetails {
PrimDefDetails::PrimFunction { simple_name: simple_name.unwrap_or(name), name }
}
match self {
PrimDef::Int32 => class("int32"),
PrimDef::Int64 => class("int64"),
PrimDef::Float => class("float"),
PrimDef::Bool => class("bool"),
PrimDef::None => class("none"),
PrimDef::Range => class("range"),
PrimDef::Str => class("str"),
PrimDef::Exception => class("Exception"),
PrimDef::UInt32 => class("uint32"),
PrimDef::UInt64 => class("uint64"),
PrimDef::Option => class("Option"),
PrimDef::OptionIsSome => fun("Option.is_some", Some("is_some")),
PrimDef::OptionIsNone => fun("Option.is_none", Some("is_none")),
PrimDef::OptionUnwrap => fun("Option.unwrap", Some("unwrap")),
PrimDef::List => class("list"),
PrimDef::NDArray => class("ndarray"),
PrimDef::NDArrayCopy => fun("ndarray.copy", Some("copy")),
PrimDef::NDArrayFill => fun("ndarray.fill", Some("fill")),
PrimDef::FunInt32 => fun("int32", None),
PrimDef::FunInt64 => fun("int64", None),
PrimDef::FunUInt32 => fun("uint32", None),
PrimDef::FunUInt64 => fun("uint64", None),
PrimDef::FunFloat => fun("float", None),
PrimDef::FunNpNDArray => fun("np_ndarray", None),
PrimDef::FunNpEmpty => fun("np_empty", None),
PrimDef::FunNpZeros => fun("np_zeros", None),
PrimDef::FunNpOnes => fun("np_ones", None),
PrimDef::FunNpFull => fun("np_full", None),
PrimDef::FunNpArray => fun("np_array", None),
PrimDef::FunNpEye => fun("np_eye", None),
PrimDef::FunNpIdentity => fun("np_identity", None),
PrimDef::FunRound => fun("round", None),
PrimDef::FunRound64 => fun("round64", None),
PrimDef::FunNpRound => fun("np_round", None),
PrimDef::FunRangeInit => fun("range.__init__", Some("__init__")),
PrimDef::FunStr => fun("str", None),
PrimDef::FunBool => fun("bool", None),
PrimDef::FunFloor => fun("floor", None),
PrimDef::FunFloor64 => fun("floor64", None),
PrimDef::FunNpFloor => fun("np_floor", None),
PrimDef::FunCeil => fun("ceil", None),
PrimDef::FunCeil64 => fun("ceil64", None),
PrimDef::FunNpCeil => fun("np_ceil", None),
PrimDef::FunLen => fun("len", None),
PrimDef::FunMin => fun("min", None),
PrimDef::FunNpMin => fun("np_min", None),
PrimDef::FunNpMinimum => fun("np_minimum", None),
PrimDef::FunMax => fun("max", None),
PrimDef::FunNpMax => fun("np_max", None),
PrimDef::FunNpMaximum => fun("np_maximum", None),
PrimDef::FunAbs => fun("abs", None),
PrimDef::FunNpIsNan => fun("np_isnan", None),
PrimDef::FunNpIsInf => fun("np_isinf", None),
PrimDef::FunNpSin => fun("np_sin", None),
PrimDef::FunNpCos => fun("np_cos", None),
PrimDef::FunNpExp => fun("np_exp", None),
PrimDef::FunNpExp2 => fun("np_exp2", None),
PrimDef::FunNpLog => fun("np_log", None),
PrimDef::FunNpLog10 => fun("np_log10", None),
PrimDef::FunNpLog2 => fun("np_log2", None),
PrimDef::FunNpFabs => fun("np_fabs", None),
PrimDef::FunNpSqrt => fun("np_sqrt", None),
PrimDef::FunNpRint => fun("np_rint", None),
PrimDef::FunNpTan => fun("np_tan", None),
PrimDef::FunNpArcsin => fun("np_arcsin", None),
PrimDef::FunNpArccos => fun("np_arccos", None),
PrimDef::FunNpArctan => fun("np_arctan", None),
PrimDef::FunNpSinh => fun("np_sinh", None),
PrimDef::FunNpCosh => fun("np_cosh", None),
PrimDef::FunNpTanh => fun("np_tanh", None),
PrimDef::FunNpArcsinh => fun("np_arcsinh", None),
PrimDef::FunNpArccosh => fun("np_arccosh", None),
PrimDef::FunNpArctanh => fun("np_arctanh", None),
PrimDef::FunNpExpm1 => fun("np_expm1", None),
PrimDef::FunNpCbrt => fun("np_cbrt", None),
PrimDef::FunSpSpecErf => fun("sp_spec_erf", None),
PrimDef::FunSpSpecErfc => fun("sp_spec_erfc", None),
PrimDef::FunSpSpecGamma => fun("sp_spec_gamma", None),
PrimDef::FunSpSpecGammaln => fun("sp_spec_gammaln", None),
PrimDef::FunSpSpecJ0 => fun("sp_spec_j0", None),
PrimDef::FunSpSpecJ1 => fun("sp_spec_j1", None),
PrimDef::FunNpArctan2 => fun("np_arctan2", None),
PrimDef::FunNpCopysign => fun("np_copysign", None),
PrimDef::FunNpFmax => fun("np_fmax", None),
PrimDef::FunNpFmin => fun("np_fmin", None),
PrimDef::FunNpLdExp => fun("np_ldexp", None),
PrimDef::FunNpHypot => fun("np_hypot", None),
PrimDef::FunNpNextAfter => fun("np_nextafter", None),
PrimDef::FunSome => fun("Some", None),
}
}
}
/// Asserts that a [`PrimDef`] is in an allowlist.
///
/// Like `debug_assert!`, this statements of this function are only
/// enabled if `cfg!(debug_assertions)` is true.
pub fn debug_assert_prim_is_allowed(prim: PrimDef, allowlist: &[PrimDef]) {
if cfg!(debug_assertions) {
let allowed = allowlist.iter().any(|p| *p == prim);
assert!(
allowed,
"Disallowed primitive definition. Got {prim:?}, but expects it to be in {allowlist:?}"
);
}
}
/// Construct the fields of class `Exception`
/// See [`TypeEnum::TObj::fields`] and [`TopLevelDef::Class::fields`]
#[must_use]
pub fn make_exception_fields(int32: Type, int64: Type, str: Type) -> Vec<(StrRef, Type, bool)> {
vec![
("__name__".into(), int32, true),
("__file__".into(), str, true),
("__line__".into(), int32, true),
("__col__".into(), int32, true),
("__func__".into(), str, true),
("__message__".into(), str, true),
("__param0__".into(), int64, true),
("__param1__".into(), int64, true),
("__param2__".into(), int64, true),
]
}
impl TopLevelDef {
pub fn to_string(&self, unifier: &mut Unifier) -> String {
match self {
@ -43,115 +331,160 @@ impl TopLevelDef {
}
impl TopLevelComposer {
pub fn make_primitives() -> (PrimitiveStore, Unifier) {
#[must_use]
pub fn make_primitives(size_t: u32) -> (PrimitiveStore, Unifier) {
let mut unifier = Unifier::new();
let int32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
obj_id: PrimDef::Int32.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
obj_id: PrimDef::Int64.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
obj_id: PrimDef::Float.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(3),
obj_id: PrimDef::Bool.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(4),
obj_id: PrimDef::None.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(5),
fields: HashMap::new(),
params: HashMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(6),
fields: HashMap::new(),
params: HashMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(7),
fields: vec![
("__name__".into(), (int32, true)),
("__file__".into(), (str, true)),
("__line__".into(), (int32, true)),
("__col__".into(), (int32, true)),
("__func__".into(), (str, true)),
("__message__".into(), (str, true)),
("__param0__".into(), (int64, true)),
("__param1__".into(), (int64, true)),
("__param2__".into(), (int64, true)),
obj_id: PrimDef::Range.id(),
fields: [
("start".into(), (int32, true)),
("stop".into(), (int32, true)),
("step".into(), (int32, true)),
]
.into_iter()
.collect::<HashMap<_, _>>(),
params: HashMap::new(),
.collect(),
params: VarMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Str.id(),
fields: HashMap::new(),
params: VarMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Exception.id(),
fields: make_exception_fields(int32, int64, str)
.into_iter()
.map(|(name, ty, mutable)| (name, (ty, mutable)))
.collect(),
params: VarMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(8),
obj_id: PrimDef::UInt32.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(9),
obj_id: PrimDef::UInt64.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let option_type_var = unifier.get_fresh_var(Some("option_type_var".into()), None);
let is_some_type_fun_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: bool,
vars: HashMap::from([(option_type_var.1, option_type_var.0)]),
vars: into_var_map([option_type_var]),
}));
let unwrap_fun_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: option_type_var.0,
vars: HashMap::from([(option_type_var.1, option_type_var.0)]),
ret: option_type_var.ty,
vars: into_var_map([option_type_var]),
}));
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(10),
obj_id: PrimDef::Option.id(),
fields: vec![
("is_some".into(), (is_some_type_fun_ty, true)),
("is_none".into(), (is_some_type_fun_ty, true)),
("unwrap".into(), (unwrap_fun_ty, true)),
(PrimDef::OptionIsSome.simple_name().into(), (is_some_type_fun_ty, true)),
(PrimDef::OptionIsNone.simple_name().into(), (is_some_type_fun_ty, true)),
(PrimDef::OptionUnwrap.simple_name().into(), (unwrap_fun_ty, true)),
]
.into_iter()
.collect::<HashMap<_, _>>(),
params: HashMap::from([(option_type_var.1, option_type_var.0)]),
params: into_var_map([option_type_var]),
});
let size_t_ty = match size_t {
32 => uint32,
64 => uint64,
_ => unreachable!(),
};
let list_elem_tvar = unifier.get_fresh_var(Some("list_elem".into()), None);
let list = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::List.id(),
fields: Mapping::new(),
params: into_var_map([list_elem_tvar]),
});
let ndarray_dtype_tvar = unifier.get_fresh_var(Some("ndarray_dtype".into()), None);
let ndarray_ndims_tvar =
unifier.get_fresh_const_generic_var(size_t_ty, Some("ndarray_ndims".into()), None);
let ndarray_copy_fun_ret_ty = unifier.get_fresh_var(None, None);
let ndarray_copy_fun_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: ndarray_copy_fun_ret_ty.ty,
vars: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
}));
let ndarray_fill_fun_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![FuncArg {
name: "value".into(),
ty: ndarray_dtype_tvar.ty,
default_value: None,
}],
ret: none,
vars: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
}));
let ndarray = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::NDArray.id(),
fields: Mapping::from([
(PrimDef::NDArrayCopy.simple_name().into(), (ndarray_copy_fun_ty, true)),
(PrimDef::NDArrayFill.simple_name().into(), (ndarray_fill_fun_ty, true)),
]),
params: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
});
unifier.unify(ndarray_copy_fun_ret_ty.ty, ndarray).unwrap();
let primitives = PrimitiveStore {
int32,
int64,
uint32,
uint64,
float,
bool,
none,
range,
str,
exception,
uint32,
uint64,
option,
list,
ndarray,
size_t,
};
unifier.put_primitive_store(&primitives);
crate::typecheck::magic_methods::set_primitives_magic_methods(&primitives, &mut unifier);
(primitives, unifier)
}
/// already include the definition_id of itself inside the ancestors vector
/// when first registering, the type_vars, fields, methods, ancestors are invalid
/// already include the `definition_id` of itself inside the ancestors vector
/// when first registering, the `type_vars`, fields, methods, ancestors are invalid
#[must_use]
pub fn make_top_level_class_def(
index: usize,
obj_id: DefinitionId,
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
name: StrRef,
constructor: Option<Type>,
@ -159,11 +492,12 @@ impl TopLevelComposer {
) -> TopLevelDef {
TopLevelDef::Class {
name,
object_id: DefinitionId(index),
type_vars: Default::default(),
fields: Default::default(),
methods: Default::default(),
ancestors: Default::default(),
object_id: obj_id,
type_vars: Vec::default(),
fields: Vec::default(),
attributes: Vec::default(),
methods: Vec::default(),
ancestors: Vec::default(),
constructor,
resolver,
loc,
@ -171,6 +505,7 @@ impl TopLevelComposer {
}
/// when first registering, the type is a invalid value
#[must_use]
pub fn make_top_level_function_def(
name: String,
simple_name: StrRef,
@ -182,15 +517,16 @@ impl TopLevelComposer {
name,
simple_name,
signature: ty,
var_id: Default::default(),
instance_to_symbol: Default::default(),
instance_to_stmt: Default::default(),
var_id: Vec::default(),
instance_to_symbol: HashMap::default(),
instance_to_stmt: HashMap::default(),
resolver,
codegen_callback: None,
loc,
}
}
#[must_use]
pub fn make_class_method_name(mut class_name: String, method_name: &str) -> String {
class_name.push('.');
class_name.push_str(method_name);
@ -200,13 +536,13 @@ impl TopLevelComposer {
pub fn get_class_method_def_info(
class_methods_def: &[(StrRef, Type, DefinitionId)],
method_name: StrRef,
) -> Result<(Type, DefinitionId), String> {
) -> Result<(Type, DefinitionId), HashSet<String>> {
for (name, ty, def_id) in class_methods_def {
if name == &method_name {
return Ok((*ty, *def_id));
}
}
Err(format!("no method {} in the current class", method_name))
Err(HashSet::from([format!("no method {method_name} in the current class")]))
}
/// get all base class def id of a class, excluding itself. \
@ -217,7 +553,7 @@ impl TopLevelComposer {
pub fn get_all_ancestors_helper(
child: &TypeAnnotation,
temp_def_list: &[Arc<RwLock<TopLevelDef>>],
) -> Result<Vec<TypeAnnotation>, String> {
) -> Result<Vec<TypeAnnotation>, HashSet<String>> {
let mut result: Vec<TypeAnnotation> = Vec::new();
let mut parent = Self::get_parent(child, temp_def_list);
while let Some(p) = parent {
@ -229,16 +565,16 @@ impl TopLevelComposer {
};
// check cycle
let no_cycle = result.iter().all(|x| {
if let TypeAnnotation::CustomClass { id, .. } = x {
id.0 != p_id.0
} else {
let TypeAnnotation::CustomClass { id, .. } = x else {
unreachable!("must be class kind annotation")
}
};
id.0 != p_id.0
});
if no_cycle {
result.push(p);
} else {
return Err("cyclic inheritance detected".into());
return Err(HashSet::from(["cyclic inheritance detected".into()]));
}
}
Ok(result)
@ -256,23 +592,23 @@ impl TopLevelComposer {
};
let child_def = temp_def_list.get(child_id.0).unwrap();
let child_def = child_def.read();
if let TopLevelDef::Class { ancestors, .. } = &*child_def {
if !ancestors.is_empty() {
Some(ancestors[0].clone())
} else {
None
}
} else {
let TopLevelDef::Class { ancestors, .. } = &*child_def else {
unreachable!("child must be top level class def")
};
if ancestors.is_empty() {
None
} else {
Some(ancestors[0].clone())
}
}
/// get the var_id of a given TVar type
pub fn get_var_id(var_ty: Type, unifier: &mut Unifier) -> Result<u32, String> {
/// get the `var_id` of a given `TVar` type
pub fn get_var_id(var_ty: Type, unifier: &mut Unifier) -> Result<TypeVarId, HashSet<String>> {
if let TypeEnum::TVar { id, .. } = unifier.get_ty(var_ty).as_ref() {
Ok(*id)
} else {
Err("not type var".to_string())
Err(HashSet::from(["not type var".to_string()]))
}
}
@ -286,13 +622,17 @@ impl TopLevelComposer {
let this = this.as_ref();
let other = unifier.get_ty(other);
let other = other.as_ref();
if let (
let (
TypeEnum::TFunc(FunSignature { args: this_args, ret: this_ret, .. }),
TypeEnum::TFunc(FunSignature { args: other_args, ret: other_ret, .. }),
) = (this, other)
{
else {
unreachable!("this function must be called with function type")
};
// check args
let args_ok = this_args
let args_ok =
this_args
.iter()
.map(|FuncArg { name, ty, .. }| (name, type_var_to_concrete_def.get(ty).unwrap()))
.zip(other_args.iter().map(|FuncArg { name, ty, .. }| {
@ -316,9 +656,6 @@ impl TopLevelComposer {
// return
args_ok && ret_ok
} else {
unreachable!("this function must be called with function type")
}
}
pub fn check_overload_field_type(
@ -334,7 +671,7 @@ impl TopLevelComposer {
)
}
pub fn get_all_assigned_field(stmts: &[ast::Stmt<()>]) -> Result<HashSet<StrRef>, String> {
pub fn get_all_assigned_field(stmts: &[Stmt<()>]) -> Result<HashSet<StrRef>, HashSet<String>> {
let mut result = HashSet::new();
for s in stmts {
match &s.node {
@ -351,10 +688,10 @@ impl TopLevelComposer {
}
} =>
{
return Err(format!(
return Err(HashSet::from([format!(
"redundant type annotation for class fields at {}",
s.location
))
)]))
}
ast::StmtKind::Assign { targets, .. } => {
for t in targets {
@ -376,14 +713,14 @@ impl TopLevelComposer {
ast::StmtKind::If { body, orelse, .. } => {
let inited_for_sure = Self::get_all_assigned_field(body.as_slice())?
.intersection(&Self::get_all_assigned_field(orelse.as_slice())?)
.cloned()
.copied()
.collect::<HashSet<_>>();
result.extend(inited_for_sure);
}
ast::StmtKind::Try { body, orelse, finalbody, .. } => {
let inited_for_sure = Self::get_all_assigned_field(body.as_slice())?
.intersection(&Self::get_all_assigned_field(orelse.as_slice())?)
.cloned()
.copied()
.collect::<HashSet<_>>();
result.extend(inited_for_sure);
result.extend(Self::get_all_assigned_field(finalbody.as_slice())?);
@ -391,9 +728,9 @@ impl TopLevelComposer {
ast::StmtKind::With { body, .. } => {
result.extend(Self::get_all_assigned_field(body.as_slice())?);
}
ast::StmtKind::Pass { .. } => {}
ast::StmtKind::Assert { .. } => {}
ast::StmtKind::Expr { .. } => {}
ast::StmtKind::Pass { .. }
| ast::StmtKind::Assert { .. }
| ast::StmtKind::Expr { .. } => {}
_ => {
unimplemented!()
@ -406,7 +743,7 @@ impl TopLevelComposer {
pub fn parse_parameter_default_value(
default: &ast::Expr,
resolver: &(dyn SymbolResolver + Send + Sync),
) -> Result<SymbolValue, String> {
) -> Result<SymbolValue, HashSet<String>> {
parse_parameter_default_value(default, resolver)
}
@ -416,40 +753,6 @@ impl TopLevelComposer {
primitive: &PrimitiveStore,
unifier: &mut Unifier,
) -> Result<(), String> {
fn type_default_param(
val: &SymbolValue,
primitive: &PrimitiveStore,
unifier: &mut Unifier,
) -> TypeAnnotation {
match val {
SymbolValue::Bool(..) => TypeAnnotation::Primitive(primitive.bool),
SymbolValue::Double(..) => TypeAnnotation::Primitive(primitive.float),
SymbolValue::I32(..) => TypeAnnotation::Primitive(primitive.int32),
SymbolValue::I64(..) => TypeAnnotation::Primitive(primitive.int64),
SymbolValue::U32(..) => TypeAnnotation::Primitive(primitive.uint32),
SymbolValue::U64(..) => TypeAnnotation::Primitive(primitive.uint64),
SymbolValue::Str(..) => TypeAnnotation::Primitive(primitive.str),
SymbolValue::Tuple(vs) => {
let vs_tys = vs
.iter()
.map(|v| type_default_param(v, primitive, unifier))
.collect::<Vec<_>>();
TypeAnnotation::Tuple(vs_tys)
}
SymbolValue::OptionNone => TypeAnnotation::CustomClass {
id: primitive.option.get_obj_id(unifier),
params: Default::default(),
},
SymbolValue::OptionSome(v) => {
let ty = type_default_param(v, primitive, unifier);
TypeAnnotation::CustomClass {
id: primitive.option.get_obj_id(unifier),
params: vec![ty],
}
}
}
}
fn is_compatible(
found: &TypeAnnotation,
expect: &TypeAnnotation,
@ -465,7 +768,7 @@ impl TopLevelComposer {
TypeAnnotation::CustomClass { id: e_id, params: e_param },
) => {
*f_id == *e_id
&& *f_id == primitive.option.get_obj_id(unifier)
&& *f_id == primitive.option.obj_id(unifier).unwrap()
&& (f_param.is_empty()
|| (f_param.len() == 1
&& e_param.len() == 1
@ -481,15 +784,15 @@ impl TopLevelComposer {
}
}
let found = type_default_param(val, primitive, unifier);
if !is_compatible(&found, ty, unifier, primitive) {
let found = val.get_type_annotation(primitive, unifier);
if is_compatible(&found, ty, unifier, primitive) {
Ok(())
} else {
Err(format!(
"incompatible default parameter type, expect {}, found {}",
ty.stringify(unifier),
found.stringify(unifier),
))
} else {
Ok(())
}
}
}
@ -497,14 +800,14 @@ impl TopLevelComposer {
pub fn parse_parameter_default_value(
default: &ast::Expr,
resolver: &(dyn SymbolResolver + Send + Sync),
) -> Result<SymbolValue, String> {
fn handle_constant(val: &Constant, loc: &Location) -> Result<SymbolValue, String> {
) -> Result<SymbolValue, HashSet<String>> {
fn handle_constant(val: &Constant, loc: &Location) -> Result<SymbolValue, HashSet<String>> {
match val {
Constant::Int(v) => {
if let Ok(v) = (*v).try_into() {
Ok(SymbolValue::I32(v))
} else {
Err(format!("integer value out of range at {}", loc))
Err(HashSet::from([format!("integer value out of range at {loc}")]))
}
}
Constant::Float(v) => Ok(SymbolValue::Double(*v)),
@ -512,74 +815,126 @@ pub fn parse_parameter_default_value(
Constant::Tuple(tuple) => Ok(SymbolValue::Tuple(
tuple.iter().map(|x| handle_constant(x, loc)).collect::<Result<Vec<_>, _>>()?,
)),
Constant::None => Err(format!(
"`None` is not supported, use `none` for option type instead ({})",
loc
)),
Constant::None => Err(HashSet::from([format!(
"`None` is not supported, use `none` for option type instead ({loc})"
)])),
_ => unimplemented!("this constant is not supported at {}", loc),
}
}
match &default.node {
ast::ExprKind::Constant { value, .. } => handle_constant(value, &default.location),
ast::ExprKind::Call { func, args, .. } if args.len() == 1 => {
match &func.node {
ast::ExprKind::Call { func, args, .. } if args.len() == 1 => match &func.node {
ast::ExprKind::Name { id, .. } if *id == "int64".into() => match &args[0].node {
ast::ExprKind::Constant { value: Constant::Int(v), .. } => {
let v: Result<i64, _> = (*v).try_into();
match v {
Ok(v) => Ok(SymbolValue::I64(v)),
_ => Err(format!("default param value out of range at {}", default.location)),
_ => Err(HashSet::from([format!(
"default param value out of range at {}",
default.location
)])),
}
}
_ => Err(format!("only allow constant integer here at {}", default.location))
}
_ => Err(HashSet::from([format!(
"only allow constant integer here at {}",
default.location
)])),
},
ast::ExprKind::Name { id, .. } if *id == "uint32".into() => match &args[0].node {
ast::ExprKind::Constant { value: Constant::Int(v), .. } => {
let v: Result<u32, _> = (*v).try_into();
match v {
Ok(v) => Ok(SymbolValue::U32(v)),
_ => Err(format!("default param value out of range at {}", default.location)),
_ => Err(HashSet::from([format!(
"default param value out of range at {}",
default.location
)])),
}
}
_ => Err(format!("only allow constant integer here at {}", default.location))
}
_ => Err(HashSet::from([format!(
"only allow constant integer here at {}",
default.location
)])),
},
ast::ExprKind::Name { id, .. } if *id == "uint64".into() => match &args[0].node {
ast::ExprKind::Constant { value: Constant::Int(v), .. } => {
let v: Result<u64, _> = (*v).try_into();
match v {
Ok(v) => Ok(SymbolValue::U64(v)),
_ => Err(format!("default param value out of range at {}", default.location)),
_ => Err(HashSet::from([format!(
"default param value out of range at {}",
default.location
)])),
}
}
_ => Err(format!("only allow constant integer here at {}", default.location))
}
ast::ExprKind::Name { id, .. } if *id == "Some".into() => Ok(
SymbolValue::OptionSome(
Box::new(parse_parameter_default_value(&args[0], resolver)?)
)
),
_ => Err(format!("unsupported default parameter at {}", default.location)),
}
}
ast::ExprKind::Tuple { elts, .. } => Ok(SymbolValue::Tuple(elts
.iter()
_ => Err(HashSet::from([format!(
"only allow constant integer here at {}",
default.location
)])),
},
ast::ExprKind::Name { id, .. } if *id == "Some".into() => Ok(SymbolValue::OptionSome(
Box::new(parse_parameter_default_value(&args[0], resolver)?),
)),
_ => Err(HashSet::from([format!(
"unsupported default parameter at {}",
default.location
)])),
},
ast::ExprKind::Tuple { elts, .. } => Ok(SymbolValue::Tuple(
elts.iter()
.map(|x| parse_parameter_default_value(x, resolver))
.collect::<Result<Vec<_>, _>>()?
.collect::<Result<Vec<_>, _>>()?,
)),
ast::ExprKind::Name { id, .. } if id == &"none".into() => Ok(SymbolValue::OptionNone),
ast::ExprKind::Name { id, .. } => {
resolver.get_default_param_value(default).ok_or_else(
|| format!(
resolver.get_default_param_value(default).ok_or_else(|| {
HashSet::from([format!(
"`{}` cannot be used as a default parameter at {} \
(not primitive type, option or tuple / not defined?)",
id,
default.location
)
)
id, default.location
)])
})
}
_ => Err(format!(
_ => Err(HashSet::from([format!(
"unsupported default parameter (not primitive type, option or tuple) at {}",
default.location
))
)])),
}
}
/// Obtains the element type of an array-like type.
pub fn arraylike_flatten_element_type(unifier: &mut Unifier, ty: Type) -> Type {
match &*unifier.get_ty(ty) {
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
unpack_ndarray_var_tys(unifier, ty).0
}
TypeEnum::TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
arraylike_flatten_element_type(unifier, iter_type_vars(params).next().unwrap().ty)
}
_ => ty,
}
}
/// Obtains the number of dimensions of an array-like type.
pub fn arraylike_get_ndims(unifier: &mut Unifier, ty: Type) -> u64 {
match &*unifier.get_ty(ty) {
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
let ndims = unpack_ndarray_var_tys(unifier, ty).1;
let TypeEnum::TLiteral { values, .. } = &*unifier.get_ty_immutable(ndims) else {
panic!("Expected TLiteral for ndarray.ndims, got {}", unifier.stringify(ndims))
};
if values.len() > 1 {
todo!("Getting num of dimensions for ndarray with more than one ndim bound is unimplemented")
}
u64::try_from(values[0].clone()).unwrap()
}
TypeEnum::TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
arraylike_get_ndims(unifier, iter_type_vars(params).next().unwrap().ty) + 1
}
_ => 0,
}
}

View File

@ -3,20 +3,24 @@ use std::{
collections::{HashMap, HashSet},
fmt::Debug,
iter::FromIterator,
ops::{Deref, DerefMut},
sync::Arc,
};
use super::codegen::CodeGenContext;
use super::typecheck::type_inferencer::PrimitiveStore;
use super::typecheck::typedef::{FunSignature, FuncArg, SharedUnifier, Type, TypeEnum, Unifier};
use super::typecheck::typedef::{
FunSignature, FuncArg, SharedUnifier, Type, TypeEnum, Unifier, VarMap,
};
use crate::{
codegen::CodeGenerator,
symbol_resolver::{SymbolResolver, ValueEnum},
typecheck::{type_inferencer::CodeLocation, typedef::CallId},
typecheck::{
type_inferencer::CodeLocation,
typedef::{CallId, TypeVarId},
},
};
use inkwell::values::BasicValueEnum;
use itertools::{izip, Itertools};
use itertools::Itertools;
use nac3parser::ast::{self, Location, Stmt, StrRef};
use parking_lot::RwLock;
@ -26,14 +30,14 @@ pub struct DefinitionId(pub usize);
pub mod builtins;
pub mod composer;
pub mod helper;
pub mod numpy;
pub mod type_annotation;
use composer::*;
use type_annotation::*;
#[cfg(test)]
mod test;
type GenCallCallback = Box<
dyn for<'ctx, 'a> Fn(
type GenCallCallback = dyn for<'ctx, 'a> Fn(
&mut CodeGenContext<'ctx, 'a>,
Option<(Type, ValueEnum<'ctx>)>,
(&FunSignature, DefinitionId),
@ -41,21 +45,28 @@ type GenCallCallback = Box<
&mut dyn CodeGenerator,
) -> Result<Option<BasicValueEnum<'ctx>>, String>
+ Send
+ Sync,
>;
+ Sync;
pub struct GenCall {
fp: GenCallCallback,
fp: Box<GenCallCallback>,
}
impl GenCall {
pub fn new(fp: GenCallCallback) -> GenCall {
#[must_use]
pub fn new(fp: Box<GenCallCallback>) -> GenCall {
GenCall { fp }
}
pub fn run<'ctx, 'a>(
/// Creates a dummy instance of [`GenCall`], which invokes [`unreachable!()`] with the given
/// `reason`.
#[must_use]
pub fn create_dummy(reason: String) -> GenCall {
Self::new(Box::new(move |_, _, _, _, _| unreachable!("{reason}")))
}
pub fn run<'ctx>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ctx: &mut CodeGenContext<'ctx, '_>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
@ -75,58 +86,66 @@ impl Debug for GenCall {
pub struct FunInstance {
pub body: Arc<Vec<Stmt<Option<Type>>>>,
pub calls: Arc<HashMap<CodeLocation, CallId>>,
pub subst: HashMap<u32, Type>,
pub subst: VarMap,
pub unifier_id: usize,
}
#[derive(Debug, Clone)]
pub enum TopLevelDef {
Class {
// name for error messages and symbols
/// Name for error messages and symbols.
name: StrRef,
// object ID used for TypeEnum
/// Object ID used for [`TypeEnum`].
object_id: DefinitionId,
/// type variables bounded to the class.
type_vars: Vec<Type>,
// class fields
// name, type, is mutable
/// Class fields.
///
/// Name and type is mutable.
fields: Vec<(StrRef, Type, bool)>,
// class methods, pointing to the corresponding function definition.
/// Class Attributes.
///
/// Name, type, value.
attributes: Vec<(StrRef, Type, ast::Constant)>,
/// Class methods, pointing to the corresponding function definition.
methods: Vec<(StrRef, Type, DefinitionId)>,
// ancestor classes, including itself.
/// Ancestor classes, including itself.
ancestors: Vec<TypeAnnotation>,
// symbol resolver of the module defined the class, none if it is built-in type
/// Symbol resolver of the module defined the class; [None] if it is built-in type.
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
// constructor type
/// Constructor type.
constructor: Option<Type>,
// definition location
/// Definition location.
loc: Option<Location>,
},
Function {
// prefix for symbol, should be unique globally
/// Prefix for symbol, should be unique globally.
name: String,
// simple name, the same as in method/function definition
/// Simple name, the same as in method/function definition.
simple_name: StrRef,
// function signature.
/// Function signature.
signature: Type,
// instantiated type variable IDs
var_id: Vec<u32>,
/// Instantiated type variable IDs.
var_id: Vec<TypeVarId>,
/// Function instance to symbol mapping
/// Key: string representation of type variable values, sorted by variable ID in ascending
///
/// * Key: String representation of type variable values, sorted by variable ID in ascending
/// order, including type variables associated with the class.
/// Value: function symbol name.
/// * Value: Function symbol name.
instance_to_symbol: HashMap<String, String>,
/// Function instances to annotated AST mapping
/// Key: string representation of type variable values, sorted by variable ID in ascending
///
/// * Key: String representation of type variable values, sorted by variable ID in ascending
/// order, including type variables associated with the class. Excluding rigid type
/// variables.
/// rigid type variables that would be substituted when the function is instantiated.
///
/// Rigid type variables that would be substituted when the function is instantiated.
instance_to_stmt: HashMap<String, FunInstance>,
// symbol resolver of the module defined the class
/// Symbol resolver of the module defined the class.
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
// custom codegen callback
/// Custom code generation callback.
codegen_callback: Option<Arc<GenCall>>,
// definition location
/// Definition location.
loc: Option<Location>,
},
}

