forked from M-Labs/nac3
1
0
Fork 0

Compare commits

...

169 Commits

Author SHA1 Message Date
lyken c845924c20
allow ListObject to have TVar item_type 2024-08-24 12:37:25 +08:00
lyken 2941e8e865
fixup! fixup! core/ndstrides: implement np_array()
fix binop +
2024-08-23 16:35:43 +08:00
lyken 4e3e490b92
fixup! core/object: add ListObject and TupleObject
fix List field items name
2024-08-23 16:27:49 +08:00
lyken aad4fafcba
fixup! core: refactor to use ListObject / List
fix gen_expr list index slice bug
2024-08-23 16:24:18 +08:00
lyken 9e005e9b07
core/model: fix Ptr::copy_from int types 2024-08-23 16:23:16 +08:00
lyken 7e45c104be
core: remove List proxy 2024-08-23 15:34:26 +08:00
lyken d2650e6979
core: refactor to use ListObject / List 2024-08-23 15:33:00 +08:00
lyken 2d799d13e2
core/model: add Int not 2024-08-23 15:28:34 +08:00
lyken 2fa3ada445
fixup! core/ndstrides: implement np_array()
fix ListObject::get_opaque_list_ptr comment
2024-08-23 12:10:51 +08:00
lyken 787fe23202
fixup! core/ndstrides: implement ndarray indexing
fix index comment
2024-08-23 11:50:45 +08:00
lyken 5a893e1c15
core: remove Range proxy 2024-08-23 10:57:31 +08:00
lyken 2b29f7b0f5
core: refactor to use RangeObject / Range 2024-08-23 10:47:30 +08:00
lyken 122f55e615
core/object: add RangeObject & RangeObject::len 2024-08-23 10:22:01 +08:00
lyken c0cace843d
core/object: remove *Object::is_instance
This makes implementations awkward.
2024-08-23 10:21:39 +08:00
lyken 5c68ef00ac
core/irrt: refactor out slice::indices and range::len 2024-08-23 10:02:45 +08:00
lyken 3782791323
core: make IRRT slice and range work for any int types 2024-08-23 09:52:08 +08:00
lyken ac6c7c5985
fixup! artiq: reimplement get_obj_value to use ndarray with strides
update symbol_resolver for general array
2024-08-22 20:57:56 +08:00
lyken 8b0305ab6b
fixup! core/model: introduce models
general array
2024-08-22 20:56:25 +08:00
lyken be4b04dbb3
core: remove old ndarray code and NDArray proxy
Nothing depends on the old ndarray implementation now.
2024-08-22 16:34:27 +08:00
lyken e652919b8e
artiq: reimplement get_obj_value to use ndarray with strides 2024-08-22 16:34:27 +08:00
lyken ef391b81aa
artiq: reimplement polymorphic_print for ndarray 2024-08-22 16:34:27 +08:00
lyken ad5506bff1
artiq: reimplement reformat_rpc_arg for ndarrays 2024-08-22 16:34:27 +08:00
lyken 31931b7b26
core/ndstrides: implement np_size() 2024-08-22 16:34:27 +08:00
lyken be6d704020
standalone: add test_ndarray_broadcast_to 2024-08-22 16:23:01 +08:00
lyken 117110dd91
standalone: add test_ndarray_subscript_assignment 2024-08-22 16:23:01 +08:00
lyken c582ffe53d
core/ndstrides: implement ndarray subscript assignment 2024-08-22 16:23:01 +08:00
lyken 51b1921e05
core/ndstrides: implement cmpop 2024-08-22 16:23:01 +08:00
lyken 50457f19e6
core/ndstrides: implement unary op 2024-08-22 16:23:01 +08:00
lyken 54f7e1edfd
core/ndstrides: implement nalgebra functions 2024-08-22 16:23:01 +08:00
lyken 1562a938a1
core/ndstrides: add NDArrayObject::to_any 2024-08-22 16:23:01 +08:00
lyken 00800ba6ee
core/ndstrides: add ContiguousNDArray
Currently this is used to interop with nalgebra.
2024-08-22 16:23:01 +08:00
lyken b9e00eb8a5
core/ndstrides: partially update builtin_fns to use ndarray with strides
nalgebra functions are not updated
2024-08-22 16:23:01 +08:00
lyken 566ce7df50
core/ndstrides: implement np_dot() for scalars and 1D 2024-08-22 16:23:01 +08:00
lyken 70c26561e1
standalone: extend test_ndarray_matmul 2024-08-22 16:23:01 +08:00
lyken 4fef633090
core/ndstrides: implement general ndarray matmul 2024-08-22 16:23:01 +08:00
lyken ae351f7678
core/ndstrides: implement binop 2024-08-22 16:23:01 +08:00
lyken 3efae534f7
core/ndstrides: add NDArrayOut, broadcast_map and map 2024-08-22 16:23:01 +08:00
lyken adca310424
core/ndstrides: add more ScalarOrNDArray and NDArrayObject utils 2024-08-22 16:23:01 +08:00
lyken e8e4801c92
core/ndstrides: implement np_transpose() (no axes argument) 2024-08-22 16:23:01 +08:00
lyken 9bdc520bbc
core/ndstrides: implement broadcasting & np_broadcast_to() 2024-08-22 16:23:01 +08:00
lyken 9bd08f8de8
core/ndstrides: implement np_reshape() 2024-08-22 16:23:01 +08:00
lyken 5f94f4a4cd
core: categorize np_{transpose,reshape} as 'view functions' 2024-08-22 16:23:01 +08:00
lyken a2a1e74be0
core/ndstrides: implement np_shape() and np_strides()
These functions are not important, but they are handy for debugging +
implementing them takes little effort.

NOTE: `np.strides()` is not an actual NumPy function. You can only(?)
access them thru `ndarray.strides`.
2024-08-22 16:23:01 +08:00
lyken 99eac99db2
core/ndstrides: implement ndarray.fill() and .copy() 2024-08-22 16:23:01 +08:00
lyken 70e1349d32
core/ndstrides: implement np_identity() and np_eye() 2024-08-22 16:23:01 +08:00
lyken ada6e95d75
core/ndstrides: implement np_array()
It also checks for inconsistent dimensions if the input is a list.
e.g., rejecting `[[1.0, 2.0], [3.0]]`. Previously it was a todo of
`np_array()`.
2024-08-22 16:23:01 +08:00
lyken 4dcbaed8ee
core/irrt: add List
Needed for implementing np_array()
2024-08-22 16:23:01 +08:00
lyken 0bb2b02daa
core/ndstrides: add NDArrayObject::atleast_nd 2024-08-22 16:23:01 +08:00
lyken df6916dbe8
core/ndstrides: add NDArrayObject::make_copy 2024-08-22 16:23:01 +08:00
lyken 0a874da5fc
core/ndstrides: implement ndarray indexing
The functionality for `...` and `np.newaxis` is there in IRRT, but there
is no implementation of them for @kernel Python expressions because of
M-Labs/nac3#486.
2024-08-22 16:23:01 +08:00
lyken c74b1eb9d8
core/irrt: rename NDIndex to NDIndexInt
The name `NDIndex` is used in later commits.
2024-08-22 16:23:01 +08:00
lyken bda003989e
core/irrt: add Slice and Range
Needed for implementing general ndarray indexing
2024-08-22 16:23:01 +08:00
lyken 5411ac5c88
core/ndstrides: implement & refactor call_len() 2024-08-22 16:23:01 +08:00
lyken 1c317f9205
core/ndstrides: implement ndarray np_{zeros,ones,full,empty} 2024-08-22 16:23:01 +08:00
lyken 701e45364c
core/object: add is_instance for ndarray, tuple and list 2024-08-22 16:23:01 +08:00
lyken fd1a9f4f77
core/model: add util::gen_for_model 2024-08-22 16:23:01 +08:00
lyken f9b8071df9
core/object: add ListObject and TupleObject
Needed for implementing other ndarray utils.
2024-08-22 16:23:00 +08:00
lyken 48fb3ff5ad
core/ndstrides: add ndarray iterator (NDIter) 2024-08-22 16:23:00 +08:00
lyken 4777909543
core/ndstrides: define ndarray with strides 2024-08-22 16:23:00 +08:00
lyken 2ae9196540
core/toplevel/helper: add {extract,create}_ndims 2024-08-22 16:23:00 +08:00
lyken 851ae28460
core/object: introduce object
Small abstraction to simplify implementations.
2024-08-22 16:23:00 +08:00
lyken 7c604237b3
core/model: introduce models 2024-08-22 16:23:00 +08:00
lyken 1c7ce22fd5
core/irrt/exceptions: add debug utils with exceptions 2024-08-22 10:06:45 +08:00
lyken 9d3d552e93
core/irrt/exceptions: allow irrt to raise exceptions 2024-08-22 10:06:45 +08:00
lyken d6deb5af26
core/irrt: split irrt.cpp into headers 2024-08-22 10:06:45 +08:00
lyken e3f3c498fe
core/irrt: build.rs capture IR defined constants 2024-08-22 10:06:45 +08:00
lyken 20781d11fd
core/irrt: build.rs capture IR defined types 2024-08-22 10:06:45 +08:00
lyken ecd72e8cad
core/irrt: reformat 2024-08-22 10:06:45 +08:00
lyken b2183fba24
core: add .clang-format 2024-08-22 10:06:45 +08:00
lyken 3c4d04e7c9
core/irrt: comment build.rs & move irrt to its own dir
To prepare for future IRRT implementations, and to also make cargo
only have to watch a single directory.
2024-08-22 10:06:45 +08:00
David Mak c5ae0e7c36 [standalone] Add tests for tuple equality 2024-08-21 16:25:32 +08:00
David Mak b8dab6cf7c [standalone] Add tests for string equality 2024-08-21 16:25:32 +08:00
David Mak 4d80ba38b7 [core] codegen/expr: Implement comparison of tuples 2024-08-21 16:25:32 +08:00
David Mak 33929bda24 [core] typecheck/typedef: Add support for tuple methods 2024-08-21 16:25:32 +08:00
David Mak a8e92212c0 [core] codegen/expr: Implement string equality 2024-08-21 16:25:32 +08:00
David Mak 908271014a [core] typecheck/magic_methods: Add equality methods to string 2024-08-21 16:25:32 +08:00
David Mak c407622f5c [core] codegen/expr: Add compilation error for unsupported cmpop 2024-08-21 15:46:13 +08:00
David Mak d7952d0629 [core] codegen/expr: Fix assertions not generated for -O0 2024-08-21 15:36:54 +08:00
David Mak ca1395aed6 [core] codegen: Remove redundant return 2024-08-21 15:36:54 +08:00
David Mak 7799aa4987 [meta] Do not specify rev in dependency version 2024-08-21 15:36:54 +08:00
David Mak 76016a26ad [meta] Apply clippy suggestions 2024-08-21 13:07:57 +08:00
lyken 8532bf5206
standalone: add missing test_ndarray_ceil() run 2024-08-21 11:39:00 +08:00
lyken 2cf64d8608
apply clippy comment changes 2024-08-21 11:21:10 +08:00
lyken 706759adb2
artiq: apply cargo fmt 2024-08-21 11:21:10 +08:00
lyken b90cf2300b
core/fix: add missing lifetime in gen_for* 2024-08-21 11:05:30 +08:00
Sebastien Bourdeauducq 0fc26df29e flake: update nixpkgs 2024-08-19 23:53:15 +08:00
David Mak 0b074c2cf2 [artiq] symbol_resolver: Set private linkage for constants 2024-08-19 14:41:43 +08:00
Sébastien Bourdeauducq a0f6961e0e cargo: update dependencies 2024-08-19 13:15:03 +08:00
David Mak b1c5c2e1d4 [artiq] Fix RPC of ndarrays to host 2024-08-15 15:41:24 +08:00
David Mak 69320a6cf1 [artiq] Fix LLVM representation of strings
Should be `%str` rather than `[N x i8]`.
2024-08-14 09:30:08 +08:00
David Mak 9e0601837a core: Add compile-time feature to disable escape analysis 2024-08-14 09:29:48 +08:00
lyken 432c81a500
core: update insta after #489 2024-08-13 15:30:34 +08:00
David Mak 6beff7a268 [artiq] Implement core_log and rtio_log in terms of polymorphic_print
Implementation mostly references the original implementation in Python.
2024-08-13 15:19:03 +08:00
David Mak 6ca7aecd4a [artiq] Add core_log and rtio_log function declarations 2024-08-13 15:19:03 +08:00
David Mak 8fd7216243 [core] toplevel/composer: Add lateinit_builtins
This is required for the new core_log and rtio_log functions, which take
a generic type as its parameter. However, in ARTIQ builtins are
initialized using one unifier and then actually used by another unifier.