View File

@ -0,0 +1,85 @@
use crate::{
toplevel::helper::PrimDef,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{Type, TypeEnum, TypeVarId, Unifier, VarMap},
},
};
use itertools::Itertools;
/// Creates a `ndarray` [`Type`] with the given type arguments.
///
/// * `dtype` - The element type of the `ndarray`, or [`None`] if the type variable is not
/// specialized.
/// * `ndims` - The number of dimensions of the `ndarray`, or [`None`] if the type variable is not
/// specialized.
pub fn make_ndarray_ty(
unifier: &mut Unifier,
primitives: &PrimitiveStore,
dtype: Option<Type>,
ndims: Option<Type>,
) -> Type {
subst_ndarray_tvars(unifier, primitives.ndarray, dtype, ndims)
}
/// Substitutes type variables in `ndarray`.
///
/// * `dtype` - The element type of the `ndarray`, or [`None`] if the type variable is not
/// specialized.
/// * `ndims` - The number of dimensions of the `ndarray`, or [`None`] if the type variable is not
/// specialized.
pub fn subst_ndarray_tvars(
unifier: &mut Unifier,
ndarray: Type,
dtype: Option<Type>,
ndims: Option<Type>,
) -> Type {
let TypeEnum::TObj { obj_id, params, .. } = &*unifier.get_ty_immutable(ndarray) else {
panic!("Expected `ndarray` to be TObj, but got {}", unifier.stringify(ndarray))
};
debug_assert_eq!(*obj_id, PrimDef::NDArray.id());
if dtype.is_none() && ndims.is_none() {
return ndarray;
}
let tvar_ids = params.iter().map(|(obj_id, _)| *obj_id).collect_vec();
debug_assert_eq!(tvar_ids.len(), 2);
let mut tvar_subst = VarMap::new();
if let Some(dtype) = dtype {
tvar_subst.insert(tvar_ids[0], dtype);
}
if let Some(ndims) = ndims {
tvar_subst.insert(tvar_ids[1], ndims);
}
unifier.subst(ndarray, &tvar_subst).unwrap_or(ndarray)
}
fn unpack_ndarray_tvars(unifier: &mut Unifier, ndarray: Type) -> Vec<(TypeVarId, Type)> {
let TypeEnum::TObj { obj_id, params, .. } = &*unifier.get_ty_immutable(ndarray) else {
panic!("Expected `ndarray` to be TObj, but got {}", unifier.stringify(ndarray))
};
debug_assert_eq!(*obj_id, PrimDef::NDArray.id());
debug_assert_eq!(params.len(), 2);
params
.iter()
.sorted_by_key(|(obj_id, _)| *obj_id)
.map(|(var_id, ty)| (*var_id, *ty))
.collect_vec()
}
/// Unpacks the type variable IDs of `ndarray` into a tuple. The elements of the tuple corresponds
/// to `dtype` (the element type) and `ndims` (the number of dimensions) of the `ndarray`
/// respectively.
pub fn unpack_ndarray_var_ids(unifier: &mut Unifier, ndarray: Type) -> (TypeVarId, TypeVarId) {
unpack_ndarray_tvars(unifier, ndarray).into_iter().map(|v| v.0).collect_tuple().unwrap()
}
/// Unpacks the type variables of `ndarray` into a tuple. The elements of the tuple corresponds to
/// `dtype` (the element type) and `ndims` (the number of dimensions) of the `ndarray` respectively.
pub fn unpack_ndarray_var_tys(unifier: &mut Unifier, ndarray: Type) -> (Type, Type) {
unpack_ndarray_tvars(unifier, ndarray).into_iter().map(|v| v.1).collect_tuple().unwrap()
}

View File

@ -1,13 +1,11 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
expression: res_vec
---
[
"Class {\nname: \"Generic_A\",\nancestors: [\"Generic_A[V]\", \"B\"],\nfields: [\"aa\", \"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"foo\", \"fn[[b:T], none]\"), (\"fun\", \"fn[[a:int32], V]\")],\ntype_vars: [\"V\"]\n}\n",
"Function {\nname: \"Generic_A.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"Generic_A.fun\",\nsig: \"fn[[a:int32], V]\",\nvar_id: [18]\n}\n",
"Function {\nname: \"Generic_A.fun\",\nsig: \"fn[[a:int32], V]\",\nvar_id: [TypeVarId(245)]\n}\n",
"Class {\nname: \"B\",\nancestors: [\"B\"],\nfields: [\"aa\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"foo\", \"fn[[b:T], none]\")],\ntype_vars: []\n}\n",
"Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"B.foo\",\nsig: \"fn[[b:T], none]\",\nvar_id: []\n}\n",

View File

@ -1,15 +1,13 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
expression: res_vec
---
[
"Class {\nname: \"A\",\nancestors: [\"A[T]\"],\nfields: [\"a\", \"b\", \"c\"],\nmethods: [(\"__init__\", \"fn[[t:T], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: [\"T\"]\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[t:T], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.foo\",\nsig: \"fn[[c:C], none]\",\nvar_id: []\n}\n",
"Class {\nname: \"B\",\nancestors: [\"B[typevar7]\", \"A[float]\"],\nfields: [\"a\", \"b\", \"c\", \"d\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: [\"typevar7\"]\n}\n",
"Class {\nname: \"B\",\nancestors: [\"B[typevar234]\", \"A[float]\"],\nfields: [\"a\", \"b\", \"c\", \"d\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: [\"typevar234\"]\n}\n",
"Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"B.fun\",\nsig: \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\",\nvar_id: []\n}\n",
"Class {\nname: \"C\",\nancestors: [\"C\", \"B[bool]\", \"A[float]\"],\nfields: [\"a\", \"b\", \"c\", \"d\", \"e\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: []\n}\n",

View File

@ -1,15 +1,13 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
expression: res_vec
---
[
"Function {\nname: \"foo\",\nsig: \"fn[[a:list[int32], b:tuple[T, float]], A[B, bool]]\",\nvar_id: []\n}\n",
"Class {\nname: \"A\",\nancestors: [\"A[T, V]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[v:V], none]\"), (\"fun\", \"fn[[a:T], V]\")],\ntype_vars: [\"T\", \"V\"]\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[v:V], none]\",\nvar_id: [20]\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [25]\n}\n",
"Function {\nname: \"gfun\",\nsig: \"fn[[a:A[int32, list[float]]], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[v:V], none]\",\nvar_id: [TypeVarId(247)]\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(252)]\n}\n",
"Function {\nname: \"gfun\",\nsig: \"fn[[a:A[list[float], int32]], none]\",\nvar_id: []\n}\n",
"Class {\nname: \"B\",\nancestors: [\"B\"],\nfields: [],\nmethods: [(\"__init__\", \"fn[[], none]\")],\ntype_vars: []\n}\n",
"Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
]

View File

@ -1,15 +1,13 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
expression: res_vec
---
[
"Class {\nname: \"A\",\nancestors: [\"A[typevar6, typevar7]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[bool, float], b:B], none]\"), (\"fun\", \"fn[[a:A[bool, float]], A[bool, int32]]\")],\ntype_vars: [\"typevar6\", \"typevar7\"]\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[a:A[bool, float], b:B], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:A[bool, float]], A[bool, int32]]\",\nvar_id: []\n}\n",
"Class {\nname: \"B\",\nancestors: [\"B\", \"A[int64, bool]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:A[bool, float]], A[bool, int32]]\"), (\"foo\", \"fn[[b:B], B]\"), (\"bar\", \"fn[[a:A[int32, list[B]]], tuple[A[bool, virtual[A[B, int32]]], B]]\")],\ntype_vars: []\n}\n",
"Class {\nname: \"A\",\nancestors: [\"A[typevar233, typevar234]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[float, bool], b:B], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\")],\ntype_vars: [\"typevar233\", \"typevar234\"]\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[a:A[float, bool], b:B], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:A[float, bool]], A[bool, int32]]\",\nvar_id: []\n}\n",
"Class {\nname: \"B\",\nancestors: [\"B\", \"A[int64, bool]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\"), (\"foo\", \"fn[[b:B], B]\"), (\"bar\", \"fn[[a:A[list[B], int32]], tuple[A[virtual[A[B, int32]], bool], B]]\")],\ntype_vars: []\n}\n",
"Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"B.foo\",\nsig: \"fn[[b:B], B]\",\nvar_id: []\n}\n",
"Function {\nname: \"B.bar\",\nsig: \"fn[[a:A[int32, list[B]]], tuple[A[bool, virtual[A[B, int32]]], B]]\",\nvar_id: []\n}\n",
"Function {\nname: \"B.bar\",\nsig: \"fn[[a:A[list[B], int32]], tuple[A[virtual[A[B, int32]], bool], B]]\",\nvar_id: []\n}\n",
]

View File

@ -1,19 +1,17 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
expression: res_vec
---
[
"Class {\nname: \"A\",\nancestors: [\"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[b:B], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.foo\",\nsig: \"fn[[a:T, b:V], none]\",\nvar_id: [26]\n}\n",
"Function {\nname: \"A.foo\",\nsig: \"fn[[a:T, b:V], none]\",\nvar_id: [TypeVarId(253)]\n}\n",
"Class {\nname: \"B\",\nancestors: [\"B\", \"C\", \"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
"Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
"Class {\nname: \"C\",\nancestors: [\"C\", \"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
"Function {\nname: \"C.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"C.fun\",\nsig: \"fn[[b:B], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"foo\",\nsig: \"fn[[a:A], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"ff\",\nsig: \"fn[[a:T], V]\",\nvar_id: [34]\n}\n",
"Function {\nname: \"ff\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(261)]\n}\n",
]

View File

@ -1,3 +1,6 @@
use super::*;
use crate::toplevel::helper::PrimDef;
use crate::typecheck::typedef::into_var_map;
use crate::{
codegen::CodeGenContext,
symbol_resolver::{SymbolResolver, ValueEnum},
@ -8,13 +11,12 @@ use crate::{
},
};
use indoc::indoc;
use nac3parser::ast::FileName;
use nac3parser::{ast::fold::Fold, parser::parse_program};
use parking_lot::Mutex;
use std::{collections::HashMap, sync::Arc};
use test_case::test_case;
use super::*;
struct ResolverInternal {
id_to_type: Mutex<HashMap<StrRef, Type>>,
id_to_def: Mutex<HashMap<StrRef, DefinitionId>>,
@ -36,7 +38,7 @@ struct Resolver(Arc<ResolverInternal>);
impl SymbolResolver for Resolver {
fn get_default_param_value(
&self,
_: &nac3parser::ast::Expr,
_: &ast::Expr,
) -> Option<crate::symbol_resolver::SymbolValue> {
unimplemented!()
}
@ -52,20 +54,25 @@ impl SymbolResolver for Resolver {
.id_to_type
.lock()
.get(&str)
.cloned()
.ok_or_else(|| format!("cannot find symbol `{}`", str))
.copied()
.ok_or_else(|| format!("cannot find symbol `{str}`"))
}
fn get_symbol_value<'ctx, 'a>(
fn get_symbol_value<'ctx>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, 'a>,
_: &mut CodeGenContext<'ctx, '_>,
) -> Option<ValueEnum<'ctx>> {
unimplemented!()
}
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, String> {
self.0.id_to_def.lock().get(&id).cloned().ok_or_else(|| "Unknown identifier".to_string())
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, HashSet<String>> {
self.0
.id_to_def
.lock()
.get(&id)
.copied()
.ok_or_else(|| HashSet::from(["Unknown identifier".to_string()]))
}
fn get_string_id(&self, _: &str) -> i32 {
@ -110,13 +117,13 @@ impl SymbolResolver for Resolver {
"register"
)]
fn test_simple_register(source: Vec<&str>) {
let mut composer: TopLevelComposer = Default::default();
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
for s in source {
let ast = parse_program(s, Default::default()).unwrap();
let ast = parse_program(s, FileName::default()).unwrap();
let ast = ast[0].clone();
composer.register_top_level(ast, None, "".into(), false).unwrap();
composer.register_top_level(ast, None, "", false).unwrap();
}
}
@ -130,14 +137,14 @@ fn test_simple_register(source: Vec<&str>) {
"register"
)]
fn test_simple_register_without_constructor(source: &str) {
let mut composer: TopLevelComposer = Default::default();
let ast = parse_program(source, Default::default()).unwrap();
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let ast = parse_program(source, FileName::default()).unwrap();
let ast = ast[0].clone();
composer.register_top_level(ast, None, "".into(), true).unwrap();
composer.register_top_level(ast, None, "", true).unwrap();
}
#[test_case(
vec![
&[
indoc! {"
def fun(a: int32) -> int32:
return a
@ -151,35 +158,35 @@ fn test_simple_register_without_constructor(source: &str) {
return 3
"},
],
vec![
&[
"fn[[a:0], 0]",
"fn[[a:2], 4]",
"fn[[b:1], 0]",
],
vec![
&[
"fun",
"foo",
"f"
];
"function compose"
)]
fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&str>) {
let mut composer: TopLevelComposer = Default::default();
fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let internal_resolver = Arc::new(ResolverInternal {
id_to_def: Default::default(),
id_to_type: Default::default(),
class_names: Default::default(),
id_to_def: Mutex::default(),
id_to_type: Mutex::default(),
class_names: Mutex::default(),
});
let resolver =
Arc::new(Resolver(internal_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
for s in source {
let ast = parse_program(s, Default::default()).unwrap();
let ast = parse_program(s, FileName::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) =
composer.register_top_level(ast, Some(resolver.clone()), "".into(), false).unwrap();
composer.register_top_level(ast, Some(resolver.clone()), "", false).unwrap();
internal_resolver.add_id_def(id, def_id);
if let Some(ty) = ty {
internal_resolver.add_id_type(id, ty);
@ -205,7 +212,7 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
}
#[test_case(
vec![
&[
indoc! {"
class A():
a: int32
@ -238,11 +245,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec![];
&[];
"simple class compose"
)]
#[test_case(
vec![
&[
indoc! {"
class Generic_A(Generic[V], B):
a: int64
@ -260,11 +267,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec![];
&[];
"generic class"
)]
#[test_case(
vec![
&[
indoc! {"
def foo(a: list[int32], b: tuple[T, float]) -> A[B, bool]:
pass
@ -289,11 +296,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec![];
&[];
"list tuple generic"
)]
#[test_case(
vec![
&[
indoc! {"
class A(Generic[T, V]):
a: A[float, bool]
@ -314,11 +321,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec![];
&[];
"self1"
)]
#[test_case(
vec![
&[
indoc! {"
class A(Generic[T]):
a: int32
@ -348,11 +355,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec![];
&[];
"inheritance_override"
)]
#[test_case(
vec![
&[
indoc! {"
class A(Generic[T]):
def __init__(self):
@ -361,11 +368,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec!["application of type vars to generic class is not currently supported (at unknown: line 4 column 24)"];
&["application of type vars to generic class is not currently supported (at unknown:4:24)"];
"err no type var in generic app"
)]
#[test_case(
vec![
&[
indoc! {"
class A(B):
def __init__(self):
@ -377,11 +384,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec!["cyclic inheritance detected"];
&["cyclic inheritance detected"];
"cyclic1"
)]
#[test_case(
vec![
&[
indoc! {"
class A(B[bool, int64]):
def __init__(self):
@ -398,30 +405,30 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"},
],
vec!["cyclic inheritance detected"];
&["cyclic inheritance detected"];
"cyclic2"
)]
#[test_case(
vec![
&[
indoc! {"
class A:
pass
"}
],
vec!["5: Class {\nname: \"A\",\ndef_id: DefinitionId(5),\nancestors: [CustomClassKind { id: DefinitionId(5), params: [] }],\nfields: [],\nmethods: [],\ntype_vars: []\n}"];
&["5: Class {\nname: \"A\",\ndef_id: DefinitionId(5),\nancestors: [CustomClassKind { id: DefinitionId(5), params: [] }],\nfields: [],\nmethods: [],\ntype_vars: []\n}"];
"simple pass in class"
)]
#[test_case(
vec![indoc! {"
&[indoc! {"
class A:
def __init__():
pass
"}],
vec!["__init__ method must have a `self` parameter (at unknown: line 2 column 5)"];
&["__init__ method must have a `self` parameter (at unknown:2:5)"];
"err no self_1"
)]
#[test_case(
vec![
&[
indoc! {"
class A(B, Generic[T], C):
def __init__(self):
@ -439,11 +446,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
"}
],
vec!["a class definition can only have at most one base class declaration and one generic declaration (at unknown: line 1 column 24)"];
&["a class definition can only have at most one base class declaration and one generic declaration (at unknown:1:24)"];
"err multiple inheritance"
)]
#[test_case(
vec![
&[
indoc! {"
class A(Generic[T]):
a: int32
@ -464,11 +471,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec!["method fun has same name as ancestors' method, but incompatible type"];
&["method fun has same name as ancestors' method, but incompatible type"];
"err_incompatible_inheritance_method"
)]
#[test_case(
vec![
&[
indoc! {"
class A(Generic[T]):
a: int32
@ -490,11 +497,11 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec!["field `a` has already declared in the ancestor classes"];
&["field `a` has already declared in the ancestor classes"];
"err_incompatible_inheritance_field"
)]
#[test_case(
vec![
&[
indoc! {"
class A:
def __init__(self):
@ -507,12 +514,12 @@ fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&s
pass
"}
],
vec!["duplicate definition of class `A` (at unknown: line 1 column 1)"];
&["duplicate definition of class `A` (at unknown:1:1)"];
"class same name"
)]
fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
fn test_analyze(source: &[&str], res: &[&str]) {
let print = false;
let mut composer: TopLevelComposer = Default::default();
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let internal_resolver = make_internal_resolver_with_tvar(
vec![
@ -527,15 +534,15 @@ fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
Arc::new(Resolver(internal_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
for s in source {
let ast = parse_program(s, Default::default()).unwrap();
let ast = parse_program(s, FileName::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) = {
match composer.register_top_level(ast, Some(resolver.clone()), "".into(), false) {
match composer.register_top_level(ast, Some(resolver.clone()), "", false) {
Ok(x) => x,
Err(msg) => {
if print {
println!("{}", msg);
println!("{msg}");
} else {
assert_eq!(res[0], msg);
}
@ -551,9 +558,9 @@ fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
if let Err(msg) = composer.start_analysis(false) {
if print {
println!("{}", msg);
println!("{}", msg.iter().sorted().join("\n----------\n"));
} else {
assert_eq!(res[0], msg);
assert_eq!(res[0], msg.iter().next().unwrap());
}
} else {
// skip 5 to skip primitives
@ -581,7 +588,7 @@ fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
return fib(n - 1)
"}
],
vec![];
&[];
"simple function"
)]
#[test_case(
@ -614,7 +621,7 @@ fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
return a.fun() + 2
"}
],
vec![];
&[];
"simple class body"
)]
#[test_case(
@ -639,7 +646,7 @@ fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
return [a, b]
"}
],
vec![];
&[];
"type var fun"
)]
#[test_case(
@ -660,7 +667,7 @@ fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
return ret if self.b else self.fun(self.a)
"}
],
vec![];
&[];
"type var class"
)]
#[test_case(
@ -684,12 +691,12 @@ fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
self.b = True
"}
],
vec![];
&[];
"no_init_inst_check"
)]
fn test_inference(source: Vec<&str>, res: Vec<&str>) {
fn test_inference(source: Vec<&str>, res: &[&str]) {
let print = true;
let mut composer: TopLevelComposer = Default::default();
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let internal_resolver = make_internal_resolver_with_tvar(
vec![
@ -711,15 +718,15 @@ fn test_inference(source: Vec<&str>, res: Vec<&str>) {
Arc::new(Resolver(internal_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
for s in source {
let ast = parse_program(s, Default::default()).unwrap();
let ast = parse_program(s, FileName::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) = {
match composer.register_top_level(ast, Some(resolver.clone()), "".into(), false) {
match composer.register_top_level(ast, Some(resolver.clone()), "", false) {
Ok(x) => x,
Err(msg) => {
if print {
println!("{}", msg);
println!("{msg}");
} else {
assert_eq!(res[0], msg);
}
@ -735,16 +742,14 @@ fn test_inference(source: Vec<&str>, res: Vec<&str>) {
if let Err(msg) = composer.start_analysis(true) {
if print {
println!("{}", msg);
println!("{}", msg.iter().sorted().join("\n----------\n"));
} else {
assert_eq!(res[0], msg);
assert_eq!(res[0], msg.iter().next().unwrap());
}
} else {
// skip 5 to skip primitives
let mut stringify_folder = TypeToStringFolder { unifier: &mut composer.unifier };
for (_i, (def, _)) in
composer.definition_ast_list.iter().skip(composer.builtin_num).enumerate()
{
for (def, _) in composer.definition_ast_list.iter().skip(composer.builtin_num) {
let def = &*def.read();
if let TopLevelDef::Function { instance_to_stmt, name, .. } = def {
@ -753,7 +758,7 @@ fn test_inference(source: Vec<&str>, res: Vec<&str>) {
name,
instance_to_stmt.len()
);
for inst in instance_to_stmt.iter() {
for inst in instance_to_stmt {
let ast = &inst.1.body;
for b in ast.iter() {
println!("{:?}", stringify_folder.fold_stmt(b.clone()).unwrap());
@ -771,22 +776,29 @@ fn make_internal_resolver_with_tvar(
unifier: &mut Unifier,
print: bool,
) -> Arc<ResolverInternal> {
let list_elem_tvar = unifier.get_fresh_var(Some("list_elem".into()), None);
let list = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::List.id(),
fields: HashMap::new(),
params: into_var_map([list_elem_tvar]),
});
let res: Arc<ResolverInternal> = ResolverInternal {
id_to_def: Default::default(),
id_to_def: Mutex::new(HashMap::from([("list".into(), PrimDef::List.id())])),
id_to_type: tvars
.into_iter()
.map(|(name, range)| {
(name, {
let (ty, id) = unifier.get_fresh_var_with_range(range.as_slice(), None, None);
let tvar = unifier.get_fresh_var_with_range(range.as_slice(), None, None);
if print {
println!("{}: {:?}, typevar{}", name, ty, id);
println!("{}: {:?}, typevar{}", name, tvar.ty, tvar.id);
}
ty
tvar.ty
})
})
.collect::<HashMap<_, _>>()
.into(),
class_names: Default::default(),
class_names: Mutex::new(HashMap::from([("list".into(), list)])),
}
.into();
if print {
@ -806,8 +818,8 @@ impl<'a> Fold<Option<Type>> for TypeToStringFolder<'a> {
Ok(if let Some(ty) = user {
self.unifier.internal_stringify(
ty,
&mut |id| format!("class{}", id.to_string()),
&mut |id| format!("typevar{}", id.to_string()),
&mut |id| format!("class{id}"),
&mut |id| format!("typevar{id}"),
&mut None,
)
} else {

View File

@ -1,4 +1,8 @@
use super::*;
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::helper::PrimDef;
use crate::typecheck::typedef::VarMap;
use nac3parser::ast::Constant;
#[derive(Clone, Debug)]
pub enum TypeAnnotation {
@ -12,7 +16,8 @@ pub enum TypeAnnotation {
// can only be CustomClassKind
Virtual(Box<TypeAnnotation>),
TypeVar(Type),
List(Box<TypeAnnotation>),
/// A `Literal` allowing a subset of literals.
Literal(Vec<Constant>),
Tuple(Vec<TypeAnnotation>),
}
@ -22,52 +27,57 @@ impl TypeAnnotation {
match self {
Primitive(ty) | TypeVar(ty) => unifier.stringify(*ty),
CustomClass { id, params } => {
let class_name = match unifier.top_level {
Some(ref top) => {
if let TopLevelDef::Class { name, .. } =
&*top.definitions.read()[id.0].read()
{
let class_name = if let Some(ref top) = unifier.top_level {
if let TopLevelDef::Class { name, .. } = &*top.definitions.read()[id.0].read() {
(*name).into()
} else {
unreachable!()
}
}
None => format!("class_def_{}", id.0),
};
format!(
"{}{}",
class_name,
{
let param_list = params.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ");
if param_list.is_empty() {
"".into()
} else {
format!("[{}]", param_list)
format!("class_def_{}", id.0)
};
format!("{}{}", class_name, {
let param_list =
params.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ");
if param_list.is_empty() {
String::new()
} else {
format!("[{param_list}]")
}
})
}
)
Literal(values) => {
format!("Literal({})", values.iter().map(|v| format!("{v:?}")).join(", "))
}
Virtual(ty) => format!("virtual[{}]", ty.stringify(unifier)),
List(ty) => format!("list[{}]", ty.stringify(unifier)),
Tuple(types) => {
format!("tuple[{}]", types.iter().map(|p| p.stringify(unifier)).collect_vec().join(", "))
format!(
"tuple[{}]",
types.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ")
)
}
}
}
}
pub fn parse_ast_to_type_annotation_kinds<T>(
/// Parses an AST expression `expr` into a [`TypeAnnotation`].
///
/// * `locked` - A [`HashMap`] containing the IDs of known definitions, mapped to a [`Vec`] of all
/// generic variables associated with the definition.
/// * `type_var` - The type variable associated with the type argument currently being parsed. Pass
/// [`None`] when this function is invoked externally.
pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
resolver: &(dyn SymbolResolver + Send + Sync),
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
unifier: &mut Unifier,
primitives: &PrimitiveStore,
expr: &ast::Expr<T>,
// the key stores the type_var of this topleveldef::class, we only need this field here
locked: HashMap<DefinitionId, Vec<Type>>,
) -> Result<TypeAnnotation, String> {
locked: HashMap<DefinitionId, Vec<Type>, S>,
) -> Result<TypeAnnotation, HashSet<String>> {
let name_handle = |id: &StrRef,
unifier: &mut Unifier,
locked: HashMap<DefinitionId, Vec<Type>>| {
locked: HashMap<DefinitionId, Vec<Type>, S>| {
if id == &"int32".into() {
Ok(TypeAnnotation::Primitive(primitives.int32))
} else if id == &"int64".into() {
@ -83,7 +93,7 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
} else if id == &"str".into() {
Ok(TypeAnnotation::Primitive(primitives.str))
} else if id == &"Exception".into() {
Ok(TypeAnnotation::CustomClass { id: DefinitionId(7), params: Default::default() })
Ok(TypeAnnotation::CustomClass { id: PrimDef::Exception.id(), params: Vec::default() })
} else if let Ok(obj_id) = resolver.get_identifier_def(*id) {
let type_vars = {
let def_read = top_level_defs[obj_id.0].try_read();
@ -91,10 +101,10 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
if let TopLevelDef::Class { type_vars, .. } = &*def_read {
type_vars.clone()
} else {
return Err(format!(
return Err(HashSet::from([format!(
"function cannot be used as a type (at {})",
expr.location
));
)]));
}
} else {
locked.get(&obj_id).unwrap().clone()
@ -102,23 +112,29 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
};
// check param number here
if !type_vars.is_empty() {
return Err(format!(
return Err(HashSet::from([format!(
"expect {} type variable parameter but got 0 (at {})",
type_vars.len(),
expr.location,
));
)]));
}
Ok(TypeAnnotation::CustomClass { id: obj_id, params: vec![] })
} else if let Ok(ty) = resolver.get_symbol_type(unifier, top_level_defs, primitives, *id) {
if let TypeEnum::TVar { .. } = unifier.get_ty(ty).as_ref() {
let var = unifier.get_fresh_var(Some(*id), Some(expr.location)).0;
let var = unifier.get_fresh_var(Some(*id), Some(expr.location)).ty;
unifier.unify(var, ty).unwrap();
Ok(TypeAnnotation::TypeVar(ty))
} else {
Err(format!("`{}` is not a valid type annotation (at {})", id, expr.location))
Err(HashSet::from([format!(
"`{}` is not a valid type annotation (at {})",
id, expr.location
)]))
}
} else {
Err(format!("`{}` is not a valid type annotation (at {})", id, expr.location))
Err(HashSet::from([format!(
"`{}` is not a valid type annotation (at {})",
id, expr.location
)]))
}
};
@ -126,20 +142,22 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
|id: &StrRef,
slice: &ast::Expr<T>,
unifier: &mut Unifier,
mut locked: HashMap<DefinitionId, Vec<Type>>| {
if vec!["virtual".into(), "Generic".into(), "list".into(), "tuple".into()].contains(id)
{
return Err(format!("keywords cannot be class name (at {})", expr.location));
mut locked: HashMap<DefinitionId, Vec<Type>, S>| {
if ["virtual".into(), "Generic".into(), "tuple".into(), "Option".into()].contains(id) {
return Err(HashSet::from([format!(
"keywords cannot be class name (at {})",
expr.location
)]));
}
let obj_id = resolver.get_identifier_def(*id)?;
let type_vars = {
let def_read = top_level_defs[obj_id.0].try_read();
if let Some(def_read) = def_read {
if let TopLevelDef::Class { type_vars, .. } = &*def_read {
type_vars.clone()
} else {
let TopLevelDef::Class { type_vars, .. } = &*def_read else {
unreachable!("must be class here")
}
};
type_vars.clone()
} else {
locked.get(&obj_id).unwrap().clone()
}
@ -152,12 +170,12 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
vec![slice]
};
if type_vars.len() != params_ast.len() {
return Err(format!(
return Err(HashSet::from([format!(
"expect {} type parameters but got {} (at {})",
type_vars.len(),
params_ast.len(),
params_ast[0].location,
));
)]));
}
let result = params_ast
.iter()
@ -181,15 +199,17 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
if no_type_var {
result
} else {
return Err(format!(
"application of type vars to generic class \
is not currently supported (at {})",
return Err(HashSet::from([
format!(
"application of type vars to generic class is not currently supported (at {})",
params_ast[0].location
));
),
]));
}
};
Ok(TypeAnnotation::CustomClass { id: obj_id, params: param_type_infos })
};
match &expr.node {
ast::ExprKind::Name { id, .. } => name_handle(id, unifier, locked),
// virtual
@ -212,23 +232,6 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
Ok(TypeAnnotation::Virtual(def.into()))
}
// list
ast::ExprKind::Subscript { value, slice, .. }
if {
matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"list".into())
} =>
{
let def_ann = parse_ast_to_type_annotation_kinds(
resolver,
top_level_defs,
unifier,
primitives,
slice.as_ref(),
locked,
)?;
Ok(TypeAnnotation::List(def_ann.into()))
}
// option
ast::ExprKind::Subscript { value, slice, .. }
if {
@ -281,16 +284,70 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
Ok(TypeAnnotation::Tuple(type_annotations))
}
// Literal
ast::ExprKind::Subscript { value, slice, .. }
if {
matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"Literal".into())
} =>
{
let tup_elts = {
if let ast::ExprKind::Tuple { elts, .. } = &slice.node {
elts.as_slice()
} else {
std::slice::from_ref(slice.as_ref())
}
};
let type_annotations = tup_elts
.iter()
.map(|e| match &e.node {
ast::ExprKind::Constant { value, .. } => {
Ok(TypeAnnotation::Literal(vec![value.clone()]))
}
_ => parse_ast_to_type_annotation_kinds(
resolver,
top_level_defs,
unifier,
primitives,
e,
locked.clone(),
),
})
.collect::<Result<Vec<_>, _>>()?
.into_iter()
.flat_map(|type_ann| match type_ann {
TypeAnnotation::Literal(values) => values,
_ => unreachable!(),
})
.collect_vec();
if type_annotations.len() == 1 {
Ok(TypeAnnotation::Literal(type_annotations))
} else {
Err(HashSet::from([format!(
"multiple literal bounds are currently unsupported (at {})",
value.location
)]))
}
}
// custom class
ast::ExprKind::Subscript { value, slice, .. } => {
if let ast::ExprKind::Name { id, .. } = &value.node {
class_name_handle(id, slice, unifier, locked)
} else {
Err(format!("unsupported expression type for class name (at {})", value.location))
Err(HashSet::from([format!(
"unsupported expression type for class name (at {})",
value.location
)]))
}
}
_ => Err(format!("unsupported expression for type annotation (at {})", expr.location)),
ast::ExprKind::Constant { value, .. } => Ok(TypeAnnotation::Literal(vec![value.clone()])),
_ => Err(HashSet::from([format!(
"unsupported expression for type annotation (at {})",
expr.location
)])),
}
}
@ -300,43 +357,46 @@ pub fn parse_ast_to_type_annotation_kinds<T>(
pub fn get_type_from_type_annotation_kinds(
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
unifier: &mut Unifier,
primitives: &PrimitiveStore,
ann: &TypeAnnotation,
subst_list: &mut Option<Vec<Type>>
) -> Result<Type, String> {
subst_list: &mut Option<Vec<Type>>,
) -> Result<Type, HashSet<String>> {
match ann {
TypeAnnotation::CustomClass { id: obj_id, params } => {
let def_read = top_level_defs[obj_id.0].read();
let class_def: &TopLevelDef = def_read.deref();
if let TopLevelDef::Class { fields, methods, type_vars, .. } = class_def {
let class_def: &TopLevelDef = &def_read;
let TopLevelDef::Class { fields, methods, type_vars, .. } = class_def else {
unreachable!("should be class def here")
};
if type_vars.len() != params.len() {
Err(format!(
return Err(HashSet::from([format!(
"unexpected number of type parameters: expected {} but got {}",
type_vars.len(),
params.len()
))
} else {
)]));
}
let param_ty = params
.iter()
.map(|x| {
get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
x,
subst_list
)
get_type_from_type_annotation_kinds(top_level_defs, unifier, x, subst_list)
})
.collect::<Result<Vec<_>, _>>()?;
let subst = {
// check for compatible range
// TODO: if allow type var to be applied(now this disallowed in the parse_to_type_annotation), need more check
let mut result: HashMap<u32, Type> = HashMap::new();
let mut result = VarMap::new();
for (tvar, p) in type_vars.iter().zip(param_ty) {
if let TypeEnum::TVar { id, range, fields: None, name, loc } =
unifier.get_ty(*tvar).as_ref()
{
match unifier.get_ty(*tvar).as_ref() {
TypeEnum::TVar {
id,
range,
fields: None,
name,
loc,
is_const_generic: false,
} => {
let ok: bool = {
// create a temp type var and unify to check compatibility
p == *tvar || {
@ -345,29 +405,51 @@ pub fn get_type_from_type_annotation_kinds(
*name,
*loc,
);
unifier.unify(temp.0, p).is_ok()
unifier.unify(temp.ty, p).is_ok()
}
};
if ok {
result.insert(*id, p);
} else {
return Err(format!(
return Err(HashSet::from([format!(
"cannot apply type {} to type variable with id {:?}",
unifier.internal_stringify(
p,
&mut |id| format!("class{}", id),
&mut |id| format!("typevar{}", id),
&mut |id| format!("class{id}"),
&mut |id| format!("typevar{id}"),
&mut None
),
*id
));
)]));
}
}
TypeEnum::TVar { id, range, name, loc, is_const_generic: true, .. } => {
let ty = range[0];
let ok: bool = {
// create a temp type var and unify to check compatibility
p == *tvar || {
let temp = unifier.get_fresh_const_generic_var(ty, *name, *loc);
unifier.unify(temp.ty, p).is_ok()
}
};
if ok {
result.insert(*id, p);
} else {
unreachable!("must be generic type var")
return Err(HashSet::from([format!(
"cannot apply type {} to type variable {}",
unifier.stringify(p),
name.unwrap_or_else(|| format!("typevar{id}").into()),
)]));
}
}
_ => unreachable!("must be generic type var"),
}
}
result
};
// Class Attributes keep a copy with Class Definition and are not added to objects
let mut tobj_fields = methods
.iter()
.map(|(name, ty, _)| {
@ -387,40 +469,37 @@ pub fn get_type_from_type_annotation_kinds(
params: subst,
});
if need_subst {
subst_list.as_mut().map(|wl| wl.push(ty));
if let Some(wl) = subst_list.as_mut() {
wl.push(ty);
}
}
Ok(ty)
}
} else {
unreachable!("should be class def here")
}
}
TypeAnnotation::Primitive(ty) | TypeAnnotation::TypeVar(ty) => Ok(*ty),
TypeAnnotation::Literal(values) => {
let values = values
.iter()
.map(SymbolValue::from_constant_inferred)
.collect::<Result<Vec<_>, _>>()
.map_err(|err| HashSet::from([err]))?;
let var = unifier.get_fresh_literal(values, None);
Ok(var)
}
TypeAnnotation::Virtual(ty) => {
let ty = get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
ty.as_ref(),
subst_list
subst_list,
)?;
Ok(unifier.add_ty(TypeEnum::TVirtual { ty }))
}
TypeAnnotation::List(ty) => {
let ty = get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
ty.as_ref(),
subst_list
)?;
Ok(unifier.add_ty(TypeEnum::TList { ty }))
}
TypeAnnotation::Tuple(tys) => {
let tys = tys
.iter()
.map(|x| {
get_type_from_type_annotation_kinds(top_level_defs, unifier, primitives, x, subst_list)
get_type_from_type_annotation_kinds(top_level_defs, unifier, x, subst_list)
})
.collect::<Result<Vec<_>, _>>()?;
Ok(unifier.add_ty(TypeEnum::TTuple { ty: tys }))
@ -437,9 +516,10 @@ pub fn get_type_from_type_annotation_kinds(
/// considered to be type variables associated with the class \
/// \
/// But note that here we do not make a duplication of `T`, `V`, we directly
/// use them as they are in the TopLevelDef::Class since those in the
/// TopLevelDef::Class.type_vars will be substitute later when seeing applications/instantiations
/// use them as they are in the [`TopLevelDef::Class`] since those in the
/// `TopLevelDef::Class.type_vars` will be substitute later when seeing applications/instantiations
/// the Type of their fields and methods will also be subst when application/instantiation
#[must_use]
pub fn make_self_type_annotation(type_vars: &[Type], object_id: DefinitionId) -> TypeAnnotation {
TypeAnnotation::CustomClass {
id: object_id,
@ -450,27 +530,25 @@ pub fn make_self_type_annotation(type_vars: &[Type], object_id: DefinitionId) ->
/// get all the occurences of type vars contained in a type annotation
/// e.g. `A[int, B[T], V, virtual[C[G]]]` => [T, V, G]
/// this function will not make a duplicate of type var
#[must_use]
pub fn get_type_var_contained_in_type_annotation(ann: &TypeAnnotation) -> Vec<TypeAnnotation> {
let mut result: Vec<TypeAnnotation> = Vec::new();
match ann {
TypeAnnotation::TypeVar(..) => result.push(ann.clone()),
TypeAnnotation::Virtual(ann) => {
result.extend(get_type_var_contained_in_type_annotation(ann.as_ref()))
result.extend(get_type_var_contained_in_type_annotation(ann.as_ref()));
}
TypeAnnotation::CustomClass { params, .. } => {
for p in params {
result.extend(get_type_var_contained_in_type_annotation(p));
}
}
TypeAnnotation::List(ann) => {
result.extend(get_type_var_contained_in_type_annotation(ann.as_ref()))
}
TypeAnnotation::Tuple(anns) => {
for a in anns {
result.extend(get_type_var_contained_in_type_annotation(a));
}
}
TypeAnnotation::Primitive(..) => {}
TypeAnnotation::Primitive(..) | TypeAnnotation::Literal { .. } => {}
}
result
}
@ -485,21 +563,20 @@ pub fn check_overload_type_annotation_compatible(
(TypeAnnotation::Primitive(a), TypeAnnotation::Primitive(b)) => a == b,
(TypeAnnotation::TypeVar(a), TypeAnnotation::TypeVar(b)) => {
let a = unifier.get_ty(*a);
let a = a.deref();
let a = &*a;
let b = unifier.get_ty(*b);
let b = b.deref();
if let (
let b = &*b;
let (
TypeEnum::TVar { id: a, fields: None, .. },
TypeEnum::TVar { id: b, fields: None, .. },
) = (a, b)
{
a == b
} else {
else {
unreachable!("must be type var")
};
a == b
}
}
(TypeAnnotation::Virtual(a), TypeAnnotation::Virtual(b))
| (TypeAnnotation::List(a), TypeAnnotation::List(b)) => {
(TypeAnnotation::Virtual(a), TypeAnnotation::Virtual(b)) => {
check_overload_type_annotation_compatible(a.as_ref(), b.as_ref(), unifier)
}