lateinit_builtins workaround this issue by deferring the initialization
of functions requiring type variables until the actual unifier is ready.
2024-08-13 15:19:03 +08:00
David Mak 4f5e417012 [core] codegen: Add function to get format constants for integers 2024-08-13 15:19:03 +08:00
David Mak a0614bad83 [core] codegen/expr: Make gen_string return `StructValue`
So that it is clear that the value itself is returned rather than a
pointer to the struct or its data.
2024-08-13 15:19:03 +08:00
David Mak 5539d144ed [core] Add `CodeGenContext::build_in_bounds_gep_and_load`
For safer accesses to `gep`-able values and faster fails.
2024-08-13 15:19:03 +08:00
David Mak b3891b9a0d standalone: Fix several issues post script refactoring
- Add helptext for check_demos.sh
- Add back support for using debug NAC3 for running tests
- Output error message when argument is not recognized
- Fixed last non-demo script argument being ignored
- Add back SSE2 requirement to NAC3 (required for mandelbrot)
2024-08-13 15:19:03 +08:00
David Mak 6fb8939179 [meta] Update dependencies 2024-08-13 15:19:03 +08:00
lyken 973dc5041a core/typecheck: Support tuple arg type in len() 2024-08-13 15:02:59 +08:00
David Mak d0da688aa7 standalone: Add tuple len test 2024-08-13 15:02:59 +08:00
David Mak 12c4e1cf48 core/toplevel/builtins: Add support for len() on tuples 2024-08-13 15:02:59 +08:00
David Mak 9b988647ed core/toplevel/builtins: Extract len() into builtin function 2024-08-13 15:02:59 +08:00
lyken 35a7cecc12
core/typecheck: fix np_array ndmin bug 2024-08-13 12:50:04 +08:00
lyken 7e3d87f841 core/codegen: fix bug in call_ceil function 2024-08-07 16:40:55 +08:00
David Mak ac0d83ef98 standalone: Add vararg.py 2024-08-06 11:48:42 +08:00
David Mak 3ff6db1a29 core/codegen: Add va_start and va_end intrinsics 2024-08-06 11:48:42 +08:00
David Mak d7b806afb4 core/codegen: Implement support for va_info on supported architectures 2024-08-06 11:48:40 +08:00
David Mak fac60c3974 core/codegen: Handle vararg in function generation 2024-08-06 11:46:00 +08:00
David Mak f5fb504a15 core/codegen/expr: Implement vararg handling in gen_call 2024-08-06 11:46:00 +08:00
David Mak faa3bb97ad core/typecheck/typedef: Add vararg to Unifier::stringify 2024-08-06 11:46:00 +08:00
David Mak 6a64c9d1de core/typecheck/typedef: Add is_vararg_ctx to TTuple 2024-08-06 11:45:54 +08:00
David Mak 3dc8498202 core/typecheck/typedef: Handle vararg parameters in unify_call 2024-08-06 11:43:13 +08:00
David Mak cbf79c5e9c core/typecheck/typedef: Add is_vararg to FuncArg, ConcreteFuncArg 2024-08-06 11:43:13 +08:00
David Mak b8aa17bf8c core/toplevel/composer: Add parsing for vararg parameter 2024-08-06 10:52:24 +08:00
David Mak f5b998cd9c core/codegen: Remove unnecessary mut from get_llvm*_type 2024-08-06 10:52:24 +08:00
David Mak c36f85ecb9 meta: Update dependencies 2024-08-06 10:52:24 +08:00
lyken 3a8c385e01 core/typecheck: fix missing ExprKind::Asterisk in fix_assignment_target_context 2024-08-05 19:30:48 +08:00
lyken 221de4d06a core/codegen: add missing comment 2024-08-05 19:30:48 +08:00
lyken fb9fe8edf2 core: reimplement assignment type inference and codegen
- distinguish between setitem and getitem
- allow starred assignment targets, but the assigned value would be a tuple
- allow both [...] and (...) to be target lists
2024-08-05 19:30:48 +08:00
lyken 894083c6a3 core/codegen: refactor gen_{for,comprehension} to match on iter type 2024-08-05 19:30:48 +08:00
Sébastien Bourdeauducq 669c6aca6b clean up and fix 32-bit demos 2024-08-05 19:04:25 +08:00
abdul124 63d2b49b09 core: remove np_linalg_matmul 2024-08-05 11:44:55 +08:00
abdul124 bf709889c4 standalone/demo: separate linalg functions from main workspace 2024-08-05 11:44:54 +08:00
abdul124 1c72698d02 core: add np_linalg_det and np_linalg_matrix_power functions 2024-07-31 18:02:54 +08:00
abdul124 54f883f0a5 core: implement np_dot using LLVM_IR 2024-07-31 15:53:51 +08:00
abdul124 4a6845dac6 standalone: add np.transpose and np.reshape functions 2024-07-31 13:23:07 +08:00
abdul124 00236f48bc core: add np.transpose and np.reshape functions 2024-07-31 13:23:07 +08:00
abdul124 a3e6bb2292 core/helper: add linalg section 2024-07-31 13:23:07 +08:00
abdul124 17171065b1 standalone: link linalg at runtime 2024-07-31 13:23:07 +08:00
abdul124 540b35ec84 standalone: move linalg functions to demo 2024-07-31 13:23:05 +08:00
abdul124 4bb00c52e3 core/builtin_fns: improve error reporting 2024-07-31 13:21:31 +08:00
abdul124 faf07527cb standalone: add runtime implementation for linalg functions 2024-07-31 13:21:28 +08:00
abdul124 d6a4d0a634 standalone: add linalg methods and tests 2024-07-29 16:48:06 +08:00
abdul124 2242c5af43 core: add linalg methods 2024-07-29 16:48:06 +08:00
David Mak 318a675ea6 standalone: Rename -m32 to -i386 2024-07-29 14:58:58 +08:00
David Mak 32e52ce198 standalone: Revert using uint32_t as slice length
Turns out list and str have always been size_t.
2024-07-29 14:58:29 +08:00
Sebastien Bourdeauducq 665ca8e32d cargo: update dependencies 2024-07-27 22:24:56 +08:00
Sebastien Bourdeauducq 12c12b1d80 flake: update nixpkgs 2024-07-27 22:22:20 +08:00
lyken 72972fa909 core/toplevel: add more numpy categories 2024-07-27 21:57:47 +08:00
lyken 142cd48594 core/toplevel: reorder PrimDef::details 2024-07-27 21:57:47 +08:00
lyken 8adfe781c5 core/toplevel: fix PrimDef method names 2024-07-27 21:57:47 +08:00
lyken 339b74161b core/toplevel: reorganize PrimDef 2024-07-27 21:57:47 +08:00
David Mak 8c5ba37d09 standalone: Add 32-bit execution tests to check_demo.sh 2024-07-26 13:35:40 +08:00
David Mak 05a8948ff2 core: Minor cleanup to use ListValue APIs 2024-07-26 13:35:40 +08:00
David Mak 6d171ec284 core: Add label name and hooks to gen_for functions 2024-07-26 13:35:40 +08:00
David Mak 0ba68f6657 core: Set target triple and datalayout for each module
Fixes an issue with inconsistent pointer sizes causing crashes.
2024-07-26 13:35:40 +08:00
David Mak 693b2a8863 core: Add support for 32-bit size_t on 64-bit targets 2024-07-26 13:35:40 +08:00
David Mak 5faeede0e5 Determine size_t using LLVM target machine 2024-07-26 13:35:38 +08:00
David Mak 266707df9d standalone: Add support for running 32-bit binaries 2024-07-26 13:32:38 +08:00
David Mak 3d3c258756 standalone: Remove support for --lli 2024-07-26 13:32:38 +08:00
David Mak ed1182cb24 standalone: Update format specifiers for exceptions
Use platform-agnostic identifiers instead.
2024-07-26 13:32:37 +08:00
David Mak fd025c1137 standalone: Use uint32_t for cslice length
Matching the expected type of string and list slices.
2024-07-26 13:32:21 +08:00
David Mak f139db9af9 meta: Update dependencies 2024-07-26 10:33:02 +08:00
lyken 44487b76ae standalone: interpret_demo.py remove duplicated section 2024-07-22 17:23:35 +08:00
lyken 1332f113e8 standalone: fix interpret_demo.py comments 2024-07-22 17:06:14 +08:00
Sébastien Bourdeauducq 7632d6f72a cargo: update dependencies 2024-07-21 11:00:25 +08:00
David Mak 4948395ca2 core/toplevel/type_annotation: Add handling for mismatching class def
Primitive types only contain fields in its Type and not its TopLevelDef.
This causes primitive object types to lack some fields.
2024-07-19 14:42:14 +08:00
David Mak 3db3061d99 artiq/symbol_resolver: Handle type of zero-length lists 2024-07-19 14:42:14 +08:00
David Mak 51c2175c80 core/codegen/stmt: Convert assertion values to i1 2024-07-19 14:42:14 +08:00
lyken 1a31a50b8a
standalone: fix __nac3_raise def in demo.c 2024-07-17 21:22:08 +08:00
lyken 6c10e3d056 core: cargo clippy 2024-07-12 21:18:53 +08:00
lyken 2dbc1ec659 cargo fmt 2024-07-12 21:16:38 +08:00
Sebastien Bourdeauducq c80378063a add np_argmin/argmax to interpret_demo environment 2024-07-12 13:27:52 +02:00
abdul124 513d30152b core: support raise exception short form 2024-07-12 18:58:34 +08:00
abdul124 45e9360c4d standalone: Add np_argmax and np_argmin tests 2024-07-12 18:19:56 +08:00
abdul124 2e01b77fc8 core: refactor np_max/np_min functions 2024-07-12 18:18:54 +08:00
abdul124 cea7cade51 core: add np_argmax/np_argmin functions 2024-07-12 18:18:28 +08:00
107 changed files with 13386 additions and 6956 deletions

3
.clang-format Normal file
View File

@ -0,0 +1,3 @@
BasedOnStyle: Microsoft
IndentWidth: 4
ReflowComments: false

1
.gitignore vendored
View File

@ -1,3 +1,4 @@
__pycache__
/target
/nac3standalone/demo/linalg/target
nix/windows/msys2

182
Cargo.lock generated
View File

@ -26,9 +26,9 @@ dependencies = [
[[package]]
name = "anstream"
version = "0.6.14"
version = "0.6.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "418c75fa768af9c03be99d17643f93f79bbba589895012a80e3452a19ddda15b"
checksum = "64e15c1ab1f89faffbf04a634d5e1962e9074f2741eef6d97f3c4e322426d526"
dependencies = [
"anstyle",
"anstyle-parse",
@ -41,36 +41,36 @@ dependencies = [
[[package]]
name = "anstyle"
version = "1.0.7"
version = "1.0.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "038dfcf04a5feb68e9c60b21c9625a54c2c0616e79b72b0fd87075a056ae1d1b"
checksum = "1bec1de6f59aedf83baf9ff929c98f2ad654b97c9510f4e70cf6f661d49fd5b1"
[[package]]
name = "anstyle-parse"
version = "0.2.4"
version = "0.2.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c03a11a9034d92058ceb6ee011ce58af4a9bf61491aa7e1e59ecd24bd40d22d4"
checksum = "eb47de1e80c2b463c735db5b217a0ddc39d612e7ac9e2e96a5aed1f57616c1cb"
dependencies = [
"utf8parse",
]
[[package]]
name = "anstyle-query"
version = "1.1.0"
version = "1.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ad186efb764318d35165f1758e7dcef3b10628e26d41a44bc5550652e6804391"
checksum = "6d36fc52c7f6c869915e99412912f22093507da8d9e942ceaf66fe4b7c14422a"
dependencies = [
"windows-sys",
"windows-sys 0.52.0",
]
[[package]]
name = "anstyle-wincon"
version = "3.0.3"
version = "3.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "61a38449feb7068f52bb06c12759005cf459ee52bb4adc1d5a7c4322d716fb19"
checksum = "5bf74e1b6e971609db8ca7a9ce79fd5768ab6ae46441c572e46cf596f59e57f8"
dependencies = [
"anstyle",
"windows-sys",
"windows-sys 0.52.0",
]
[[package]]
@ -117,9 +117,12 @@ checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
[[package]]
name = "cc"
version = "1.1.0"
version = "1.1.13"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "eaff6f8ce506b9773fa786672d63fc7a191ffea1be33f72bbd4aeacefca9ffc8"
checksum = "72db2f7947ecee9b03b510377e8bb9077afa27176fdbff55c51027e976fdcc48"
dependencies = [
"shlex",
]
[[package]]
name = "cfg-if"
@ -129,9 +132,9 @@ checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
[[package]]
name = "clap"
version = "4.5.9"
version = "4.5.16"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "64acc1846d54c1fe936a78dc189c34e28d3f5afc348403f28ecf53660b9b8462"
checksum = "ed6719fffa43d0d87e5fd8caeab59be1554fb028cd30edc88fc4369b17971019"
dependencies = [
"clap_builder",
"clap_derive",
@ -139,9 +142,9 @@ dependencies = [
[[package]]
name = "clap_builder"
version = "4.5.9"
version = "4.5.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6fb8393d67ba2e7bfaf28a23458e4e2b543cc73a99595511eb207fdb8aede942"
checksum = "216aec2b177652e3846684cbfe25c9964d18ec45234f0f5da5157b207ed1aab6"
dependencies = [
"anstream",
"anstyle",
@ -151,27 +154,27 @@ dependencies = [
[[package]]
name = "clap_derive"
version = "4.5.8"
version = "4.5.13"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2bac35c6dafb060fd4d275d9a4ffae97917c13a6327903a8be2153cd964f7085"
checksum = "501d359d5f3dcaf6ecdeee48833ae73ec6e42723a1e52419c79abf9507eec0a0"
dependencies = [
"heck 0.5.0",
"proc-macro2",
"quote",
"syn 2.0.70",
"syn 2.0.75",
]
[[package]]
name = "clap_lex"
version = "0.7.1"
version = "0.7.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4b82cf0babdbd58558212896d1a4272303a57bdb245c2bf1147185fb45640e70"
checksum = "1462739cb27611015575c0c11df5df7601141071f07518d56fcc1be504cbec97"
[[package]]
name = "colorchoice"
version = "1.0.1"
version = "1.0.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0b6a852b24ab71dffc585bcb46eaf7959d175cb865a7152e35b348d1b2960422"
checksum = "d3fd119d74b830634cea2a0f58bbd0d54540518a14397557951e79340abc28c0"
[[package]]
name = "console"
@ -182,7 +185,7 @@ dependencies = [
"encode_unicode",
"lazy_static",
"libc",
"windows-sys",
"windows-sys 0.52.0",
]
[[package]]
@ -302,7 +305,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "534c5cf6194dfab3db3242765c03bbe257cf92f22b38f6bc0c58d59108a820ba"
dependencies = [
"libc",
"windows-sys",
"windows-sys 0.52.0",
]
[[package]]
@ -385,9 +388,9 @@ dependencies = [
[[package]]
name = "indexmap"
version = "2.2.6"
version = "2.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "168fb715dda47215e360912c096649d23d58bf392ac62f73919e831745e40f26"
checksum = "93ead53efc7ea8ed3cfb0c79fc8023fbb782a5432b52830b6518941cebe6505c"
dependencies = [
"equivalent",
"hashbrown 0.14.5",
@ -421,7 +424,7 @@ checksum = "4fa4d8d74483041a882adaa9a29f633253a66dde85055f0495c121620ac484b2"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.70",
"syn 2.0.75",
]
[[package]]
@ -440,9 +443,9 @@ dependencies = [
[[package]]
name = "is_terminal_polyfill"
version = "1.70.0"
version = "1.70.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f8478577c03552c21db0e2724ffb8986a5ce7af88107e6be5d2ee6e158c12800"
checksum = "7943c866cc5cd64cbc25b2e01621d07fa8eb2a1a23160ee81ce38704e97b8ecf"
[[package]]
name = "itertools"
@ -507,15 +510,15 @@ checksum = "bbd2bcb4c963f2ddae06a2efc7e9f3591312473c50c6685e1f298068316e66fe"
[[package]]
name = "libc"
version = "0.2.155"
version = "0.2.157"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "97b3888a4aecf77e811145cadf6eef5901f4782c53886191b2f693f24761847c"
checksum = "374af5f94e54fa97cf75e945cce8a6b201e88a1a07e688b47dfd2a59c66dbd86"
[[package]]
name = "libloading"
version = "0.8.4"
version = "0.8.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e310b3a6b5907f99202fcdb4960ff45b93735d7c7d96b760fcff8db2dc0e103d"
checksum = "4979f22fdb869068da03c9f7528f8297c6fd2606bc3a4affe42e6a823fdb8da4"
dependencies = [
"cfg-if",
"windows-targets",
@ -616,7 +619,7 @@ name = "nac3core"
version = "0.1.0"
dependencies = [
"crossbeam",
"indexmap 2.2.6",
"indexmap 2.4.0",
"indoc",
"inkwell",
"insta",
@ -706,7 +709,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b4c5cc86750666a3ed20bdaf5ca2a0344f9c67674cae0515bec2da16fbaa47db"
dependencies = [
"fixedbitset",
"indexmap 2.2.6",
"indexmap 2.4.0",
]
[[package]]
@ -749,7 +752,7 @@ dependencies = [
"phf_shared 0.11.2",
"proc-macro2",
"quote",
"syn 2.0.70",
"syn 2.0.75",
]
[[package]]
@ -778,15 +781,18 @@ checksum = "5be167a7af36ee22fe3115051bc51f6e6c7054c9348e28deb4f49bd6f705a315"
[[package]]
name = "portable-atomic"
version = "1.6.0"
version = "1.7.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7170ef9988bc169ba16dd36a7fa041e5c4cbeb6a35b76d4c03daded371eae7c0"
checksum = "da544ee218f0d287a911e9c99a39a8c9bc8fcad3cb8db5959940044ecfc67265"
[[package]]
name = "ppv-lite86"
version = "0.2.17"
version = "0.2.20"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5b40af805b3121feab8a3c29f04d8ad262fa8e0561883e7653e024ae4479e6de"
checksum = "77957b295656769bb8ad2b6a6b09d897d94f05c41b069aede1fcdaa675eaea04"
dependencies = [
"zerocopy",
]
[[package]]
name = "precomputed-hash"
@ -850,7 +856,7 @@ dependencies = [
"proc-macro2",
"pyo3-macros-backend",
"quote",
"syn 2.0.70",
"syn 2.0.75",
]
[[package]]
@ -863,7 +869,7 @@ dependencies = [
"proc-macro2",
"pyo3-build-config",
"quote",
"syn 2.0.70",
"syn 2.0.75",
]
[[package]]
@ -927,9 +933,9 @@ dependencies = [
[[package]]
name = "redox_syscall"
version = "0.5.2"
version = "0.5.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c82cf8cff14456045f55ec4241383baeff27af886adb72ffb2162f99911de0fd"
checksum = "2a908a6e00f1fdd0dfd9c0eb08ce85126f6d8bbda50017e74bc4a4b7d4a926a4"
dependencies = [
"bitflags",
]
@ -947,9 +953,9 @@ dependencies = [
[[package]]
name = "regex"
version = "1.10.5"
version = "1.10.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b91213439dad192326a0d7c6ee3955910425f441d7038e0d6933b0aec5c4517f"
checksum = "4219d74c6b67a3654a9fbebc4b419e22126d13d2f3c4a07ee0cb61ff79a79619"
dependencies = [
"aho-corasick",
"memchr",
@ -991,7 +997,7 @@ dependencies = [
"errno",
"libc",
"linux-raw-sys",
"windows-sys",
"windows-sys 0.52.0",
]
[[package]]
@ -1029,31 +1035,32 @@ checksum = "61697e0a1c7e512e84a621326239844a24d8207b4669b41bc18b32ea5cbf988b"
[[package]]
name = "serde"
version = "1.0.204"
version = "1.0.208"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "bc76f558e0cbb2a839d37354c575f1dc3fdc6546b5be373ba43d95f231bf7c12"
checksum = "cff085d2cb684faa248efb494c39b68e522822ac0de72ccf08109abde717cfb2"
dependencies = [
"serde_derive",
]
[[package]]
name = "serde_derive"
version = "1.0.204"
version = "1.0.208"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e0cd7e117be63d3c3678776753929474f3b04a43a080c744d6b0ae2a8c28e222"
checksum = "24008e81ff7613ed8e5ba0cfaf24e2c2f1e5b8a0495711e44fcd4882fca62bcf"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.70",
"syn 2.0.75",
]
[[package]]
name = "serde_json"
version = "1.0.120"
version = "1.0.125"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4e0d21c9a8cae1235ad58a00c11cb40d4b1e5c784f1ef2c537876ed6ffd8b7c5"
checksum = "83c8e735a073ccf5be70aa8066aa984eaf2fa000db6c8d0100ae605b366d31ed"
dependencies = [
"itoa",
"memchr",
"ryu",
"serde",
]
@ -1071,10 +1078,16 @@ dependencies = [
]
[[package]]
name = "similar"
version = "2.5.0"
name = "shlex"
version = "1.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fa42c91313f1d05da9b26f267f931cf178d4aba455b4c4622dd7355eb80c6640"
checksum = "0fda2ff0d084019ba4d7c6f371c95d8fd75ce3524c3cb8fb653a3023f6323e64"
[[package]]
name = "similar"
version = "2.6.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1de1d4f81173b03af4c0cbed3c898f6bff5b870e4a7f5d6f4057d62a7a4b686e"
[[package]]
name = "siphasher"
@ -1134,7 +1147,7 @@ dependencies = [
"proc-macro2",
"quote",
"rustversion",
"syn 2.0.70",
"syn 2.0.75",
]
[[package]]
@ -1150,9 +1163,9 @@ dependencies = [
[[package]]
name = "syn"
version = "2.0.70"
version = "2.0.75"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2f0209b68b3613b093e0ec905354eccaedcfe83b8cb37cbdeae64026c3064c16"
checksum = "f6af063034fc1935ede7be0122941bafa9bacb949334d090b77ca98b5817c7d9"
dependencies = [
"proc-macro2",
"quote",
@ -1161,20 +1174,21 @@ dependencies = [
[[package]]
name = "target-lexicon"
version = "0.12.15"
version = "0.12.16"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4873307b7c257eddcb50c9bedf158eb669578359fb28428bef438fec8e6ba7c2"
checksum = "61c41af27dd6d1e27b1b16b489db798443478cef1f06a660c96db617ba5de3b1"
[[package]]
name = "tempfile"
version = "3.10.1"
version = "3.12.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "85b77fafb263dd9d05cbeac119526425676db3784113aa9295c88498cbf8bff1"
checksum = "04cbcdd0c794ebb0d4cf35e88edd2f7d2c4c3e9a5a6dab322839b321c6a87a64"
dependencies = [
"cfg-if",
"fastrand",
"once_cell",
"rustix",
"windows-sys",
"windows-sys 0.59.0",
]
[[package]]
@ -1203,22 +1217,22 @@ dependencies = [
[[package]]
name = "thiserror"
version = "1.0.61"
version = "1.0.63"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c546c80d6be4bc6a00c0f01730c08df82eaa7a7a61f11d656526506112cc1709"
checksum = "c0342370b38b6a11b6cc11d6a805569958d54cfa061a29969c3b5ce2ea405724"
dependencies = [
"thiserror-impl",
]
[[package]]
name = "thiserror-impl"
version = "1.0.61"
version = "1.0.63"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "46c3384250002a6d5af4d114f2845d37b57521033f30d5c3f46c4d70e1197533"
checksum = "a4558b58466b9ad7ca0f102865eccc95938dca1a74a856f2b57b6629050da261"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.70",
"syn 2.0.75",
]
[[package]]
@ -1336,9 +1350,9 @@ checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821"
[[package]]
name = "version_check"
version = "0.9.4"
version = "0.9.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f"
checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
[[package]]
name = "walkdir"
@ -1374,11 +1388,11 @@ checksum = "ac3b87c63620426dd9b991e5ce0329eff545bccbbb34f3be09ff6fb6ab51b7b6"
[[package]]
name = "winapi-util"
version = "0.1.8"
version = "0.1.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4d4cc384e1e73b93bafa6fb4f1df8c41695c8a91cf9c4c64358067d15a7b6c6b"
checksum = "cf221c93e13a30d793f7645a0e7762c55d169dbb0a49671918a2319d289b10bb"
dependencies = [
"windows-sys",
"windows-sys 0.59.0",
]
[[package]]
@ -1396,6 +1410,15 @@ dependencies = [
"windows-targets",
]
[[package]]
name = "windows-sys"
version = "0.59.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1e38bc4d79ed67fd075bcc251a1c39b32a1776bbe92e5bef1f0bf1f8c531853b"
dependencies = [
"windows-targets",
]
[[package]]
name = "windows-targets"
version = "0.52.6"
@ -1475,6 +1498,7 @@ version = "0.7.35"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1b9b4fd18abc82b8136838da5d50bae7bdea537c574d8dc1a34ed098d6c166f0"
dependencies = [
"byteorder",
"zerocopy-derive",
]
@ -1486,5 +1510,5 @@ checksum = "fa4f8080344d4671fb4e831a13ad1e68092748387dfc4f55e356242fae12ce3e"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.70",
"syn 2.0.75",
]

View File

@ -2,11 +2,11 @@
"nodes": {
"nixpkgs": {
"locked": {
"lastModified": 1720418205,
"narHash": "sha256-cPJoFPXU44GlhWg4pUk9oUPqurPlCFZ11ZQPk21GTPU=",
"lastModified": 1723637854,
"narHash": "sha256-med8+5DSWa2UnOqtdICndjDAEjxr5D7zaIiK4pn0Q7c=",
"owner": "NixOS",
"repo": "nixpkgs",
"rev": "655a58a72a6601292512670343087c2d75d859c1",
"rev": "c3aa7b8938b17aebd2deecf7be0636000d62a2b9",
"type": "github"
},
"original": {

View File

@ -6,6 +6,7 @@
outputs = { self, nixpkgs }:
let
pkgs = import nixpkgs { system = "x86_64-linux"; };
pkgs32 = import nixpkgs { system = "i686-linux"; };
in rec {
packages.x86_64-linux = rec {
llvm-nac3 = pkgs.callPackage ./nix/llvm {};
@ -15,6 +16,22 @@
ln -s ${pkgs.llvmPackages_14.clang-unwrapped}/bin/clang $out/bin/clang-irrt
ln -s ${pkgs.llvmPackages_14.llvm.out}/bin/llvm-as $out/bin/llvm-as-irrt
'';
demo-linalg-stub = pkgs.rustPlatform.buildRustPackage {
name = "demo-linalg-stub";
src = ./nac3standalone/demo/linalg;
cargoLock = {
lockFile = ./nac3standalone/demo/linalg/Cargo.lock;
};
doCheck = false;
};
demo-linalg-stub32 = pkgs32.rustPlatform.buildRustPackage {
name = "demo-linalg-stub32";
src = ./nac3standalone/demo/linalg;
cargoLock = {
lockFile = ./nac3standalone/demo/linalg/Cargo.lock;
};
doCheck = false;
};
nac3artiq = pkgs.python3Packages.toPythonModule (
pkgs.rustPlatform.buildRustPackage rec {
name = "nac3artiq";
@ -24,7 +41,7 @@
lockFile = ./Cargo.lock;
};
passthru.cargoLock = cargoLock;
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_14.clang llvm-tools-irrt pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
nativeBuildInputs = [ pkgs.python3 (pkgs.wrapClangMulti pkgs.llvmPackages_14.clang) llvm-tools-irrt pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
buildInputs = [ pkgs.python3 llvm-nac3 ];
checkInputs = [ (pkgs.python3.withPackages(ps: [ ps.numpy ps.scipy ])) ];
checkPhase =
@ -32,7 +49,9 @@
echo "Checking nac3standalone demos..."
pushd nac3standalone/demo
patchShebangs .
./check_demos.sh
export DEMO_LINALG_STUB=${demo-linalg-stub}/lib/liblinalg.a
export DEMO_LINALG_STUB32=${demo-linalg-stub32}/lib/liblinalg.a
./check_demos.sh -i686
popd
echo "Running Cargo tests..."
cargoCheckHook
@ -149,7 +168,7 @@
buildInputs = with pkgs; [
# build dependencies
packages.x86_64-linux.llvm-nac3
llvmPackages_14.clang llvmPackages_14.llvm.out # for running nac3standalone demos
(pkgs.wrapClangMulti llvmPackages_14.clang) llvmPackages_14.llvm.out # for running nac3standalone demos
packages.x86_64-linux.llvm-tools-irrt
cargo
rustc
@ -162,6 +181,11 @@
pre-commit
rustfmt
];
shellHook =
''
export DEMO_LINALG_STUB=${packages.x86_64-linux.demo-linalg-stub}/lib/liblinalg.a
export DEMO_LINALG_STUB32=${packages.x86_64-linux.demo-linalg-stub32}/lib/liblinalg.a
'';
};
devShells.x86_64-linux.msys2 = pkgs.mkShell {
name = "nac3-dev-shell-msys2";

View File

@ -24,3 +24,4 @@ features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-l
[features]
init-llvm-profile = []
no-escape-analysis = ["nac3core/no-escape-analysis"]

View File

@ -0,0 +1,24 @@
from min_artiq import *
from numpy import int32
@nac3
class EmptyList:
core: KernelInvariant[Core]
def __init__(self):
self.core = Core()
@rpc
def get_empty(self) -> list[int32]:
return []
@kernel
def run(self):
a: list[int32] = self.get_empty()
if a != []:
raise ValueError
if __name__ == "__main__":
EmptyList().run()

26
nac3artiq/demo/str_abi.py Normal file
View File

@ -0,0 +1,26 @@
from min_artiq import *
from numpy import ndarray, zeros as np_zeros
@nac3
class StrFail:
core: KernelInvariant[Core]
def __init__(self):
self.core = Core()
@kernel
def hello(self, arg: str):
pass
@kernel
def consume_ndarray(self, arg: ndarray[str, 1]):
pass
def run(self):
self.hello("world")
self.consume_ndarray(np_zeros([10], dtype=str))
if __name__ == "__main__":
StrFail().run()