View File

@ -1,14 +1,18 @@
use crate::typecheck::typedef::TypeEnum;
use crate::toplevel::helper::PrimDef;
use super::type_inferencer::Inferencer;
use super::typedef::Type;
use nac3parser::ast::{self, Expr, ExprKind, Stmt, StmtKind, StrRef};
use super::typedef::{Type, TypeEnum};
use nac3parser::ast::{
self, Constant, Expr, ExprKind,
Operator::{LShift, RShift},
Stmt, StmtKind, StrRef,
};
use std::{collections::HashSet, iter::once};
impl<'a> Inferencer<'a> {
fn should_have_value(&mut self, expr: &Expr<Option<Type>>) -> Result<(), String> {
fn should_have_value(&mut self, expr: &Expr<Option<Type>>) -> Result<(), HashSet<String>> {
if matches!(expr.custom, Some(ty) if self.unifier.unioned(ty, self.primitives.none)) {
Err(format!("Error at {}: cannot have value none", expr.location))
Err(HashSet::from([format!("Error at {}: cannot have value none", expr.location)]))
} else {
Ok(())
}
@ -18,10 +22,11 @@ impl<'a> Inferencer<'a> {
&mut self,
pattern: &Expr<Option<Type>>,
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<(), String> {
) -> Result<(), HashSet<String>> {
match &pattern.node {
ast::ExprKind::Name { id, .. } if id == &"none".into() =>
Err(format!("cannot assign to a `none` (at {})", pattern.location)),
ExprKind::Name { id, .. } if id == &"none".into() => {
Err(HashSet::from([format!("cannot assign to a `none` (at {})", pattern.location)]))
}
ExprKind::Name { id, .. } => {
if !defined_identifiers.contains(id) {
defined_identifiers.insert(*id);
@ -30,7 +35,7 @@ impl<'a> Inferencer<'a> {
Ok(())
}
ExprKind::Tuple { elts, .. } => {
for elt in elts.iter() {
for elt in elts {
self.check_pattern(elt, defined_identifiers)?;
self.should_have_value(elt)?;
}
@ -41,16 +46,17 @@ impl<'a> Inferencer<'a> {
self.should_have_value(value)?;
self.check_expr(slice, defined_identifiers)?;
if let TypeEnum::TTuple { .. } = &*self.unifier.get_ty(value.custom.unwrap()) {
return Err(format!(
return Err(HashSet::from([format!(
"Error at {}: cannot assign to tuple element",
value.location
));
)]));
}
Ok(())
}
ExprKind::Constant { .. } => {
Err(format!("cannot assign to a constant (at {})", pattern.location))
}
ExprKind::Constant { .. } => Err(HashSet::from([format!(
"cannot assign to a constant (at {})",
pattern.location
)])),
_ => self.check_expr(pattern, defined_identifiers),
}
}
@ -59,15 +65,18 @@ impl<'a> Inferencer<'a> {
&mut self,
expr: &Expr<Option<Type>>,
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<(), String> {
) -> Result<(), HashSet<String>> {
// there are some cases where the custom field is None
if let Some(ty) = &expr.custom {
if !self.unifier.is_concrete(*ty, &self.function_data.bound_variables) {
return Err(format!(
if !matches!(&expr.node, ExprKind::Constant { value: Constant::Ellipsis, .. })
&& !ty.obj_id(self.unifier).is_some_and(|id| id == PrimDef::List.id())
&& !self.unifier.is_concrete(*ty, &self.function_data.bound_variables)
{
return Err(HashSet::from([format!(
"expected concrete type at {} but got {}",
expr.location,
self.unifier.get_ty(*ty).get_type_name()
));
)]));
}
}
match &expr.node {
@ -87,10 +96,10 @@ impl<'a> Inferencer<'a> {
self.defined_identifiers.insert(*id);
}
Err(e) => {
return Err(format!(
return Err(HashSet::from([format!(
"type error at identifier `{}` ({}) at {}",
id, e, expr.location
));
)]))
}
}
}
@ -98,7 +107,7 @@ impl<'a> Inferencer<'a> {
ExprKind::List { elts, .. }
| ExprKind::Tuple { elts, .. }
| ExprKind::BoolOp { values: elts, .. } => {
for elt in elts.iter() {
for elt in elts {
self.check_expr(elt, defined_identifiers)?;
self.should_have_value(elt)?;
}
@ -107,11 +116,25 @@ impl<'a> Inferencer<'a> {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
}
ExprKind::BinOp { left, right, .. } => {
ExprKind::BinOp { left, op, right } => {
self.check_expr(left, defined_identifiers)?;
self.check_expr(right, defined_identifiers)?;
self.should_have_value(left)?;
self.should_have_value(right)?;
// Check whether a bitwise shift has a negative RHS constant value
if *op == LShift || *op == RShift {
if let ExprKind::Constant { value, .. } = &right.node {
let Constant::Int(rhs_val) = value else { unreachable!() };
if *rhs_val < 0 {
return Err(HashSet::from([format!(
"shift count is negative at {}",
right.location
)]));
}
}
}
}
ExprKind::UnaryOp { operand, .. } => {
self.check_expr(operand, defined_identifiers)?;
@ -141,7 +164,7 @@ impl<'a> Inferencer<'a> {
}
ExprKind::Lambda { args, body } => {
let mut defined_identifiers = defined_identifiers.clone();
for arg in args.args.iter() {
for arg in &args.args {
// TODO: should we check the types here?
if !defined_identifiers.contains(&arg.node.arg) {
defined_identifiers.insert(arg.node.arg);
@ -179,24 +202,45 @@ impl<'a> Inferencer<'a> {
Ok(())
}
/// Check that the return value is a non-`alloca` type, effectively only allowing primitive types.
///
/// This is a workaround preventing the caller from using a variable `alloca`-ed in the body, which
/// is freed when the function returns.
fn check_return_value_ty(&mut self, ret_ty: Type) -> bool {
match &*self.unifier.get_ty_immutable(ret_ty) {
TypeEnum::TObj { .. } => [
self.primitives.int32,
self.primitives.int64,
self.primitives.uint32,
self.primitives.uint64,
self.primitives.float,
self.primitives.bool,
]
.iter()
.any(|allowed_ty| self.unifier.unioned(ret_ty, *allowed_ty)),
TypeEnum::TTuple { ty } => ty.iter().all(|t| self.check_return_value_ty(*t)),
_ => false,
}
}
// check statements for proper identifier def-use and return on all paths
fn check_stmt(
&mut self,
stmt: &Stmt<Option<Type>>,
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<bool, String> {
) -> Result<bool, HashSet<String>> {
match &stmt.node {
StmtKind::For { target, iter, body, orelse, .. } => {
self.check_expr(iter, defined_identifiers)?;
self.should_have_value(iter)?;
let mut local_defined_identifiers = defined_identifiers.clone();
for stmt in orelse.iter() {
for stmt in orelse {
self.check_stmt(stmt, &mut local_defined_identifiers)?;
}
let mut local_defined_identifiers = defined_identifiers.clone();
self.check_pattern(target, &mut local_defined_identifiers)?;
self.should_have_value(target)?;
for stmt in body.iter() {
for stmt in body {
self.check_stmt(stmt, &mut local_defined_identifiers)?;
}
Ok(false)
@ -209,7 +253,7 @@ impl<'a> Inferencer<'a> {
let body_returned = self.check_block(body, &mut body_identifiers)?;
let orelse_returned = self.check_block(orelse, &mut orelse_identifiers)?;
for ident in body_identifiers.iter() {
for ident in &body_identifiers {
if !defined_identifiers.contains(ident) && orelse_identifiers.contains(ident) {
defined_identifiers.insert(*ident);
}
@ -226,7 +270,7 @@ impl<'a> Inferencer<'a> {
}
StmtKind::With { items, body, .. } => {
let mut new_defined_identifiers = defined_identifiers.clone();
for item in items.iter() {
for item in items {
self.check_expr(&item.context_expr, defined_identifiers)?;
if let Some(var) = item.optional_vars.as_ref() {
self.check_pattern(var, &mut new_defined_identifiers)?;
@ -238,7 +282,7 @@ impl<'a> Inferencer<'a> {
StmtKind::Try { body, handlers, orelse, finalbody, .. } => {
self.check_block(body, &mut defined_identifiers.clone())?;
self.check_block(orelse, &mut defined_identifiers.clone())?;
for handler in handlers.iter() {
for handler in handlers {
let mut defined_identifiers = defined_identifiers.clone();
let ast::ExcepthandlerKind::ExceptHandler { name, body, .. } = &handler.node;
if let Some(name) = name {
@ -273,6 +317,30 @@ impl<'a> Inferencer<'a> {
if let Some(value) = value {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
// Check that the return value is a non-`alloca` type, effectively only allowing primitive types.
// This is a workaround preventing the caller from using a variable `alloca`-ed in the body, which
// is freed when the function returns.
if let Some(ret_ty) = value.custom {
// Explicitly allow ellipsis as a return value, as the type of the ellipsis is contextually
// inferred and just generates an unconditional assertion
if matches!(
value.node,
ExprKind::Constant { value: Constant::Ellipsis, .. }
) {
return Ok(true);
}
if !self.check_return_value_ty(ret_ty) {
return Err(HashSet::from([
format!(
"return value of type {} must be a primitive or a tuple of primitives at {}",
self.unifier.stringify(ret_ty),
value.location,
),
]));
}
}
}
Ok(true)
}
@ -291,11 +359,11 @@ impl<'a> Inferencer<'a> {
&mut self,
block: &[Stmt<Option<Type>>],
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<bool, String> {
) -> Result<bool, HashSet<String>> {
let mut ret = false;
for stmt in block {
if ret {
return Err(format!("dead code at {:?}", stmt.location));
eprintln!("warning: dead code at {}\n", stmt.location);
}
if self.check_stmt(stmt, defined_identifiers)? {
ret = true;

View File

@ -1,69 +1,150 @@
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::helper::PrimDef;
use crate::toplevel::numpy::{make_ndarray_ty, unpack_ndarray_var_tys};
use crate::typecheck::{
type_inferencer::*,
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
};
use nac3parser::ast::{self, StrRef};
use itertools::{iproduct, Itertools};
use nac3parser::ast::StrRef;
use nac3parser::ast::{Cmpop, Operator, Unaryop};
use std::cmp::max;
use std::collections::HashMap;
use std::rc::Rc;
use strum::IntoEnumIterator;
pub fn binop_name(op: &Operator) -> &'static str {
match op {
Operator::Add => "__add__",
Operator::Sub => "__sub__",
Operator::Div => "__truediv__",
Operator::Mod => "__mod__",
Operator::Mult => "__mul__",
Operator::Pow => "__pow__",
Operator::BitOr => "__or__",
Operator::BitXor => "__xor__",
Operator::BitAnd => "__and__",
Operator::LShift => "__lshift__",
Operator::RShift => "__rshift__",
Operator::FloorDiv => "__floordiv__",
Operator::MatMult => "__matmul__",
/// The variant of a binary operator.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum BinopVariant {
/// The normal variant.
/// For addition, it would be `+`.
Normal,
/// The "Augmented Assigning Operator" variant.
/// For addition, it would be `+=`.
AugAssign,
}
/// A binary operator with its variant.
#[derive(Debug, Clone, Copy)]
pub struct Binop {
/// The base [`Operator`] of this binary operator.
pub base: Operator,
/// The variant of this binary operator.
pub variant: BinopVariant,
}
impl Binop {
/// Make a [`Binop`] of the normal variant from an [`Operator`].
#[must_use]
pub fn normal(base: Operator) -> Self {
Binop { base, variant: BinopVariant::Normal }
}
/// Make a [`Binop`] of the aug assign variant from an [`Operator`].
#[must_use]
pub fn aug_assign(base: Operator) -> Self {
Binop { base, variant: BinopVariant::AugAssign }
}
}
pub fn binop_assign_name(op: &Operator) -> &'static str {
match op {
Operator::Add => "__iadd__",
Operator::Sub => "__isub__",
Operator::Div => "__itruediv__",
Operator::Mod => "__imod__",
Operator::Mult => "__imul__",
Operator::Pow => "__ipow__",
Operator::BitOr => "__ior__",
Operator::BitXor => "__ixor__",
Operator::BitAnd => "__iand__",
Operator::LShift => "__ilshift__",
Operator::RShift => "__irshift__",
Operator::FloorDiv => "__ifloordiv__",
Operator::MatMult => "__imatmul__",
}
/// Details about an operator (unary, binary, etc...) in Python
#[derive(Debug, Clone, Copy)]
pub struct OpInfo {
/// The method name of the binary operator.
/// For addition, this would be `__add__`, and `__iadd__` if
/// it is the augmented assigning variant.
pub method_name: &'static str,
/// The symbol of the binary operator.
/// For addition, this would be `+`, and `+=` if
/// it is the augmented assigning variant.
pub symbol: &'static str,
}
pub fn unaryop_name(op: &Unaryop) -> &'static str {
match op {
Unaryop::UAdd => "__pos__",
Unaryop::USub => "__neg__",
Unaryop::Not => "__not__",
Unaryop::Invert => "__inv__",
}
/// Helper macro to conveniently build an [`OpInfo`].
///
/// Example usage: `make_info("add", "+")` generates `OpInfo { name: "__add__", symbol: "+" }`
macro_rules! make_op_info {
($name:expr, $symbol:expr) => {
OpInfo { method_name: concat!("__", $name, "__"), symbol: $symbol }
};
}
pub fn comparison_name(op: &Cmpop) -> Option<&'static str> {
pub trait HasOpInfo {
fn op_info(&self) -> OpInfo;
}
fn try_get_cmpop_info(op: Cmpop) -> Option<OpInfo> {
match op {
Cmpop::Lt => Some("__lt__"),
Cmpop::LtE => Some("__le__"),
Cmpop::Gt => Some("__gt__"),
Cmpop::GtE => Some("__ge__"),
Cmpop::Eq => Some("__eq__"),
Cmpop::NotEq => Some("__ne__"),
Cmpop::Lt => Some(make_op_info!("lt", "<")),
Cmpop::LtE => Some(make_op_info!("le", "<=")),
Cmpop::Gt => Some(make_op_info!("gt", ">")),
Cmpop::GtE => Some(make_op_info!("ge", ">=")),
Cmpop::Eq => Some(make_op_info!("eq", "==")),
Cmpop::NotEq => Some(make_op_info!("ne", "!=")),
_ => None,
}
}
impl OpInfo {
#[must_use]
pub fn supports_cmpop(op: Cmpop) -> bool {
try_get_cmpop_info(op).is_some()
}
}
impl HasOpInfo for Cmpop {
fn op_info(&self) -> OpInfo {
try_get_cmpop_info(*self).expect("{self:?} is not supported")
}
}
impl HasOpInfo for Binop {
fn op_info(&self) -> OpInfo {
// Helper macro to generate both the normal variant [`OpInfo`] and the
// augmented assigning variant [`OpInfo`] for a binary operator conveniently.
macro_rules! info {
($name:literal, $symbol:literal) => {
(
make_op_info!($name, $symbol),
make_op_info!(concat!("i", $name), concat!($symbol, "=")),
)
};
}
let (normal_variant, aug_assign_variant) = match self.base {
Operator::Add => info!("add", "+"),
Operator::Sub => info!("sub", "-"),
Operator::Div => info!("truediv", "/"),
Operator::Mod => info!("mod", "%"),
Operator::Mult => info!("mul", "*"),
Operator::Pow => info!("pow", "**"),
Operator::BitOr => info!("or", "|"),
Operator::BitXor => info!("xor", "^"),
Operator::BitAnd => info!("and", "&"),
Operator::LShift => info!("lshift", "<<"),
Operator::RShift => info!("rshift", ">>"),
Operator::FloorDiv => info!("floordiv", "//"),
Operator::MatMult => info!("matmul", "@"),
};
match self.variant {
BinopVariant::Normal => normal_variant,
BinopVariant::AugAssign => aug_assign_variant,
}
}
}
impl HasOpInfo for Unaryop {
fn op_info(&self) -> OpInfo {
match self {
Unaryop::UAdd => make_op_info!("pos", "+"),
Unaryop::USub => make_op_info!("neg", "-"),
Unaryop::Not => make_op_info!("not", "not"), // i.e., `not False`, so the symbol is just `not`.
Unaryop::Invert => make_op_info!("inv", "~"),
}
}
}
pub(super) fn with_fields<F>(unifier: &mut Unifier, ty: Type, f: F)
where
F: FnOnce(&mut Unifier, &mut HashMap<StrRef, (Type, bool)>),
@ -86,38 +167,28 @@ pub fn impl_binop(
_store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
ops: &[ast::Operator],
ret_ty: Option<Type>,
ops: &[Operator],
) {
with_fields(unifier, ty, |unifier, fields| {
let (other_ty, other_var_id) = if other_ty.len() == 1 {
(other_ty[0], None)
} else {
let (ty, var_id) = unifier.get_fresh_var_with_range(other_ty, Some("N".into()), None);
(ty, Some(var_id))
let tvar = unifier.get_fresh_var_with_range(other_ty, Some("N".into()), None);
(tvar.ty, Some(tvar.id))
};
let function_vars = if let Some(var_id) = other_var_id {
vec![(var_id, other_ty)].into_iter().collect::<HashMap<_, _>>()
vec![(var_id, other_ty)].into_iter().collect::<VarMap>()
} else {
HashMap::new()
VarMap::new()
};
for op in ops {
fields.insert(binop_name(op).into(), {
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty,
vars: function_vars.clone(),
args: vec![FuncArg {
ty: other_ty,
default_value: None,
name: "other".into(),
}],
})),
false,
)
});
fields.insert(binop_assign_name(op).into(), {
let ret_ty = ret_ty.unwrap_or_else(|| unifier.get_fresh_var(None, None).ty);
for (base_op, variant) in iproduct!(ops, [BinopVariant::Normal, BinopVariant::AugAssign]) {
let op = Binop { base: *base_op, variant };
fields.insert(op.op_info().method_name.into(), {
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty,
@ -135,15 +206,17 @@ pub fn impl_binop(
});
}
pub fn impl_unaryop(unifier: &mut Unifier, ty: Type, ret_ty: Type, ops: &[ast::Unaryop]) {
pub fn impl_unaryop(unifier: &mut Unifier, ty: Type, ret_ty: Option<Type>, ops: &[Unaryop]) {
with_fields(unifier, ty, |unifier, fields| {
let ret_ty = ret_ty.unwrap_or_else(|| unifier.get_fresh_var(None, None).ty);
for op in ops {
fields.insert(
unaryop_name(op).into(),
op.op_info().method_name.into(),
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty,
vars: HashMap::new(),
vars: VarMap::new(),
args: vec![],
})),
false,
@ -155,19 +228,35 @@ pub fn impl_unaryop(unifier: &mut Unifier, ty: Type, ret_ty: Type, ops: &[ast::U
pub fn impl_cmpop(
unifier: &mut Unifier,
store: &PrimitiveStore,
_store: &PrimitiveStore,
ty: Type,
other_ty: Type,
ops: &[ast::Cmpop],
other_ty: &[Type],
ops: &[Cmpop],
ret_ty: Option<Type>,
) {
with_fields(unifier, ty, |unifier, fields| {
let (other_ty, other_var_id) = if other_ty.len() == 1 {
(other_ty[0], None)
} else {
let tvar = unifier.get_fresh_var_with_range(other_ty, Some("N".into()), None);
(tvar.ty, Some(tvar.id))
};
let function_vars = if let Some(var_id) = other_var_id {
vec![(var_id, other_ty)].into_iter().collect::<VarMap>()
} else {
VarMap::new()
};
let ret_ty = ret_ty.unwrap_or_else(|| unifier.get_fresh_var(None, None).ty);
for op in ops {
fields.insert(
comparison_name(op).unwrap().into(),
op.op_info().method_name.into(),
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: store.bool,
vars: HashMap::new(),
ret: ret_ty,
vars: function_vars.clone(),
args: vec![FuncArg {
ty: other_ty,
default_value: None,
@ -181,13 +270,13 @@ pub fn impl_cmpop(
});
}
/// Add, Sub, Mult
/// `Add`, `Sub`, `Mult`
pub fn impl_basic_arithmetic(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
ret_ty: Option<Type>,
) {
impl_binop(
unifier,
@ -195,94 +284,390 @@ pub fn impl_basic_arithmetic(
ty,
other_ty,
ret_ty,
&[ast::Operator::Add, ast::Operator::Sub, ast::Operator::Mult],
)
&[Operator::Add, Operator::Sub, Operator::Mult],
);
}
/// Pow
/// `Pow`
pub fn impl_pow(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
ret_ty: Option<Type>,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::Pow])
impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::Pow]);
}
/// BitOr, BitXor, BitAnd
/// `BitOr`, `BitXor`, `BitAnd`
pub fn impl_bitwise_arithmetic(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_binop(
unifier,
store,
ty,
&[ty],
ty,
&[ast::Operator::BitAnd, ast::Operator::BitOr, ast::Operator::BitXor],
)
Some(ty),
&[Operator::BitAnd, Operator::BitOr, Operator::BitXor],
);
}
/// LShift, RShift
/// `LShift`, `RShift`
pub fn impl_bitwise_shift(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_binop(unifier, store, ty, &[ty], ty, &[ast::Operator::LShift, ast::Operator::RShift])
impl_binop(
unifier,
store,
ty,
&[store.int32, store.uint32],
Some(ty),
&[Operator::LShift, Operator::RShift],
);
}
/// Div
pub fn impl_div(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: &[Type]) {
impl_binop(unifier, store, ty, other_ty, store.float, &[ast::Operator::Div])
/// `Div`
pub fn impl_div(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::Div]);
}
/// FloorDiv
/// `FloorDiv`
pub fn impl_floordiv(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
ret_ty: Option<Type>,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::FloorDiv])
impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::FloorDiv]);
}
/// Mod
/// `Mod`
pub fn impl_mod(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
ret_ty: Option<Type>,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::Mod])
impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::Mod]);
}
/// UAdd, USub
pub fn impl_sign(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, ty, ty, &[ast::Unaryop::UAdd, ast::Unaryop::USub])
/// [`Operator::MatMult`]
pub fn impl_matmul(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::MatMult]);
}
/// Invert
pub fn impl_invert(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, ty, ty, &[ast::Unaryop::Invert])
/// `UAdd`, `USub`
pub fn impl_sign(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, ret_ty: Option<Type>) {
impl_unaryop(unifier, ty, ret_ty, &[Unaryop::UAdd, Unaryop::USub]);
}
/// Not
pub fn impl_not(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, ty, store.bool, &[ast::Unaryop::Not])
/// `Invert`
pub fn impl_invert(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, ret_ty: Option<Type>) {
impl_unaryop(unifier, ty, ret_ty, &[Unaryop::Invert]);
}
/// Lt, LtE, Gt, GtE
pub fn impl_comparison(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: Type) {
/// `Not`
pub fn impl_not(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, ret_ty: Option<Type>) {
impl_unaryop(unifier, ty, ret_ty, &[Unaryop::Not]);
}
/// `Lt`, `LtE`, `Gt`, `GtE`
pub fn impl_comparison(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
) {
impl_cmpop(
unifier,
store,
ty,
other_ty,
&[ast::Cmpop::Lt, ast::Cmpop::Gt, ast::Cmpop::LtE, ast::Cmpop::GtE],
)
&[Cmpop::Lt, Cmpop::Gt, Cmpop::LtE, Cmpop::GtE],
ret_ty,
);
}
/// Eq, NotEq
pub fn impl_eq(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_cmpop(unifier, store, ty, ty, &[ast::Cmpop::Eq, ast::Cmpop::NotEq])
/// `Eq`, `NotEq`
pub fn impl_eq(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
) {
impl_cmpop(unifier, store, ty, other_ty, &[Cmpop::Eq, Cmpop::NotEq], ret_ty);
}
/// Returns the expected return type of binary operations with at least one `ndarray` operand.
pub fn typeof_ndarray_broadcast(
unifier: &mut Unifier,
primitives: &PrimitiveStore,
left: Type,
right: Type,
) -> Result<Type, String> {
let is_left_ndarray = left.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
let is_right_ndarray = right.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
assert!(is_left_ndarray || is_right_ndarray);
if is_left_ndarray && is_right_ndarray {
// Perform broadcasting on two ndarray operands.
let (left_ty_dtype, left_ty_ndims) = unpack_ndarray_var_tys(unifier, left);
let (right_ty_dtype, right_ty_ndims) = unpack_ndarray_var_tys(unifier, right);
assert!(unifier.unioned(left_ty_dtype, right_ty_dtype));
let left_ty_ndims = match &*unifier.get_ty_immutable(left_ty_ndims) {
TypeEnum::TLiteral { values, .. } => values.clone(),
_ => unreachable!(),
};
let right_ty_ndims = match &*unifier.get_ty_immutable(right_ty_ndims) {
TypeEnum::TLiteral { values, .. } => values.clone(),
_ => unreachable!(),
};
let res_ndims = left_ty_ndims
.into_iter()
.cartesian_product(right_ty_ndims)
.map(|(left, right)| {
let left_val = u64::try_from(left).unwrap();
let right_val = u64::try_from(right).unwrap();
max(left_val, right_val)
})
.unique()
.map(SymbolValue::U64)
.collect_vec();
let res_ndims = unifier.get_fresh_literal(res_ndims, None);
Ok(make_ndarray_ty(unifier, primitives, Some(left_ty_dtype), Some(res_ndims)))
} else {
let (ndarray_ty, scalar_ty) = if is_left_ndarray { (left, right) } else { (right, left) };
let (ndarray_ty_dtype, _) = unpack_ndarray_var_tys(unifier, ndarray_ty);
if unifier.unioned(ndarray_ty_dtype, scalar_ty) {
Ok(ndarray_ty)
} else {
let (expected_ty, actual_ty) = if is_left_ndarray {
(ndarray_ty_dtype, scalar_ty)
} else {
(scalar_ty, ndarray_ty_dtype)
};
Err(format!(
"Expected right-hand side operand to be {}, got {}",
unifier.stringify(expected_ty),
unifier.stringify(actual_ty),
))
}
}
}
/// Returns the return type given a binary operator and its primitive operands.
pub fn typeof_binop(
unifier: &mut Unifier,
primitives: &PrimitiveStore,
op: Operator,
lhs: Type,
rhs: Type,
) -> Result<Option<Type>, String> {
let op = Binop { base: op, variant: BinopVariant::Normal };
let is_left_list = lhs.obj_id(unifier).is_some_and(|id| id == PrimDef::List.id());
let is_right_list = rhs.obj_id(unifier).is_some_and(|id| id == PrimDef::List.id());
let is_left_ndarray = lhs.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
let is_right_ndarray = rhs.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
Ok(Some(match op.base {
Operator::Add | Operator::Sub | Operator::Mult | Operator::Mod | Operator::FloorDiv => {
if is_left_list || is_right_list {
if ![Operator::Add, Operator::Mult].contains(&op.base) {
return Err(format!(
"Binary operator {} not supported for list",
op.op_info().symbol
));
}
if is_left_list {
lhs
} else {
rhs
}
} else if is_left_ndarray || is_right_ndarray {
typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?
} else if unifier.unioned(lhs, rhs) {
lhs
} else {
return Ok(None);
}
}
Operator::MatMult => {
let (_, lhs_ndims) = unpack_ndarray_var_tys(unifier, lhs);
let lhs_ndims = match &*unifier.get_ty_immutable(lhs_ndims) {
TypeEnum::TLiteral { values, .. } => {
assert_eq!(values.len(), 1);
u64::try_from(values[0].clone()).unwrap()
}
_ => unreachable!(),
};
let (_, rhs_ndims) = unpack_ndarray_var_tys(unifier, rhs);
let rhs_ndims = match &*unifier.get_ty_immutable(rhs_ndims) {
TypeEnum::TLiteral { values, .. } => {
assert_eq!(values.len(), 1);
u64::try_from(values[0].clone()).unwrap()
}
_ => unreachable!(),
};
match (lhs_ndims, rhs_ndims) {
(2, 2) => typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?,
(lhs, rhs) if lhs == 0 || rhs == 0 => {
return Err(format!(
"Input operand {} does not have enough dimensions (has {lhs}, requires {rhs})",
u8::from(rhs == 0)
))
}
(lhs, rhs) => {
return Err(format!(
"ndarray.__matmul__ on {lhs}D and {rhs}D operands not supported"
))
}
}
}
Operator::Div => {
if is_left_ndarray || is_right_ndarray {
typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?
} else if unifier.unioned(lhs, rhs) {
primitives.float
} else {
return Ok(None);
}
}
Operator::Pow => {
if is_left_ndarray || is_right_ndarray {
typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?
} else if [
primitives.int32,
primitives.int64,
primitives.uint32,
primitives.uint64,
primitives.float,
]
.into_iter()
.any(|ty| unifier.unioned(lhs, ty))
{
lhs
} else {
return Ok(None);
}
}
Operator::LShift | Operator::RShift => lhs,
Operator::BitOr | Operator::BitXor | Operator::BitAnd => {
if unifier.unioned(lhs, rhs) {
lhs
} else {
return Ok(None);
}
}
}))
}
pub fn typeof_unaryop(
unifier: &mut Unifier,
primitives: &PrimitiveStore,
op: Unaryop,
operand: Type,
) -> Result<Option<Type>, String> {
let operand_obj_id = operand.obj_id(unifier);
if op == Unaryop::Not
&& operand_obj_id.is_some_and(|id| id == primitives.ndarray.obj_id(unifier).unwrap())
{
return Err(
"The truth value of an array with more than one element is ambiguous".to_string()
);
}
Ok(match op {
Unaryop::Not => match operand_obj_id {
Some(v) if v == PrimDef::NDArray.id() => Some(operand),
Some(_) => Some(primitives.bool),
_ => None,
},
Unaryop::Invert => {
if operand_obj_id.is_some_and(|id| id == PrimDef::Bool.id()) {
Some(primitives.int32)
} else if operand_obj_id.is_some_and(|id| PrimDef::iter().any(|prim| id == prim.id())) {
Some(operand)
} else {
None
}
}
Unaryop::UAdd | Unaryop::USub => {
if operand_obj_id.is_some_and(|id| id == PrimDef::NDArray.id()) {
let (dtype, _) = unpack_ndarray_var_tys(unifier, operand);
if dtype.obj_id(unifier).is_some_and(|id| id == PrimDef::Bool.id()) {
return Err(if op == Unaryop::UAdd {
"The ufunc 'positive' cannot be applied to ndarray[bool, N]".to_string()
} else {
"The numpy boolean negative, the `-` operator, is not supported, use the `~` operator function instead.".to_string()
});
}
Some(operand)
} else if operand_obj_id.is_some_and(|id| id == PrimDef::Bool.id()) {
Some(primitives.int32)
} else if operand_obj_id.is_some_and(|id| PrimDef::iter().any(|prim| id == prim.id())) {
Some(operand)
} else {
None
}
}
})
}
/// Returns the return type given a comparison operator and its primitive operands.
pub fn typeof_cmpop(
unifier: &mut Unifier,
primitives: &PrimitiveStore,
_op: Cmpop,
lhs: Type,
rhs: Type,
) -> Result<Option<Type>, String> {
let is_left_ndarray = lhs.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
let is_right_ndarray = rhs.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
Ok(Some(if is_left_ndarray || is_right_ndarray {
let brd = typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?;
let (_, ndims) = unpack_ndarray_var_tys(unifier, brd);
make_ndarray_ty(unifier, primitives, Some(primitives.bool), Some(ndims))
} else if unifier.unioned(lhs, rhs) {
primitives.bool
} else {
return Ok(None);
}))
}
pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifier) {
@ -293,38 +678,77 @@ pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifie
bool: bool_t,
uint32: uint32_t,
uint64: uint64_t,
list: list_t,
ndarray: ndarray_t,
..
} = *store;
let size_t = store.usize();
/* int ======== */
for t in [int32_t, int64_t, uint32_t, uint64_t] {
impl_basic_arithmetic(unifier, store, t, &[t], t);
impl_pow(unifier, store, t, &[t], t);
let ndarray_int_t = make_ndarray_ty(unifier, store, Some(t), None);
impl_basic_arithmetic(unifier, store, t, &[t, ndarray_int_t], None);
impl_pow(unifier, store, t, &[t, ndarray_int_t], None);
impl_bitwise_arithmetic(unifier, store, t);
impl_bitwise_shift(unifier, store, t);
impl_div(unifier, store, t, &[t]);
impl_floordiv(unifier, store, t, &[t], t);
impl_mod(unifier, store, t, &[t], t);
impl_invert(unifier, store, t);
impl_not(unifier, store, t);
impl_comparison(unifier, store, t, t);
impl_eq(unifier, store, t);
impl_div(unifier, store, t, &[t, ndarray_int_t], None);
impl_floordiv(unifier, store, t, &[t, ndarray_int_t], None);
impl_mod(unifier, store, t, &[t, ndarray_int_t], None);
impl_invert(unifier, store, t, Some(t));
impl_not(unifier, store, t, Some(bool_t));
impl_comparison(unifier, store, t, &[t, ndarray_int_t], None);
impl_eq(unifier, store, t, &[t, ndarray_int_t], None);
}
for t in [int32_t, int64_t] {
impl_sign(unifier, store, t);
impl_sign(unifier, store, t, Some(t));
}
/* float ======== */
impl_basic_arithmetic(unifier, store, float_t, &[float_t], float_t);
impl_pow(unifier, store, float_t, &[int32_t, float_t], float_t);
impl_div(unifier, store, float_t, &[float_t]);
impl_floordiv(unifier, store, float_t, &[float_t], float_t);
impl_mod(unifier, store, float_t, &[float_t], float_t);
impl_sign(unifier, store, float_t);
impl_not(unifier, store, float_t);
impl_comparison(unifier, store, float_t, float_t);
impl_eq(unifier, store, float_t);
let ndarray_float_t = make_ndarray_ty(unifier, store, Some(float_t), None);
let ndarray_int32_t = make_ndarray_ty(unifier, store, Some(int32_t), None);
impl_basic_arithmetic(unifier, store, float_t, &[float_t, ndarray_float_t], None);
impl_pow(unifier, store, float_t, &[int32_t, float_t, ndarray_int32_t, ndarray_float_t], None);
impl_div(unifier, store, float_t, &[float_t, ndarray_float_t], None);
impl_floordiv(unifier, store, float_t, &[float_t, ndarray_float_t], None);
impl_mod(unifier, store, float_t, &[float_t, ndarray_float_t], None);
impl_sign(unifier, store, float_t, Some(float_t));
impl_not(unifier, store, float_t, Some(bool_t));
impl_comparison(unifier, store, float_t, &[float_t, ndarray_float_t], None);
impl_eq(unifier, store, float_t, &[float_t, ndarray_float_t], None);
/* bool ======== */
impl_not(unifier, store, bool_t);
impl_eq(unifier, store, bool_t);
let ndarray_bool_t = make_ndarray_ty(unifier, store, Some(bool_t), None);
impl_invert(unifier, store, bool_t, Some(int32_t));
impl_not(unifier, store, bool_t, Some(bool_t));
impl_sign(unifier, store, bool_t, Some(int32_t));
impl_eq(unifier, store, bool_t, &[bool_t, ndarray_bool_t], None);
/* list ======== */
impl_binop(unifier, store, list_t, &[list_t], Some(list_t), &[Operator::Add]);
impl_binop(unifier, store, list_t, &[int32_t, int64_t], Some(list_t), &[Operator::Mult]);
impl_cmpop(unifier, store, list_t, &[list_t], &[Cmpop::Eq, Cmpop::NotEq], Some(bool_t));
/* ndarray ===== */
let ndarray_usized_ndims_tvar =
unifier.get_fresh_const_generic_var(size_t, Some("ndarray_ndims".into()), None);
let ndarray_unsized_t =
make_ndarray_ty(unifier, store, None, Some(ndarray_usized_ndims_tvar.ty));
let (ndarray_dtype_t, _) = unpack_ndarray_var_tys(unifier, ndarray_t);
let (ndarray_unsized_dtype_t, _) = unpack_ndarray_var_tys(unifier, ndarray_unsized_t);
impl_basic_arithmetic(
unifier,
store,
ndarray_t,
&[ndarray_unsized_t, ndarray_unsized_dtype_t],
None,
);
impl_pow(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
impl_div(unifier, store, ndarray_t, &[ndarray_t, ndarray_dtype_t], None);
impl_floordiv(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
impl_mod(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
impl_matmul(unifier, store, ndarray_t, &[ndarray_t], Some(ndarray_t));
impl_sign(unifier, store, ndarray_t, Some(ndarray_t));
impl_invert(unifier, store, ndarray_t, Some(ndarray_t));
impl_eq(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
impl_comparison(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
}