View File

@ -2,7 +2,9 @@ use nac3core::{
codegen::{
expr::gen_call,
llvm_intrinsics::{call_int_smax, call_stackrestore, call_stacksave},
stmt::{gen_block, gen_with},
model::*,
object::{any::AnyObject, list::ListObject, ndarray::NDArrayObject, range::RangeObject},
stmt::{gen_block, gen_for_callback_incrementing, gen_if_callback, gen_with},
CodeGenContext, CodeGenerator,
},
symbol_resolver::ValueEnum,
@ -13,7 +15,11 @@ use nac3core::{
use nac3parser::ast::{Expr, ExprKind, Located, Stmt, StmtKind, StrRef};
use inkwell::{
context::Context, module::Linkage, types::IntType, values::BasicValueEnum, AddressSpace,
context::Context,
module::Linkage,
types::IntType,
values::{BasicValueEnum, PointerValue, StructValue},
AddressSpace, IntPredicate,
};
use pyo3::{
@ -23,10 +29,12 @@ use pyo3::{
use crate::{symbol_resolver::InnerResolver, timeline::TimeFns};
use itertools::Itertools;
use std::{
collections::hash_map::DefaultHasher,
collections::HashMap,
collections::{hash_map::DefaultHasher, HashMap},
hash::{Hash, Hasher},
iter::once,
mem,
sync::Arc,
};
@ -386,7 +394,7 @@ fn gen_rpc_tag(
} else {
let ty_enum = ctx.unifier.get_ty(ty);
match &*ty_enum {
TTuple { ty } => {
TTuple { ty, is_vararg_ctx: false } => {
buffer.push(b't');
buffer.push(ty.len() as u8);
for ty in ty {
@ -414,7 +422,10 @@ fn gen_rpc_tag(
} else {
unreachable!()
};
assert!((0u64..=u64::from(u8::MAX)).contains(&ndarray_ndims));
assert!(
(0u64..=u64::from(u8::MAX)).contains(&ndarray_ndims),
"Only NDArrays of sizes between 0 and 255 can be RPCed"
);
buffer.push(b'a');
buffer.push((ndarray_ndims & 0xFF) as u8);
@ -426,6 +437,77 @@ fn gen_rpc_tag(
Ok(())
}
/// Formats an RPC argument to conform to the expected format required by `send_value`.
///
/// See `artiq/firmware/libproto_artiq/rpc_proto.rs` for the expected format.
fn format_rpc_arg<'ctx>(
generator: &mut dyn CodeGenerator,
ctx: &mut CodeGenContext<'ctx, '_>,
(arg, arg_ty, arg_idx): (BasicValueEnum<'ctx>, Type, usize),
) -> PointerValue<'ctx> {
let llvm_i8 = ctx.ctx.i8_type();
let llvm_pi8 = llvm_i8.ptr_type(AddressSpace::default());
let arg_slot = match &*ctx.unifier.get_ty_immutable(arg_ty) {
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
// NAC3: NDArray = { usize, usize*, T* }
// libproto_artiq: NDArray = [data[..], dim_sz[..]]
let ndarray = AnyObject { ty: arg_ty, value: arg };
let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
let dtype = ctx.get_llvm_type(generator, ndarray.dtype);
let ndims = ndarray.ndims_llvm(generator, ctx.ctx);
// `ndarray.data` is possibly not contiguous. We need to force it to be continuous,
// and we might have to copy the whole ndarray.
let carray = ndarray.make_contiguous_ndarray(generator, ctx, Any(dtype));
let sizeof_sizet = Int(SizeT).sizeof(generator, ctx.ctx);
let sizeof_sizet = Int(SizeT).truncate_or_bit_cast(generator, ctx, sizeof_sizet);
let sizeof_pdata = Ptr(Any(dtype)).sizeof(generator, ctx.ctx);
let sizeof_pdata = Int(SizeT).truncate_or_bit_cast(generator, ctx, sizeof_pdata);
let sizeof_buf_shape = sizeof_sizet.mul(ctx, ndims);
let sizeof_buf = sizeof_buf_shape.add(ctx, sizeof_pdata);
// buf = { data: void*, shape: [size_t; ndims]; }
let buf = Int(Byte).array_alloca(generator, ctx, sizeof_buf.value);
let buf_data = buf;
let buf_shape = buf_data.offset(ctx, sizeof_pdata.value);
// Write to `buf->data`
let carray_data = carray.get(generator, ctx, |f| f.data); // has type Ptr<Any>
let carray_data = carray_data.pointer_cast(generator, ctx, Int(Byte));
buf_data.copy_from(generator, ctx, carray_data, sizeof_pdata.value);
// Write to `buf->shape`
let carray_shape = ndarray.instance.get(generator, ctx, |f| f.shape);
let carray_shape_i8 = carray_shape.pointer_cast(generator, ctx, Int(Byte));
buf_shape.copy_from(generator, ctx, carray_shape_i8, sizeof_buf_shape.value);
buf.value
}
_ => {
let arg_slot = generator
.gen_var_alloc(ctx, arg.get_type(), Some(&format!("rpc.arg{arg_idx}")))
.unwrap();
ctx.builder.build_store(arg_slot, arg).unwrap();
ctx.builder
.build_bitcast(arg_slot, llvm_pi8, "rpc.arg")
.map(BasicValueEnum::into_pointer_value)
.unwrap()
}
};
debug_assert_eq!(arg_slot.get_type(), llvm_pi8);
arg_slot
}
fn rpc_codegen_callback_fn<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
obj: Option<(Type, ValueEnum<'ctx>)>,
@ -433,10 +515,10 @@ fn rpc_codegen_callback_fn<'ctx>(
args: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
generator: &mut dyn CodeGenerator,
) -> Result<Option<BasicValueEnum<'ctx>>, String> {
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();
let size_type = generator.get_size_type(ctx.ctx);
let ptr_type = int8.ptr_type(AddressSpace::default());
let tag_ptr_type = ctx.ctx.struct_type(&[ptr_type.into(), size_type.into()], false);
let service_id = int32.const_int(fun.1 .0 as u64, false);
@ -509,22 +591,25 @@ fn rpc_codegen_callback_fn<'ctx>(
.0
.args
.iter()
.map(|arg| mapping.remove(&arg.name).unwrap().to_basic_value_enum(ctx, generator, arg.ty))
.collect::<Result<Vec<_>, _>>()?;
.map(|arg| {
mapping
.remove(&arg.name)
.unwrap()
.to_basic_value_enum(ctx, generator, arg.ty)
.map(|llvm_val| (llvm_val, arg.ty))
})
.collect::<Result<Vec<(_, _)>, _>>()?;
if let Some(obj) = obj {
if let ValueEnum::Static(obj) = obj.1 {
real_params.insert(0, obj.get_const_obj(ctx, generator));
if let ValueEnum::Static(obj_val) = obj.1 {
real_params.insert(0, (obj_val.get_const_obj(ctx, generator), obj.0));
} else {
// should be an error here...
panic!("only host object is allowed");
}
}
for (i, arg) in real_params.iter().enumerate() {
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();
for (i, (arg, arg_ty)) in real_params.iter().enumerate() {
let arg_slot = format_rpc_arg(generator, ctx, (*arg, *arg_ty, i));
let arg_ptr = unsafe {
ctx.builder.build_gep(
args_ptr,
@ -700,6 +785,7 @@ pub fn attributes_writeback(
name: i.to_string().into(),
ty: *ty,
default_value: None,
is_vararg: false,
})
.collect(),
ret: ctx.primitives.none,
@ -723,3 +809,468 @@ pub fn rpc_codegen_callback() -> Arc<GenCall> {
rpc_codegen_callback_fn(ctx, obj, fun, args, generator)
})))
}
/// Returns the `fprintf` format constant for the given [`llvm_int_t`][`IntType`] on a platform with
/// [`llvm_usize`] as its native word size.
///
/// Note that, similar to format constants in `<inttypes.h>`, these constants need to be prepended
/// with `%`.
#[must_use]
fn get_fprintf_format_constant<'ctx>(
llvm_usize: IntType<'ctx>,
llvm_int_t: IntType<'ctx>,
is_unsigned: bool,
) -> String {
debug_assert!(matches!(llvm_usize.get_bit_width(), 8 | 16 | 32 | 64));
let conv_spec = if is_unsigned { 'u' } else { 'd' };
// https://en.cppreference.com/w/c/language/arithmetic_types
// Note that NAC3 does **not** support LP32 and LLP64 configurations
match llvm_int_t.get_bit_width() {
8 => format!("hh{conv_spec}"),
16 => format!("h{conv_spec}"),
32 => conv_spec.to_string(),
64 => format!("{}{conv_spec}", if llvm_usize.get_bit_width() == 64 { "l" } else { "ll" }),
_ => todo!(
"Not yet implemented for i{} on {}-bit platform",
llvm_int_t.get_bit_width(),
llvm_usize.get_bit_width()
),
}
}
/// Prints one or more `values` to `core_log` or `rtio_log`.
///
/// * `separator` - The separator between multiple values.
/// * `suffix` - String to terminate the printed string, if any.
/// * `as_repr` - Whether the `repr()` output of values instead of `str()`.
/// * `as_rtio` - Whether to print to `rtio_log` instead of `core_log`.
fn polymorphic_print<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
generator: &mut dyn CodeGenerator,
values: &[(Type, ValueEnum<'ctx>)],
separator: &str,
suffix: Option<&str>,
as_repr: bool,
as_rtio: bool,
) -> Result<(), String> {
let printf = |ctx: &mut CodeGenContext<'ctx, '_>,
generator: &mut dyn CodeGenerator,
fmt: String,
args: Vec<BasicValueEnum<'ctx>>| {
debug_assert!(!fmt.is_empty());
debug_assert_eq!(fmt.as_bytes().last().unwrap(), &0u8);
let fn_name = if as_rtio { "rtio_log" } else { "core_log" };
let print_fn = ctx.module.get_function(fn_name).unwrap_or_else(|| {
let llvm_pi8 = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
let fn_t = if as_rtio {
let llvm_void = ctx.ctx.void_type();
llvm_void.fn_type(&[llvm_pi8.into()], true)
} else {
let llvm_i32 = ctx.ctx.i32_type();
llvm_i32.fn_type(&[llvm_pi8.into()], true)
};
ctx.module.add_function(fn_name, fn_t, None)
});
let fmt = ctx.gen_string(generator, fmt);
let fmt = unsafe { fmt.get_field_at_index_unchecked(0) }.into_pointer_value();
ctx.builder
.build_call(
print_fn,
&once(fmt.into()).chain(args).map(BasicValueEnum::into).collect_vec(),
"",
)
.unwrap();
};
let llvm_i32 = ctx.ctx.i32_type();
let llvm_i64 = ctx.ctx.i64_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let suffix = suffix.unwrap_or_default();
let mut fmt = String::new();
let mut args = Vec::new();
let flush = |ctx: &mut CodeGenContext<'ctx, '_>,
generator: &mut dyn CodeGenerator,
fmt: &mut String,
args: &mut Vec<BasicValueEnum<'ctx>>| {
if !fmt.is_empty() {
fmt.push('\0');
printf(ctx, generator, mem::take(fmt), mem::take(args));
}
};
for (ty, value) in values {
let ty = *ty;
let value = value.clone().to_basic_value_enum(ctx, generator, ty).unwrap();
if !fmt.is_empty() {
fmt.push_str(separator);
}
match &*ctx.unifier.get_ty_immutable(ty) {
TypeEnum::TTuple { ty: tys, is_vararg_ctx: false } => {
let pvalue = {
let pvalue = generator.gen_var_alloc(ctx, value.get_type(), None).unwrap();
ctx.builder.build_store(pvalue, value).unwrap();
pvalue
};
fmt.push('(');
flush(ctx, generator, &mut fmt, &mut args);
let tuple_vals = tys
.iter()
.enumerate()
.map(|(i, ty)| {
(*ty, {
let pfield =
ctx.builder.build_struct_gep(pvalue, i as u32, "").unwrap();
ValueEnum::from(ctx.builder.build_load(pfield, "").unwrap())
})
})
.collect_vec();
polymorphic_print(ctx, generator, &tuple_vals, ", ", None, true, as_rtio)?;
if tuple_vals.len() == 1 {
fmt.push_str(",)");
} else {
fmt.push(')');
}
}
TypeEnum::TFunc { .. } => todo!(),
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::None.id() => {
fmt.push_str("None");
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Bool.id() => {
fmt.push_str("%.*s");
let true_str = ctx.gen_string(generator, "True");
let true_data =
unsafe { true_str.get_field_at_index_unchecked(0) }.into_pointer_value();
let true_len = unsafe { true_str.get_field_at_index_unchecked(1) }.into_int_value();
let false_str = ctx.gen_string(generator, "False");
let false_data =
unsafe { false_str.get_field_at_index_unchecked(0) }.into_pointer_value();
let false_len =
unsafe { false_str.get_field_at_index_unchecked(1) }.into_int_value();
let bool_val = generator.bool_to_i1(ctx, value.into_int_value());
args.extend([
ctx.builder.build_select(bool_val, true_len, false_len, "").unwrap(),
ctx.builder.build_select(bool_val, true_data, false_data, "").unwrap(),
]);
}
TypeEnum::TObj { obj_id, .. }
if *obj_id == PrimDef::Int32.id()
|| *obj_id == PrimDef::Int64.id()
|| *obj_id == PrimDef::UInt32.id()
|| *obj_id == PrimDef::UInt64.id() =>
{
let is_unsigned =
*obj_id == PrimDef::UInt32.id() || *obj_id == PrimDef::UInt64.id();
let llvm_int_t = value.get_type().into_int_type();
debug_assert!(matches!(llvm_usize.get_bit_width(), 32 | 64));
debug_assert!(matches!(llvm_int_t.get_bit_width(), 32 | 64));
let fmt_spec = format!(
"%{}",
get_fprintf_format_constant(llvm_usize, llvm_int_t, is_unsigned)
);
fmt.push_str(fmt_spec.as_str());
args.push(value);
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Float.id() => {
fmt.push_str("%g");
args.push(value);
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Str.id() => {
if as_repr {
fmt.push_str("\"%.*s\"");
} else {
fmt.push_str("%.*s");
}
let str = value.into_struct_value();
let str_data = unsafe { str.get_field_at_index_unchecked(0) }.into_pointer_value();
let str_len = unsafe { str.get_field_at_index_unchecked(1) }.into_int_value();
args.extend(&[str_len.into(), str_data.into()]);
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::List.id() => {
fmt.push('[');
flush(ctx, generator, &mut fmt, &mut args);
let list = AnyObject { ty, value };
let list = ListObject::from_object(generator, ctx, list);
let items = list.instance.get(generator, ctx, |f| f.items);
let len = list.instance.get(generator, ctx, |f| f.len).value;
let last =
ctx.builder.build_int_sub(len, llvm_usize.const_int(1, false), "").unwrap();
gen_for_callback_incrementing(
generator,
ctx,
None,
llvm_usize.const_zero(),
(len, false),
|generator, ctx, _, i| {
let item = items.get_index(generator, ctx, i).value;
polymorphic_print(
ctx,
generator,
&[(list.item_type, item.into())],
"",
None,
true,
as_rtio,
)?;
gen_if_callback(
generator,
ctx,
|_, ctx| {
Ok(ctx
.builder
.build_int_compare(IntPredicate::ULT, i, last, "")
.unwrap())
},
|generator, ctx| {
printf(ctx, generator, ", \0".into(), Vec::default());
Ok(())
},
|_, _| Ok(()),
)?;
Ok(())
},
llvm_usize.const_int(1, false),
)?;
fmt.push(']');
flush(ctx, generator, &mut fmt, &mut args);
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
fmt.push_str("array([");
flush(ctx, generator, &mut fmt, &mut args);
let ndarray = AnyObject { ty, value };
let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
let num_0 = Int(SizeT).const_0(generator, ctx.ctx);
// Print `ndarray` as a flat list delimited by interspersed with ", \0"
ndarray.foreach(generator, ctx, |generator, ctx, _, hdl| {
let i = hdl.get_index(generator, ctx);
let scalar = hdl.get_scalar(generator, ctx);
// if (i != 0) { puts(", "); }
gen_if_callback(
generator,
ctx,
|_, ctx| {
let not_first = i.compare(ctx, IntPredicate::NE, num_0);
Ok(not_first.value)
},
|generator, ctx| {
printf(ctx, generator, ", \0".into(), Vec::default());
Ok(())
},
|_, _| Ok(()),
)?;
// Print element
polymorphic_print(
ctx,
generator,
&[(scalar.ty, scalar.value.into())],
"",
None,
true,
as_rtio,
)?;
Ok(())
})?;
fmt.push_str(")]");
flush(ctx, generator, &mut fmt, &mut args);
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Range.id() => {
fmt.push_str("range(");
flush(ctx, generator, &mut fmt, &mut args);
let range = AnyObject { ty, value };
let range = RangeObject::from_object(generator, ctx, range);
let (start, stop, step) = range.instance.destructure(generator, ctx);
let start = start.value;
let stop = stop.value;
let step = step.value;
polymorphic_print(
ctx,
generator,
&[
(ctx.primitives.int32, start.into()),
(ctx.primitives.int32, stop.into()),
(ctx.primitives.int32, step.into()),
],
", ",
None,
false,
as_rtio,
)?;
fmt.push(')');
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Exception.id() => {
let fmt_str = format!(
"%{}(%{}, %{1:}, %{1:})",
get_fprintf_format_constant(llvm_usize, llvm_i32, false),
get_fprintf_format_constant(llvm_usize, llvm_i64, false),
);
let exn = value.into_pointer_value();
let name = ctx
.build_in_bounds_gep_and_load(
exn,
&[llvm_i32.const_zero(), llvm_i32.const_zero()],
None,
)
.into_int_value();
let param0 = ctx
.build_in_bounds_gep_and_load(
exn,
&[llvm_i32.const_zero(), llvm_i32.const_int(6, false)],
None,
)
.into_int_value();
let param1 = ctx
.build_in_bounds_gep_and_load(
exn,
&[llvm_i32.const_zero(), llvm_i32.const_int(7, false)],
None,
)
.into_int_value();
let param2 = ctx
.build_in_bounds_gep_and_load(
exn,
&[llvm_i32.const_zero(), llvm_i32.const_int(8, false)],
None,
)
.into_int_value();
fmt.push_str(fmt_str.as_str());
args.extend_from_slice(&[name.into(), param0.into(), param1.into(), param2.into()]);
}
_ => unreachable!(
"Unsupported object type for polymorphic_print: {}",
ctx.unifier.stringify(ty)
),
}
}
fmt.push_str(suffix);
flush(ctx, generator, &mut fmt, &mut args);
Ok(())
}
/// Invokes the `core_log` intrinsic function.
pub fn call_core_log_impl<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
generator: &mut dyn CodeGenerator,
arg: (Type, BasicValueEnum<'ctx>),
) -> Result<(), String> {
let (arg_ty, arg_val) = arg;
polymorphic_print(ctx, generator, &[(arg_ty, arg_val.into())], " ", Some("\n"), false, false)?;
Ok(())
}
/// Invokes the `rtio_log` intrinsic function.
pub fn call_rtio_log_impl<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
generator: &mut dyn CodeGenerator,
channel: StructValue<'ctx>,
arg: (Type, BasicValueEnum<'ctx>),
) -> Result<(), String> {
let (arg_ty, arg_val) = arg;
polymorphic_print(
ctx,
generator,
&[(ctx.primitives.str, channel.into())],
" ",
Some("\x1E"),
false,
true,
)?;
polymorphic_print(ctx, generator, &[(arg_ty, arg_val.into())], " ", Some("\x1D"), false, true)?;
Ok(())
}
/// Generates a call to `core_log`.
pub fn gen_core_log<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
obj: &Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: &[(Option<StrRef>, ValueEnum<'ctx>)],
generator: &mut dyn CodeGenerator,
) -> Result<(), String> {
assert!(obj.is_none());
assert_eq!(args.len(), 1);
let value_ty = fun.0.args[0].ty;
let value_arg = args[0].1.clone().to_basic_value_enum(ctx, generator, value_ty)?;
call_core_log_impl(ctx, generator, (value_ty, value_arg))
}
/// Generates a call to `rtio_log`.
pub fn gen_rtio_log<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
obj: &Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: &[(Option<StrRef>, ValueEnum<'ctx>)],
generator: &mut dyn CodeGenerator,
) -> Result<(), String> {
assert!(obj.is_none());
assert_eq!(args.len(), 2);
let channel_ty = fun.0.args[0].ty;
assert!(ctx.unifier.unioned(channel_ty, ctx.primitives.str));
let channel_arg =
args[0].1.clone().to_basic_value_enum(ctx, generator, channel_ty)?.into_struct_value();
let value_ty = fun.0.args[1].ty;
let value_arg = args[1].1.clone().to_basic_value_enum(ctx, generator, value_ty)?;
call_rtio_log_impl(ctx, generator, channel_arg, (value_ty, value_arg))
}