View File

@ -1,24 +1,46 @@
use std::collections::HashMap;
use std::fmt::Display;
use crate::typecheck::typedef::TypeEnum;
use crate::typecheck::{magic_methods::HasOpInfo, typedef::TypeEnum};
use super::typedef::{RecordKey, Type, Unifier};
use nac3parser::ast::{Location, StrRef};
use super::{
magic_methods::Binop,
typedef::{RecordKey, Type, Unifier},
};
use itertools::Itertools;
use nac3parser::ast::{Cmpop, Location, StrRef};
#[derive(Debug, Clone)]
pub enum TypeErrorKind {
TooManyArguments {
expected: usize,
got: usize,
GotMultipleValues {
name: StrRef,
},
TooManyArguments {
expected_min_count: usize,
expected_max_count: usize,
got_count: usize,
},
MissingArgs {
missing_arg_names: Vec<StrRef>,
},
MissingArgs(String),
UnknownArgName(StrRef),
IncorrectArgType {
name: StrRef,
expected: Type,
got: Type,
},
UnsupportedBinaryOpTypes {
operator: Binop,
lhs_type: Type,
rhs_type: Type,
expected_rhs_type: Type,
},
UnsupportedComparsionOpTypes {
operator: Cmpop,
lhs_type: Type,
rhs_type: Type,
expected_rhs_type: Type,
},
FieldUnificationError {
field: RecordKey,
types: (Type, Type),
@ -34,6 +56,7 @@ pub enum TypeErrorKind {
},
RequiresTypeAnn,
PolymorphicFunctionPointer,
NoSuchAttribute(RecordKey, Type),
}
#[derive(Debug, Clone)]
@ -43,15 +66,18 @@ pub struct TypeError {
}
impl TypeError {
#[must_use]
pub fn new(kind: TypeErrorKind, loc: Option<Location>) -> TypeError {
TypeError { kind, loc }
}
#[must_use]
pub fn at(mut self, loc: Option<Location>) -> TypeError {
self.loc = self.loc.or(loc);
self
}
#[must_use]
pub fn to_display(self, unifier: &Unifier) -> DisplayTypeError {
DisplayTypeError { err: self, unifier }
}
@ -64,8 +90,8 @@ pub struct DisplayTypeError<'a> {
fn loc_to_str(loc: Option<Location>) -> String {
match loc {
Some(loc) => format!("(in {})", loc),
None => "".to_string(),
Some(loc) => format!("(in {loc})"),
None => String::new(),
}
}
@ -74,23 +100,49 @@ impl<'a> Display for DisplayTypeError<'a> {
use TypeErrorKind::*;
let mut notes = Some(HashMap::new());
match &self.err.kind {
TooManyArguments { expected, got } => {
write!(f, "Too many arguments. Expected {} but got {}", expected, got)
GotMultipleValues { name } => {
write!(f, "For multiple values for parameter {name}")
}
MissingArgs(args) => {
write!(f, "Missing arguments: {}", args)
TooManyArguments { expected_min_count, expected_max_count, got_count } => {
debug_assert!(expected_min_count <= expected_max_count);
if expected_min_count == expected_max_count {
let expected_count = expected_min_count; // or expected_max_count
write!(f, "Too many arguments. Expected {expected_count} but got {got_count}")
} else {
write!(f, "Too many arguments. Expected {expected_min_count} to {expected_max_count} arguments but got {got_count}")
}
}
MissingArgs { missing_arg_names } => {
let args = missing_arg_names.iter().join(", ");
write!(f, "Missing arguments: {args}")
}
UnsupportedBinaryOpTypes { operator, lhs_type, rhs_type, expected_rhs_type } => {
let op_symbol = operator.op_info().symbol;
let lhs_type_str = self.unifier.stringify_with_notes(*lhs_type, &mut notes);
let rhs_type_str = self.unifier.stringify_with_notes(*rhs_type, &mut notes);
let expected_rhs_type_str =
self.unifier.stringify_with_notes(*expected_rhs_type, &mut notes);
write!(f, "Unsupported operand type(s) for {op_symbol}: '{lhs_type_str}' and '{rhs_type_str}' (right operand should have type {expected_rhs_type_str})")
}
UnsupportedComparsionOpTypes { operator, lhs_type, rhs_type, expected_rhs_type } => {
let op_symbol = operator.op_info().symbol;
let lhs_type_str = self.unifier.stringify_with_notes(*lhs_type, &mut notes);
let rhs_type_str = self.unifier.stringify_with_notes(*rhs_type, &mut notes);
let expected_rhs_type_str =
self.unifier.stringify_with_notes(*expected_rhs_type, &mut notes);
write!(f, "'{op_symbol}' not supported between instances of '{lhs_type_str}' and '{rhs_type_str}' (right operand should have type {expected_rhs_type_str})")
}
UnknownArgName(name) => {
write!(f, "Unknown argument name: {}", name)
write!(f, "Unknown argument name: {name}")
}
IncorrectArgType { name, expected, got } => {
let expected = self.unifier.stringify_with_notes(*expected, &mut notes);
let got = self.unifier.stringify_with_notes(*got, &mut notes);
write!(
f,
"Incorrect argument type for {}. Expected {}, but got {}",
name, expected, got
)
write!(f, "Incorrect argument type for parameter {name}. Expected {expected}, but got {got}")
}
FieldUnificationError { field, types, loc } => {
let lhs = self.unifier.stringify_with_notes(types.0, &mut notes);
@ -126,7 +178,7 @@ impl<'a> Display for DisplayTypeError<'a> {
);
if let Some(loc) = loc {
result?;
write!(f, " (in {})", loc)?;
write!(f, " (in {loc})")?;
return Ok(());
}
result
@ -136,12 +188,12 @@ impl<'a> Display for DisplayTypeError<'a> {
{
let t1 = self.unifier.stringify_with_notes(*t1, &mut notes);
let t2 = self.unifier.stringify_with_notes(*t2, &mut notes);
write!(f, "Tuple length mismatch: got {} and {}", t1, t2)
write!(f, "Tuple length mismatch: got {t1} and {t2}")
}
_ => {
let t1 = self.unifier.stringify_with_notes(*t1, &mut notes);
let t2 = self.unifier.stringify_with_notes(*t2, &mut notes);
write!(f, "Incompatible types: {} and {}", t1, t2)
write!(f, "Incompatible types: {t1} and {t2}")
}
}
}
@ -150,18 +202,21 @@ impl<'a> Display for DisplayTypeError<'a> {
write!(f, "Cannot assign to an element of a tuple")
} else {
let t = self.unifier.stringify_with_notes(*t, &mut notes);
write!(f, "Cannot assign to field {} of {}, which is immutable", name, t)
write!(f, "Cannot assign to field {name} of {t}, which is immutable")
}
}
NoSuchField(name, t) => {
let t = self.unifier.stringify_with_notes(*t, &mut notes);
write!(f, "`{}::{}` field/method does not exist", t, name)
write!(f, "`{t}::{name}` field/method does not exist")
}
NoSuchAttribute(name, t) => {
let t = self.unifier.stringify_with_notes(*t, &mut notes);
write!(f, "`{t}::{name}` is not a class attribute")
}
TupleIndexOutOfBounds { index, len } => {
write!(
f,
"Tuple index out of bounds. Got {} but tuple has only {} elements",
index, len
"Tuple index out of bounds. Got {index} but tuple has only {len} elements"
)
}
RequiresTypeAnn => {
@ -172,13 +227,13 @@ impl<'a> Display for DisplayTypeError<'a> {
}
}?;
if let Some(loc) = self.err.loc {
write!(f, " at {}", loc)?;
write!(f, " at {loc}")?;
}
let notes = notes.unwrap();
if !notes.is_empty() {
write!(f, "\n\nNotes:")?;
for line in notes.values() {
write!(f, "\n {}", line)?;
write!(f, "\n {line}")?;
}
}
Ok(())

File diff suppressed because it is too large Load Diff

View File

@ -3,12 +3,14 @@ use super::*;
use crate::{
codegen::CodeGenContext,
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, TopLevelDef},
toplevel::{helper::PrimDef, DefinitionId, TopLevelDef},
};
use indexmap::IndexMap;
use indoc::indoc;
use itertools::zip;
use nac3parser::ast::FileName;
use nac3parser::parser::parse_program;
use parking_lot::RwLock;
use std::iter::zip;
use test_case::test_case;
struct Resolver {
@ -20,7 +22,7 @@ struct Resolver {
impl SymbolResolver for Resolver {
fn get_default_param_value(
&self,
_: &nac3parser::ast::Expr,
_: &ast::Expr,
) -> Option<crate::symbol_resolver::SymbolValue> {
unimplemented!()
}
@ -32,19 +34,22 @@ impl SymbolResolver for Resolver {
_: &PrimitiveStore,
str: StrRef,
) -> Result<Type, String> {
self.id_to_type.get(&str).cloned().ok_or_else(|| format!("cannot find symbol `{}`", str))
self.id_to_type.get(&str).copied().ok_or_else(|| format!("cannot find symbol `{str}`"))
}
fn get_symbol_value<'ctx, 'a>(
fn get_symbol_value<'ctx>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, 'a>,
_: &mut CodeGenContext<'ctx, '_>,
) -> Option<ValueEnum<'ctx>> {
unimplemented!()
}
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, String> {
self.id_to_def.get(&id).cloned().ok_or_else(|| "Unknown identifier".to_string())
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, HashSet<String>> {
self.id_to_def
.get(&id)
.copied()
.ok_or_else(|| HashSet::from(["Unknown identifier".to_string()]))
}
fn get_string_id(&self, _: &str) -> i32 {
@ -62,7 +67,7 @@ struct TestEnvironment {
pub primitives: PrimitiveStore,
pub id_to_name: HashMap<usize, StrRef>,
pub identifier_mapping: HashMap<StrRef, Type>,
pub virtual_checks: Vec<(Type, Type, nac3parser::ast::Location)>,
pub virtual_checks: Vec<(Type, Type, Location)>,
pub calls: HashMap<CodeLocation, CallId>,
pub top_level: TopLevelContext,
}
@ -72,67 +77,81 @@ impl TestEnvironment {
let mut unifier = Unifier::new();
let int32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
obj_id: PrimDef::Int32.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
with_fields(&mut unifier, int32, |unifier, fields| {
let add_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![FuncArg { name: "other".into(), ty: int32, default_value: None }],
ret: int32,
vars: HashMap::new(),
vars: VarMap::new(),
}));
fields.insert("__add__".into(), (add_ty, false));
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
obj_id: PrimDef::Int64.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
obj_id: PrimDef::Float.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(3),
obj_id: PrimDef::Bool.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(4),
obj_id: PrimDef::None.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(5),
obj_id: PrimDef::Range.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(6),
obj_id: PrimDef::Str.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(7),
obj_id: PrimDef::Exception.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(8),
obj_id: PrimDef::UInt32.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(9),
obj_id: PrimDef::UInt64.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(10),
obj_id: PrimDef::Option.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let list_elem_tvar = unifier.get_fresh_var(Some("list_elem".into()), None);
let list = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::List.id(),
fields: HashMap::new(),
params: into_var_map([list_elem_tvar]),
});
let ndarray_dtype_tvar = unifier.get_fresh_var(Some("ndarray_dtype".into()), None);
let ndarray_ndims_tvar =
unifier.get_fresh_const_generic_var(uint64, Some("ndarray_ndims".into()), None);
let ndarray = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::NDArray.id(),
fields: HashMap::new(),
params: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
});
let primitives = PrimitiveStore {
int32,
@ -146,10 +165,14 @@ impl TestEnvironment {
uint32,
uint64,
option,
list,
ndarray,
size_t: 64,
};
unifier.put_primitive_store(&primitives);
set_primitives_magic_methods(&primitives, &mut unifier);
let id_to_name = [
let id_to_name: HashMap<_, _> = [
(0, "int32".into()),
(1, "int64".into()),
(2, "float".into()),
@ -159,23 +182,21 @@ impl TestEnvironment {
(6, "str".into()),
(7, "exception".into()),
]
.iter()
.cloned()
.collect();
.into();
let mut identifier_mapping = HashMap::new();
identifier_mapping.insert("None".into(), none);
let resolver = Arc::new(Resolver {
id_to_type: identifier_mapping.clone(),
id_to_def: Default::default(),
class_names: Default::default(),
id_to_def: HashMap::default(),
class_names: HashMap::default(),
}) as Arc<dyn SymbolResolver + Send + Sync>;
TestEnvironment {
top_level: TopLevelContext {
definitions: Default::default(),
unifiers: Default::default(),
definitions: Arc::default(),
unifiers: Arc::default(),
personality_symbol: None,
},
unifier,
@ -197,70 +218,95 @@ impl TestEnvironment {
let mut identifier_mapping = HashMap::new();
let mut top_level_defs: Vec<Arc<RwLock<TopLevelDef>>> = Vec::new();
let int32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
obj_id: PrimDef::Int32.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
with_fields(&mut unifier, int32, |unifier, fields| {
let add_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![FuncArg { name: "other".into(), ty: int32, default_value: None }],
ret: int32,
vars: HashMap::new(),
vars: VarMap::new(),
}));
fields.insert("__add__".into(), (add_ty, false));
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
obj_id: PrimDef::Int64.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
obj_id: PrimDef::Float.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(3),
obj_id: PrimDef::Bool.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(4),
obj_id: PrimDef::None.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(5),
obj_id: PrimDef::Range.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(6),
obj_id: PrimDef::Str.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(7),
obj_id: PrimDef::Exception.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(8),
obj_id: PrimDef::UInt32.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(9),
obj_id: PrimDef::UInt64.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(10),
obj_id: PrimDef::Option.id(),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
});
let list_elem_tvar = unifier.get_fresh_var(Some("list_elem".into()), None);
let list = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::List.id(),
fields: HashMap::new(),
params: into_var_map([list_elem_tvar]),
});
let ndarray = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::NDArray.id(),
fields: HashMap::new(),
params: VarMap::new(),
});
identifier_mapping.insert("None".into(), none);
for (i, name) in ["int32", "int64", "float", "bool", "none", "range", "str", "Exception"]
for (i, name) in [
"int32",
"int64",
"float",
"bool",
"none",
"range",
"str",
"Exception",
"uint32",
"uint64",
"Option",
"list",
"ndarray",
]
.iter()
.enumerate()
{
@ -268,10 +314,11 @@ impl TestEnvironment {
RwLock::new(TopLevelDef::Class {
name: (*name).into(),
object_id: DefinitionId(i),
type_vars: Default::default(),
fields: Default::default(),
methods: Default::default(),
ancestors: Default::default(),
type_vars: Vec::default(),
fields: Vec::default(),
attributes: Vec::default(),
methods: Vec::default(),
ancestors: Vec::default(),
resolver: None,
constructor: None,
loc: None,
@ -279,7 +326,7 @@ impl TestEnvironment {
.into(),
);
}
let defs = 7;
let defs = 12;
let primitives = PrimitiveStore {
int32,
@ -293,23 +340,29 @@ impl TestEnvironment {
uint32,
uint64,
option,
list,
ndarray,
size_t: 64,
};
let (v0, id) = unifier.get_dummy_var();
unifier.put_primitive_store(&primitives);
let tvar = unifier.get_dummy_var();
let foo_ty = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(defs + 1),
fields: [("a".into(), (v0, true))].iter().cloned().collect::<HashMap<_, _>>(),
params: [(id, v0)].iter().cloned().collect::<HashMap<_, _>>(),
fields: [("a".into(), (tvar.ty, true))].into(),
params: into_var_map([tvar]),
});
top_level_defs.push(
RwLock::new(TopLevelDef::Class {
name: "Foo".into(),
object_id: DefinitionId(defs + 1),
type_vars: vec![v0],
fields: [("a".into(), v0, true)].into(),
methods: Default::default(),
ancestors: Default::default(),
type_vars: vec![tvar.ty],
fields: [("a".into(), tvar.ty, true)].into(),
attributes: Vec::default(),
methods: Vec::default(),
ancestors: Vec::default(),
resolver: None,
constructor: None,
loc: None,
@ -322,31 +375,29 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: foo_ty,
vars: [(id, v0)].iter().cloned().collect(),
vars: into_var_map([tvar]),
})),
);
let fun = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: int32,
vars: Default::default(),
vars: IndexMap::default(),
}));
let bar = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(defs + 2),
fields: [("a".into(), (int32, true)), ("b".into(), (fun, true))]
.iter()
.cloned()
.collect::<HashMap<_, _>>(),
params: Default::default(),
fields: [("a".into(), (int32, true)), ("b".into(), (fun, true))].into(),
params: IndexMap::default(),
});
top_level_defs.push(
RwLock::new(TopLevelDef::Class {
name: "Bar".into(),
object_id: DefinitionId(defs + 2),
type_vars: Default::default(),
type_vars: Vec::default(),
fields: [("a".into(), int32, true), ("b".into(), fun, true)].into(),
methods: Default::default(),
ancestors: Default::default(),
attributes: Vec::default(),
methods: Vec::default(),
ancestors: Vec::default(),
resolver: None,
constructor: None,
loc: None,
@ -358,26 +409,24 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: bar,
vars: Default::default(),
vars: IndexMap::default(),
})),
);
let bar2 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(defs + 3),
fields: [("a".into(), (bool, true)), ("b".into(), (fun, false))]
.iter()
.cloned()
.collect::<HashMap<_, _>>(),
params: Default::default(),
fields: [("a".into(), (bool, true)), ("b".into(), (fun, false))].into(),
params: IndexMap::default(),
});
top_level_defs.push(
RwLock::new(TopLevelDef::Class {
name: "Bar2".into(),
object_id: DefinitionId(defs + 3),
type_vars: Default::default(),
type_vars: Vec::default(),
fields: [("a".into(), bool, true), ("b".into(), fun, false)].into(),
methods: Default::default(),
ancestors: Default::default(),
attributes: Vec::default(),
methods: Vec::default(),
ancestors: Vec::default(),
resolver: None,
constructor: None,
loc: None,
@ -389,10 +438,10 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: bar2,
vars: Default::default(),
vars: IndexMap::default(),
})),
);
let class_names = [("Bar".into(), bar), ("Bar2".into(), bar2)].iter().cloned().collect();
let class_names: HashMap<_, _> = [("Bar".into(), bar), ("Bar2".into(), bar2)].into();
let id_to_name = [
"int32".into(),
@ -403,18 +452,22 @@ impl TestEnvironment {
"range".into(),
"str".into(),
"exception".into(),
"uint32".into(),
"uint64".into(),
"option".into(),
"list".into(),
"ndarray".into(),
"Foo".into(),
"Bar".into(),
"Bar2".into(),
]
.iter()
.into_iter()
.enumerate()
.map(|(a, b)| (a, *b))
.collect();
let top_level = TopLevelContext {
definitions: Arc::new(top_level_defs.into()),
unifiers: Default::default(),
unifiers: Arc::default(),
personality_symbol: None,
};
@ -425,9 +478,7 @@ impl TestEnvironment {
("Bar".into(), DefinitionId(defs + 2)),
("Bar2".into(), DefinitionId(defs + 3)),
]
.iter()
.cloned()
.collect(),
.into(),
class_names,
}) as Arc<dyn SymbolResolver + Send + Sync>;
@ -452,11 +503,11 @@ impl TestEnvironment {
top_level: &self.top_level,
function_data: &mut self.function_data,
unifier: &mut self.unifier,
variable_mapping: Default::default(),
variable_mapping: HashMap::default(),
primitives: &mut self.primitives,
virtual_checks: &mut self.virtual_checks,
calls: &mut self.calls,
defined_identifiers: Default::default(),
defined_identifiers: HashSet::default(),
in_handler: false,
}
}
@ -468,7 +519,7 @@ impl TestEnvironment {
c = 1.234
d = True
"},
[("a", "int32"), ("b", "int64"), ("c", "float"), ("d", "bool")].iter().cloned().collect(),
&[("a", "int32"), ("b", "int64"), ("c", "float"), ("d", "bool")].into(),
&[]
; "primitives test")]
#[test_case(indoc! {"
@ -477,7 +528,7 @@ impl TestEnvironment {
c = 1.234
d = b(c)
"},
[("a", "fn[[x:float, y:float], float]"), ("b", "fn[[x:float], float]"), ("c", "float"), ("d", "float")].iter().cloned().collect(),
&[("a", "fn[[x:float, y:float], float]"), ("b", "fn[[x:float], float]"), ("c", "float"), ("d", "float")].into(),
&[]
; "lambda test")]
#[test_case(indoc! {"
@ -486,7 +537,7 @@ impl TestEnvironment {
a = b
c = b(1)
"},
[("a", "fn[[x:int32], int32]"), ("b", "fn[[x:int32], int32]"), ("c", "int32")].iter().cloned().collect(),
&[("a", "fn[[x:int32], int32]"), ("b", "fn[[x:int32], int32]"), ("c", "int32")].into(),
&[]
; "lambda test 2")]
#[test_case(indoc! {"
@ -502,15 +553,15 @@ impl TestEnvironment {
b(123)
"},
[("a", "fn[[x:bool], bool]"), ("b", "fn[[x:int32], int32]"), ("c", "bool"),
("d", "int32"), ("foo1", "Foo[bool]"), ("foo2", "Foo[int32]")].iter().cloned().collect(),
&[("a", "fn[[x:bool], bool]"), ("b", "fn[[x:int32], int32]"), ("c", "bool"),
("d", "int32"), ("foo1", "Foo[bool]"), ("foo2", "Foo[int32]")].into(),
&[]
; "obj test")]
#[test_case(indoc! {"
a = [1, 2, 3]
b = [x + x for x in a]
"},
[("a", "list[int32]"), ("b", "list[int32]")].iter().cloned().collect(),
&[("a", "list[int32]"), ("b", "list[int32]")].into(),
&[]
; "listcomp test")]
#[test_case(indoc! {"
@ -518,25 +569,25 @@ impl TestEnvironment {
b = a.b()
a = virtual(Bar2())
"},
[("a", "virtual[Bar]"), ("b", "int32")].iter().cloned().collect(),
&[("a", "virtual[Bar]"), ("b", "int32")].into(),
&[("Bar", "Bar"), ("Bar2", "Bar")]
; "virtual test")]
#[test_case(indoc! {"
a = [virtual(Bar(), Bar), virtual(Bar2())]
b = [x.b() for x in a]
"},
[("a", "list[virtual[Bar]]"), ("b", "list[int32]")].iter().cloned().collect(),
&[("a", "list[virtual[Bar]]"), ("b", "list[int32]")].into(),
&[("Bar", "Bar"), ("Bar2", "Bar")]
; "virtual list test")]
fn test_basic(source: &str, mapping: HashMap<&str, &str>, virtuals: &[(&str, &str)]) {
println!("source:\n{}", source);
fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &str)]) {
println!("source:\n{source}");
let mut env = TestEnvironment::new();
let id_to_name = std::mem::take(&mut env.id_to_name);
let mut defined_identifiers: HashSet<_> = env.identifier_mapping.keys().cloned().collect();
let mut defined_identifiers: HashSet<_> = env.identifier_mapping.keys().copied().collect();
defined_identifiers.insert("virtual".into());
let mut inferencer = env.get_inferencer();
inferencer.defined_identifiers = defined_identifiers.clone();
let statements = parse_program(source, Default::default()).unwrap();
inferencer.defined_identifiers.clone_from(&defined_identifiers);
let statements = parse_program(source, FileName::default()).unwrap();
let statements = statements
.into_iter()
.map(|v| inferencer.fold_stmt(v))
@ -545,37 +596,37 @@ fn test_basic(source: &str, mapping: HashMap<&str, &str>, virtuals: &[(&str, &st
inferencer.check_block(&statements, &mut defined_identifiers).unwrap();
for (k, v) in inferencer.variable_mapping.iter() {
for (k, v) in &inferencer.variable_mapping {
let name = inferencer.unifier.internal_stringify(
*v,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{}", v),
&mut |v| format!("v{v}"),
&mut None,
);
println!("{}: {}", k, name);
println!("{k}: {name}");
}
for (k, v) in mapping.iter() {
for (k, v) in mapping {
let ty = inferencer.variable_mapping.get(&(*k).into()).unwrap();
let name = inferencer.unifier.internal_stringify(
*ty,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{}", v),
&mut |v| format!("v{v}"),
&mut None,
);
assert_eq!(format!("{}: {}", k, v), format!("{}: {}", k, name));
assert_eq!(format!("{k}: {v}"), format!("{k}: {name}"));
}
assert_eq!(inferencer.virtual_checks.len(), virtuals.len());
for ((a, b, _), (x, y)) in zip(inferencer.virtual_checks.iter(), virtuals) {
let a = inferencer.unifier.internal_stringify(
*a,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{}", v),
&mut |v| format!("v{v}"),
&mut None,
);
let b = inferencer.unifier.internal_stringify(
*b,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{}", v),
&mut |v| format!("v{v}"),
&mut None,
);
@ -594,14 +645,14 @@ fn test_basic(source: &str, mapping: HashMap<&str, &str>, virtuals: &[(&str, &st
g = a // b
h = a % b
"},
[("a", "int32"),
&[("a", "int32"),
("b", "int32"),
("c", "int32"),
("d", "int32"),
("e", "int32"),
("f", "float"),
("g", "int32"),
("h", "int32")].iter().cloned().collect()
("h", "int32")].into()
; "int32")]
#[test_case(
indoc! {"
@ -617,7 +668,7 @@ fn test_basic(source: &str, mapping: HashMap<&str, &str>, virtuals: &[(&str, &st
ii = 3
j = a ** b
"},
[("a", "float"),
&[("a", "float"),
("b", "float"),
("c", "float"),
("d", "float"),
@ -627,7 +678,7 @@ fn test_basic(source: &str, mapping: HashMap<&str, &str>, virtuals: &[(&str, &st
("h", "float"),
("i", "float"),
("ii", "int32"),
("j", "float")].iter().cloned().collect()
("j", "float")].into()
; "float"
)]
#[test_case(
@ -645,7 +696,7 @@ fn test_basic(source: &str, mapping: HashMap<&str, &str>, virtuals: &[(&str, &st
k = a < b
l = a != b
"},
[("a", "int64"),
&[("a", "int64"),
("b", "int64"),
("c", "int64"),
("d", "int64"),
@ -656,7 +707,7 @@ fn test_basic(source: &str, mapping: HashMap<&str, &str>, virtuals: &[(&str, &st
("i", "bool"),
("j", "bool"),
("k", "bool"),
("l", "bool")].iter().cloned().collect()
("l", "bool")].into()
; "int64"
)]
#[test_case(
@ -667,22 +718,22 @@ fn test_basic(source: &str, mapping: HashMap<&str, &str>, virtuals: &[(&str, &st
d = not a
e = a != b
"},
[("a", "bool"),
&[("a", "bool"),
("b", "bool"),
("c", "bool"),
("d", "bool"),
("e", "bool")].iter().cloned().collect()
("e", "bool")].into()
; "boolean"
)]
fn test_primitive_magic_methods(source: &str, mapping: HashMap<&str, &str>) {
println!("source:\n{}", source);
fn test_primitive_magic_methods(source: &str, mapping: &HashMap<&str, &str>) {
println!("source:\n{source}");
let mut env = TestEnvironment::basic_test_env();
let id_to_name = std::mem::take(&mut env.id_to_name);
let mut defined_identifiers: HashSet<_> = env.identifier_mapping.keys().cloned().collect();
let mut defined_identifiers: HashSet<_> = env.identifier_mapping.keys().copied().collect();
defined_identifiers.insert("virtual".into());
let mut inferencer = env.get_inferencer();
inferencer.defined_identifiers = defined_identifiers.clone();
let statements = parse_program(source, Default::default()).unwrap();
inferencer.defined_identifiers.clone_from(&defined_identifiers);
let statements = parse_program(source, FileName::default()).unwrap();
let statements = statements
.into_iter()
.map(|v| inferencer.fold_stmt(v))
@ -691,23 +742,23 @@ fn test_primitive_magic_methods(source: &str, mapping: HashMap<&str, &str>) {
inferencer.check_block(&statements, &mut defined_identifiers).unwrap();
for (k, v) in inferencer.variable_mapping.iter() {
for (k, v) in &inferencer.variable_mapping {
let name = inferencer.unifier.internal_stringify(
*v,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{}", v),
&mut |v| format!("v{v}"),
&mut None,
);
println!("{}: {}", k, name);
println!("{k}: {name}");
}
for (k, v) in mapping.iter() {
for (k, v) in mapping {
let ty = inferencer.variable_mapping.get(&(*k).into()).unwrap();
let name = inferencer.unifier.internal_stringify(
*ty,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{}", v),
&mut |v| format!("v{v}"),
&mut None,
);
assert_eq!(format!("{}: {}", k, v), format!("{}: {}", k, name));
assert_eq!(format!("{k}: {v}"), format!("{k}: {name}"));
}
}

File diff suppressed because it is too large Load Diff

View File

@ -32,28 +32,25 @@ impl Unifier {
ty1.len() == ty2.len()
&& ty1.iter().zip(ty2.iter()).all(|(t1, t2)| self.eq(*t1, *t2))
}
(TypeEnum::TList { ty: ty1 }, TypeEnum::TList { ty: ty2 })
| (TypeEnum::TVirtual { ty: ty1 }, TypeEnum::TVirtual { ty: ty2 }) => {
self.eq(*ty1, *ty2)
}
(TypeEnum::TVirtual { ty: ty1 }, TypeEnum::TVirtual { ty: ty2 }) => self.eq(*ty1, *ty2),
(
TypeEnum::TObj { obj_id: id1, params: params1, .. },
TypeEnum::TObj { obj_id: id2, params: params2, .. },
) => id1 == id2 && self.map_eq(params1, params2),
// TCall and TFunc are not yet implemented
// TLiteral, TCall and TFunc are not yet implemented
_ => false,
}
}
fn map_eq<K>(&mut self, map1: &Mapping<K>, map2: &Mapping<K>) -> bool
fn map_eq<K>(&mut self, map1: &IndexMapping<K>, map2: &IndexMapping<K>) -> bool
where
K: std::hash::Hash + std::cmp::Eq + std::clone::Clone,
K: std::hash::Hash + Eq + Clone,
{
if map1.len() != map2.len() {
return false;
}
for (k, v) in map1.iter() {
if !map2.get(k).map(|v1| self.eq(*v, *v1)).unwrap_or(false) {
for (k, v) in map1 {
if !map2.get(k).is_some_and(|v1| self.eq(*v, *v1)) {
return false;
}
}
@ -62,13 +59,13 @@ impl Unifier {
fn map_eq2<K>(&mut self, map1: &Mapping<K, RecordField>, map2: &Mapping<K, RecordField>) -> bool
where
K: std::hash::Hash + std::cmp::Eq + std::clone::Clone,
K: std::hash::Hash + Eq + Clone,
{
if map1.len() != map2.len() {
return false;
}
for (k, v) in map1.iter() {
if !map2.get(k).map(|v1| self.eq(v.ty, v1.ty)).unwrap_or(false) {
for (k, v) in map1 {
if !map2.get(k).is_some_and(|v1| self.eq(v.ty, v1.ty)) {
return false;
}
}
@ -91,7 +88,7 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
}),
);
type_mapping.insert(
@ -99,7 +96,7 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
}),
);
type_mapping.insert(
@ -107,16 +104,25 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: HashMap::new(),
params: VarMap::new(),
}),
);
let (v0, id) = unifier.get_dummy_var();
let tvar = unifier.get_dummy_var();
type_mapping.insert(
"Foo".into(),
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(3),
fields: [("a".into(), (v0, true))].iter().cloned().collect::<HashMap<_, _>>(),
params: [(id, v0)].iter().cloned().collect::<HashMap<_, _>>(),
fields: [("a".into(), (tvar.ty, true))].into(),
params: into_var_map([tvar]),
}),
);
let tvar = unifier.get_dummy_var();
type_mapping.insert(
"list".into(),
unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::List.id(),
fields: HashMap::new(),
params: into_var_map([tvar]),
}),
);
@ -129,17 +135,43 @@ impl TestEnvironment {
result.0
}
fn internal_parse<'a, 'b>(
&'a mut self,
typ: &'b str,
mapping: &Mapping<String>,
) -> (Type, &'b str) {
fn internal_parse<'b>(&mut self, typ: &'b str, mapping: &Mapping<String>) -> (Type, &'b str) {
// for testing only, so we can just panic when the input is malformed
let end = typ.find(|c| ['[', ',', ']', '='].contains(&c)).unwrap_or_else(|| typ.len());
let end = typ.find(|c| ['[', ',', ']', '='].contains(&c)).unwrap_or(typ.len());
match &typ[..end] {
"list" => {
let mut s = &typ[end..];
assert_eq!(&s[0..1], "[");
let mut ty = Vec::new();
while &s[0..1] != "]" {
let result = self.internal_parse(&s[1..], mapping);
ty.push(result.0);
s = result.1;
}
assert_eq!(ty.len(), 1);
let list_elem_tvar = if let TypeEnum::TObj { params, .. } =
&*self.unifier.get_ty_immutable(self.type_mapping["list"])
{
iter_type_vars(params).next().unwrap()
} else {
unreachable!()
};
(
self.unifier
.subst(
self.type_mapping["list"],
&into_var_map([TypeVar { id: list_elem_tvar.id, ty: ty[0] }]),
)
.unwrap(),
&s[1..],
)
}
"tuple" => {
let mut s = &typ[end..];
assert!(&s[0..1] == "[");
assert_eq!(&s[0..1], "[");
let mut ty = Vec::new();
while &s[0..1] != "]" {
let result = self.internal_parse(&s[1..], mapping);
@ -148,15 +180,9 @@ impl TestEnvironment {
}
(self.unifier.add_ty(TypeEnum::TTuple { ty }), &s[1..])
}
"list" => {
assert!(&typ[end..end + 1] == "[");
let (ty, s) = self.internal_parse(&typ[end + 1..], mapping);
assert!(&s[0..1] == "]");
(self.unifier.add_ty(TypeEnum::TList { ty }), &s[1..])
}
"Record" => {
let mut s = &typ[end..];
assert!(&s[0..1] == "[");
assert_eq!(&s[0..1], "[");
let mut fields = HashMap::new();
while &s[0..1] != "]" {
let eq = s.find('=').unwrap();
@ -169,14 +195,14 @@ impl TestEnvironment {
}
x => {
let mut s = &typ[end..];
let ty = mapping.get(x).cloned().unwrap_or_else(|| {
let ty = mapping.get(x).copied().unwrap_or_else(|| {
// mapping should be type variables, type_mapping should be concrete types
// we should not resolve the type of type variables.
let mut ty = *self.type_mapping.get(x).unwrap();
let te = self.unifier.get_ty(ty);
if let TypeEnum::TObj { params, .. } = &*te.as_ref() {
if let TypeEnum::TObj { params, .. } = &*te {
if !params.is_empty() {
assert!(&s[0..1] == "[");
assert_eq!(&s[0..1], "[");
let mut p = Vec::new();
while &s[0..1] != "]" {
let result = self.internal_parse(&s[1..], mapping);
@ -186,7 +212,7 @@ impl TestEnvironment {
s = &s[1..];
ty = self
.unifier
.subst(ty, &params.keys().cloned().zip(p.into_iter()).collect())
.subst(ty, &params.keys().copied().zip(p).collect())
.unwrap_or(ty);
}
}
@ -250,12 +276,12 @@ fn test_unify(
let mut mapping = HashMap::new();
for i in 1..=variable_count {
let v = env.unifier.get_dummy_var();
mapping.insert(format!("v{}", i), v.0);
mapping.insert(format!("v{i}"), v.ty);
}
// unification may have side effect when we do type resolution, so freeze the types
// before doing unification.
let mut pairs = Vec::new();
for (a, b) in perm.iter() {
for (a, b) in &perm {
let t1 = env.parse(a, &mapping);
let t2 = env.parse(b, &mapping);
pairs.push((t1, t2));
@ -263,8 +289,8 @@ fn test_unify(
for (t1, t2) in pairs {
env.unifier.unify(t1, t2).unwrap();
}
for (a, b) in verify_pairs.iter() {
println!("{} = {}", a, b);
for (a, b) in verify_pairs {
println!("{a} = {b}");
let t1 = env.parse(a, &mapping);
let t2 = env.parse(b, &mapping);
println!("a = {}, b = {}", env.unifier.stringify(t1), env.unifier.stringify(t2));
@ -278,7 +304,7 @@ fn test_unify(
("v1", "tuple[int]"),
("v2", "list[int]"),
],
(("v1", "v2"), "Incompatible types: list[0] and tuple[0]")
(("v1", "v2"), "Incompatible types: 11[0] and tuple[0]")
; "type mismatch"
)]
#[test_case(2,
@ -302,7 +328,7 @@ fn test_unify(
("v1", "Record[a=float,b=int]"),
("v2", "Foo[v3]"),
],
(("v1", "v2"), "`3[typevar4]::b` field/method does not exist")
(("v1", "v2"), "`3[typevar5]::b` field/method does not exist")
; "record obj merge"
)]
/// Test cases for invalid unifications.
@ -315,12 +341,12 @@ fn test_invalid_unification(
let mut mapping = HashMap::new();
for i in 1..=variable_count {
let v = env.unifier.get_dummy_var();
mapping.insert(format!("v{}", i), v.0);
mapping.insert(format!("v{i}"), v.ty);
}
// unification may have side effect when we do type resolution, so freeze the types
// before doing unification.
let mut pairs = Vec::new();
for (a, b) in unify_pairs.iter() {
for (a, b) in unify_pairs {
let t1 = env.parse(a, &mapping);
let t2 = env.parse(b, &mapping);
pairs.push((t1, t2));
@ -339,23 +365,17 @@ fn test_recursive_subst() {
let int = *env.type_mapping.get("int").unwrap();
let foo_id = *env.type_mapping.get("Foo").unwrap();
let foo_ty = env.unifier.get_ty(foo_id);
let mapping: HashMap<_, _>;
with_fields(&mut env.unifier, foo_id, |_unifier, fields| {
fields.insert("rec".into(), (foo_id, true));
});
if let TypeEnum::TObj { params, .. } = &*foo_ty {
mapping = params.iter().map(|(id, _)| (*id, int)).collect();
} else {
unreachable!()
}
let TypeEnum::TObj { params, .. } = &*foo_ty else { unreachable!() };
let mapping = params.iter().map(|(id, _)| (*id, int)).collect();
let instantiated = env.unifier.subst(foo_id, &mapping).unwrap();
let instantiated_ty = env.unifier.get_ty(instantiated);
if let TypeEnum::TObj { fields, .. } = &*instantiated_ty {
let TypeEnum::TObj { fields, .. } = &*instantiated_ty else { unreachable!() };
assert!(env.unifier.unioned(fields.get(&"a".into()).unwrap().0, int));
assert!(env.unifier.unioned(fields.get(&"rec".into()).unwrap().0, instantiated));
} else {
unreachable!()
}
}
#[test]
@ -365,36 +385,27 @@ fn test_virtual() {
let fun = env.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: int,
vars: HashMap::new(),
vars: VarMap::new(),
}));
let bar = env.unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(5),
fields: [("f".into(), (fun, false)), ("a".into(), (int, false))]
.iter()
.cloned()
.collect::<HashMap<StrRef, _>>(),
params: HashMap::new(),
fields: [("f".into(), (fun, false)), ("a".into(), (int, false))].into(),
params: VarMap::new(),
});
let v0 = env.unifier.get_dummy_var().0;
let v1 = env.unifier.get_dummy_var().0;
let v0 = env.unifier.get_dummy_var().ty;
let v1 = env.unifier.get_dummy_var().ty;
let a = env.unifier.add_ty(TypeEnum::TVirtual { ty: bar });
let b = env.unifier.add_ty(TypeEnum::TVirtual { ty: v0 });
let c = env
.unifier
.add_record([("f".into(), RecordField::new(v1, false, None))].iter().cloned().collect());
let c = env.unifier.add_record([("f".into(), RecordField::new(v1, false, None))].into());
env.unifier.unify(a, b).unwrap();
env.unifier.unify(b, c).unwrap();
assert!(env.unifier.eq(v1, fun));
let d = env
.unifier
.add_record([("a".into(), RecordField::new(v1, true, None))].iter().cloned().collect());
let d = env.unifier.add_record([("a".into(), RecordField::new(v1, true, None))].into());
assert_eq!(env.unify(b, d), Err("`virtual[5]::a` field/method does not exist".to_string()));
let d = env
.unifier
.add_record([("b".into(), RecordField::new(v1, true, None))].iter().cloned().collect());
let d = env.unifier.add_record([("b".into(), RecordField::new(v1, true, None))].into());
assert_eq!(env.unify(b, d), Err("`virtual[5]::b` field/method does not exist".to_string()));
}
@ -407,86 +418,132 @@ fn test_typevar_range() {
let int_list = env.parse("list[int]", &HashMap::new());
let float_list = env.parse("list[float]", &HashMap::new());
let list_elem_tvar = if let TypeEnum::TObj { params, .. } =
&*env.unifier.get_ty_immutable(env.type_mapping["list"])
{
iter_type_vars(params).next().unwrap()
} else {
unreachable!()
};
// unification between v and int
// where v in (int, bool)
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
env.unifier.unify(int, v).unwrap();
// unification between v and list[int]
// where v in (int, bool)
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
assert_eq!(
env.unify(int_list, v),
Err("Expected any one of these types: 0, 2, but got list[0]".to_string())
Err("Expected any one of these types: 0, 2, but got 11[0]".to_string())
);
// unification between v and float
// where v in (int, bool)
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
assert_eq!(
env.unify(float, v),
Err("Expected any one of these types: 0, 2, but got 1".to_string())
);
let v1 = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
let v1_list = env.unifier.add_ty(TypeEnum::TList { ty: v1 });
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).0;
let v1 = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
let v1_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: v1 }]),
});
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).ty;
// unification between v and int
// where v in (int, list[v1]), v1 in (int, bool)
env.unifier.unify(int, v).unwrap();
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).0;
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).ty;
// unification between v and list[int]
// where v in (int, list[v1]), v1 in (int, bool)
env.unifier.unify(int_list, v).unwrap();
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).0;
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).ty;
// unification between v and list[float]
// where v in (int, list[v1]), v1 in (int, bool)
println!("float_list: {}, v: {}", env.unifier.stringify(float_list), env.unifier.stringify(v));
assert_eq!(
env.unify(float_list, v),
Err("Expected any one of these types: 0, list[typevar5], but got list[1]\n\nNotes:\n typevar5 ∈ {0, 2}".to_string())
Err("Expected any one of these types: 0, 11[typevar6], but got 11[1]\n\nNotes:\n typevar6 ∈ {0, 2}".to_string())
);
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).0;
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).ty;
env.unifier.unify(a, b).unwrap();
env.unifier.unify(a, float).unwrap();
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).0;
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).ty;
env.unifier.unify(a, b).unwrap();
assert_eq!(env.unify(a, int), Err("Expected any one of these types: 1, but got 0".into()));
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).0;
let a_list = env.unifier.add_ty(TypeEnum::TList { ty: a });
let a_list = env.unifier.get_fresh_var_with_range(&[a_list], None, None).0;
let b_list = env.unifier.add_ty(TypeEnum::TList { ty: b });
let b_list = env.unifier.get_fresh_var_with_range(&[b_list], None, None).0;
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).ty;
let a_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: a }]),
});
let a_list = env.unifier.get_fresh_var_with_range(&[a_list], None, None).ty;
let b_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: b }]),
});
let b_list = env.unifier.get_fresh_var_with_range(&[b_list], None, None).ty;
env.unifier.unify(a_list, b_list).unwrap();
let float_list = env.unifier.add_ty(TypeEnum::TList { ty: float });
let float_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: float }]),
});
env.unifier.unify(a_list, float_list).unwrap();
// previous unifications should not affect a and b
env.unifier.unify(a, int).unwrap();
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).0;
let a_list = env.unifier.add_ty(TypeEnum::TList { ty: a });
let b_list = env.unifier.add_ty(TypeEnum::TList { ty: b });
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).ty;
let a_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: a }]),
});
let b_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: b }]),
});
env.unifier.unify(a_list, b_list).unwrap();
let int_list = env.unifier.add_ty(TypeEnum::TList { ty: int });
let int_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: int }]),
});
assert_eq!(
env.unify(a_list, int_list),
Err("Incompatible types: list[typevar22] and list[0]\
\n\nNotes:\n typevar22 {1}".into())
Err("Incompatible types: 11[typevar23] and 11[0]\
\n\nNotes:\n typevar23 {1}"
.into())
);
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_dummy_var().0;
let a_list = env.unifier.add_ty(TypeEnum::TList { ty: a });
let a_list = env.unifier.get_fresh_var_with_range(&[a_list], None, None).0;
let b_list = env.unifier.add_ty(TypeEnum::TList { ty: b });
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
let b = env.unifier.get_dummy_var().ty;
let a_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: a }]),
});
let a_list = env.unifier.get_fresh_var_with_range(&[a_list], None, None).ty;
let b_list = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: b }]),
});
env.unifier.unify(a_list, b_list).unwrap();
assert_eq!(
env.unify(b, boolean),
@ -497,17 +554,29 @@ fn test_typevar_range() {
#[test]
fn test_rigid_var() {
let mut env = TestEnvironment::new();
let a = env.unifier.get_fresh_rigid_var(None, None).0;
let b = env.unifier.get_fresh_rigid_var(None, None).0;
let x = env.unifier.get_dummy_var().0;
let list_a = env.unifier.add_ty(TypeEnum::TList { ty: a });
let list_x = env.unifier.add_ty(TypeEnum::TList { ty: x });
let a = env.unifier.get_fresh_rigid_var(None, None).ty;
let b = env.unifier.get_fresh_rigid_var(None, None).ty;
let x = env.unifier.get_dummy_var().ty;
let list_elem_tvar = env.unifier.get_fresh_var(Some("list_elem".into()), None);
let list_a = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: a }]),
});
let list_x = env.unifier.add_ty(TypeEnum::TObj {
obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
fields: Mapping::default(),
params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: x }]),
});
let int = env.parse("int", &HashMap::new());
let list_int = env.parse("list[int]", &HashMap::new());
assert_eq!(env.unify(a, b), Err("Incompatible types: typevar3 and typevar2".to_string()));
assert_eq!(env.unify(a, b), Err("Incompatible types: typevar4 and typevar3".to_string()));
env.unifier.unify(list_a, list_x).unwrap();
assert_eq!(env.unify(list_x, list_int), Err("Incompatible types: list[typevar2] and list[0]".to_string()));
assert_eq!(
env.unify(list_x, list_int),
Err("Incompatible types: 11[typevar3] and 11[0]".to_string())
);
env.unifier.replace_rigid_var(a, int);
env.unifier.unify(list_x, list_int).unwrap();
@ -521,16 +590,26 @@ fn test_instantiation() {
let float = env.parse("float", &HashMap::new());
let list_int = env.parse("list[int]", &HashMap::new());
let obj_map: HashMap<_, _> =
[(0usize, "int"), (1, "float"), (2, "bool")].iter().cloned().collect();
let list_elem_tvar = if let TypeEnum::TObj { params, .. } =
&*env.unifier.get_ty_immutable(env.type_mapping["list"])
{
iter_type_vars(params).next().unwrap()
} else {
unreachable!()
};
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
let list_v = env.unifier.add_ty(TypeEnum::TList { ty: v });
let v1 = env.unifier.get_fresh_var_with_range(&[list_v, int], None, None).0;
let v2 = env.unifier.get_fresh_var_with_range(&[list_int, float], None, None).0;
let t = env.unifier.get_dummy_var().0;
let obj_map: HashMap<_, _> = [(0usize, "int"), (1, "float"), (2, "bool"), (11, "list")].into();
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
let list_v = env
.unifier
.subst(env.type_mapping["list"], &into_var_map([TypeVar { id: list_elem_tvar.id, ty: v }]))
.unwrap();
let v1 = env.unifier.get_fresh_var_with_range(&[list_v, int], None, None).ty;
let v2 = env.unifier.get_fresh_var_with_range(&[list_int, float], None, None).ty;
let t = env.unifier.get_dummy_var().ty;
let tuple = env.unifier.add_ty(TypeEnum::TTuple { ty: vec![v, v1, v2] });
let v3 = env.unifier.get_fresh_var_with_range(&[tuple, t], None, None).0;
let v3 = env.unifier.get_fresh_var_with_range(&[tuple, t], None, None).ty;
// t = TypeVar('t')
// v = TypeVar('v', int, bool)
// v1 = TypeVar('v1', 'list[v]', int)
@ -552,7 +631,7 @@ fn test_instantiation() {
tuple[int, list[bool], list[int]]
tuple[int, list[int], float]
tuple[int, list[int], list[int]]
v5"
v6"
}
.split('\n')
.collect_vec();
@ -561,8 +640,8 @@ fn test_instantiation() {
.map(|ty| {
env.unifier.internal_stringify(
*ty,
&mut |i| obj_map.get(&i).unwrap().to_string(),
&mut |i| format!("v{}", i),
&mut |i| (*obj_map.get(&i).unwrap()).to_string(),
&mut |i| format!("v{i}"),
&mut None,
)
})

View File

@ -16,21 +16,10 @@ pub struct UnificationTable<V> {
#[derive(Clone, Debug)]
enum Action<V> {
Parent {
key: usize,
original_parent: usize,
},
Value {
key: usize,
original_value: Option<V>,
},
Rank {
key: usize,
original_rank: u32,
},
Marker {
generation: u32,
}
Parent { key: usize, original_parent: usize },
Value { key: usize, original_value: Option<V> },
Rank { key: usize, original_rank: u32 },
Marker { generation: u32 },
}
impl<V> Default for UnificationTable<V> {
@ -41,7 +30,13 @@ impl<V> Default for UnificationTable<V> {
impl<V> UnificationTable<V> {
pub fn new() -> UnificationTable<V> {
UnificationTable { parents: Vec::new(), ranks: Vec::new(), values: Vec::new(), log: Vec::new(), generation: 0 }
UnificationTable {
parents: Vec::new(),
ranks: Vec::new(),
values: Vec::new(),
log: Vec::new(),
generation: 0,
}
}
pub fn new_key(&mut self, v: V) -> UnificationKey {
@ -125,7 +120,10 @@ impl<V> UnificationTable<V> {
pub fn restore_snapshot(&mut self, snapshot: (usize, u32)) {
let (log_len, generation) = snapshot;
assert!(self.log.len() >= log_len, "snapshot restoration error");
assert!(matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation), "snapshot restoration error");
assert!(
matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation),
"snapshot restoration error"
);
for action in self.log.drain(log_len - 1..).rev() {
match action {
Action::Parent { key, original_parent } => {
@ -145,7 +143,10 @@ impl<V> UnificationTable<V> {
pub fn discard_snapshot(&mut self, snapshot: (usize, u32)) {
let (log_len, generation) = snapshot;
assert!(self.log.len() >= log_len, "snapshot discard error");
assert!(matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation), "snapshot discard error");
assert!(
matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation),
"snapshot discard error"
);
self.log.clear();
}
}
@ -159,11 +160,23 @@ where
.enumerate()
.map(|(i, (v, p))| if *p == i { v.as_ref().map(|v| v.as_ref().clone()) } else { None })
.collect();
UnificationTable { parents: self.parents.clone(), ranks: self.ranks.clone(), values, log: Vec::new(), generation: 0 }
UnificationTable {
parents: self.parents.clone(),
ranks: self.ranks.clone(),
values,
log: Vec::new(),
generation: 0,
}
}
pub fn from_send(table: &UnificationTable<V>) -> UnificationTable<Rc<V>> {
let values = table.values.iter().cloned().map(|v| v.map(Rc::new)).collect();
UnificationTable { parents: table.parents.clone(), ranks: table.ranks.clone(), values, log: Vec::new(), generation: 0 }
UnificationTable {
parents: table.parents.clone(),
ranks: table.ranks.clone(),
values,
log: Vec::new(),
generation: 0,
}
}
}

View File

@ -2,7 +2,7 @@
name = "nac3ld"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2018"
edition = "2021"
[dependencies]
byteorder = { version = "1.4", default-features = false }
byteorder = { version = "1.5", default-features = false }