View File

@ -24,6 +24,7 @@ use std::rc::Rc;
use std::sync::Arc;
use inkwell::{
context::Context,
memory_buffer::MemoryBuffer,
module::{Linkage, Module},
passes::PassBuilderOptions,
@ -32,9 +33,10 @@ use inkwell::{
OptimizationLevel,
};
use itertools::Itertools;
use nac3core::codegen::irrt::setup_irrt_exceptions;
use nac3core::codegen::{gen_func_impl, CodeGenLLVMOptions, CodeGenTargetMachineOptions};
use nac3core::toplevel::builtins::get_exn_constructor;
use nac3core::typecheck::typedef::{TypeEnum, Unifier, VarMap};
use nac3core::typecheck::typedef::{into_var_map, TypeEnum, Unifier, VarMap};
use nac3parser::{
ast::{ExprKind, Stmt, StmtKind, StrRef},
parser::parse_program,
@ -50,7 +52,7 @@ use nac3core::{
codegen::{concrete_type::ConcreteTypeStore, CodeGenTask, WithCall, WorkerRegistry},
symbol_resolver::SymbolResolver,
toplevel::{
composer::{ComposerConfig, TopLevelComposer},
composer::{BuiltinFuncCreator, BuiltinFuncSpec, ComposerConfig, TopLevelComposer},
DefinitionId, GenCall, TopLevelDef,
},
typecheck::typedef::{FunSignature, FuncArg},
@ -59,13 +61,13 @@ use nac3core::{
use nac3ld::Linker;
use tempfile::{self, TempDir};
use crate::codegen::attributes_writeback;
use crate::{
codegen::{rpc_codegen_callback, ArtiqCodeGenerator},
codegen::{
attributes_writeback, gen_core_log, gen_rtio_log, rpc_codegen_callback, ArtiqCodeGenerator,
},
symbol_resolver::{DeferredEvaluationStore, InnerResolver, PythonHelper, Resolver},
};
use tempfile::{self, TempDir};
mod codegen;
mod symbol_resolver;
@ -126,7 +128,7 @@ struct Nac3 {
isa: Isa,
time_fns: &'static (dyn TimeFns + Sync),
primitive: PrimitiveStore,
builtins: Vec<(StrRef, FunSignature, Arc<GenCall>)>,
builtins: Vec<BuiltinFuncSpec>,
pyid_to_def: Arc<RwLock<HashMap<u64, DefinitionId>>>,
primitive_ids: PrimitivePythonId,
working_directory: TempDir,
@ -264,7 +266,7 @@ impl Nac3 {
arg_names.len(),
));
}
for (i, FuncArg { ty, default_value, name }) in args.iter().enumerate() {
for (i, FuncArg { ty, default_value, name, .. }) in args.iter().enumerate() {
let in_name = match arg_names.get(i) {
Some(n) => n,
None if default_value.is_none() => {
@ -300,6 +302,64 @@ impl Nac3 {
None
}
/// Returns a [`Vec`] of builtins that needs to be initialized during method compilation time.
fn get_lateinit_builtins() -> Vec<Box<BuiltinFuncCreator>> {
vec![
Box::new(|primitives, unifier| {
let arg_ty = unifier.get_fresh_var(Some("T".into()), None);
(
"core_log".into(),
FunSignature {
args: vec![FuncArg {
name: "arg".into(),
ty: arg_ty.ty,
default_value: None,
is_vararg: false,
}],
ret: primitives.none,
vars: into_var_map([arg_ty]),
},
Arc::new(GenCall::new(Box::new(move |ctx, obj, fun, args, generator| {
gen_core_log(ctx, &obj, fun, &args, generator)?;
Ok(None)
}))),
)
}),
Box::new(|primitives, unifier| {
let arg_ty = unifier.get_fresh_var(Some("T".into()), None);
(
"rtio_log".into(),
FunSignature {
args: vec![
FuncArg {
name: "channel".into(),
ty: primitives.str,
default_value: None,
is_vararg: false,
},
FuncArg {
name: "arg".into(),
ty: arg_ty.ty,
default_value: None,
is_vararg: false,
},
],
ret: primitives.none,
vars: into_var_map([arg_ty]),
},
Arc::new(GenCall::new(Box::new(move |ctx, obj, fun, args, generator| {
gen_rtio_log(ctx, &obj, fun, &args, generator)?;
Ok(None)
}))),
)
}),
]
}
fn compile_method<T>(
&self,
obj: &PyAny,
@ -312,6 +372,7 @@ impl Nac3 {
let size_t = self.isa.get_size_type();
let (mut composer, mut builtins_def, mut builtins_ty) = TopLevelComposer::new(
self.builtins.clone(),
Self::get_lateinit_builtins(),
ComposerConfig { kernel_ann: Some("Kernel"), kernel_invariant_ann: "KernelInvariant" },
size_t,
);
@ -497,6 +558,11 @@ impl Nac3 {
.register_top_level(synthesized.pop().unwrap(), Some(resolver.clone()), "", false)
.unwrap();
// Process IRRT
let context = inkwell::context::Context::create();
let irrt = load_irrt(&context);
setup_irrt_exceptions(&context, &irrt, resolver.as_ref());
let fun_signature =
FunSignature { args: vec![], ret: self.primitive.none, vars: VarMap::new() };
let mut store = ConcreteTypeStore::new();
@ -625,7 +691,9 @@ impl Nac3 {
let buffer = buffer.as_slice().into();
membuffer.lock().push(buffer);
})));
let size_t = if self.isa == Isa::Host { 64 } else { 32 };
let size_t = Context::create()
.ptr_sized_int_type(&self.get_llvm_target_machine().get_target_data(), None)
.get_bit_width();
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
@ -644,6 +712,9 @@ impl Nac3 {
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 target_machine = self.llvm_options.create_target_machine().unwrap();
module.set_data_layout(&target_machine.get_target_data().get_data_layout());
module.set_triple(&target_machine.get_triple());
let builder = context.create_builder();
let (_, module, _) = gen_func_impl(
&context,
@ -662,7 +733,7 @@ impl Nac3 {
membuffer.lock().push(buffer);
});
let context = inkwell::context::Context::create();
// Link all modules into `main`.
let buffers = membuffers.lock();
let main = context
.create_module_from_ir(MemoryBuffer::create_from_memory_range(&buffers[0], "main"))
@ -691,8 +762,7 @@ impl Nac3 {
)
.unwrap();
main.link_in_module(load_irrt(&context))
.map_err(|err| CompileError::new_err(err.to_string()))?;
main.link_in_module(irrt).map_err(|err| CompileError::new_err(err.to_string()))?;
let mut function_iter = main.get_first_function();
while let Some(func) = function_iter {
@ -847,7 +917,7 @@ impl Nac3 {
Isa::RiscV32IMA => &timeline::NOW_PINNING_TIME_FNS,
Isa::CortexA9 | Isa::Host => &timeline::EXTERN_TIME_FNS,
};
let primitive: PrimitiveStore = TopLevelComposer::make_primitives(isa.get_size_type()).0;
let (primitive, _) = TopLevelComposer::make_primitives(isa.get_size_type());
let builtins = vec![
(
"now_mu".into(),
@ -863,6 +933,7 @@ impl Nac3 {
name: "t".into(),
ty: primitive.int64,
default_value: None,
is_vararg: false,
}],
ret: primitive.none,
vars: VarMap::new(),
@ -882,6 +953,7 @@ impl Nac3 {
name: "dt".into(),
ty: primitive.int64,
default_value: None,
is_vararg: false,
}],
ret: primitive.none,
vars: VarMap::new(),