View File

@ -27,27 +27,39 @@ pub const DW_EH_PE_indirect: u8 = 0x80;
pub struct DwarfReader<'a> {
pub slice: &'a [u8],
pub virt_addr: u32,
base_slice: &'a [u8],
base_virt_addr: u32,
}
impl<'a> DwarfReader<'a> {
pub fn new(slice: &[u8], virt_addr: u32) -> DwarfReader {
DwarfReader { slice, virt_addr }
DwarfReader { slice, virt_addr, base_slice: slice, base_virt_addr: virt_addr }
}
pub fn offset(&mut self, offset: i32) {
/// Creates a new instance from another instance of [DwarfReader], optionally removing any
/// offsets previously applied to the other instance.
pub fn from_reader(other: &DwarfReader<'a>, reset_offset: bool) -> DwarfReader<'a> {
if reset_offset {
DwarfReader::new(other.base_slice, other.base_virt_addr)
} else {
DwarfReader::new(other.slice, other.virt_addr)
}
}
pub fn offset(&mut self, offset: u32) {
self.slice = &self.slice[offset as usize..];
self.virt_addr = self.virt_addr.wrapping_add(offset as u32);
self.virt_addr = self.virt_addr.wrapping_add(offset);
}
// ULEB128 and SLEB128 encodings are defined in Section 7.6 - "Variable
// Length Data".
/// ULEB128 and SLEB128 encodings are defined in Section 7.6 - "Variable Length Data" of the
/// [DWARF-4 Manual](https://dwarfstd.org/doc/DWARF4.pdf).
pub fn read_uleb128(&mut self) -> u64 {
let mut shift: usize = 0;
let mut result: u64 = 0;
let mut byte: u8;
loop {
byte = self.read_u8();
result |= ((byte & 0x7F) as u64) << shift;
result |= u64::from(byte & 0x7F) << shift;
shift += 7;
if byte & 0x80 == 0 {
break;
@ -62,7 +74,7 @@ impl<'a> DwarfReader<'a> {
let mut byte: u8;
loop {
byte = self.read_u8();
result |= ((byte & 0x7F) as u64) << shift;
result |= u64::from(byte & 0x7F) << shift;
shift += 7;
if byte & 0x80 == 0 {
break;
@ -70,7 +82,7 @@ impl<'a> DwarfReader<'a> {
}
// sign-extend
if shift < u64::BITS && (byte & 0x40) != 0 {
result |= (!0 as u64) << shift;
result |= (!0u64) << shift;
}
result as i64
}
@ -144,10 +156,9 @@ fn read_encoded_pointer(reader: &mut DwarfReader, encoding: u8) -> Result<usize,
}
match encoding & 0x0F {
DW_EH_PE_absptr => Ok(reader.read_u32() as usize),
DW_EH_PE_absptr | DW_EH_PE_udata4 => Ok(reader.read_u32() as usize),
DW_EH_PE_uleb128 => Ok(reader.read_uleb128() as usize),
DW_EH_PE_udata2 => Ok(reader.read_u16() as usize),
DW_EH_PE_udata4 => Ok(reader.read_u32() as usize),
DW_EH_PE_udata8 => Ok(reader.read_u64() as usize),
DW_EH_PE_sleb128 => Ok(reader.read_sleb128() as usize),
DW_EH_PE_sdata2 => Ok(reader.read_i16() as usize),
@ -157,10 +168,7 @@ fn read_encoded_pointer(reader: &mut DwarfReader, encoding: u8) -> Result<usize,
}
}
fn read_encoded_pointer_with_pc(
reader: &mut DwarfReader,
encoding: u8,
) -> Result<usize, ()> {
fn read_encoded_pointer_with_pc(reader: &mut DwarfReader, encoding: u8) -> Result<usize, ()> {
let entry_virt_addr = reader.virt_addr;
let mut result = read_encoded_pointer(reader, encoding)?;
@ -200,38 +208,64 @@ fn round_up(unrounded: usize, align: usize) -> Result<usize, ()> {
}
}
// Minimalistic structure to store everything needed for parsing FDEs to synthesize
// .eh_frame_hdr section. Since we are only linking 1 object file, there should only be 1 call
// frame information (CFI) record, so there should be only 1 common information entry (CIE).
// So the class parses the only CIE on init, cache the encoding info, then parse the FDE on
// iterations based on the cached encoding format.
/// Minimalistic structure to store everything needed for parsing FDEs to synthesize `.eh_frame_hdr`
/// section.
///
/// Refer to [The Linux Standard Base Core Specification, Generic Part](https://refspecs.linuxfoundation.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/ehframechpt.html)
/// for more information.
pub struct EH_Frame<'a> {
// It refers to the augmentation data that corresponds to 'R' in the augmentation string
pub fde_pointer_encoding: u8,
pub fde_reader: DwarfReader<'a>,
pub fde_sz: usize,
reader: DwarfReader<'a>,
}
impl<'a> EH_Frame<'a> {
pub fn new(eh_frame_slice: &[u8], eh_frame_addr: u32) -> Result<EH_Frame, ()> {
let mut cie_reader = DwarfReader::new(eh_frame_slice, eh_frame_addr);
let eh_frame_size = eh_frame_slice.len();
/// Creates an [EH_Frame] using the bytes in the `.eh_frame` section and its address in the ELF
/// file.
pub fn new(eh_frame_slice: &[u8], eh_frame_addr: u32) -> EH_Frame {
EH_Frame { reader: DwarfReader::new(eh_frame_slice, eh_frame_addr) }
}
/// Returns an [Iterator] over all Call Frame Information (CFI) records.
pub fn cfi_records(&self) -> CFI_Records<'a> {
let reader = DwarfReader::from_reader(&self.reader, true);
let len = reader.slice.len();
CFI_Records { reader, available: len }
}
}
/// A single Call Frame Information (CFI) record.
///
/// From the [specification](https://refspecs.linuxfoundation.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/ehframechpt.html):
///
/// > Each CFI record contains a Common Information Entry (CIE) record followed by 1 or more Frame
/// Description Entry (FDE) records.
pub struct CFI_Record<'a> {
// It refers to the augmentation data that corresponds to 'R' in the augmentation string
fde_pointer_encoding: u8,
fde_reader: DwarfReader<'a>,
}
impl<'a> CFI_Record<'a> {
pub fn from_reader(cie_reader: &mut DwarfReader<'a>) -> Result<CFI_Record<'a>, ()> {
let length = cie_reader.read_u32();
let fde_reader = match length {
// eh_frame with 0 lengths means the CIE is terminated
// while length == u32::MAX means that the length is only representable with 64 bits,
0 => panic!("Cannot create an EH_Frame from a termination CIE"),
// length == u32::MAX means that the length is only representable with 64 bits,
// which does not make sense in a system with 32-bit address.
0 | 0xFFFFFFFF => unimplemented!(),
0xFFFF_FFFF => unimplemented!(),
_ => {
let mut fde_reader = DwarfReader::new(cie_reader.slice, cie_reader.virt_addr);
fde_reader.offset(length as i32);
let mut fde_reader = DwarfReader::from_reader(cie_reader, false);
fde_reader.offset(length);
fde_reader
}
};
let fde_sz = eh_frame_size - mem::size_of::<u32>() - length as usize;
// Routine check on the .eh_frame well-formness, in terms of CIE ID & Version args.
assert_eq!(cie_reader.read_u32(), 0);
let cie_ptr = cie_reader.read_u32();
assert_eq!(cie_ptr, 0);
assert_eq!(cie_reader.read_u8(), 1);
// Parse augmentation string
@ -242,7 +276,7 @@ impl<'a> EH_Frame<'a> {
// Skip code/data alignment factors & return address register along the way as well
// We only tackle the case where 'z' and 'R' are part of the augmentation string, otherwise
// we cannot get the addresses to make .eh_frame_hdr
let mut aug_data_reader = DwarfReader::new(cie_reader.slice, cie_reader.virt_addr);
let mut aug_data_reader = DwarfReader::from_reader(cie_reader, false);
let mut aug_str_len = 0;
loop {
if aug_data_reader.read_u8() == b'\0' {
@ -279,43 +313,114 @@ impl<'a> EH_Frame<'a> {
}
assert_ne!(fde_pointer_encoding, DW_EH_PE_omit);
Ok(EH_Frame { fde_pointer_encoding, fde_reader, fde_sz })
Ok(CFI_Record { fde_pointer_encoding, fde_reader })
}
pub fn iterate_fde(&self, callback: &mut dyn FnMut(u32, u32)) -> Result<(), ()> {
/// Returns a [DwarfReader] initialized to the first Frame Description Entry (FDE) of this CFI
/// record.
pub fn get_fde_reader(&self) -> DwarfReader<'a> {
DwarfReader::from_reader(&self.fde_reader, true)
}
/// Returns an [Iterator] over all Frame Description Entries (FDEs).
pub fn fde_records(&self) -> FDE_Records<'a> {
let reader = self.get_fde_reader();
let len = reader.slice.len();
FDE_Records { pointer_encoding: self.fde_pointer_encoding, reader, available: len }
}
}
/// [Iterator] over Call Frame Information (CFI) records in an
/// [Exception Handling (EH) frame][EH_Frame].
pub struct CFI_Records<'a> {
reader: DwarfReader<'a>,
available: usize,
}
impl<'a> Iterator for CFI_Records<'a> {
type Item = CFI_Record<'a>;
fn next(&mut self) -> Option<Self::Item> {
loop {
if self.available == 0 {
return None;
}
let mut this_reader = DwarfReader::from_reader(&self.reader, false);
// Remove the length of the header and the content from the counter
let length = self.reader.read_u32();
let length = match length {
// eh_frame with 0-length means the CIE is terminated
0 => return None,
0xFFFF_FFFF => unimplemented!("CIE entries larger than 4 bytes not supported"),
other => other,
} as usize;
// Remove the length of the header and the content from the counter
self.available -= length + mem::size_of::<u32>();
let mut next_reader = DwarfReader::from_reader(&self.reader, false);
next_reader.offset(length as u32);
let cie_ptr = self.reader.read_u32();
self.reader = next_reader;
// Skip this record if it is a FDE
if cie_ptr == 0 {
// Rewind back to the start of the CFI Record
return Some(CFI_Record::from_reader(&mut this_reader).ok().unwrap());
}
}
}
}
/// [Iterator] over Frame Description Entries (FDEs) in an
/// [Exception Handling (EH) frame][EH_Frame].
pub struct FDE_Records<'a> {
pointer_encoding: u8,
reader: DwarfReader<'a>,
available: usize,
}
impl<'a> Iterator for FDE_Records<'a> {
type Item = (u32, u32);
fn next(&mut self) -> Option<Self::Item> {
// Parse each FDE to obtain the starting address that the FDE applies to
// Send the FDE offset and the mentioned address to a callback that write up the
// .eh_frame_hdr section
let mut remaining_len = self.fde_sz;
let mut reader = DwarfReader::new(self.fde_reader.slice, self.fde_reader.virt_addr);
loop {
if remaining_len == 0 {
break;
}
let fde_virt_addr = reader.virt_addr;
let length = match reader.read_u32() {
0 | 0xFFFFFFFF => unimplemented!(),
other => other,
};
if self.available == 0 {
return None;
}
// Remove the length of the header and the content from the counter
remaining_len -= length as usize + mem::size_of::<u32>();
let mut next_fde_reader = DwarfReader::new(reader.slice, reader.virt_addr);
next_fde_reader.offset(length as i32);
let length = match self.reader.read_u32() {
// eh_frame with 0-length means the CIE is terminated
0 => return None,
0xFFFF_FFFF => unimplemented!("CIE entries larger than 4 bytes not supported"),
other => other,
} as usize;
// Skip CIE pointer offset
reader.read_u32();
// Remove the length of the header and the content from the counter
self.available -= length + mem::size_of::<u32>();
let mut next_fde_reader = DwarfReader::from_reader(&self.reader, false);
next_fde_reader.offset(length as u32);
// Parse PC Begin using the encoding scheme mentioned in the CIE
let pc_begin = read_encoded_pointer_with_pc(&mut reader, self.fde_pointer_encoding)?;
let cie_ptr = self.reader.read_u32();
let next_val = if cie_ptr != 0 {
let pc_begin = read_encoded_pointer_with_pc(&mut self.reader, self.pointer_encoding)
.expect("Failed to read PC Begin");
Some((pc_begin as u32, self.reader.virt_addr))
} else {
None
};
callback(pc_begin as u32, fde_virt_addr);
self.reader = next_fde_reader;
reader = next_fde_reader;
}
Ok(())
next_val
}
}
@ -326,29 +431,32 @@ pub struct EH_Frame_Hdr<'a> {
}
impl<'a> EH_Frame_Hdr<'a> {
// Create a EH_Frame_Hdr object, and write out the fixed fields of .eh_frame_hdr to memory
// eh_frame_ptr_enc will be 0x1B (PC-relative, 4 bytes)
// table_enc will be 0x3B (Relative to the start of .eh_frame_hdr, 4 bytes)
// Load address is not known at this point.
/// Create a [EH_Frame_Hdr] object, and write out the fixed fields of `.eh_frame_hdr` to memory.
///
/// Load address is not known at this point.
pub fn new(
eh_frame_hdr_slice: &mut [u8],
eh_frame_hdr_addr: u32,
eh_frame_addr: u32,
) -> EH_Frame_Hdr {
let mut writer = DwarfWriter::new(eh_frame_hdr_slice);
writer.write_u8(1);
writer.write_u8(0x1B);
writer.write_u8(0x03);
writer.write_u8(0x3B);
let eh_frame_offset =
(eh_frame_addr).wrapping_sub(eh_frame_hdr_addr + ((mem::size_of::<u8>() as u32) * 4));
writer.write_u32(eh_frame_offset);
writer.write_u32(0);
writer.write_u8(1); // version
writer.write_u8(0x1B); // eh_frame_ptr_enc - PC-relative 4-byte signed value
writer.write_u8(0x03); // fde_count_enc - 4-byte unsigned value
writer.write_u8(0x3B); // table_enc - .eh_frame_hdr section-relative 4-byte signed value
let eh_frame_offset = eh_frame_addr.wrapping_sub(
eh_frame_hdr_addr + writer.offset as u32 + ((mem::size_of::<u8>() as u32) * 4),
);
writer.write_u32(eh_frame_offset); // eh_frame_ptr
writer.write_u32(0); // `fde_count`, will be written in finalize_fde
EH_Frame_Hdr { fde_writer: writer, eh_frame_hdr_addr, fdes: Vec::new() }
}
/// The offset of the `fde_count` value relative to the start of the `.eh_frame_hdr` section in
/// bytes.
fn fde_count_offset() -> usize {
8
}
@ -367,7 +475,10 @@ impl<'a> EH_Frame_Hdr<'a> {
self.fde_writer.write_u32(*init_loc);
self.fde_writer.write_u32(*addr);
}
LittleEndian::write_u32(&mut self.fde_writer.slice[Self::fde_count_offset()..], self.fdes.len() as u32);
LittleEndian::write_u32(
&mut self.fde_writer.slice[Self::fde_count_offset()..],
self.fdes.len() as u32,
);
}
pub fn size_from_eh_frame(eh_frame: &[u8]) -> usize {
@ -379,14 +490,20 @@ impl<'a> EH_Frame_Hdr<'a> {
// The original length field should be able to hold the entire value.
// The device memory space is limited to 32-bits addresses anyway.
let entry_length = reader.read_u32();
if entry_length == 0 || entry_length == 0xFFFFFFFF {
if entry_length == 0 || entry_length == 0xFFFF_FFFF {
unimplemented!()
}
if reader.read_u32() != 0 {
// This slot stores the CIE ID (for CIE)/CIE Pointer (for FDE).
// This value must be non-zero for FDEs.
let cie_ptr = reader.read_u32();
if cie_ptr != 0 {
fde_count += 1;
}
reader.offset(entry_length as i32 - mem::size_of::<u32>() as i32)
reader.offset(entry_length - mem::size_of::<u32>() as u32);
}
12 + fde_count * 8
}
}

View File

@ -1,5 +1,5 @@
/* generated from elf.h with rust-bindgen and then manually altered */
#![allow(non_camel_case_types, non_snake_case, non_upper_case_globals, dead_code)]
#![allow(non_camel_case_types, non_snake_case, non_upper_case_globals, dead_code, clippy::pedantic)]
pub const EI_NIDENT: usize = 16;
pub const EI_MAG0: usize = 0;
@ -10,7 +10,7 @@ pub const EI_MAG2: usize = 2;
pub const ELFMAG2: u8 = b'L';
pub const EI_MAG3: usize = 3;
pub const ELFMAG3: u8 = b'F';
pub const ELFMAG: &'static [u8; 5usize] = b"\x7fELF\x00";
pub const ELFMAG: &[u8; 5usize] = b"\x7fELF\x00";
pub const SELFMAG: usize = 4;
pub const EI_CLASS: usize = 4;
pub const ELFCLASSNONE: u8 = 0;
@ -428,8 +428,8 @@ pub const VER_NDX_ELIMINATE: usize = 65281;
pub const VER_NEED_NONE: usize = 0;
pub const VER_NEED_CURRENT: usize = 1;
pub const VER_NEED_NUM: usize = 2;
pub const ELF_NOTE_SOLARIS: &'static [u8; 13usize] = b"SUNW Solaris\x00";
pub const ELF_NOTE_GNU: &'static [u8; 4usize] = b"GNU\x00";
pub const ELF_NOTE_SOLARIS: &[u8; 13usize] = b"SUNW Solaris\x00";
pub const ELF_NOTE_GNU: &[u8; 4usize] = b"GNU\x00";
pub const ELF_NOTE_PAGESIZE_HINT: usize = 1;
pub const NT_GNU_ABI_TAG: usize = 1;
pub const ELF_NOTE_ABI: usize = 1;

View File

@ -1,7 +1,30 @@
#![deny(
future_incompatible,
let_underscore,
nonstandard_style,
rust_2024_compatibility,
clippy::all
)]
#![warn(clippy::pedantic)]
#![allow(
clippy::cast_possible_truncation,
clippy::cast_possible_wrap,
clippy::cast_sign_loss,
clippy::doc_markdown,
clippy::enum_glob_use,
clippy::missing_errors_doc,
clippy::missing_panics_doc,
clippy::module_name_repetitions,
clippy::similar_names,
clippy::struct_field_names,
clippy::too_many_lines,
clippy::wildcard_imports
)]
use dwarf::*;
use elf::*;
use std::collections::HashMap;
use std::{convert, mem, ptr, slice, str};
use std::{mem, ptr, slice, str};
extern crate byteorder;
use byteorder::{ByteOrder, LittleEndian};
@ -21,7 +44,7 @@ pub enum Error {
Lookup(&'static str),
}
impl convert::From<&'static str> for Error {
impl From<&'static str> for Error {
fn from(desc: &'static str) -> Error {
Error::Parsing(desc)
}
@ -70,45 +93,45 @@ struct SectionRecord<'a> {
data: Vec<u8>,
}
fn read_unaligned<T: Copy>(data: &[u8], offset: usize) -> Result<T, ()> {
fn read_unaligned<T: Copy>(data: &[u8], offset: usize) -> Option<T> {
if data.len() < offset + mem::size_of::<T>() {
Err(())
None
} else {
let ptr = data.as_ptr().wrapping_offset(offset as isize) as *const T;
Ok(unsafe { ptr::read_unaligned(ptr) })
let ptr = data.as_ptr().wrapping_add(offset).cast();
Some(unsafe { ptr::read_unaligned(ptr) })
}
}
pub fn get_ref_slice<T: Copy>(data: &[u8], offset: usize, len: usize) -> Result<&[T], ()> {
#[must_use]
pub fn get_ref_slice<T: Copy>(data: &[u8], offset: usize, len: usize) -> Option<&[T]> {
if data.len() < offset + mem::size_of::<T>() * len {
Err(())
None
} else {
let ptr = data.as_ptr().wrapping_offset(offset as isize) as *const T;
Ok(unsafe { slice::from_raw_parts(ptr, len) })
let ptr = data.as_ptr().wrapping_add(offset).cast();
Some(unsafe { slice::from_raw_parts(ptr, len) })
}
}
fn from_struct_vec<T>(struct_vec: Vec<T>) -> Vec<u8> {
fn from_struct_slice<T>(struct_vec: &[T]) -> Vec<u8> {
let ptr = struct_vec.as_ptr();
unsafe { slice::from_raw_parts(ptr as *const u8, struct_vec.len() * mem::size_of::<T>()) }
.to_vec()
unsafe { slice::from_raw_parts(ptr.cast(), mem::size_of_val(struct_vec)) }.to_vec()
}
fn to_struct_slice<T>(bytes: &[u8]) -> &[T] {
unsafe { slice::from_raw_parts(bytes.as_ptr() as *const T, bytes.len() / mem::size_of::<T>()) }
unsafe { slice::from_raw_parts(bytes.as_ptr().cast(), bytes.len() / mem::size_of::<T>()) }
}
fn to_struct_mut_slice<T>(bytes: &mut [u8]) -> &mut [T] {
unsafe {
slice::from_raw_parts_mut(bytes.as_mut_ptr() as *mut T, bytes.len() / mem::size_of::<T>())
slice::from_raw_parts_mut(bytes.as_mut_ptr().cast(), bytes.len() / mem::size_of::<T>())
}
}
fn elf_hash(name: &[u8]) -> u32 {
let mut h: u32 = 0;
for c in name {
h = (h << 4) + *c as u32;
let g = h & 0xf0000000;
h = (h << 4) + u32::from(*c);
let g = h & 0xf000_0000;
if g != 0 {
h ^= g >> 24;
h &= !g;
@ -177,7 +200,7 @@ impl<'a> Linker<'a> {
}
fn load_section(&mut self, shdr: &Elf32_Shdr, sh_name_str: &'a str, data: Vec<u8>) -> usize {
let mut elf_shdr = shdr.clone();
let mut elf_shdr = *shdr;
// Maintain alignment requirement specified in sh_addralign
let align = shdr.sh_addralign;
@ -202,22 +225,26 @@ impl<'a> Linker<'a> {
relocs: &[R],
target_section: Elf32_Word,
) -> Result<(), Error> {
type RelocateFn = dyn Fn(&mut [u8], Elf32_Word);
struct RelocInfo<'a, R> {
pub defined_val: bool,
pub indirect_reloc: Option<&'a R>,
pub pc_relative: bool,
pub relocate: Option<Box<RelocateFn>>,
}
for reloc in relocs {
let sym = match reloc.sym_info() as usize {
STN_UNDEF => None,
sym_index => Some(
self.symtab
.get(sym_index as usize)
.ok_or("symbol out of bounds of symbol table")?,
),
sym_index => {
Some(self.symtab.get(sym_index).ok_or("symbol out of bounds of symbol table")?)
}
};
let resolve_symbol_addr =
|sym_option: Option<&Elf32_Sym>| -> Result<Elf32_Word, Error> {
let sym = match sym_option {
Some(sym) => sym,
None => return Ok(0),
};
let Some(sym) = sym_option else { return Ok(0) };
match sym.st_shndx {
SHN_UNDEF => Err(Error::Lookup("undefined symbol")),
@ -240,17 +267,10 @@ impl<'a> Linker<'a> {
let get_target_section_index = || -> Result<usize, Error> {
self.section_map
.get(&(target_section as usize))
.map(|&index| index)
.copied()
.ok_or(Error::Parsing("Cannot find section with matching sh_index"))
};
struct RelocInfo<'a, R> {
pub defined_val: bool,
pub indirect_reloc: Option<&'a R>,
pub pc_relative: bool,
pub relocate: Option<Box<dyn Fn(&mut [u8], Elf32_Word)>>,
}
let classify = |reloc: &R, sym_option: Option<&Elf32_Sym>| -> Option<RelocInfo<R>> {
let defined_val = sym_option.map_or(true, |sym| {
sym.st_shndx != SHN_UNDEF || ELF32_ST_BIND(sym.st_info) == STB_LOCAL
@ -262,7 +282,7 @@ impl<'a> Linker<'a> {
indirect_reloc: None,
pc_relative: true,
relocate: Some(Box::new(|target_word, value| {
LittleEndian::write_u32(target_word, value)
LittleEndian::write_u32(target_word, value);
})),
}),
@ -273,9 +293,9 @@ impl<'a> Linker<'a> {
relocate: Some(Box::new(|target_word, value| {
LittleEndian::write_u32(
target_word,
(LittleEndian::read_u32(target_word) & 0x80000000)
| value & 0x7FFFFFFF,
)
(LittleEndian::read_u32(target_word) & 0x8000_0000)
| value & 0x7FFF_FFFF,
);
})),
}),
@ -297,8 +317,8 @@ impl<'a> Linker<'a> {
relocate: Some(Box::new(|target_word, value| {
let auipc_raw = LittleEndian::read_u32(target_word);
let auipc_insn =
(auipc_raw & 0xFFF) | ((value + 0x800) & 0xFFFFF000);
LittleEndian::write_u32(target_word, auipc_insn)
(auipc_raw & 0xFFF) | ((value + 0x800) & 0xFFFF_F000);
LittleEndian::write_u32(target_word, auipc_insn);
})),
})
}
@ -308,19 +328,14 @@ impl<'a> Linker<'a> {
indirect_reloc: None,
pc_relative: true,
relocate: Some(Box::new(|target_word, value| {
LittleEndian::write_u32(target_word, value)
LittleEndian::write_u32(target_word, value);
})),
}),
R_RISCV_PCREL_LO12_I => {
let expected_offset = sym_option.map_or(0, |sym| sym.st_value);
let indirect_reloc = if let Some(reloc) =
relocs.iter().find(|reloc| reloc.offset() == expected_offset)
{
reloc
} else {
return None;
};
let indirect_reloc =
relocs.iter().find(|reloc| reloc.offset() == expected_offset)?;
Some(RelocInfo {
defined_val: {
let indirect_sym =
@ -334,14 +349,14 @@ impl<'a> Linker<'a> {
// Here, we convert to direct addressing
// GOT reloc (indirect) -> lw + addi
// PCREL reloc (direct) -> addi
let (lo_opcode, lo_funct3) = (0b0010011, 0b000);
let (lo_opcode, lo_funct3) = (0b001_0011, 0b000);
let addi_lw_raw = LittleEndian::read_u32(target_word);
let addi_insn = lo_opcode
| (addi_lw_raw & 0xF8F80)
| (lo_funct3 << 12)
| ((value & 0xFFF) << 20);
LittleEndian::write_u32(target_word, addi_insn)
LittleEndian::write_u32(target_word, addi_insn);
})),
})
}
@ -358,10 +373,7 @@ impl<'a> Linker<'a> {
indirect_reloc: None,
pc_relative: false,
relocate: Some(Box::new(|target_word, value| {
LittleEndian::write_u32(
target_word,
value,
)
LittleEndian::write_u32(target_word, value);
})),
}),
@ -371,7 +383,7 @@ impl<'a> Linker<'a> {
pc_relative: false,
relocate: Some(Box::new(|target_word, value| {
let old_value = LittleEndian::read_u32(target_word);
LittleEndian::write_u32(target_word, old_value.wrapping_add(value))
LittleEndian::write_u32(target_word, old_value.wrapping_add(value));
})),
}),
@ -381,7 +393,7 @@ impl<'a> Linker<'a> {
pc_relative: false,
relocate: Some(Box::new(|target_word, value| {
let old_value = LittleEndian::read_u32(target_word);
LittleEndian::write_u32(target_word, old_value.wrapping_sub(value))
LittleEndian::write_u32(target_word, old_value.wrapping_sub(value));
})),
}),
@ -390,10 +402,7 @@ impl<'a> Linker<'a> {
indirect_reloc: None,
pc_relative: false,
relocate: Some(Box::new(|target_word, value| {
LittleEndian::write_u16(
target_word,
value as u16,
)
LittleEndian::write_u16(target_word, value as u16);
})),
}),
@ -406,7 +415,7 @@ impl<'a> Linker<'a> {
LittleEndian::write_u16(
target_word,
old_value.wrapping_add(value as u16),
)
);
})),
}),
@ -419,7 +428,7 @@ impl<'a> Linker<'a> {
LittleEndian::write_u16(
target_word,
old_value.wrapping_sub(value as u16),
)
);
})),
}),
@ -501,7 +510,7 @@ impl<'a> Linker<'a> {
if let Some(relocate) = reloc_info.relocate {
let target_word = &mut target_sec_image[reloc.offset() as usize..];
relocate(target_word, value)
relocate(target_word, value);
} else {
self.rela_dyn_relas.push(Elf32_Rela {
r_offset: rela_off,
@ -549,17 +558,18 @@ impl<'a> Linker<'a> {
let eh_frame_slice = eh_frame_rec.data.as_slice();
// Prepare a new buffer to dodge borrow check
let mut eh_frame_hdr_vec: Vec<u8> = vec![0; eh_frame_hdr_rec.shdr.sh_size as usize];
let eh_frame = EH_Frame::new(eh_frame_slice, eh_frame_rec.shdr.sh_offset)
.map_err(|()| "cannot read EH frame")?;
let eh_frame = EH_Frame::new(eh_frame_slice, eh_frame_rec.shdr.sh_offset);
let mut eh_frame_hdr = EH_Frame_Hdr::new(
eh_frame_hdr_vec.as_mut_slice(),
eh_frame_hdr_rec.shdr.sh_offset,
eh_frame_rec.shdr.sh_offset,
);
let mut fde_callback = |init_pos, virt_addr| eh_frame_hdr.add_fde(init_pos, virt_addr);
eh_frame
.iterate_fde(&mut fde_callback)
.map_err(|()| "failed to add FDE to .eh_frame_hdr while iterating .eh_frame")?;
eh_frame.cfi_records().flat_map(|cfi| cfi.fde_records()).for_each(&mut |(
init_pos,
virt_addr,
)| {
eh_frame_hdr.add_fde(init_pos, virt_addr);
});
// Sort FDE entries in .eh_frame_hdr
eh_frame_hdr.finalize_fde();
@ -573,39 +583,114 @@ impl<'a> Linker<'a> {
}
pub fn ld(data: &'a [u8]) -> Result<Vec<u8>, Error> {
let ehdr = read_unaligned::<Elf32_Ehdr>(data, 0).map_err(|()| "cannot read ELF header")?;
fn allocate_rela_dyn<R: Relocatable>(
linker: &Linker,
relocs: &[R],
) -> Result<(usize, Vec<u32>), Error> {
let mut alloc_size = 0;
let mut rela_dyn_sym_indices = Vec::new();
for reloc in relocs {
if reloc.sym_info() as usize == STN_UNDEF {
continue;
}
let sym: &Elf32_Sym = linker
.symtab
.get(reloc.sym_info() as usize)
.ok_or("symbol out of bounds of symbol table")?;
match (linker.isa, reloc.type_info()) {
// Absolute address relocations
// A runtime relocation is needed to find the loading address
(Isa::CortexA9, R_ARM_ABS32) | (Isa::RiscV32, R_RISCV_32) => {
alloc_size += mem::size_of::<Elf32_Rela>(); // FIXME: RELA vs REL
if ELF32_ST_BIND(sym.st_info) == STB_GLOBAL && sym.st_shndx == SHN_UNDEF {
rela_dyn_sym_indices.push(reloc.sym_info());
}
}
// Relative address relocations
// Relay the relocation to the runtime linker only if the symbol is not defined
(Isa::CortexA9, R_ARM_REL32 | R_ARM_PREL31 | R_ARM_TARGET2)
| (
Isa::RiscV32,
R_RISCV_CALL_PLT | R_RISCV_PCREL_HI20 | R_RISCV_GOT_HI20 | R_RISCV_32_PCREL
| R_RISCV_SET32 | R_RISCV_ADD32 | R_RISCV_SUB32 | R_RISCV_SET16
| R_RISCV_ADD16 | R_RISCV_SUB16 | R_RISCV_SET8 | R_RISCV_ADD8
| R_RISCV_SUB8 | R_RISCV_SET6 | R_RISCV_SUB6,
) => {
if ELF32_ST_BIND(sym.st_info) == STB_GLOBAL && sym.st_shndx == SHN_UNDEF {
alloc_size += mem::size_of::<Elf32_Rela>(); // FIXME: RELA vs REL
rela_dyn_sym_indices.push(reloc.sym_info());
}
}
// RISC-V: Lower 12-bits relocations
// If the upper 20-bits relocation cannot be resolved,
// this relocation will be relayed to the runtime linker.
(Isa::RiscV32, R_RISCV_PCREL_LO12_I) => {
// Find the HI20 relocation
let indirect_reloc = relocs
.iter()
.find(|reloc| reloc.offset() == sym.st_value)
.ok_or("malformatted LO12 relocation")?;
let indirect_sym = linker.symtab[indirect_reloc.sym_info() as usize];
if ELF32_ST_BIND(indirect_sym.st_info) == STB_GLOBAL
&& indirect_sym.st_shndx == SHN_UNDEF
{
alloc_size += mem::size_of::<Elf32_Rela>(); // FIXME: RELA vs REL
rela_dyn_sym_indices.push(reloc.sym_info());
}
}
_ => {
println!("Relocation type 0x{:X?} is not supported", reloc.type_info());
unimplemented!()
}
}
}
Ok((alloc_size, rela_dyn_sym_indices))
}
let Some(ehdr) = read_unaligned::<Elf32_Ehdr>(data, 0) else {
Err("cannot read ELF header")?
};
let isa = match ehdr.e_machine {
EM_ARM => Isa::CortexA9,
EM_RISCV => Isa::RiscV32,
_ => return Err(Error::Parsing("unsupported architecture")),
};
let shdrs = get_ref_slice::<Elf32_Shdr>(data, ehdr.e_shoff as usize, ehdr.e_shnum as usize)
.map_err(|()| "cannot read section header table")?;
let Some(shdrs) =
get_ref_slice::<Elf32_Shdr>(data, ehdr.e_shoff as usize, ehdr.e_shnum as usize)
else {
Err("cannot read section header table")?
};
// Read .strtab
let strtab_shdr = shdrs[ehdr.e_shstrndx as usize];
let strtab =
let Some(strtab) =
get_ref_slice::<u8>(data, strtab_shdr.sh_offset as usize, strtab_shdr.sh_size as usize)
.map_err(|()| "cannot read the string table from data")?;
else {
Err("cannot read the string table from data")?
};
// Read .symtab
let symtab_shdr = shdrs
.iter()
.find(|shdr| shdr.sh_type as usize == SHT_SYMTAB)
.ok_or(Error::Parsing("cannot find the symbol table"))?;
let symtab = get_ref_slice::<Elf32_Sym>(
let Some(symtab) = get_ref_slice::<Elf32_Sym>(
data,
symtab_shdr.sh_offset as usize,
symtab_shdr.sh_size as usize / mem::size_of::<Elf32_Sym>(),
)
.map_err(|()| "cannot read the symbol table from data")?;
) else {
Err("cannot read the symbol table from data")?
};
// Section table for the .elf paired with the section name
// To be formalized incrementally
// Very hashmap-like structure, but the order matters, so it is a vector
let mut elf_shdrs = Vec::new();
elf_shdrs.push(SectionRecord {
let elf_shdrs = vec![SectionRecord {
shdr: Elf32_Shdr {
sh_name: 0,
sh_type: 0,
@ -620,7 +705,7 @@ impl<'a> Linker<'a> {
},
name: "",
data: vec![0; 0],
});
}];
let elf_sh_data_off = mem::size_of::<Elf32_Ehdr>() + mem::size_of::<Elf32_Phdr>() * 5;
// Image of the linked dynamic library, to be formalized incrementally
@ -660,8 +745,8 @@ impl<'a> Linker<'a> {
linker.load_section(
&text_shdr,
".text",
(&data[text_shdr.sh_offset as usize
..text_shdr.sh_offset as usize + text_shdr.sh_size as usize])
data[text_shdr.sh_offset as usize
..text_shdr.sh_offset as usize + text_shdr.sh_size as usize]
.to_vec(),
);
linker.section_map.insert(text_shdr_index, 1);
@ -679,8 +764,8 @@ impl<'a> Linker<'a> {
let loaded_index = linker.load_section(
&arm_exidx_shdr,
".ARM.exidx",
(&data[arm_exidx_shdr.sh_offset as usize
..arm_exidx_shdr.sh_offset as usize + arm_exidx_shdr.sh_size as usize])
data[arm_exidx_shdr.sh_offset as usize
..arm_exidx_shdr.sh_offset as usize + arm_exidx_shdr.sh_size as usize]
.to_vec(),
);
linker.section_map.insert(arm_exidx_shdr_index, loaded_index);
@ -699,7 +784,7 @@ impl<'a> Linker<'a> {
let elf_shdrs_index = linker.load_section(
shdr,
str::from_utf8(section_name).unwrap(),
(&data[shdr.sh_offset as usize..(shdr.sh_offset + shdr.sh_size) as usize]).to_vec(),
data[shdr.sh_offset as usize..(shdr.sh_offset + shdr.sh_size) as usize].to_vec(),
);
linker.section_map.insert(i, elf_shdrs_index);
}
@ -756,21 +841,27 @@ impl<'a> Linker<'a> {
($shdr: expr, $stmt: expr) => {
match $shdr.sh_type as usize {
SHT_RELA => {
let relocs = get_ref_slice::<Elf32_Rela>(
let Some(relocs) = get_ref_slice::<Elf32_Rela>(
data,
$shdr.sh_offset as usize,
$shdr.sh_size as usize / mem::size_of::<Elf32_Rela>(),
)
.map_err(|()| "cannot parse relocations")?;
) else {
Err("cannot parse relocations")?
};
#[allow(clippy::redundant_closure_call)]
$stmt(relocs)
}
SHT_REL => {
let relocs = get_ref_slice::<Elf32_Rel>(
let Some(relocs) = get_ref_slice::<Elf32_Rel>(
data,
$shdr.sh_offset as usize,
$shdr.sh_size as usize / mem::size_of::<Elf32_Rel>(),
)
.map_err(|()| "cannot parse relocations")?;
) else {
Err("cannot parse relocations")?
};
#[allow(clippy::redundant_closure_call)]
$stmt(relocs)
}
_ => unreachable!(),
@ -778,84 +869,6 @@ impl<'a> Linker<'a> {
};
}
fn allocate_rela_dyn<R: Relocatable>(
linker: &Linker,
relocs: &[R],
) -> Result<(usize, Vec<u32>), Error> {
let mut alloc_size = 0;
let mut rela_dyn_sym_indices = Vec::new();
for reloc in relocs {
if reloc.sym_info() as usize == STN_UNDEF {
continue;
}
let sym: &Elf32_Sym = linker
.symtab
.get(reloc.sym_info() as usize)
.ok_or("symbol out of bounds of symbol table")?;
match (linker.isa, reloc.type_info()) {
// Absolute address relocations
// A runtime relocation is needed to find the loading address
(Isa::CortexA9, R_ARM_ABS32) | (Isa::RiscV32, R_RISCV_32) => {
alloc_size += mem::size_of::<Elf32_Rela>(); // FIXME: RELA vs REL
if ELF32_ST_BIND(sym.st_info) == STB_GLOBAL && sym.st_shndx == SHN_UNDEF {
rela_dyn_sym_indices.push(reloc.sym_info());
}
}
// Relative address relocations
// Relay the relocation to the runtime linker only if the symbol is not defined
(Isa::CortexA9, R_ARM_REL32)
| (Isa::CortexA9, R_ARM_PREL31)
| (Isa::CortexA9, R_ARM_TARGET2)
| (Isa::RiscV32, R_RISCV_CALL_PLT)
| (Isa::RiscV32, R_RISCV_PCREL_HI20)
| (Isa::RiscV32, R_RISCV_GOT_HI20)
| (Isa::RiscV32, R_RISCV_32_PCREL)
| (Isa::RiscV32, R_RISCV_SET32)
| (Isa::RiscV32, R_RISCV_ADD32)
| (Isa::RiscV32, R_RISCV_SUB32)
| (Isa::RiscV32, R_RISCV_SET16)
| (Isa::RiscV32, R_RISCV_ADD16)
| (Isa::RiscV32, R_RISCV_SUB16)
| (Isa::RiscV32, R_RISCV_SET8)
| (Isa::RiscV32, R_RISCV_ADD8)
| (Isa::RiscV32, R_RISCV_SUB8)
| (Isa::RiscV32, R_RISCV_SET6)
| (Isa::RiscV32, R_RISCV_SUB6) => {
if ELF32_ST_BIND(sym.st_info) == STB_GLOBAL && sym.st_shndx == SHN_UNDEF {
alloc_size += mem::size_of::<Elf32_Rela>(); // FIXME: RELA vs REL
rela_dyn_sym_indices.push(reloc.sym_info());
}
}
// RISC-V: Lower 12-bits relocations
// If the upper 20-bits relocation cannot be resolved,
// this relocation will be relayed to the runtime linker.
(Isa::RiscV32, R_RISCV_PCREL_LO12_I) => {
// Find the HI20 relocation
let indirect_reloc = relocs
.iter()
.find(|reloc| reloc.offset() == sym.st_value)
.ok_or("malformatted LO12 relocation")?;
let indirect_sym = linker.symtab[indirect_reloc.sym_info() as usize];
if ELF32_ST_BIND(indirect_sym.st_info) == STB_GLOBAL
&& indirect_sym.st_shndx == SHN_UNDEF
{
alloc_size += mem::size_of::<Elf32_Rela>(); // FIXME: RELA vs REL
rela_dyn_sym_indices.push(reloc.sym_info());
}
}
_ => {
println!("Relocation type 0x{:X?} is not supported", reloc.type_info());
unimplemented!()
}
}
}
Ok((alloc_size, rela_dyn_sym_indices))
}
for shdr in shdrs
.iter()
.filter(|shdr| shdr.sh_type as usize == SHT_REL || shdr.sh_type as usize == SHT_RELA)
@ -883,7 +896,7 @@ impl<'a> Linker<'a> {
}
// Avoid symbol duplication
rela_dyn_sym_indices.sort();
rela_dyn_sym_indices.sort_unstable();
rela_dyn_sym_indices.dedup();
if rela_dyn_size != 0 {
@ -919,7 +932,7 @@ impl<'a> Linker<'a> {
dynsym_names.push((0, 0));
for rela_dyn_sym_index in rela_dyn_sym_indices {
let mut sym = linker.symtab[rela_dyn_sym_index as usize].clone();
let mut sym = linker.symtab[rela_dyn_sym_index as usize];
let sym_name = name_starting_at_slice(strtab, sym.st_name as usize)
.map_err(|_| "cannot read symbol name from the original .strtab")?;
let dynstr_start_index = dynstr.len();
@ -929,7 +942,7 @@ impl<'a> Linker<'a> {
let elf_shdr_index = linker
.section_map
.get(&(sym.st_shndx as usize))
.map(|&index| index)
.copied()
.ok_or(Error::Parsing("Cannot find section with matching sh_index"))?;
let elf_shdr_offset = linker.elf_shdrs[elf_shdr_index].shdr.sh_offset;
sym.st_value += elf_shdr_offset;
@ -956,7 +969,7 @@ impl<'a> Linker<'a> {
let modinit_shdr_index = linker
.section_map
.get(&(modinit_sym.st_shndx as usize))
.map(|&index| index)
.copied()
.ok_or(Error::Parsing("Cannot find section with matching sh_index"))?;
let modinit_shdr = linker.elf_shdrs[modinit_shdr_index].shdr;
@ -1014,9 +1027,10 @@ impl<'a> Linker<'a> {
let mut hash_bucket: Vec<u32> = vec![0; dynsym.len()];
let mut hash_chain: Vec<u32> = vec![0; dynsym.len()];
for sym_index in 1..dynsym.len() {
let (str_start, str_end) = dynsym_names[sym_index];
let hash = elf_hash(&dynstr[str_start..str_end]);
for (sym_index, (str_start, str_end)) in
dynsym_names.iter().enumerate().take(dynsym.len()).skip(1)
{
let hash = elf_hash(&dynstr[*str_start..*str_end]);
let mut hash_index = hash as usize % hash_bucket.len();
if hash_bucket[hash_index] == 0 {
@ -1067,7 +1081,7 @@ impl<'a> Linker<'a> {
sh_entsize: mem::size_of::<Elf32_Sym>() as Elf32_Word,
},
".dynsym",
from_struct_vec(dynsym),
from_struct_slice(&dynsym),
);
let hash_elf_index = linker.load_section(
&Elf32_Shdr {
@ -1083,7 +1097,7 @@ impl<'a> Linker<'a> {
sh_entsize: 4,
},
".hash",
from_struct_vec(hash),
from_struct_slice(&hash),
);
// Link .rela.dyn header to the .dynsym header
@ -1105,7 +1119,7 @@ impl<'a> Linker<'a> {
let elf_shdrs_index = linker.load_section(
shdr,
str::from_utf8(section_name).unwrap(),
(&data[shdr.sh_offset as usize..(shdr.sh_offset + shdr.sh_size) as usize])
data[shdr.sh_offset as usize..(shdr.sh_offset + shdr.sh_size) as usize]
.to_vec(),
);
linker.section_map.insert(i, elf_shdrs_index);
@ -1182,7 +1196,7 @@ impl<'a> Linker<'a> {
};
let dynamic_elf_index =
linker.load_section(&dynamic_shdr, ".dynamic", from_struct_vec(dyn_entries));
linker.load_section(&dynamic_shdr, ".dynamic", from_struct_slice(&dyn_entries));
let last_w_sec_elf_index = linker.elf_shdrs.len() - 1;
@ -1209,7 +1223,7 @@ impl<'a> Linker<'a> {
let elf_shdrs_index = linker.load_section(
shdr,
section_name,
(&data[shdr.sh_offset as usize..(shdr.sh_offset + shdr.sh_size) as usize])
data[shdr.sh_offset as usize..(shdr.sh_offset + shdr.sh_size) as usize]
.to_vec(),
);
linker.section_map.insert(i, elf_shdrs_index);
@ -1258,12 +1272,14 @@ impl<'a> Linker<'a> {
update_dynsym_record!(b"__bss_start", bss_offset, bss_elf_index as Elf32_Section);
update_dynsym_record!(b"_end", bss_offset, bss_elf_index as Elf32_Section);
} else {
for (bss_iter_index, &(bss_section_index, section_name)) in bss_index_vec.iter().enumerate() {
for (bss_iter_index, &(bss_section_index, section_name)) in
bss_index_vec.iter().enumerate()
{
let shdr = &shdrs[bss_section_index];
let bss_elf_index = linker.load_section(
shdr,
section_name,
(&data[shdr.sh_offset as usize..(shdr.sh_offset + shdr.sh_size) as usize])
data[shdr.sh_offset as usize..(shdr.sh_offset + shdr.sh_size) as usize]
.to_vec(),
);
linker.section_map.insert(bss_section_index, bss_elf_index);
@ -1331,7 +1347,7 @@ impl<'a> Linker<'a> {
// Prepare a STRTAB to hold the names of section headers
// Fix the sh_name field of the section headers
let mut shstrtab = Vec::new();
for shdr_rec in linker.elf_shdrs.iter_mut() {
for shdr_rec in &mut linker.elf_shdrs {
let shstrtab_index = shstrtab.len();
shstrtab.extend(shdr_rec.name.as_bytes());
shstrtab.push(0);
@ -1372,20 +1388,17 @@ impl<'a> Linker<'a> {
let alignment = (4 - (linker.image.len() % 4)) % 4;
let sec_headers_offset = linker.image.len() + alignment;
linker.image.extend(vec![0; alignment]);
for rec in linker.elf_shdrs.iter() {
for rec in &linker.elf_shdrs {
let shdr = rec.shdr;
linker.image.extend(unsafe {
slice::from_raw_parts(
&shdr as *const Elf32_Shdr as *const u8,
mem::size_of::<Elf32_Shdr>(),
)
slice::from_raw_parts(ptr::addr_of!(shdr).cast(), mem::size_of::<Elf32_Shdr>())
});
}
// Update the PHDRs
let phdr_offset = mem::size_of::<Elf32_Ehdr>();
unsafe {
let phdr_ptr = linker.image.as_mut_ptr().add(phdr_offset) as *mut Elf32_Phdr;
let phdr_ptr = linker.image.as_mut_ptr().add(phdr_offset).cast();
let phdr_slice = slice::from_raw_parts_mut(phdr_ptr, 5);
// List of program headers:
// 1. ELF headers & program headers
@ -1462,9 +1475,9 @@ impl<'a> Linker<'a> {
}
// Update the EHDR
let ehdr_ptr = linker.image.as_mut_ptr() as *mut Elf32_Ehdr;
let ehdr_ptr = linker.image.as_mut_ptr().cast();
unsafe {
(*ehdr_ptr) = Elf32_Ehdr {
*ehdr_ptr = Elf32_Ehdr {
e_ident: ehdr.e_ident,
e_type: ET_DYN,
e_machine: ehdr.e_machine,

View File

@ -5,20 +5,20 @@ description = "Parser for python code."
authors = [ "RustPython Team", "M-Labs" ]
build = "build.rs"
license = "MIT"
edition = "2018"
edition = "2021"
[build-dependencies]
lalrpop = "0.19"
lalrpop = "0.20"
[dependencies]
nac3ast = { path = "../nac3ast" }
lalrpop-util = "0.19"
lalrpop-util = "0.20"
log = "0.4"
unic-emoji-char = "0.9"
unic-ucd-ident = "0.9"
unicode_names2 = "0.5"
unicode_names2 = "1.2"
phf = { version = "0.11", features = ["macros"] }
ahash = "0.7"
ahash = "0.8"
[dev-dependencies]
insta = "=1.11.0"

View File

@ -1,15 +1,15 @@
use lalrpop_util::ParseError;
use nac3ast::*;
use crate::ast::Ident;
use crate::ast::Location;
use crate::token::Tok;
use crate::error::*;
use crate::token::Tok;
use lalrpop_util::ParseError;
use nac3ast::*;
pub fn make_config_comment(
com_loc: Location,
stmt_loc: Location,
nac3com_above: Vec<(Ident, Tok)>,
nac3com_end: Option<Ident>
nac3com_end: Option<Ident>,
) -> Result<Vec<Ident>, ParseError<Location, Tok, LexicalError>> {
if com_loc.column() != stmt_loc.column() && !nac3com_above.is_empty() {
return Err(ParseError::User {
@ -17,24 +17,25 @@ pub fn make_config_comment(
location: com_loc,
error: LexicalErrorType::OtherError(
format!(
"config comment at top must have the same indentation with what it applies (comment at {}, statement at {})",
com_loc,
stmt_loc,
"config comment at top must have the same indentation with what it applies (comment at {com_loc}, statement at {stmt_loc})",
)
)
}
})
});
};
Ok(
nac3com_above
Ok(nac3com_above
.into_iter()
.map(|(com, _)| com)
.chain(nac3com_end.map_or_else(|| vec![].into_iter(), |com| vec![com].into_iter()))
.collect()
)
.collect())
}
pub fn handle_small_stmt<U>(stmts: &mut [Stmt<U>], nac3com_above: Vec<(Ident, Tok)>, nac3com_end: Option<Ident>, com_above_loc: Location) -> Result<(), ParseError<Location, Tok, LexicalError>> {
pub fn handle_small_stmt<U>(
stmts: &mut [Stmt<U>],
nac3com_above: Vec<(Ident, Tok)>,
nac3com_end: Option<Ident>,
com_above_loc: Location,
) -> Result<(), ParseError<Location, Tok, LexicalError>> {
if com_above_loc.column() != stmts[0].location.column() && !nac3com_above.is_empty() {
return Err(ParseError::User {
error: LexicalError {
@ -47,17 +48,12 @@ pub fn handle_small_stmt<U>(stmts: &mut [Stmt<U>], nac3com_above: Vec<(Ident, To
)
)
}
})
});
}
apply_config_comments(
&mut stmts[0],
nac3com_above
.into_iter()
.map(|(com, _)| com).collect()
);
apply_config_comments(&mut stmts[0], nac3com_above.into_iter().map(|(com, _)| com).collect());
apply_config_comments(
stmts.last_mut().unwrap(),
nac3com_end.map_or_else(Vec::new, |com| vec![com])
nac3com_end.map_or_else(Vec::new, |com| vec![com]),
);
Ok(())
}
@ -80,6 +76,8 @@ fn apply_config_comments<U>(stmt: &mut Stmt<U>, comments: Vec<Ident>) {
| StmtKind::Nonlocal { config_comment, .. }
| StmtKind::Assert { config_comment, .. } => config_comment.extend(comments),
_ => { unreachable!("only small statements should call this function") }
_ => {
unreachable!("only small statements should call this function")
}
}
}

View File

@ -37,7 +37,7 @@ impl fmt::Display for LexicalErrorType {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
LexicalErrorType::StringError => write!(f, "Got unexpected string"),
LexicalErrorType::FStringError(error) => write!(f, "Got error in f-string: {}", error),
LexicalErrorType::FStringError(error) => write!(f, "Got error in f-string: {error}"),
LexicalErrorType::UnicodeError => write!(f, "Got unexpected unicode"),
LexicalErrorType::NestingError => write!(f, "Got unexpected nesting"),
LexicalErrorType::IndentationError => {
@ -59,13 +59,13 @@ impl fmt::Display for LexicalErrorType {
write!(f, "positional argument follows keyword argument")
}
LexicalErrorType::UnrecognizedToken { tok } => {
write!(f, "Got unexpected token {}", tok)
write!(f, "Got unexpected token {tok}")
}
LexicalErrorType::LineContinuationError => {
write!(f, "unexpected character after line continuation character")
}
LexicalErrorType::Eof => write!(f, "unexpected EOF while parsing"),
LexicalErrorType::OtherError(msg) => write!(f, "{}", msg),
LexicalErrorType::OtherError(msg) => write!(f, "{msg}"),
}
}
}
@ -96,7 +96,7 @@ impl fmt::Display for FStringErrorType {
FStringErrorType::UnopenedRbrace => write!(f, "Unopened '}}'"),
FStringErrorType::ExpectedRbrace => write!(f, "Expected '}}' after conversion flag."),
FStringErrorType::InvalidExpression(error) => {
write!(f, "Invalid expression: {}", error)
write!(f, "Invalid expression: {error}")
}
FStringErrorType::InvalidConversionFlag => write!(f, "Invalid conversion flag"),
FStringErrorType::EmptyExpression => write!(f, "Empty expression"),
@ -144,36 +144,27 @@ pub enum ParseErrorType {
impl From<LalrpopError<Location, Tok, LexicalError>> for ParseError {
fn from(err: LalrpopError<Location, Tok, LexicalError>) -> Self {
match err {
// TODO: Are there cases where this isn't an EOF?
LalrpopError::InvalidToken { location } => ParseError {
error: ParseErrorType::Eof,
location,
},
LalrpopError::ExtraToken { token } => ParseError {
error: ParseErrorType::ExtraToken(token.1),
location: token.0,
},
LalrpopError::User { error } => ParseError {
error: ParseErrorType::Lexical(error.error),
location: error.location,
},
LalrpopError::ExtraToken { token } => {
ParseError { error: ParseErrorType::ExtraToken(token.1), location: token.0 }
}
LalrpopError::User { error } => {
ParseError { error: ParseErrorType::Lexical(error.error), location: error.location }
}
LalrpopError::UnrecognizedToken { token, expected } => {
// Hacky, but it's how CPython does it. See PyParser_AddToken,
// in particular "Only one possible expected token" comment.
let expected = if expected.len() == 1 {
Some(expected[0].clone())
} else {
None
};
let expected = if expected.len() == 1 { Some(expected[0].clone()) } else { None };
ParseError {
error: ParseErrorType::UnrecognizedToken(token.1, expected),
location: token.0,
}
}
LalrpopError::UnrecognizedEOF { location, .. } => ParseError {
error: ParseErrorType::Eof,
location,
},
LalrpopError::UnrecognizedEof { location, .. }
// TODO: Are there cases where this isn't an EOF?
| LalrpopError::InvalidToken { location } => {
ParseError { error: ParseErrorType::Eof, location }
}
}
}
}
@ -188,7 +179,7 @@ impl fmt::Display for ParseErrorType {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match *self {
ParseErrorType::Eof => write!(f, "Got unexpected EOF"),
ParseErrorType::ExtraToken(ref tok) => write!(f, "Got extraneous token: {:?}", tok),
ParseErrorType::ExtraToken(ref tok) => write!(f, "Got extraneous token: {tok:?}"),
ParseErrorType::InvalidToken => write!(f, "Got invalid token"),
ParseErrorType::UnrecognizedToken(ref tok, ref expected) => {
if *tok == Tok::Indent {
@ -196,10 +187,10 @@ impl fmt::Display for ParseErrorType {
} else if expected.as_deref() == Some("Indent") {
write!(f, "expected an indented block")
} else {
write!(f, "Got unexpected token {}", tok)
write!(f, "Got unexpected token {tok}")
}
}
ParseErrorType::Lexical(ref error) => write!(f, "{}", error),
ParseErrorType::Lexical(ref error) => write!(f, "{error}"),
}
}
}
@ -207,6 +198,7 @@ impl fmt::Display for ParseErrorType {
impl Error for ParseErrorType {}
impl ParseErrorType {
#[must_use]
pub fn is_indentation_error(&self) -> bool {
match self {
ParseErrorType::Lexical(LexicalErrorType::IndentationError) => true,
@ -216,11 +208,11 @@ impl ParseErrorType {
_ => false,
}
}
#[must_use]
pub fn is_tab_error(&self) -> bool {
matches!(
self,
ParseErrorType::Lexical(LexicalErrorType::TabError)
| ParseErrorType::Lexical(LexicalErrorType::TabsAfterSpaces)
ParseErrorType::Lexical(LexicalErrorType::TabError | LexicalErrorType::TabsAfterSpaces)
)
}
}

View File

@ -15,10 +15,7 @@ struct FStringParser<'a> {
impl<'a> FStringParser<'a> {
fn new(source: &'a str, str_location: Location) -> Self {
Self {
chars: source.chars().peekable(),
str_location,
}
Self { chars: source.chars().peekable(), str_location }
}
#[inline]
@ -133,10 +130,10 @@ impl<'a> FStringParser<'a> {
)
} else {
Box::new(self.expr(ExprKind::Constant {
value: spec_expression.to_owned().into(),
value: spec_expression.clone().into(),
kind: None,
}))
})
});
}
'(' | '{' | '[' => {
expression.push(ch);
@ -251,17 +248,11 @@ impl<'a> FStringParser<'a> {
}
if !content.is_empty() {
values.push(self.expr(ExprKind::Constant {
value: content.into(),
kind: None,
}))
values.push(self.expr(ExprKind::Constant { value: content.into(), kind: None }));
}
let s = match values.len() {
0 => self.expr(ExprKind::Constant {
value: String::new().into(),
kind: None,
}),
0 => self.expr(ExprKind::Constant { value: String::new().into(), kind: None }),
1 => values.into_iter().next().unwrap(),
_ => self.expr(ExprKind::JoinedStr { values }),
};
@ -270,16 +261,14 @@ impl<'a> FStringParser<'a> {
}
fn parse_fstring_expr(source: &str) -> Result<Expr, ParseError> {
let fstring_body = format!("({})", source);
let fstring_body = format!("({source})");
parse_expression(&fstring_body)
}
/// Parse an fstring from a string, located at a certain position in the sourcecode.
/// In case of errors, we will get the location and the error returned.
pub fn parse_located_fstring(source: &str, location: Location) -> Result<Expr, FStringError> {
FStringParser::new(source, location)
.parse()
.map_err(|error| FStringError { error, location })
FStringParser::new(source, location).parse().map_err(|error| FStringError { error, location })
}
#[cfg(test)]
@ -293,7 +282,7 @@ mod tests {
#[test]
fn test_parse_fstring() {
let source = "{a}{ b }{{foo}}";
let parse_ast = parse_fstring(&source).unwrap();
let parse_ast = parse_fstring(source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -301,7 +290,7 @@ mod tests {
#[test]
fn test_parse_fstring_nested_spec() {
let source = "{foo:{spec}}";
let parse_ast = parse_fstring(&source).unwrap();
let parse_ast = parse_fstring(source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -309,7 +298,7 @@ mod tests {
#[test]
fn test_parse_fstring_not_nested_spec() {
let source = "{foo:spec}";
let parse_ast = parse_fstring(&source).unwrap();
let parse_ast = parse_fstring(source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -322,7 +311,7 @@ mod tests {
#[test]
fn test_fstring_parse_selfdocumenting_base() {
let src = "{user=}";
let parse_ast = parse_fstring(&src).unwrap();
let parse_ast = parse_fstring(src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -330,7 +319,7 @@ mod tests {
#[test]
fn test_fstring_parse_selfdocumenting_base_more() {
let src = "mix {user=} with text and {second=}";
let parse_ast = parse_fstring(&src).unwrap();
let parse_ast = parse_fstring(src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -338,7 +327,7 @@ mod tests {
#[test]
fn test_fstring_parse_selfdocumenting_format() {
let src = "{user=:>10}";
let parse_ast = parse_fstring(&src).unwrap();
let parse_ast = parse_fstring(src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -371,35 +360,35 @@ mod tests {
#[test]
fn test_parse_fstring_not_equals() {
let source = "{1 != 2}";
let parse_ast = parse_fstring(&source).unwrap();
let parse_ast = parse_fstring(source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_equals() {
let source = "{42 == 42}";
let parse_ast = parse_fstring(&source).unwrap();
let parse_ast = parse_fstring(source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_selfdoc_prec_space() {
let source = "{x =}";
let parse_ast = parse_fstring(&source).unwrap();
let parse_ast = parse_fstring(source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_selfdoc_trailing_space() {
let source = "{x= }";
let parse_ast = parse_fstring(&source).unwrap();
let parse_ast = parse_fstring(source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_yield_expr() {
let source = "{yield}";
let parse_ast = parse_fstring(&source).unwrap();
let parse_ast = parse_fstring(source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
}

View File

@ -54,8 +54,7 @@ pub fn parse_args(func_args: Vec<FunctionArgument>) -> Result<ArgumentList, Lexi
let mut keyword_names = HashSet::with_capacity_and_hasher(func_args.len(), RandomState::new());
for (name, value) in func_args {
match name {
Some((location, name)) => {
if let Some((location, name)) = name {
if let Some(keyword_name) = &name {
if keyword_names.contains(keyword_name) {
return Err(LexicalError {
@ -69,13 +68,9 @@ pub fn parse_args(func_args: Vec<FunctionArgument>) -> Result<ArgumentList, Lexi
keywords.push(ast::Keyword::new(
location,
ast::KeywordData {
arg: name.map(|name| name.into()),
value: Box::new(value),
},
ast::KeywordData { arg: name.map(String::into), value: Box::new(value) },
));
}
None => {
} else {
// Allow starred args after keyword arguments.
if !keywords.is_empty() && !is_starred(&value) {
return Err(LexicalError {
@ -87,7 +82,6 @@ pub fn parse_args(func_args: Vec<FunctionArgument>) -> Result<ArgumentList, Lexi
args.push(value);
}
}
}
Ok(ArgumentList { args, keywords })
}

View File

@ -3,12 +3,12 @@
//! This means source code is translated into separate tokens.
pub use super::token::Tok;
use crate::ast::{Location, FileName};
use crate::ast::{FileName, Location};
use crate::error::{LexicalError, LexicalErrorType};
use std::char;
use std::cmp::Ordering;
use std::str::FromStr;
use std::num::IntErrorKind;
use std::str::FromStr;
use unic_emoji_char::is_emoji_presentation;
use unic_ucd_ident::{is_xid_continue, is_xid_start};
@ -32,20 +32,14 @@ impl IndentationLevel {
if self.spaces <= other.spaces {
Ok(Ordering::Less)
} else {
Err(LexicalError {
location,
error: LexicalErrorType::TabError,
})
Err(LexicalError { location, error: LexicalErrorType::TabError })
}
}
Ordering::Greater => {
if self.spaces >= other.spaces {
Ok(Ordering::Greater)
} else {
Err(LexicalError {
location,
error: LexicalErrorType::TabError,
})
Err(LexicalError { location, error: LexicalErrorType::TabError })
}
}
Ordering::Equal => Ok(self.spaces.cmp(&other.spaces)),
@ -63,7 +57,7 @@ pub struct Lexer<T: Iterator<Item = char>> {
chr1: Option<char>,
chr2: Option<char>,
location: Location,
config_comment_prefix: Option<&'static str>
config_comment_prefix: Option<&'static str>,
}
pub static KEYWORDS: phf::Map<&'static str, Tok> = phf::phf_map! {
@ -136,11 +130,7 @@ where
T: Iterator<Item = char>,
{
pub fn new(source: T) -> Self {
let mut nlh = NewlineHandler {
source,
chr0: None,
chr1: None,
};
let mut nlh = NewlineHandler { source, chr0: None, chr1: None };
nlh.shift();
nlh.shift();
nlh
@ -169,7 +159,7 @@ where
self.shift();
} else {
// Transform MAC EOL into \n
self.chr0 = Some('\n')
self.chr0 = Some('\n');
}
} else {
break;
@ -189,13 +179,13 @@ where
chars: input,
at_begin_of_line: true,
nesting: 0,
indentation_stack: vec![Default::default()],
indentation_stack: vec![IndentationLevel::default()],
pending: Vec::new(),
chr0: None,
location: start,
chr1: None,
chr2: None,
config_comment_prefix: Some(" nac3:")
config_comment_prefix: Some(" nac3:"),
};
lxr.next_char();
lxr.next_char();
@ -217,11 +207,9 @@ where
let mut saw_f = false;
loop {
// Detect r"", f"", b"" and u""
if !(saw_b || saw_u || saw_f) && matches!(self.chr0, Some('b') | Some('B')) {
if !(saw_b || saw_u || saw_f) && matches!(self.chr0, Some('b' | 'B')) {
saw_b = true;
} else if !(saw_b || saw_r || saw_u || saw_f)
&& matches!(self.chr0, Some('u') | Some('U'))
{
} else if !(saw_b || saw_r || saw_u || saw_f) && matches!(self.chr0, Some('u' | 'U')) {
saw_u = true;
} else if !(saw_r || saw_u) && (self.chr0 == Some('r') || self.chr0 == Some('R')) {
saw_r = true;
@ -287,15 +275,15 @@ where
let end_pos = self.get_pos();
let value = match i128::from_str_radix(&value_text, radix) {
Ok(value) => value,
Err(e) => {
match e.kind() {
Err(e) => match e.kind() {
IntErrorKind::PosOverflow | IntErrorKind::NegOverflow => i128::MAX,
_ => return Err(LexicalError {
error: LexicalErrorType::OtherError(format!("{:?}", e)),
_ => {
return Err(LexicalError {
error: LexicalErrorType::OtherError(format!("{e:?}")),
location: start_pos,
}),
}
})
}
},
};
Ok((start_pos, Tok::Int { value }, end_pos))
}
@ -338,14 +326,7 @@ where
if self.chr0 == Some('j') || self.chr0 == Some('J') {
self.next_char();
let end_pos = self.get_pos();
Ok((
start_pos,
Tok::Complex {
real: 0.0,
imag: value,
},
end_pos,
))
Ok((start_pos, Tok::Complex { real: 0.0, imag: value }, end_pos))
} else {
let end_pos = self.get_pos();
Ok((start_pos, Tok::Float { value }, end_pos))
@ -364,7 +345,7 @@ where
let value = value_text.parse::<i128>().ok();
let nonzero = match value {
Some(value) => value != 0i128,
None => true
None => true,
};
if start_is_zero && nonzero {
return Err(LexicalError {
@ -379,7 +360,7 @@ where
/// Consume a sequence of numbers with the given radix,
/// the digits can be decorated with underscores
/// like this: '1_2_3_4' == '1234'
/// like this: `'1_2_3_4'` == `'1234'`
fn radix_run(&mut self, radix: u32) -> String {
let mut value_text = String::new();
@ -412,7 +393,7 @@ where
2 => matches!(c, Some('0'..='1')),
8 => matches!(c, Some('0'..='7')),
10 => matches!(c, Some('0'..='9')),
16 => matches!(c, Some('0'..='9') | Some('a'..='f') | Some('A'..='F')),
16 => matches!(c, Some('0'..='9' | 'a'..='f' | 'A'..='F')),
other => unimplemented!("Radix not implemented: {}", other),
}
}
@ -420,8 +401,8 @@ where
/// Test if we face '[eE][-+]?[0-9]+'
fn at_exponent(&self) -> bool {
match self.chr0 {
Some('e') | Some('E') => match self.chr1 {
Some('+') | Some('-') => matches!(self.chr2, Some('0'..='9')),
Some('e' | 'E') => match self.chr1 {
Some('+' | '-') => matches!(self.chr2, Some('0'..='9')),
Some('0'..='9') => true,
_ => false,
},
@ -433,19 +414,17 @@ where
fn lex_comment(&mut self) -> Option<Spanned> {
self.next_char();
// if possibly nac3 pseudocomment, special handling for `# nac3:`
let (mut prefix, mut is_comment) = self
.config_comment_prefix
.map_or_else(|| ("".chars(), false), |v| (v.chars(), true));
let (mut prefix, mut is_comment) =
self.config_comment_prefix.map_or_else(|| ("".chars(), false), |v| (v.chars(), true));
// for the correct location of config comment
let mut start_loc = self.location;
start_loc.go_left();
loop {
match self.chr0 {
Some('\n') => return None,
None => return None,
Some('\n') | None => return None,
Some(c) => {
if let (true, Some(p)) = (is_comment, prefix.next()) {
is_comment = is_comment && c == p
is_comment = is_comment && c == p;
} else {
// done checking prefix, if is comment then return the spanned
if is_comment {
@ -460,22 +439,20 @@ where
return Some((
start_loc,
Tok::ConfigComment { content: content.trim().into() },
self.location
self.location,
));
}
}
}
}
self.next_char();
};
}
}
fn unicode_literal(&mut self, literal_number: usize) -> Result<char, LexicalError> {
let mut p: u32 = 0u32;
let unicode_error = LexicalError {
error: LexicalErrorType::UnicodeError,
location: self.get_pos(),
};
let unicode_error =
LexicalError { error: LexicalErrorType::UnicodeError, location: self.get_pos() };
for i in 1..=literal_number {
match self.next_char() {
Some(c) => match c.to_digit(16) {
@ -486,8 +463,8 @@ where
}
}
match p {
0xD800..=0xDFFF => Ok(std::char::REPLACEMENT_CHARACTER),
_ => std::char::from_u32(p).ok_or(unicode_error),
0xD800..=0xDFFF => Ok(char::REPLACEMENT_CHARACTER),
_ => char::from_u32(p).ok_or(unicode_error),
}
}
@ -496,7 +473,7 @@ where
octet_content.push(first);
while octet_content.len() < 3 {
if let Some('0'..='7') = self.chr0 {
octet_content.push(self.next_char().unwrap())
octet_content.push(self.next_char().unwrap());
} else {
break;
}
@ -530,10 +507,8 @@ where
}
}
}
unicode_names2::character(&name).ok_or(LexicalError {
error: LexicalErrorType::UnicodeError,
location: start_pos,
})
unicode_names2::character(&name)
.ok_or(LexicalError { error: LexicalErrorType::UnicodeError, location: start_pos })
}
fn lex_string(
@ -566,7 +541,7 @@ where
} else if is_raw {
string_content.push('\\');
if let Some(c) = self.next_char() {
string_content.push(c)
string_content.push(c);
} else {
return Err(LexicalError {
error: LexicalErrorType::StringError,
@ -599,7 +574,7 @@ where
Some('u') if !is_bytes => string_content.push(self.unicode_literal(4)?),
Some('U') if !is_bytes => string_content.push(self.unicode_literal(8)?),
Some('N') if !is_bytes => {
string_content.push(self.parse_unicode_name()?)
string_content.push(self.parse_unicode_name()?);
}
Some(c) => {
string_content.push('\\');
@ -650,20 +625,15 @@ where
let end_pos = self.get_pos();
let tok = if is_bytes {
Tok::Bytes {
value: string_content.chars().map(|c| c as u8).collect(),
}
Tok::Bytes { value: string_content.chars().map(|c| c as u8).collect() }
} else {
Tok::String {
value: string_content,
is_fstring,
}
Tok::String { value: string_content, is_fstring }
};
Ok((start_pos, tok, end_pos))
}
fn is_identifier_start(&self, c: char) -> bool {
fn is_identifier_start(c: char) -> bool {
match c {
'_' | 'a'..='z' | 'A'..='Z' => true,
'+' | '-' | '*' | '/' | '=' | ' ' | '<' | '>' => false,
@ -835,18 +805,14 @@ where
// Check if we have some character:
if let Some(c) = self.chr0 {
// First check identifier:
if self.is_identifier_start(c) {
if Self::is_identifier_start(c) {
let identifier = self.lex_identifier()?;
self.emit(identifier);
} else if is_emoji_presentation(c) {
let tok_start = self.get_pos();
self.next_char();
let tok_end = self.get_pos();
self.emit((
tok_start,
Tok::Name { name: c.to_string().into() },
tok_end,
));
self.emit((tok_start, Tok::Name { name: c.to_string().into() }, tok_end));
} else {
self.consume_character(c)?;
}
@ -899,18 +865,15 @@ where
'=' => {
let tok_start = self.get_pos();
self.next_char();
match self.chr0 {
Some('=') => {
if let Some('=') = self.chr0 {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::EqEqual, tok_end));
}
_ => {
} else {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::Equal, tok_end));
}
}
}
'+' => {
let tok_start = self.get_pos();
self.next_char();
@ -934,18 +897,15 @@ where
}
Some('*') => {
self.next_char();
match self.chr0 {
Some('=') => {
if let Some('=') = self.chr0 {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::DoubleStarEqual, tok_end));
}
_ => {
} else {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::DoubleStar, tok_end));
}
}
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::Star, tok_end));
@ -963,18 +923,15 @@ where
}
Some('/') => {
self.next_char();
match self.chr0 {
Some('=') => {
if let Some('=') = self.chr0 {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::DoubleSlashEqual, tok_end));
}
_ => {
} else {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::DoubleSlash, tok_end));
}
}
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::Slash, tok_end));
@ -1141,18 +1098,15 @@ where
match self.chr0 {
Some('<') => {
self.next_char();
match self.chr0 {
Some('=') => {
if let Some('=') = self.chr0 {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::LeftShiftEqual, tok_end));
}
_ => {
} else {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::LeftShift, tok_end));
}
}
}
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
@ -1170,18 +1124,15 @@ where
match self.chr0 {
Some('>') => {
self.next_char();
match self.chr0 {
Some('=') => {
if let Some('=') = self.chr0 {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::RightShiftEqual, tok_end));
}
_ => {
} else {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::RightShift, tok_end));
}
}
}
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
@ -1333,13 +1284,14 @@ where
#[cfg(test)]
mod tests {
use super::{make_tokenizer, NewlineHandler, Tok};
use nac3ast::FileName;
const WINDOWS_EOL: &str = "\r\n";
const MAC_EOL: &str = "\r";
const UNIX_EOL: &str = "\n";
pub fn lex_source(source: &str) -> Vec<Tok> {
let lexer = make_tokenizer(source, Default::default());
let lexer = make_tokenizer(source, FileName::default());
lexer.map(|x| x.unwrap().1).collect()
}
@ -1419,7 +1371,7 @@ class Foo(A, B):
Dedent,
Dedent
]
)
);
}
#[test]
@ -1439,14 +1391,8 @@ class Foo(A, B):
assert_eq!(
tokens,
vec![
Tok::String {
value: "\\\\".to_owned(),
is_fstring: false,
},
Tok::String {
value: "\\".to_owned(),
is_fstring: false,
},
Tok::String { value: "\\\\".to_owned(), is_fstring: false },
Tok::String { value: "\\".to_owned(), is_fstring: false },
Tok::Newline,
]
);
@ -1459,27 +1405,13 @@ class Foo(A, B):
assert_eq!(
tokens,
vec![
Tok::Int {
value: 47i128,
},
Tok::Int {
value: 13i128,
},
Tok::Int {
value: 0i128,
},
Tok::Int {
value: 123i128,
},
Tok::Int { value: 47i128 },
Tok::Int { value: 13i128 },
Tok::Int { value: 0i128 },
Tok::Int { value: 123i128 },
Tok::Float { value: 0.2 },
Tok::Complex {
real: 0.0,
imag: 2.0,
},
Tok::Complex {
real: 0.0,
imag: 2.2,
},
Tok::Complex { real: 0.0, imag: 2.0 },
Tok::Complex { real: 0.0, imag: 2.2 },
Tok::Newline,
]
);
@ -1539,21 +1471,13 @@ class Foo(A, B):
assert_eq!(
tokens,
vec![
Tok::Name {
name: String::from("avariable").into(),
},
Tok::Name { name: String::from("avariable").into() },
Tok::Equal,
Tok::Int {
value: 99i128
},
Tok::Int { value: 99i128 },
Tok::Plus,
Tok::Int {
value: 2i128
},
Tok::Int { value: 2i128 },
Tok::Minus,
Tok::Int {
value: 0i128
},
Tok::Int { value: 0i128 },
Tok::Newline,
]
);
@ -1740,42 +1664,15 @@ class Foo(A, B):
assert_eq!(
tokens,
vec![
Tok::String {
value: String::from("double"),
is_fstring: false,
},
Tok::String {
value: String::from("single"),
is_fstring: false,
},
Tok::String {
value: String::from("can't"),
is_fstring: false,
},
Tok::String {
value: String::from("\\\""),
is_fstring: false,
},
Tok::String {
value: String::from("\t\r\n"),
is_fstring: false,
},
Tok::String {
value: String::from("\\g"),
is_fstring: false,
},
Tok::String {
value: String::from("raw\\'"),
is_fstring: false,
},
Tok::String {
value: String::from("Đ"),
is_fstring: false,
},
Tok::String {
value: String::from("\u{80}\u{0}a"),
is_fstring: false,
},
Tok::String { value: String::from("double"), is_fstring: false },
Tok::String { value: String::from("single"), is_fstring: false },
Tok::String { value: String::from("can't"), is_fstring: false },
Tok::String { value: String::from("\\\""), is_fstring: false },
Tok::String { value: String::from("\t\r\n"), is_fstring: false },
Tok::String { value: String::from("\\g"), is_fstring: false },
Tok::String { value: String::from("raw\\'"), is_fstring: false },
Tok::String { value: String::from("Đ"), is_fstring: false },
Tok::String { value: String::from("\u{80}\u{0}a"), is_fstring: false },
Tok::Newline,
]
);
@ -1830,7 +1727,7 @@ class Foo(A, B):
#[test]
fn test_escape_char_in_byte_literal() {
// backslash does not escape
let source = r##"b"omkmok\Xaa""##;
let source = r#"b"omkmok\Xaa""#;
let tokens = lex_source(source);
let res = vec![111, 109, 107, 109, 111, 107, 92, 88, 97, 97];
assert_eq!(tokens, vec![Tok::Bytes { value: res }, Tok::Newline]);
@ -1840,41 +1737,17 @@ class Foo(A, B):
fn test_raw_byte_literal() {
let source = r"rb'\x1z'";
let tokens = lex_source(source);
assert_eq!(
tokens,
vec![
Tok::Bytes {
value: b"\\x1z".to_vec()
},
Tok::Newline
]
);
assert_eq!(tokens, vec![Tok::Bytes { value: b"\\x1z".to_vec() }, Tok::Newline]);
let source = r"rb'\\'";
let tokens = lex_source(source);
assert_eq!(
tokens,
vec![
Tok::Bytes {
value: b"\\\\".to_vec()
},
Tok::Newline
]
)
assert_eq!(tokens, vec![Tok::Bytes { value: b"\\\\".to_vec() }, Tok::Newline]);
}
#[test]
fn test_escape_octet() {
let source = r##"b'\43a\4\1234'"##;
let source = r"b'\43a\4\1234'";
let tokens = lex_source(source);
assert_eq!(
tokens,
vec![
Tok::Bytes {
value: b"#a\x04S4".to_vec()
},
Tok::Newline
]
)
assert_eq!(tokens, vec![Tok::Bytes { value: b"#a\x04S4".to_vec() }, Tok::Newline]);
}
#[test]
@ -1883,13 +1756,7 @@ class Foo(A, B):
let tokens = lex_source(source);
assert_eq!(
tokens,
vec![
Tok::String {
value: "\u{2002}".to_owned(),
is_fstring: false,
},
Tok::Newline
]
)
vec![Tok::String { value: "\u{2002}".to_owned(), is_fstring: false }, Tok::Newline]
);
}
}