View File

@ -1,12 +1,15 @@
use crate::PrimitivePythonId;
use inkwell::{
types::{BasicType, BasicTypeEnum},
values::BasicValueEnum,
module::Linkage,
types::BasicType,
values::{BasicValue, BasicValueEnum},
AddressSpace,
};
use itertools::Itertools;
use nac3core::{
codegen::{
classes::{NDArrayType, ProxyType},
model::*,
object::ndarray::{make_contiguous_strides, NDArray},
CodeGenContext, CodeGenerator,
},
symbol_resolver::{StaticValue, SymbolResolver, SymbolValue, ValueEnum},
@ -24,7 +27,7 @@ use nac3parser::ast::{self, StrRef};
use parking_lot::{Mutex, RwLock};
use pyo3::{
types::{PyDict, PyTuple},
PyAny, PyObject, PyResult, Python,
PyAny, PyErr, PyObject, PyResult, Python,
};
use std::{
collections::{HashMap, HashSet},
@ -34,8 +37,6 @@ use std::{
},
};
use crate::PrimitivePythonId;
pub enum PrimitiveValue {
I32(i32),
I64(i64),
@ -133,6 +134,8 @@ impl StaticValue for PythonValue {
format!("{}_const", self.id).as_str(),
);
global.set_constant(true);
// Set linkage of global to private to avoid name collisions
global.set_linkage(Linkage::Private);
global.set_initializer(&ctx.ctx.const_struct(
&[ctx.ctx.i32_type().const_int(u64::from(id), false).into()],
false,
@ -163,7 +166,7 @@ impl StaticValue for PythonValue {
PrimitiveValue::Bool(val) => {
ctx.ctx.i8_type().const_int(u64::from(*val), false).into()
}
PrimitiveValue::Str(val) => ctx.ctx.const_string(val.as_bytes(), true).into(),
PrimitiveValue::Str(val) => ctx.gen_string(generator, val).into(),
});
}
if let Some(global) = ctx.module.get_global(&self.id.to_string()) {
@ -351,7 +354,7 @@ impl InnerResolver {
Ok(Ok((ndarray, false)))
} else if ty_id == self.primitive_ids.tuple {
// do not handle type var param and concrete check here
Ok(Ok((unifier.add_ty(TypeEnum::TTuple { ty: vec![] }), false)))
Ok(Ok((unifier.add_ty(TypeEnum::TTuple { ty: vec![], is_vararg_ctx: false }), false)))
} else if ty_id == self.primitive_ids.option {
Ok(Ok((primitives.option, false)))
} else if ty_id == self.primitive_ids.none {
@ -555,7 +558,10 @@ impl InnerResolver {
Err(err) => return Ok(Err(err)),
_ => return Ok(Err("tuple type needs at least 1 type parameters".to_string()))
};
Ok(Ok((unifier.add_ty(TypeEnum::TTuple { ty: args }), true)))
Ok(Ok((
unifier.add_ty(TypeEnum::TTuple { ty: args, is_vararg_ctx: false }),
true,
)))
}
TypeEnum::TObj { params, obj_id, .. } => {
let subst = {
@ -797,7 +803,9 @@ impl InnerResolver {
.map(|elem| self.get_obj_type(py, elem, unifier, defs, primitives))
.collect();
let types = types?;
Ok(types.map(|types| unifier.add_ty(TypeEnum::TTuple { ty: types })))
Ok(types.map(|types| {
unifier.add_ty(TypeEnum::TTuple { ty: types, is_vararg_ctx: false })
}))
}
// special handling for option type since its class member layout in python side
// is special and cannot be mapped directly to a nac3 type as below
@ -972,7 +980,7 @@ impl InnerResolver {
} else if ty_id == self.primitive_ids.string || ty_id == self.primitive_ids.np_str_ {
let val: String = obj.extract().unwrap();
self.id_to_primitive.write().insert(id, PrimitiveValue::Str(val.clone()));
Ok(Some(ctx.ctx.const_string(val.as_bytes(), true).into()))
Ok(Some(ctx.gen_string(generator, val).into()))
} else if ty_id == self.primitive_ids.float || ty_id == self.primitive_ids.float64 {
let val: f64 = obj.extract().unwrap();
self.id_to_primitive.write().insert(id, PrimitiveValue::F64(val));
@ -991,8 +999,15 @@ impl InnerResolver {
}
_ => unreachable!("must be list"),
};
let ty = ctx.get_llvm_type(generator, elem_ty);
let size_t = generator.get_size_type(ctx.ctx);
let ty = if len == 0
&& matches!(&*ctx.unifier.get_ty_immutable(elem_ty), TypeEnum::TVar { .. })
{
// The default type for zero-length lists of unknown element type is size_t
size_t.into()
} else {
ctx.get_llvm_type(generator, elem_ty)
};
let arr_ty = ctx
.ctx
.struct_type(&[ty.ptr_type(AddressSpace::default()).into(), size_t.into()], false);
@ -1072,15 +1087,12 @@ impl InnerResolver {
let (ndarray_dtype, ndarray_ndims) =
unpack_ndarray_var_tys(&mut ctx.unifier, ndarray_ty);
let llvm_usize = generator.get_size_type(ctx.ctx);
let ndarray_dtype_llvm_ty = ctx.get_llvm_type(generator, ndarray_dtype);
let ndarray_llvm_ty = NDArrayType::new(generator, ctx.ctx, ndarray_dtype_llvm_ty);
let dtype = Any(ctx.get_llvm_type(generator, ndarray_dtype));
{
if self.global_value_ids.read().contains_key(&id) {
let global = ctx.module.get_global(&id_str).unwrap_or_else(|| {
ctx.module.add_global(
ndarray_llvm_ty.as_underlying_type(),
Struct(NDArray).get_type(generator, ctx.ctx),
Some(AddressSpace::default()),
&id_str,
)
@ -1100,103 +1112,143 @@ impl InnerResolver {
} else {
todo!("Unpacking literal of more than one element unimplemented")
};
let Ok(ndarray_ndims) = u64::try_from(ndarray_ndims) else {
let Ok(ndims) = u64::try_from(ndarray_ndims) else {
unreachable!("Expected u64 value for ndarray_ndims")
};
// Obtain the shape of the ndarray
let shape_tuple: &PyTuple = obj.getattr("shape")?.downcast()?;
assert_eq!(shape_tuple.len(), ndarray_ndims as usize);
let shape_values: Result<Option<Vec<_>>, _> = shape_tuple
assert_eq!(shape_tuple.len(), ndims as usize);
// The Rust type inferencer cannot figure this out
let shape_values: Result<Vec<Instance<'ctx, Int<SizeT>>>, PyErr> = shape_tuple
.iter()
.enumerate()
.map(|(i, elem)| {
self.get_obj_value(py, elem, ctx, generator, ctx.primitives.usize()).map_err(
|e| super::CompileError::new_err(format!("Error getting element {i}: {e}")),
)
let value = self
.get_obj_value(py, elem, ctx, generator, ctx.primitives.usize())
.map_err(|e| {
super::CompileError::new_err(format!("Error getting element {i}: {e}"))
})?
.unwrap();
let value = Int(SizeT).check_value(generator, ctx.ctx, value).unwrap();
Ok(value)
})
.collect();
let shape_values = shape_values?.unwrap();
let shape_values = llvm_usize.const_array(
&shape_values.into_iter().map(BasicValueEnum::into_int_value).collect_vec(),
);
let shape_values = shape_values?;
// Also use this opportunity to get the constant values of `shape_values` for calculating strides.
let shape_u64s = shape_values
.iter()
.map(|dim| {
assert!(dim.value.is_const());
dim.value.get_zero_extended_constant().unwrap()
})
.collect_vec();
let shape_values = Int(SizeT).const_array(generator, ctx.ctx, &shape_values);
// create a global for ndarray.shape and initialize it using the shape
let shape_global = ctx.module.add_global(
llvm_usize.array_type(ndarray_ndims as u32),
Array { len: AnyLen(ndims as u32), item: Int(SizeT) }.get_type(generator, ctx.ctx),
Some(AddressSpace::default()),
&(id_str.clone() + ".shape"),
);
shape_global.set_initializer(&shape_values);
shape_global.set_initializer(&shape_values.value);
// Obtain the (flattened) elements of the ndarray
let sz: usize = obj.getattr("size")?.extract()?;
let data: Result<Option<Vec<_>>, _> = (0..sz)
let data_values: Vec<Instance<'ctx, Any>> = (0..sz)
.map(|i| {
obj.getattr("flat")?.get_item(i).and_then(|elem| {
self.get_obj_value(py, elem, ctx, generator, ndarray_dtype).map_err(|e| {
super::CompileError::new_err(format!("Error getting element {i}: {e}"))
let value = self
.get_obj_value(py, elem, ctx, generator, ndarray_dtype)
.map_err(|e| {
super::CompileError::new_err(format!(
"Error getting element {i}: {e}"
))
})?
.unwrap();
let value = dtype.check_value(generator, ctx.ctx, value).unwrap();
Ok(value)
})
})
})
.collect();
let data = data?.unwrap().into_iter();
let data = match ndarray_dtype_llvm_ty {
BasicTypeEnum::ArrayType(ty) => {
ty.const_array(&data.map(BasicValueEnum::into_array_value).collect_vec())
}
BasicTypeEnum::FloatType(ty) => {
ty.const_array(&data.map(BasicValueEnum::into_float_value).collect_vec())
}
BasicTypeEnum::IntType(ty) => {
ty.const_array(&data.map(BasicValueEnum::into_int_value).collect_vec())
}
BasicTypeEnum::PointerType(ty) => {
ty.const_array(&data.map(BasicValueEnum::into_pointer_value).collect_vec())
}
BasicTypeEnum::StructType(ty) => {
ty.const_array(&data.map(BasicValueEnum::into_struct_value).collect_vec())
}
BasicTypeEnum::VectorType(_) => unreachable!(),
};
.try_collect()?;
let data = dtype.const_array(generator, ctx.ctx, &data_values);
// create a global for ndarray.data and initialize it using the elements
//
// NOTE: NDArray's `data` is `u8*`. Here, `data_global` is an array of `dtype`.
// We will have to cast it to an `u8*` later.
let data_global = ctx.module.add_global(
ndarray_dtype_llvm_ty.array_type(sz as u32),
Array { len: AnyLen(sz as u32), item: dtype }.get_type(generator, ctx.ctx),
Some(AddressSpace::default()),
&(id_str.clone() + ".data"),
);
data_global.set_initializer(&data);
data_global.set_initializer(&data.value);
// Get the constant itemsize.
let itemsize = dtype.get_type(generator, ctx.ctx).size_of().unwrap();
let itemsize = itemsize.get_zero_extended_constant().unwrap();
// Create the strides needed for ndarray.strides
let strides = make_contiguous_strides(itemsize, ndims, &shape_u64s);
let strides = strides
.into_iter()
.map(|stride| Int(SizeT).const_int(generator, ctx.ctx, stride))
.collect_vec();
let strides = Int(SizeT).const_array(generator, ctx.ctx, &strides);
// create a global for ndarray.strides and initialize it
let strides_global = ctx.module.add_global(
Array { len: AnyLen(ndims as u32), item: Int(Byte) }.get_type(generator, ctx.ctx),
Some(AddressSpace::default()),
&(id_str.clone() + ".strides"),
);
strides_global.set_initializer(&strides.value);
// create a global for the ndarray object and initialize it
let value = ndarray_llvm_ty.as_underlying_type().const_named_struct(&[
llvm_usize.const_int(ndarray_ndims, false).into(),
shape_global
.as_pointer_value()
.const_cast(llvm_usize.ptr_type(AddressSpace::default()))
.into(),
data_global
.as_pointer_value()
.const_cast(ndarray_dtype_llvm_ty.ptr_type(AddressSpace::default()))
.into(),
]);
// We are also doing [`Model::check_value`] instead of [`Model::believe_value`] to catch bugs.
let ndarray = ctx.module.add_global(
ndarray_llvm_ty.as_underlying_type(),
// NOTE: data_global is an array of dtype, we want a `u8*`.
let ndarray_data = Ptr(dtype).check_value(generator, ctx.ctx, data_global).unwrap();
let ndarray_data = Ptr(Int(Byte)).pointer_cast(generator, ctx, ndarray_data.value);
let ndarray_itemsize = Int(SizeT).const_int(generator, ctx.ctx, itemsize);
let ndarray_ndims = Int(SizeT).const_int(generator, ctx.ctx, ndims as u64);
let ndarray_shape =
Ptr(Int(SizeT)).check_value(generator, ctx.ctx, shape_global).unwrap();
let ndarray_strides =
Ptr(Int(SizeT)).check_value(generator, ctx.ctx, strides_global).unwrap();
let ndarray = Struct(NDArray).const_struct(
generator,
ctx.ctx,
&[
ndarray_data.value.as_basic_value_enum(),
ndarray_itemsize.value.as_basic_value_enum(),
ndarray_ndims.value.as_basic_value_enum(),
ndarray_shape.value.as_basic_value_enum(),
ndarray_strides.value.as_basic_value_enum(),
],
);
let ndarray_global = ctx.module.add_global(
Struct(NDArray).get_type(generator, ctx.ctx),
Some(AddressSpace::default()),
&id_str,
);
ndarray.set_initializer(&value);
ndarray_global.set_initializer(&ndarray.value);
Ok(Some(ndarray.as_pointer_value().into()))
Ok(Some(ndarray_global.as_pointer_value().into()))
} else if ty_id == self.primitive_ids.tuple {
let expected_ty_enum = ctx.unifier.get_ty_immutable(expected_ty);
let TypeEnum::TTuple { ty } = expected_ty_enum.as_ref() else { unreachable!() };
let TypeEnum::TTuple { ty, is_vararg_ctx: false } = expected_ty_enum.as_ref() else {
unreachable!()
};
let tup_tys = ty.iter();
let elements: &PyTuple = obj.downcast()?;

View File

@ -4,6 +4,9 @@ version = "0.1.0"
authors = ["M-Labs"]
edition = "2021"
[features]
no-escape-analysis = []
[dependencies]
itertools = "0.13"
crossbeam = "0.8"
@ -11,8 +14,8 @@ indexmap = "2.2"
parking_lot = "0.12"
rayon = "1.8"
nac3parser = { path = "../nac3parser" }
strum = "0.26.2"
strum_macros = "0.26.4"
strum = "0.26"
strum_macros = "0.26"
[dependencies.inkwell]
version = "0.4"

View File

@ -8,37 +8,50 @@ use std::{
};
fn main() {
const FILE: &str = "src/codegen/irrt/irrt.cpp";
// Define relevant directories
let out_dir = env::var("OUT_DIR").unwrap();
let out_dir = Path::new(&out_dir);
let irrt_dir = Path::new("irrt");
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.
*/
let flags: &[&str] = &[
let mut flags: Vec<&str> = vec![
"--target=wasm32",
FILE,
"-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",
"-o",
"-",
"-I",
irrt_dir.to_str().unwrap(),
irrt_cpp_path.to_str().unwrap(),
];
println!("cargo:rerun-if-changed={FILE}");
let out_dir = env::var("OUT_DIR").unwrap();
let out_path = Path::new(&out_dir);
match env::var("PROFILE").as_deref() {
Ok("debug") => {
flags.push("-O0");
flags.push("-DIRRT_DEBUG_ASSERT");
}
Ok("release") => {
flags.push("-O3");
}
flavor => panic!("Unknown or missing build flavor {flavor:?}"),
}
// Tell Cargo to rerun if any file under `irrt_dir` (recursive) changes
println!("cargo:rerun-if-changed={}", irrt_dir.to_str().unwrap());
// Compile IRRT and capture the LLVM IR output
let output = Command::new("clang-irrt")
.args(flags)
.output()
@ -52,7 +65,17 @@ 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::new(r"(?ms:^define.*?\}$)|(?m:^declare.*?$)").unwrap();
// Filter out irrelevant IR
//
// Regex:
// - `(?ms:^define.*?\}$)` captures LLVM `define` blocks
// - `(?m:^declare.*?$)` captures LLVM `declare` lines
// - `(?m:^%.+?=\s*type\s*\{.+?\}$)` captures LLVM `type` declarations
// - `(?m:^@.+?=.+$)` captures global constants
let regex_filter = Regex::new(
r"(?ms:^define.*?\}$)|(?m:^declare.*?$)|(?m:^%.+?=\s*type\s*\{.+?\}$)|(?m:^@.+?=.+$)",
)
.unwrap();
for f in regex_filter.captures_iter(&output) {
assert_eq!(f.len(), 1);
filtered_output.push_str(&f[0]);
@ -63,18 +86,24 @@ fn main() {
.unwrap()
.replace_all(&filtered_output, "");
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();
// For debugging
// Doing `DEBUG_DUMP_IRRT=1 cargo build -p nac3core` dumps the LLVM IR generated
const DEBUG_DUMP_IRRT: &str = "DEBUG_DUMP_IRRT";
println!("cargo:rerun-if-env-changed={DEBUG_DUMP_IRRT}");
if env::var(DEBUG_DUMP_IRRT).is_ok() {
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();
}
// Assemble the emitted and filtered IR to .bc
// That .bc will be integrated into nac3core's codegen
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();

15
nac3core/irrt/irrt.cpp Normal file
View File

@ -0,0 +1,15 @@
#include <irrt/exception.hpp>
#include <irrt/int_types.hpp>
#include <irrt/list.hpp>
#include <irrt/math_util.hpp>
#include <irrt/ndarray/array.hpp>
#include <irrt/ndarray/basic.hpp>
#include <irrt/ndarray/broadcast.hpp>
#include <irrt/ndarray/def.hpp>
#include <irrt/ndarray/indexing.hpp>
#include <irrt/ndarray/iter.hpp>
#include <irrt/ndarray/matmul.hpp>
#include <irrt/ndarray/reshape.hpp>
#include <irrt/ndarray/transpose.hpp>
#include <irrt/original.hpp>
#include <irrt/slice.hpp>

View File

@ -0,0 +1,9 @@
#pragma once
#include <irrt/int_types.hpp>
template <typename SizeT> struct CSlice
{
uint8_t *base;
SizeT len;
};

View File

@ -0,0 +1,20 @@
#pragma once
#include <irrt/int_types.hpp>
namespace cstr
{
/**
* @brief Implementation of `strlen()`.
*/
uint32_t length(const char *str)
{
uint32_t length = 0;
while (*str != '\0')
{
length++;
str++;
}
return length;
}
} // namespace cstr

View File

@ -0,0 +1,21 @@
#pragma once
#ifdef IRRT_DEBUG_ASSERT
#define IRRT_DEBUG_ASSERT_BOOL true
#else
#define IRRT_DEBUG_ASSERT_BOOL false
#endif
#define raise_debug_assert(SizeT, msg, param1, param2, param3) \
raise_exception(SizeT, EXN_ASSERTION_ERROR, \
"IRRT debug assert failed: " msg, param1, param2, param3);
#define debug_assert_eq(SizeT, lhs, rhs) \
if (IRRT_DEBUG_ASSERT_BOOL && (lhs) != (rhs)) { \
raise_debug_assert(SizeT, "LHS = {0}. RHS = {1}", lhs, rhs, NO_PARAM); \
}
#define debug_assert(SizeT, expr) \
if (IRRT_DEBUG_ASSERT_BOOL && !(expr)) { \
raise_debug_assert(SizeT, "Got false.", NO_PARAM, NO_PARAM, NO_PARAM); \
}