View File

@ -15,6 +15,24 @@
//!
//! ```
#![deny(
future_incompatible,
let_underscore,
nonstandard_style,
rust_2024_compatibility,
clippy::all
)]
#![warn(clippy::pedantic)]
#![allow(
clippy::enum_glob_use,
clippy::fn_params_excessive_bools,
clippy::missing_errors_doc,
clippy::missing_panics_doc,
clippy::module_name_repetitions,
clippy::too_many_lines,
clippy::wildcard_imports
)]
#[macro_use]
extern crate log;
use lalrpop_util::lalrpop_mod;
@ -27,9 +45,16 @@ pub mod lexer;
pub mod mode;
pub mod parser;
lalrpop_mod!(
#[allow(clippy::all)]
#[allow(unused)]
#[allow(
future_incompatible,
let_underscore,
nonstandard_style,
rust_2024_compatibility,
unused,
clippy::all,
clippy::pedantic
)]
python
);
pub mod token;
pub mod config_comment_helper;
pub mod token;

View File

@ -5,6 +5,7 @@
//! parse a whole program, a single statement, or a single
//! expression.
use nac3ast::Location;
use std::iter;
use crate::ast::{self, FileName};
@ -63,7 +64,7 @@ pub fn parse_program(source: &str, file: FileName) -> Result<ast::Suite, ParseEr
///
/// ```
pub fn parse_expression(source: &str) -> Result<ast::Expr, ParseError> {
parse(source, Mode::Expression, Default::default()).map(|top| match top {
parse(source, Mode::Expression, FileName::default()).map(|top| match top {
ast::Mod::Expression { body } => *body,
_ => unreachable!(),
})
@ -72,12 +73,10 @@ pub fn parse_expression(source: &str) -> Result<ast::Expr, ParseError> {
// Parse a given source code
pub fn parse(source: &str, mode: Mode, file: FileName) -> Result<ast::Mod, ParseError> {
let lxr = lexer::make_tokenizer(source, file);
let marker_token = (Default::default(), mode.to_marker(), Default::default());
let marker_token = (Location::default(), mode.to_marker(), Location::default());
let tokenizer = iter::once(Ok(marker_token)).chain(lxr);
python::TopParser::new()
.parse(tokenizer)
.map_err(ParseError::from)
python::TopParser::new().parse(tokenizer).map_err(ParseError::from)
}
#[cfg(test)]
@ -86,42 +85,42 @@ mod tests {
#[test]
fn test_parse_empty() {
let parse_ast = parse_program("", Default::default()).unwrap();
let parse_ast = parse_program("", FileName::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_print_hello() {
let source = String::from("print('Hello world')");
let parse_ast = parse_program(&source, Default::default()).unwrap();
let parse_ast = parse_program(&source, FileName::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_print_2() {
let source = String::from("print('Hello world', 2)");
let parse_ast = parse_program(&source, Default::default()).unwrap();
let parse_ast = parse_program(&source, FileName::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_kwargs() {
let source = String::from("my_func('positional', keyword=2)");
let parse_ast = parse_program(&source, Default::default()).unwrap();
let parse_ast = parse_program(&source, FileName::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_if_elif_else() {
let source = String::from("if 1: 10\nelif 2: 20\nelse: 30");
let parse_ast = parse_program(&source, Default::default()).unwrap();
let parse_ast = parse_program(&source, FileName::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_lambda() {
let source = "lambda x, y: x * y"; // lambda(x, y): x * y";
let parse_ast = parse_program(source, Default::default()).unwrap();
let parse_ast = parse_program(source, FileName::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -129,7 +128,7 @@ mod tests {
fn test_parse_tuples() {
let source = "a, b = 4, 5";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
}
#[test]
@ -140,7 +139,7 @@ class Foo(A, B):
pass
def method_with_default(self, arg='default'):
pass";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
}
#[test]
@ -183,7 +182,7 @@ while i < 2: # nac3: 4
# nac3: if1
if 1: # nac3: if2
3";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
}
#[test]
@ -196,7 +195,7 @@ while test: # nac3: while3
# nac3: simple assign0
a = 3 # nac3: simple assign1
";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
}
#[test]
@ -215,7 +214,7 @@ if a: # nac3: small2
for i in a: # nac3: for1
pass
";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
}
#[test]
@ -224,6 +223,6 @@ for i in a: # nac3: for1
if a: # nac3: something
a = 3
";
assert!(parse_program(source, Default::default()).is_err());
assert!(parse_program(source, FileName::default()).is_err());
}
}

View File

@ -1,7 +1,7 @@
//! Different token definitions.
//! Loosely based on token.h from CPython source:
use std::fmt::{self, Write};
use crate::ast;
use std::fmt::{self, Write};
/// Python source code can be tokenized in a sequence of these tokens.
#[derive(Clone, Debug, PartialEq)]
@ -111,15 +111,23 @@ impl fmt::Display for Tok {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
use Tok::*;
match self {
Name { name } => write!(f, "'{}'", ast::get_str_from_ref(&ast::get_str_ref_lock(), *name)),
Int { value } => if *value != i128::MAX { write!(f, "'{}'", value) } else { write!(f, "'#OFL#'") },
Float { value } => write!(f, "'{}'", value),
Complex { real, imag } => write!(f, "{}j{}", real, imag),
Name { name } => {
write!(f, "'{}'", ast::get_str_from_ref(&ast::get_str_ref_lock(), *name))
}
Int { value } => {
if *value == i128::MAX {
write!(f, "'#OFL#'")
} else {
write!(f, "'{value}'")
}
}
Float { value } => write!(f, "'{value}'"),
Complex { real, imag } => write!(f, "{real}j{imag}"),
String { value, is_fstring } => {
if *is_fstring {
write!(f, "f")?
write!(f, "f")?;
}
write!(f, "{:?}", value)
write!(f, "{value:?}")
}
Bytes { value } => {
write!(f, "b\"")?;
@ -129,12 +137,16 @@ impl fmt::Display for Tok {
10 => f.write_str("\\n")?,
13 => f.write_str("\\r")?,
32..=126 => f.write_char(*i as char)?,
_ => write!(f, "\\x{:02x}", i)?,
_ => write!(f, "\\x{i:02x}")?,
}
}
f.write_str("\"")
}
ConfigComment { content } => write!(f, "ConfigComment: '{}'", ast::get_str_from_ref(&ast::get_str_ref_lock(), *content)),
ConfigComment { content } => write!(
f,
"ConfigComment: '{}'",
ast::get_str_from_ref(&ast::get_str_ref_lock(), *content)
),
Newline => f.write_str("Newline"),
Indent => f.write_str("Indent"),
Dedent => f.write_str("Dedent"),

View File

@ -2,14 +2,18 @@
name = "nac3standalone"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2018"
edition = "2021"
[dependencies]
parking_lot = "0.12"
nac3parser = { path = "../nac3parser" }
nac3core = { path = "../nac3core" }
[dependencies.clap]
version = "4.5"
features = ["derive"]
[dependencies.inkwell]
git = "https://github.com/TheDan64/inkwell.git"
version = "0.4"
default-features = false
features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]

4
nac3standalone/demo/.gitignore vendored Normal file
View File

@ -0,0 +1,4 @@
*.bc
*.ll
*.o
/demo

View File

@ -0,0 +1,25 @@
#!/usr/bin/env bash
set -e
if [ -z "$1" ]; then
echo "Requires at least one argument"
exit 1
fi
declare -a nac3args
while [ $# -gt 1 ]; do
nac3args+=("$1")
shift
done
demo="$1"
echo -n "Checking $demo... "
./interpret_demo.py "$demo" > interpreted.log
./run_demo.sh --out run.log "${nac3args[@]}" "$demo"
./run_demo.sh --lli --out run_lli.log "${nac3args[@]}" "$demo"
diff -Nau interpreted.log run.log
diff -Nau interpreted.log run_lli.log
echo "ok"
rm -f interpreted.log run.log run_lli.log

View File

@ -4,12 +4,8 @@ set -e
count=0
for demo in src/*.py; do
echo -n "checking $demo... "
./interpret_demo.py $demo > interpreted.log
./run_demo.sh $demo > run.log
diff -Nau interpreted.log run.log
echo "ok"
let "count+=1"
./check_demo.sh "$@" "$demo"
((count += 1))
done
echo "Ran $count demo checks - PASSED"

122
nac3standalone/demo/demo.c Normal file
View File

@ -0,0 +1,122 @@
#include <inttypes.h>
#include <math.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#define usize size_t
double dbl_nan(void) {
return NAN;
}
double dbl_inf(void) {
return INFINITY;
}
void output_bool(bool x) {
puts(x ? "True" : "False");
}
void output_int32(int32_t x) {
printf("%"PRId32"\n", x);
}
void output_int64(int64_t x) {
printf("%"PRId64"\n", x);
}
void output_uint32(uint32_t x) {
printf("%"PRIu32"\n", x);
}
void output_uint64(uint64_t x) {
printf("%"PRIu64"\n", x);
}
void output_float64(double x) {
if (isnan(x)) {
puts("nan");
} else {
printf("%f\n", x);
}
}
void output_range(int32_t range[3]) {
printf("range(");
printf("%d, %d", range[0], range[1]);
if (range[2] != 1) {
printf(", %d", range[2]);
}
puts(")");
}
void output_asciiart(int32_t x) {
static const char *chars = " .,-:;i+hHM$*#@ ";
if (x < 0) {
putchar('\n');
} else {
putchar(chars[x]);
}
}
struct cslice {
void *data;
usize len;
};
void output_int32_list(struct cslice *slice) {
const int32_t *data = (int32_t *) slice->data;
putchar('[');
for (usize i = 0; i < slice->len; ++i) {
if (i == slice->len - 1) {
printf("%d", data[i]);
} else {
printf("%d, ", data[i]);
}
}
putchar(']');
putchar('\n');
}
void output_str(struct cslice *slice) {
const char *data = (const char *) slice->data;
for (usize i = 0; i < slice->len; ++i) {
putchar(data[i]);
}
}
void output_strln(struct cslice *slice) {
output_str(slice);
putchar('\n');
}
uint64_t dbg_stack_address(__attribute__((unused)) struct cslice *slice) {
int i;
void *ptr = (void *) &i;
return (uintptr_t) ptr;
}
uint32_t __nac3_personality(uint32_t state, uint32_t exception_object, uint32_t context) {
printf("__nac3_personality(state: %u, exception_object: %u, context: %u)\n", state, exception_object, context);
exit(101);
__builtin_unreachable();
}
uint32_t __nac3_raise(uint32_t state, uint32_t exception_object, uint32_t context) {
printf("__nac3_raise(state: %u, exception_object: %u, context: %u)\n", state, exception_object, context);
exit(101);
__builtin_unreachable();
}
void __nac3_end_catch(void) {}
extern int32_t run(void);
int main(void) {
run();
}

View File

@ -1,90 +0,0 @@
mod cslice {
// copied from https://github.com/dherman/cslice
use std::marker::PhantomData;
use std::slice;
#[repr(C)]
#[derive(Clone, Copy)]
pub struct CSlice<'a, T> {
base: *const T,
len: usize,
marker: PhantomData<&'a ()>,
}
impl<'a, T> AsRef<[T]> for CSlice<'a, T> {
fn as_ref(&self) -> &[T] {
unsafe { slice::from_raw_parts(self.base, self.len) }
}
}
}
#[no_mangle]
pub extern "C" fn output_int32(x: i32) {
println!("{}", x);
}
#[no_mangle]
pub extern "C" fn output_int64(x: i64) {
println!("{}", x);
}
#[no_mangle]
pub extern "C" fn output_uint32(x: u32) {
println!("{}", x);
}
#[no_mangle]
pub extern "C" fn output_uint64(x: u64) {
println!("{}", x);
}
#[no_mangle]
pub extern "C" fn output_float64(x: f64) {
// debug output to preserve the digits after the decimal points
// to match python `print` function
println!("{:?}", x);
}
#[no_mangle]
pub extern "C" fn output_asciiart(x: i32) {
let chars = " .,-:;i+hHM$*#@ ";
if x < 0 {
println!("");
} else {
print!("{}", chars.chars().nth(x as usize).unwrap());
}
}
#[no_mangle]
pub extern "C" fn output_int32_list(x: &cslice::CSlice<i32>) {
print!("[");
let mut it = x.as_ref().iter().peekable();
while let Some(e) = it.next() {
if it.peek().is_none() {
print!("{}", e);
} else {
print!("{}, ", e);
}
}
println!("]");
}
#[no_mangle]
pub extern "C" fn __nac3_personality(_state: u32, _exception_object: u32, _context: u32) -> u32 {
unimplemented!();
}
#[no_mangle]
pub extern "C" fn __nac3_raise(_state: u32, _exception_object: u32, _context: u32) -> u32 {
unimplemented!();
}
extern "C" {
fn run() -> i32;
}
fn main() {
unsafe {
run();
}
}

View File

@ -3,10 +3,14 @@
import sys
import importlib.util
import importlib.machinery
import math
import numpy as np
import numpy.typing as npt
import pathlib
from numpy import int32, int64, uint32, uint64
from typing import TypeVar, Generic
from scipy import special
from typing import TypeVar, Generic, Literal, Union
T = TypeVar('T')
class Option(Generic[T]):
@ -41,26 +45,99 @@ def Some(v: T) -> Option[T]:
none = Option(None)
class _ConstGenericMarker:
pass
def ConstGeneric(name, constraint):
return TypeVar(name, _ConstGenericMarker, constraint)
N = TypeVar("N", bound=np.uint64)
class _NDArrayDummy(Generic[T, N]):
pass
# https://stackoverflow.com/questions/67803260/how-to-create-a-type-alias-with-a-throw-away-generic
NDArray = Union[npt.NDArray[T], _NDArrayDummy[T, N]]
def _bool(x):
if isinstance(x, np.ndarray):
return np.bool_(x)
else:
return bool(x)
def _float(x):
if isinstance(x, np.ndarray):
return np.float_(x)
else:
return float(x)
def round_away_zero(x):
if isinstance(x, np.ndarray):
return np.vectorize(round_away_zero)(x)
else:
if x >= 0.0:
return math.floor(x + 0.5)
else:
return math.ceil(x - 0.5)
def _floor(x):
if isinstance(x, np.ndarray):
return np.vectorize(_floor)(x)
else:
return math.floor(x)
def _ceil(x):
if isinstance(x, np.ndarray):
return np.vectorize(_ceil)(x)
else:
return math.ceil(x)
def patch(module):
def dbl_nan():
return np.nan
def dbl_inf():
return np.inf
def output_asciiart(x):
if x < 0:
sys.stdout.write("\n")
else:
sys.stdout.write(" .,-:;i+hHM$*#@ "[x])
def output_float(x):
print("%f" % x)
def output_strln(x):
print(x, end='')
def dbg_stack_address(_):
return 0
def extern(fun):
name = fun.__name__
if name == "output_asciiart":
if name == "dbl_nan":
return dbl_nan
elif name == "dbl_inf":
return dbl_inf
elif name == "output_asciiart":
return output_asciiart
elif name == "output_float64":
return output_float
elif name == "output_str":
return output_strln
elif name in {
"output_bool",
"output_int32",
"output_int64",
"output_int32_list",
"output_uint32",
"output_uint64",
"output_float64"
"output_strln",
"output_range",
}:
return print
elif name == "dbg_stack_address":
return dbg_stack_address
else:
raise NotImplementedError
@ -68,13 +145,93 @@ def patch(module):
module.int64 = int64
module.uint32 = uint32
module.uint64 = uint64
module.bool = _bool
module.float = _float
module.TypeVar = TypeVar
module.ConstGeneric = ConstGeneric
module.Generic = Generic
module.Literal = Literal
module.extern = extern
module.Option = Option
module.Some = Some
module.none = none
# Builtin Math functions
module.round = round_away_zero
module.round64 = round_away_zero
module.np_round = np.round
module.floor = _floor
module.floor64 = _floor
module.np_floor = np.floor
module.ceil = _ceil
module.ceil64 = _ceil
module.np_ceil = np.ceil
# NumPy ndarray functions
module.ndarray = NDArray
module.np_ndarray = np.ndarray
module.np_empty = np.empty
module.np_zeros = np.zeros
module.np_ones = np.ones
module.np_full = np.full
module.np_eye = np.eye
module.np_identity = np.identity
module.np_array = np.array
# NumPy Math functions
module.np_isnan = np.isnan
module.np_isinf = np.isinf
module.np_min = np.min
module.np_minimum = np.minimum
module.np_max = np.max
module.np_maximum = np.maximum
module.np_sin = np.sin
module.np_cos = np.cos
module.np_exp = np.exp
module.np_exp2 = np.exp2
module.np_log = np.log
module.np_log10 = np.log10
module.np_log2 = np.log2
module.np_fabs = np.fabs
module.np_trunc = np.trunc
module.np_sqrt = np.sqrt
module.np_rint = np.rint
module.np_tan = np.tan
module.np_arcsin = np.arcsin
module.np_arccos = np.arccos
module.np_arctan = np.arctan
module.np_sinh = np.sinh
module.np_cosh = np.cosh
module.np_tanh = np.tanh
module.np_arcsinh = np.arcsinh
module.np_arccosh = np.arccosh
module.np_arctanh = np.arctanh
module.np_expm1 = np.expm1
module.np_cbrt = np.cbrt
module.np_arctan2 = np.arctan2
module.np_copysign = np.copysign
module.np_fmax = np.fmax
module.np_fmin = np.fmin
module.np_ldexp = np.ldexp
module.np_hypot = np.hypot
module.np_nextafter = np.nextafter
# SciPy Math Functions
module.sp_spec_erf = special.erf
module.sp_spec_erfc = special.erfc
module.sp_spec_gamma = special.gamma
module.sp_spec_gammaln = special.gammaln
module.sp_spec_j0 = special.j0
module.sp_spec_j1 = special.j1
# NumPy NDArray Functions
module.np_ndarray = np.ndarray
module.np_empty = np.empty
module.np_zeros = np.zeros
module.np_ones = np.ones
module.np_full = np.full
module.np_eye = np.eye
module.np_identity = np.identity
def file_import(filename, prefix="file_import_"):
filename = pathlib.Path(filename)

View File

@ -7,14 +7,72 @@ if [ -z "$1" ]; then
exit 1
fi
if [ -e ../../target/release/nac3standalone ]; then
declare -a nac3args
while [ $# -ge 1 ]; do
case "$1" in
--help)
echo "Usage: run_demo.sh [--help] [--out OUTFILE] [--lli] [--debug] -- [NAC3ARGS...]"
exit
;;
--out)
shift
outfile="$1"
;;
--lli)
use_lli=1
;;
--debug)
debug=1
;;
--)
shift
break
;;
*)
break
;;
esac
shift
done
while [ $# -ge 1 ]; do
nac3args+=("$1")
shift
done
if [ -n "$debug" ] && [ -e ../../target/debug/nac3standalone ]; then
nac3standalone=../../target/debug/nac3standalone
elif [ -e ../../target/release/nac3standalone ]; then
nac3standalone=../../target/release/nac3standalone
else
# used by Nix builds
nac3standalone=../../target/x86_64-unknown-linux-gnu/release/nac3standalone
fi
rm -f *.o
$nac3standalone $1
rustc -o demo demo.rs -Crelocation-model=static -Clink-arg=./module.o
rm -f ./*.o ./*.bc demo
if [ -z "$use_lli" ]; then
$nac3standalone "${nac3args[@]}"
clang -c -std=gnu11 -Wall -Wextra -O3 -o demo.o demo.c
clang -lm -o demo module.o demo.o
if [ -z "$outfile" ]; then
./demo
else
./demo > "$outfile"
fi
else
$nac3standalone --emit-llvm "${nac3args[@]}"
clang -c -std=gnu11 -Wall -Wextra -O3 -emit-llvm -o demo.bc demo.c
shopt -s nullglob
llvm-link -o nac3out.bc module*.bc main.bc
shopt -u nullglob
if [ -z "$outfile" ]; then
lli --extra-module demo.bc --extra-module irrt.bc nac3out.bc
else
lli --extra-module demo.bc --extra-module irrt.bc nac3out.bc > "$outfile"
fi
fi

View File

@ -0,0 +1,30 @@
# Different cases for using boolean variables in boolean contexts.
# Tests whether all boolean variables (expressed as i8s) are lowered into i1s before used in branching instruction (`br`)
def bfunc(b: bool) -> bool:
return not b
def run() -> int32:
b1 = True
b2 = False
if b1:
pass
if not b2:
pass
while b2:
pass
l = [i for i in range(10) if b2]
b_and = True and False
b_or = True or False
b_and = b1 and b2
b_or = b1 or b2
bfunc(b1)
return 0

View File

@ -6,6 +6,10 @@ def output_int32(x: int32):
def output_int64(x: int64):
...
@extern
def output_strln(x: str):
...
class B:
b: int32
@ -23,10 +27,14 @@ class A:
def get_a(self) -> int32:
return self.a
def get_b(self) -> B:
return self.b
# def get_b(self) -> B:
# return self.b
class Initless:
def foo(self):
output_strln("hello")
def run() -> int32:
a = A(10)
output_int32(a.a)
@ -35,4 +43,8 @@ def run() -> int32:
output_int32(a.a)
output_int32(a.get_a())
# output_int32(a.get_b().b) FIXME: NAC3 prints garbage
initless = Initless()
initless.foo()
return 0

View File

@ -0,0 +1,50 @@
A = ConstGeneric("A", int32)
B = ConstGeneric("B", uint32)
T = TypeVar("T")
class ConstGenericClass(Generic[A]):
def __init__(self):
pass
class ConstGeneric2Class(Generic[A, B]):
def __init__(self):
pass
class HybridGenericClass2(Generic[A, T]):
pass
class HybridGenericClass3(Generic[T, A, B]):
pass
def make_generic_2() -> ConstGenericClass[Literal[2]]:
return ...
def make_generic2_1_2() -> ConstGeneric2Class[Literal[1], Literal[2]]:
return ...
def make_hybrid_class_2_int32() -> HybridGenericClass2[Literal[2], int32]:
return ...
def make_hybrid_class_i32_0_1() -> HybridGenericClass3[int32, Literal[0], Literal[1]]:
return ...
def consume_generic_2(instance: ConstGenericClass[Literal[2]]):
pass
def consume_generic2_1_2(instance: ConstGeneric2Class[Literal[1], Literal[2]]):
pass
def consume_hybrid_class_2_i32(instance: HybridGenericClass2[Literal[2], int32]):
pass
def consume_hybrid_class_i32_0_1(instance: HybridGenericClass3[int32, Literal[0], Literal[1]]):
pass
def f():
consume_generic_2(make_generic_2())
consume_generic2_1_2(make_generic2_1_2())
consume_hybrid_class_2_i32(make_hybrid_class_2_int32())
consume_hybrid_class_i32_0_1(make_hybrid_class_i32_0_1())
def run() -> int32:
return 0

View File

@ -0,0 +1,8 @@
def f():
return
return
def run() -> int32:
f()
return 0

View File

@ -0,0 +1,102 @@
@extern
def output_bool(x: bool):
...
@extern
def output_int32(x: int32):
...
@extern
def output_int64(x: int64):
...
@extern
def output_uint32(x: uint32):
...
@extern
def output_uint64(x: uint64):
...
@extern
def output_float64(x: float):
...
@extern
def output_range(x: range):
...
@extern
def output_int32_list(x: list[int32]):
...
@extern
def output_asciiart(x: int32):
...
@extern
def output_str(x: str):
...
@extern
def output_strln(x: str):
...
def test_output_bool():
output_bool(True)
output_bool(False)
def test_output_int32():
output_int32(-128)
def test_output_int64():
output_int64(int64(-256))
def test_output_uint32():
output_uint32(uint32(128))
def test_output_uint64():
output_uint64(uint64(256))
def test_output_float64():
output_float64(0.0)
output_float64(1.0)
output_float64(-1.0)
output_float64(128.0)
output_float64(-128.0)
output_float64(16.25)
output_float64(-16.25)
def test_output_range():
r = range(1, 100, 5)
output_int32(r.start)
output_int32(r.stop)
output_int32(r.step)
output_range(range(10))
output_range(range(1, 10))
output_range(range(1, 10, 2))
def test_output_asciiart():
for i in range(17):
output_asciiart(i)
output_asciiart(0)
def test_output_int32_list():
output_int32_list([0, 1, 3, 5, 10])
def test_output_str_family():
output_str("hello")
output_strln(" world")
def run() -> int32:
test_output_bool()
test_output_int32()
test_output_int64()
test_output_uint32()
test_output_uint64()
test_output_float64()
test_output_range()
test_output_asciiart()
test_output_int32_list()
test_output_str_family()
return 0

View File

@ -0,0 +1,17 @@
@extern
def output_int32(x: int32):
...
@extern
def output_int32_list(x: list[int32]):
...
def run() -> int32:
bl = [True, False]
bl1 = bl[:]
bl1[1:] = [True]
output_int32_list([int32(b) for b in bl1])
output_int32_list([int32(b) for b in bl1])
return 0

View File

@ -1,3 +1,7 @@
@extern
def output_bool(x: bool):
...
@extern
def output_int32_list(x: list[int32]):
...
@ -30,6 +34,32 @@ def run() -> int32:
get_list_slice()
list_slice_assignment()
output_int32_list([1, 2, 3] + [4, 5, 6])
output_int32_list([1, 2, 3] * 3)
output_bool([] == [])
output_bool([0] == [])
output_bool([0] == [0])
output_bool([0, 1] == [0])
output_bool([0, 1] == [0, 1])
output_bool([] != [])
output_bool([0] != [])
output_bool([0] != [0])
output_bool([0] != [0, 1])
output_bool([0, 1] != [0, 1])
output_bool([] == [] == [])
output_bool([0] == [0] == [0])
output_bool([0, 1] == [0] == [0, 1])
output_bool([0, 1] == [0, 1] == [0])
output_bool([0] == [0, 1] == [0, 1])
output_bool([0, 1] == [0, 1] == [0, 1])
output_bool([] != [] != [])
output_bool([0] != [0] != [0])
output_bool([0, 1] != [0] != [0, 1])
output_bool([0, 1] != [0, 1] != [0])
output_bool([0] != [0, 1] != [0, 1])
output_bool([0, 1] != [0, 1] != [0, 1])
return 0
def get_list_slice():

View File

@ -1,9 +1,12 @@
# For Loop using an increasing range() expression as its iterable
@extern
def output_int32(x: int32):
...
def run() -> int32:
for _ in range(10):
output_int32(_)
_ = 0
i = 0
for i in range(10):
output_int32(i)
output_int32(i)
return 0

Some files were not shown because too many files have changed in this diff Show More