View File

@ -0,0 +1,80 @@
#pragma once
#include <irrt/cslice.hpp>
#include <irrt/cstr_util.hpp>
#include <irrt/int_types.hpp>
/**
* @brief The int type of ARTIQ exception IDs.
*/
typedef int32_t ExceptionId;
/*
* Set of exceptions C++ IRRT can use.
* Must be synchronized with `setup_irrt_exceptions` in `nac3core/src/codegen/irrt/mod.rs`.
*/
extern "C"
{
ExceptionId EXN_INDEX_ERROR;
ExceptionId EXN_VALUE_ERROR;
ExceptionId EXN_ASSERTION_ERROR;
ExceptionId EXN_TYPE_ERROR;
}
/**
* @brief Extern function to `__nac3_raise`
*
* The parameter `err` could be `Exception<int32_t>` or `Exception<int64_t>`. The caller
* must make sure to pass `Exception`s with the correct `SizeT` depending on the `size_t` of the runtime.
*/
extern "C" void __nac3_raise(void *err);
namespace
{
/**
* @brief NAC3's Exception struct
*/
template <typename SizeT> struct Exception
{
ExceptionId id;
CSlice<SizeT> filename;
int32_t line;
int32_t column;
CSlice<SizeT> function;
CSlice<SizeT> msg;
int64_t params[3];
};
const int64_t NO_PARAM = 0;
template <typename SizeT>
void _raise_exception_helper(ExceptionId id, const char *filename, int32_t line, const char *function, const char *msg,
int64_t param0, int64_t param1, int64_t param2)
{
Exception<SizeT> e = {
.id = id,
.filename = {.base = (uint8_t *)filename, .len = (int32_t)cstr::length(filename)},
.line = line,
.column = 0,
.function = {.base = (uint8_t *)function, .len = (int32_t)cstr::length(function)},
.msg = {.base = (uint8_t *)msg, .len = (int32_t)cstr::length(msg)},
};
e.params[0] = param0;
e.params[1] = param1;
e.params[2] = param2;
__nac3_raise((void *)&e);
__builtin_unreachable();
}
/**
* @brief Raise an exception with location details (location in the IRRT source files).
* @param SizeT The runtime `size_t` type.
* @param id The ID of the exception to raise.
* @param msg A global constant C-string of the error message.
*
* `param0` and `param2` are optional format arguments of `msg`. They should be set to
* `NO_PARAM` to indicate they are unused.
*/
#define raise_exception(SizeT, id, msg, param0, param1, param2) \
_raise_exception_helper<SizeT>(id, __FILE__, __LINE__, __FUNCTION__, msg, param0, param1, param2)
} // namespace

View File

@ -0,0 +1,8 @@
#pragma once
using int8_t = _BitInt(8);
using uint8_t = unsigned _BitInt(8);
using int32_t = _BitInt(32);
using uint32_t = unsigned _BitInt(32);
using int64_t = _BitInt(64);
using uint64_t = unsigned _BitInt(64);

View File

@ -0,0 +1,19 @@
#pragma once
#include <irrt/int_types.hpp>
#include <irrt/slice.hpp>
namespace
{
/**
* @brief A list in NAC3.
*
* The `items` field is opaque. You must rely on external contexts to
* know how to interpret it.
*/
template <typename SizeT> struct List
{
uint8_t *items;
SizeT len;
};
} // namespace

View File

@ -0,0 +1,14 @@
#pragma once
namespace
{
template <typename T> const T &max(const T &a, const T &b)
{
return a > b ? a : b;
}
template <typename T> const T &min(const T &a, const T &b)
{
return a > b ? b : a;
}
} // namespace

View File

@ -0,0 +1,130 @@
#pragma once
#include <irrt/debug.hpp>
#include <irrt/exception.hpp>
#include <irrt/int_types.hpp>
#include <irrt/list.hpp>
#include <irrt/ndarray/basic.hpp>
#include <irrt/ndarray/def.hpp>
namespace
{
namespace ndarray
{
namespace array
{
template <typename SizeT>
void set_and_validate_list_shape_helper(SizeT axis, List<SizeT> *list, SizeT ndims, SizeT *shape)
{
if (shape[axis] == -1)
{
// Dimension is unspecified. Set it.
shape[axis] = list->len;
}
else
{
// Dimension is specified. Check.
if (shape[axis] != list->len)
{
// Mismatch, throw an error.
// NOTE: NumPy's error message is more complex and needs more PARAMS to display.
raise_exception(SizeT, EXN_VALUE_ERROR,
"The requested array has an inhomogenous shape "
"after {0} dimension(s).",
axis, shape[axis], list->len);
}
}
if (axis + 1 == ndims)
{
// `list` has type `list[ItemType]`
// Do nothing
}
else
{
// `list` has type `list[list[...]]`
List<SizeT> **lists = (List<SizeT> **)(list->items);
for (SizeT i = 0; i < list->len; i++)
{
set_and_validate_list_shape_helper<SizeT>(axis + 1, lists[i], ndims, shape);
}
}
}
// TODO: Document me
template <typename SizeT> void set_and_validate_list_shape(List<SizeT> *list, SizeT ndims, SizeT *shape)
{
for (SizeT axis = 0; axis < ndims; axis++)
{
shape[axis] = -1; // Sentinel to say this dimension is unspecified.
}
set_and_validate_list_shape_helper<SizeT>(0, list, ndims, shape);
}
template <typename SizeT>
void write_list_to_array_helper(SizeT axis, SizeT *index, List<SizeT> *list, NDArray<SizeT> *ndarray)
{
debug_assert_eq(SizeT, list->len, ndarray->shape[axis]);
if (IRRT_DEBUG_ASSERT_BOOL)
{
if (!ndarray::basic::is_c_contiguous(ndarray))
{
raise_debug_assert(SizeT, "ndarray is not C-contiguous", ndarray->strides[0], ndarray->strides[1],
NO_PARAM);
}
}
if (axis + 1 == ndarray->ndims)
{
// `list` has type `list[ItemType]`
// `ndarray` is contiguous, so we can do this, and this is fast.
uint8_t *dst = ndarray->data + (ndarray->itemsize * (*index));
__builtin_memcpy(dst, list->items, ndarray->itemsize * list->len);
*index += list->len;
}
else
{
// `list` has type `list[list[...]]`
List<SizeT> **lists = (List<SizeT> **)(list->items);
for (SizeT i = 0; i < list->len; i++)
{
write_list_to_array_helper<SizeT>(axis + 1, index, lists[i], ndarray);
}
}
}
// TODO: Document me
template <typename SizeT> void write_list_to_array(List<SizeT> *list, NDArray<SizeT> *ndarray)
{
SizeT index = 0;
write_list_to_array_helper<SizeT>((SizeT)0, &index, list, ndarray);
}
} // namespace array
} // namespace ndarray
} // namespace
extern "C"
{
using namespace ndarray::array;
void __nac3_ndarray_array_set_and_validate_list_shape(List<int32_t> *list, int32_t ndims, int32_t *shape)
{
set_and_validate_list_shape(list, ndims, shape);
}
void __nac3_ndarray_array_set_and_validate_list_shape64(List<int64_t> *list, int64_t ndims, int64_t *shape)
{
set_and_validate_list_shape(list, ndims, shape);
}
void __nac3_ndarray_array_write_list_to_array(List<int32_t> *list, NDArray<int32_t> *ndarray)
{
write_list_to_array(list, ndarray);
}
void __nac3_ndarray_array_write_list_to_array64(List<int64_t> *list, NDArray<int64_t> *ndarray)
{
write_list_to_array(list, ndarray);
}
}

View File

@ -0,0 +1,380 @@
#pragma once
#include <irrt/debug.hpp>
#include <irrt/exception.hpp>
#include <irrt/int_types.hpp>
#include <irrt/ndarray/def.hpp>
namespace
{
namespace ndarray
{
namespace basic
{
/**
* @brief Asserts that `shape` does not contain negative dimensions.
*
* @param ndims Number of dimensions in `shape`
* @param shape The shape to check on
*/
template <typename SizeT> void assert_shape_no_negative(SizeT ndims, const SizeT *shape)
{
for (SizeT axis = 0; axis < ndims; axis++)
{
if (shape[axis] < 0)
{
raise_exception(SizeT, EXN_VALUE_ERROR,
"negative dimensions are not allowed; axis {0} "
"has dimension {1}",
axis, shape[axis], NO_PARAM);
}
}
}
/**
* @brief Check two shapes are the same in the context of writing outputting to an ndarray.
*
* This function throws error messages for output shape mismatches.
*/
template <typename SizeT>
void assert_output_shape_same(SizeT ndarray_ndims, const SizeT *ndarray_shape, SizeT output_ndims,
const SizeT *output_shape)
{
if (ndarray_ndims != output_ndims)
{
// There is no corresponding NumPy error message like this.
raise_exception(SizeT, EXN_VALUE_ERROR, "Cannot write output of ndims {0} to an ndarray with ndims {1}",
output_ndims, ndarray_ndims, NO_PARAM);
}
for (SizeT axis = 0; axis < ndarray_ndims; axis++)
{
if (ndarray_shape[axis] != output_shape[axis])
{
// There is no corresponding NumPy error message like this.
raise_exception(SizeT, EXN_VALUE_ERROR,
"Mismatched dimensions on axis {0}, output has "
"dimension {1}, but destination ndarray has dimension {2}.",
axis, output_shape[axis], ndarray_shape[axis]);
}
}
}
/**
* @brief Returns the number of elements of an ndarray given its shape.
*
* @param ndims Number of dimensions in `shape`
* @param shape The shape of the ndarray
*/
template <typename SizeT> SizeT calc_size_from_shape(SizeT ndims, const SizeT *shape)
{
SizeT size = 1;
for (SizeT axis = 0; axis < ndims; axis++)
size *= shape[axis];
return size;
}
/**
* @brief Compute the array indices of the `nth` (0-based) element of an ndarray given only its shape.
*
* @param ndims Number of elements in `shape` and `indices`
* @param shape The shape of the ndarray
* @param indices The returned indices indexing the ndarray with shape `shape`.
* @param nth The index of the element of interest.
*/
template <typename SizeT> void set_indices_by_nth(SizeT ndims, const SizeT *shape, SizeT *indices, SizeT nth)
{
for (SizeT i = 0; i < ndims; i++)
{
SizeT axis = ndims - i - 1;
SizeT dim = shape[axis];
indices[axis] = nth % dim;
nth /= dim;
}
}
/**
* @brief Return the number of elements of an `ndarray`
*
* This function corresponds to `<an_ndarray>.size`
*/
template <typename SizeT> SizeT size(const NDArray<SizeT> *ndarray)
{
return calc_size_from_shape(ndarray->ndims, ndarray->shape);
}
/**
* @brief Return of the number of its content of an `ndarray`.
*
* This function corresponds to `<an_ndarray>.nbytes`.
*/
template <typename SizeT> SizeT nbytes(const NDArray<SizeT> *ndarray)
{
return size(ndarray) * ndarray->itemsize;
}
/**
* @brief Get the `len()` of an ndarray, and asserts that `ndarray` is a sized object.
*
* This function corresponds to `<an_ndarray>.__len__`.
*
* @param dst_length The returned result
*/
template <typename SizeT> SizeT len(const NDArray<SizeT> *ndarray)
{
// numpy prohibits `__len__` on unsized objects
if (ndarray->ndims == 0)
{
raise_exception(SizeT, EXN_TYPE_ERROR, "len() of unsized object", NO_PARAM, NO_PARAM, NO_PARAM);
}
else
{
return ndarray->shape[0];
}
}
/**
* @brief Return a boolean indicating if `ndarray` is (C-)contiguous.
*
* You may want to see: ndarray's rules for C-contiguity: https://github.com/numpy/numpy/blob/df256d0d2f3bc6833699529824781c58f9c6e697/numpy/core/src/multiarray/flagsobject.c#L95C1-L99C45
*/
template <typename SizeT> bool is_c_contiguous(const NDArray<SizeT> *ndarray)
{
// Other references:
// - tinynumpy's implementation: https://github.com/wadetb/tinynumpy/blob/0d23d22e07062ffab2afa287374c7b366eebdda1/tinynumpy/tinynumpy.py#L102
// - ndarray's flags["C_CONTIGUOUS"]: https://numpy.org/doc/stable/reference/generated/numpy.ndarray.flags.html#numpy.ndarray.flags
// - ndarray's rules for C-contiguity: https://github.com/numpy/numpy/blob/df256d0d2f3bc6833699529824781c58f9c6e697/numpy/core/src/multiarray/flagsobject.c#L95C1-L99C45
// From https://github.com/numpy/numpy/blob/df256d0d2f3bc6833699529824781c58f9c6e697/numpy/core/src/multiarray/flagsobject.c#L95C1-L99C45:
//
// The traditional rule is that for an array to be flagged as C contiguous,
// the following must hold:
//
// strides[-1] == itemsize
// strides[i] == shape[i+1] * strides[i + 1]
// [...]
// According to these rules, a 0- or 1-dimensional array is either both
// C- and F-contiguous, or neither; and an array with 2+ dimensions
// can be C- or F- contiguous, or neither, but not both. Though there
// there are exceptions for arrays with zero or one item, in the first
// case the check is relaxed up to and including the first dimension
// with shape[i] == 0. In the second case `strides == itemsize` will
// can be true for all dimensions and both flags are set.
if (ndarray->ndims == 0)
{
return true;
}
if (ndarray->strides[ndarray->ndims - 1] != ndarray->itemsize)
{
return false;
}
for (SizeT i = 1; i < ndarray->ndims; i++)
{
SizeT axis_i = ndarray->ndims - i - 1;
if (ndarray->strides[axis_i] != ndarray->shape[axis_i + 1] * ndarray->strides[axis_i + 1])
{
return false;
}
}
return true;
}
/**
* @brief Return the pointer to the element indexed by `indices`.
*/
template <typename SizeT> uint8_t *get_pelement_by_indices(const NDArray<SizeT> *ndarray, const SizeT *indices)
{
uint8_t *element = ndarray->data;
for (SizeT dim_i = 0; dim_i < ndarray->ndims; dim_i++)
element += indices[dim_i] * ndarray->strides[dim_i];
return element;
}
/**
* @brief Convenience function. Like `get_pelement_by_indices` but
* reinterprets the element pointer.
*/
template <typename SizeT, typename T> T *get_ptr(const NDArray<SizeT> *ndarray, const SizeT *indices)
{
return (T *)get_pelement_by_indices(ndarray, indices);
}
/**
* @brief Return the pointer to the nth (0-based) element in a flattened view of `ndarray`.
*
* This function does no bound check.
*/
template <typename SizeT> uint8_t *get_nth_pelement(const NDArray<SizeT> *ndarray, SizeT nth)
{
uint8_t *element = ndarray->data;
for (SizeT i = 0; i < ndarray->ndims; i++)
{
SizeT axis = ndarray->ndims - i - 1;
SizeT dim = ndarray->shape[axis];
element += ndarray->strides[axis] * (nth % dim);
nth /= dim;
}
return element;
}
/**
* @brief Update the strides of an ndarray given an ndarray `shape`
* and assuming that the ndarray is fully c-contagious.
*
* You might want to read https://ajcr.net/stride-guide-part-1/.
*/
template <typename SizeT> void set_strides_by_shape(NDArray<SizeT> *ndarray)
{
SizeT stride_product = 1;
for (SizeT i = 0; i < ndarray->ndims; i++)
{
SizeT axis = ndarray->ndims - i - 1;
ndarray->strides[axis] = stride_product * ndarray->itemsize;
stride_product *= ndarray->shape[axis];
}
}
/**
* @brief Set an element in `ndarray`.
*
* @param pelement Pointer to the element in `ndarray` to be set.
* @param pvalue Pointer to the value `pelement` will be set to.
*/
template <typename SizeT> void set_pelement_value(NDArray<SizeT> *ndarray, uint8_t *pelement, const uint8_t *pvalue)
{
__builtin_memcpy(pelement, pvalue, ndarray->itemsize);
}
/**
* @brief Copy data from one ndarray to another of the exact same size and itemsize.
*
* Both ndarrays will be viewed in their flatten views when copying the elements.
*/
template <typename SizeT> void copy_data(const NDArray<SizeT> *src_ndarray, NDArray<SizeT> *dst_ndarray)
{
// TODO: Make this faster with memcpy
debug_assert_eq(SizeT, src_ndarray->itemsize, dst_ndarray->itemsize);
for (SizeT i = 0; i < size(src_ndarray); i++)
{
auto src_element = ndarray::basic::get_nth_pelement(src_ndarray, i);
auto dst_element = ndarray::basic::get_nth_pelement(dst_ndarray, i);
ndarray::basic::set_pelement_value(dst_ndarray, dst_element, src_element);
}
}
} // namespace basic
} // namespace ndarray
} // namespace
extern "C"
{
using namespace ndarray::basic;
void __nac3_ndarray_util_assert_shape_no_negative(int32_t ndims, int32_t *shape)
{
assert_shape_no_negative(ndims, shape);
}
void __nac3_ndarray_util_assert_shape_no_negative64(int64_t ndims, int64_t *shape)
{
assert_shape_no_negative(ndims, shape);
}
void __nac3_ndarray_util_assert_output_shape_same(int32_t ndarray_ndims, const int32_t *ndarray_shape,
int32_t output_ndims, const int32_t *output_shape)
{
assert_output_shape_same(ndarray_ndims, ndarray_shape, output_ndims, output_shape);
}
void __nac3_ndarray_util_assert_output_shape_same64(int64_t ndarray_ndims, const int64_t *ndarray_shape,
int64_t output_ndims, const int64_t *output_shape)
{
assert_output_shape_same(ndarray_ndims, ndarray_shape, output_ndims, output_shape);
}
uint32_t __nac3_ndarray_size(NDArray<int32_t> *ndarray)
{
return size(ndarray);
}
uint64_t __nac3_ndarray_size64(NDArray<int64_t> *ndarray)
{
return size(ndarray);
}
uint32_t __nac3_ndarray_nbytes(NDArray<int32_t> *ndarray)
{
return nbytes(ndarray);
}
uint64_t __nac3_ndarray_nbytes64(NDArray<int64_t> *ndarray)
{
return nbytes(ndarray);
}
int32_t __nac3_ndarray_len(NDArray<int32_t> *ndarray)
{
return len(ndarray);
}
int64_t __nac3_ndarray_len64(NDArray<int64_t> *ndarray)
{
return len(ndarray);
}
bool __nac3_ndarray_is_c_contiguous(NDArray<int32_t> *ndarray)
{
return is_c_contiguous(ndarray);
}
bool __nac3_ndarray_is_c_contiguous64(NDArray<int64_t> *ndarray)
{
return is_c_contiguous(ndarray);
}
uint8_t *__nac3_ndarray_get_nth_pelement(const NDArray<int32_t> *ndarray, int32_t nth)
{
return get_nth_pelement(ndarray, nth);
}
uint8_t *__nac3_ndarray_get_nth_pelement64(const NDArray<int64_t> *ndarray, int64_t nth)
{
return get_nth_pelement(ndarray, nth);
}
uint8_t *__nac3_ndarray_get_pelement_by_indices(const NDArray<int32_t> *ndarray, int32_t *indices)
{
return get_pelement_by_indices(ndarray, indices);
}
uint8_t *__nac3_ndarray_get_pelement_by_indices64(const NDArray<int64_t> *ndarray, int64_t *indices)
{
return get_pelement_by_indices(ndarray, indices);
}
void __nac3_ndarray_set_strides_by_shape(NDArray<int32_t> *ndarray)
{
set_strides_by_shape(ndarray);
}
void __nac3_ndarray_set_strides_by_shape64(NDArray<int64_t> *ndarray)
{
set_strides_by_shape(ndarray);
}
void __nac3_ndarray_copy_data(NDArray<int32_t> *src_ndarray, NDArray<int32_t> *dst_ndarray)
{
copy_data(src_ndarray, dst_ndarray);
}
void __nac3_ndarray_copy_data64(NDArray<int64_t> *src_ndarray, NDArray<int64_t> *dst_ndarray)
{
copy_data(src_ndarray, dst_ndarray);
}
}

View File

@ -0,0 +1,188 @@
#pragma once
#include <irrt/int_types.hpp>
#include <irrt/ndarray/def.hpp>
#include <irrt/slice.hpp>
namespace
{
template <typename SizeT> struct ShapeEntry
{
SizeT ndims;
SizeT *shape;
};
} // namespace
namespace
{
namespace ndarray
{
namespace broadcast
{
/**
* @brief Return true if `src_shape` can broadcast to `dst_shape`.
*
* See https://numpy.org/doc/stable/user/basics.broadcasting.html
*/
template <typename SizeT>
bool can_broadcast_shape_to(SizeT target_ndims, const SizeT *target_shape, SizeT src_ndims, const SizeT *src_shape)
{
if (src_ndims > target_ndims)
{
return false;
}
for (SizeT i = 0; i < src_ndims; i++)
{
SizeT target_dim = target_shape[target_ndims - i - 1];
SizeT src_dim = src_shape[src_ndims - i - 1];
if (!(src_dim == 1 || target_dim == src_dim))
{
return false;
}
}
return true;
}
/**
* @brief Performs `np.broadcast_shapes(<shapes>)`
*
* @param num_shapes Number of entries in `shapes`
* @param shapes The list of shape to do `np.broadcast_shapes` on.
* @param dst_ndims The length of `dst_shape`.
* `dst_ndims` must be `max([shape.ndims for shape in shapes])`, but the caller has to calculate it/provide it.
* for this function since they should already know in order to allocate `dst_shape` in the first place.
* @param dst_shape The resulting shape. Must be pre-allocated by the caller. This function calculate the result
* of `np.broadcast_shapes` and write it here.
*/
template <typename SizeT>
void broadcast_shapes(SizeT num_shapes, const ShapeEntry<SizeT> *shapes, SizeT dst_ndims, SizeT *dst_shape)
{
for (SizeT dst_axis = 0; dst_axis < dst_ndims; dst_axis++)
{
dst_shape[dst_axis] = 1;
}
#ifdef IRRT_DEBUG_ASSERT
SizeT max_ndims_found = 0;
#endif
for (SizeT i = 0; i < num_shapes; i++)
{
ShapeEntry<SizeT> entry = shapes[i];
// Check pre-condition: `dst_ndims` must be `max([shape.ndims for shape in shapes])`
debug_assert(SizeT, entry.ndims <= dst_ndims);
#ifdef IRRT_DEBUG_ASSERT
max_ndims_found = max(max_ndims_found, entry.ndims);
#endif
for (SizeT j = 0; j < entry.ndims; j++)
{
SizeT entry_axis = entry.ndims - j - 1;
SizeT dst_axis = dst_ndims - j - 1;
SizeT entry_dim = entry.shape[entry_axis];
SizeT dst_dim = dst_shape[dst_axis];
if (dst_dim == 1)
{
dst_shape[dst_axis] = entry_dim;
}
else if (entry_dim == 1 || entry_dim == dst_dim)
{
// Do nothing
}
else
{
raise_exception(SizeT, EXN_VALUE_ERROR,
"shape mismatch: objects cannot be broadcast "
"to a single shape.",
NO_PARAM, NO_PARAM, NO_PARAM);
}
}
}
// Check pre-condition: `dst_ndims` must be `max([shape.ndims for shape in shapes])`
debug_assert_eq(SizeT, max_ndims_found, dst_ndims);
}
/**
* @brief Perform `np.broadcast_to(<ndarray>, <target_shape>)` and appropriate assertions.
*
* This function attempts to broadcast `src_ndarray` to a new shape defined by `dst_ndarray.shape`,
* and return the result by modifying `dst_ndarray`.
*
* # Notes on `dst_ndarray`
* The caller is responsible for allocating space for the resulting ndarray.
* Here is what this function expects from `dst_ndarray` when called:
* - `dst_ndarray->data` does not have to be initialized.
* - `dst_ndarray->itemsize` does not have to be initialized.
* - `dst_ndarray->ndims` must be initialized, determining the length of `dst_ndarray->shape`
* - `dst_ndarray->shape` must be allocated, and must contain the desired target broadcast shape.
* - `dst_ndarray->strides` must be allocated, through it can contain uninitialized values.
* When this function call ends:
* - `dst_ndarray->data` is set to `src_ndarray->data` (`dst_ndarray` is just a view to `src_ndarray`)
* - `dst_ndarray->itemsize` is set to `src_ndarray->itemsize`
* - `dst_ndarray->ndims` is unchanged.
* - `dst_ndarray->shape` is unchanged.
* - `dst_ndarray->strides` is updated accordingly by how ndarray broadcast_to works.
*/
template <typename SizeT> void broadcast_to(const NDArray<SizeT> *src_ndarray, NDArray<SizeT> *dst_ndarray)
{
if (!ndarray::broadcast::can_broadcast_shape_to(dst_ndarray->ndims, dst_ndarray->shape, src_ndarray->ndims,
src_ndarray->shape))
{
raise_exception(SizeT, EXN_VALUE_ERROR, "operands could not be broadcast together", NO_PARAM, NO_PARAM,
NO_PARAM);
}
dst_ndarray->data = src_ndarray->data;
dst_ndarray->itemsize = src_ndarray->itemsize;
for (SizeT i = 0; i < dst_ndarray->ndims; i++)
{
SizeT src_axis = src_ndarray->ndims - i - 1;
SizeT dst_axis = dst_ndarray->ndims - i - 1;
if (src_axis < 0 || (src_ndarray->shape[src_axis] == 1 && dst_ndarray->shape[dst_axis] != 1))
{
// Freeze the steps in-place
dst_ndarray->strides[dst_axis] = 0;
}
else
{
dst_ndarray->strides[dst_axis] = src_ndarray->strides[src_axis];
}
}
}
} // namespace broadcast
} // namespace ndarray
} // namespace
extern "C"
{
using namespace ndarray::broadcast;
void __nac3_ndarray_broadcast_to(NDArray<int32_t> *src_ndarray, NDArray<int32_t> *dst_ndarray)
{
broadcast_to(src_ndarray, dst_ndarray);
}
void __nac3_ndarray_broadcast_to64(NDArray<int64_t> *src_ndarray, NDArray<int64_t> *dst_ndarray)
{
broadcast_to(src_ndarray, dst_ndarray);
}
void __nac3_ndarray_broadcast_shapes(int32_t num_shapes, const ShapeEntry<int32_t> *shapes, int32_t dst_ndims,
int32_t *dst_shape)
{
broadcast_shapes(num_shapes, shapes, dst_ndims, dst_shape);
}
void __nac3_ndarray_broadcast_shapes64(int64_t num_shapes, const ShapeEntry<int64_t> *shapes, int64_t dst_ndims,
int64_t *dst_shape)
{
broadcast_shapes(num_shapes, shapes, dst_ndims, dst_shape);
}
}

View File

@ -0,0 +1,47 @@
#pragma once
#include <irrt/int_types.hpp>
namespace
{
/**
* @brief The NDArray object
*
* The official numpy implementations: https://github.com/numpy/numpy/blob/735a477f0bc2b5b84d0e72d92f224bde78d4e069/doc/source/reference/c-api/types-and-structures.rst
*/
template <typename SizeT> struct NDArray
{
/**
* @brief The underlying data this `ndarray` is pointing to.
*
* Must be set to `nullptr` to indicate that this NDArray's `data` is uninitialized.
*/
uint8_t *data;
/**
* @brief The number of bytes of a single element in `data`.
*/
SizeT itemsize;
/**
* @brief The number of dimensions of this shape.
*/
SizeT ndims;
/**
* @brief The NDArray shape, with length equal to `ndims`.
*
* Note that it may contain 0.
*/
SizeT *shape;
/**
* @brief Array strides, with length equal to `ndims`
*
* The stride values are in units of bytes, not number of elements.
*
* Note that `strides` can have negative values.
*/
SizeT *strides;
};
} // namespace

View File

@ -0,0 +1,242 @@
#pragma once
#include <irrt/exception.hpp>
#include <irrt/int_types.hpp>
#include <irrt/ndarray/basic.hpp>
#include <irrt/ndarray/def.hpp>
#include <irrt/slice.hpp>
namespace
{
typedef uint8_t NDIndexType;
/**
* @brief A single element index
*
* `data` points to a `int32_t`.
*/
const NDIndexType ND_INDEX_TYPE_SINGLE_ELEMENT = 0;
/**
* @brief A slice index
*
* `data` points to a `Slice<int32_t>`.
*/
const NDIndexType ND_INDEX_TYPE_SLICE = 1;
/**
* @brief `np.newaxis` / `None`
*
* `data` is unused.
*/
const NDIndexType ND_INDEX_TYPE_NEWAXIS = 2;
/**
* @brief `Ellipsis` / `...`
*
* `data` is unused.
*/
const NDIndexType ND_INDEX_TYPE_ELLIPSIS = 3;
/**
* @brief An index used in ndarray indexing
*/
struct NDIndex
{
/**
* @brief Enum tag to specify the type of index.
*
* Please see comments of each enum constant.
*/
NDIndexType type;
/**
* @brief The accompanying data associated with `type`.
*
* Please see comments of each enum constant.
*/
uint8_t *data;
};
} // namespace
namespace
{
namespace ndarray
{
namespace indexing
{
/**
* @brief Perform ndarray "basic indexing" (https://numpy.org/doc/stable/user/basics.indexing.html#basic-indexing)
*
* This function is very similar to performing `dst_ndarray = src_ndarray[indices]` in Python.
*
* This function also does proper assertions on `indices` to check for out of bounds access.
*
* # Notes on `dst_ndarray`
* The caller is responsible for allocating space for the resulting ndarray.
* Here is what this function expects from `dst_ndarray` when called:
* - `dst_ndarray->data` does not have to be initialized.
* - `dst_ndarray->itemsize` does not have to be initialized.
* - `dst_ndarray->ndims` must be initialized, and it must be equal to the expected `ndims` of the `dst_ndarray` after
* indexing `src_ndarray` with `indices`.
* - `dst_ndarray->shape` must be allocated, through it can contain uninitialized values.
* - `dst_ndarray->strides` must be allocated, through it can contain uninitialized values.
* When this function call ends:
* - `dst_ndarray->data` is set to `src_ndarray->data` (`dst_ndarray` is just a view to `src_ndarray`)
* - `dst_ndarray->itemsize` is set to `src_ndarray->itemsize`
* - `dst_ndarray->ndims` is unchanged.
* - `dst_ndarray->shape` is updated according to how `src_ndarray` is indexed.
* - `dst_ndarray->strides` is updated accordingly by how ndarray indexing works.
*
* @param indices indices to index `src_ndarray`, ordered in the same way you would write them in Python.
* @param src_ndarray The NDArray to be indexed.
* @param dst_ndarray The resulting NDArray after indexing. Further details in the comments above,
*/
template <typename SizeT>
void index(SizeT num_indices, const NDIndex *indices, const NDArray<SizeT> *src_ndarray, NDArray<SizeT> *dst_ndarray)
{
// Validate `indices`.
// Expected value of `dst_ndarray->ndims`.
SizeT expected_dst_ndims = src_ndarray->ndims;
// To check for "too many indices for array: array is ?-dimensional, but ? were indexed"
SizeT num_indexed = 0;
// There may be ellipsis `...` in `indices`. There can only be 0 or 1 ellipsis.
SizeT num_ellipsis = 0;
for (SizeT i = 0; i < num_indices; i++)
{
if (indices[i].type == ND_INDEX_TYPE_SINGLE_ELEMENT)
{
expected_dst_ndims--;
num_indexed++;
}
else if (indices[i].type == ND_INDEX_TYPE_SLICE)
{
num_indexed++;
}
else if (indices[i].type == ND_INDEX_TYPE_NEWAXIS)
{
expected_dst_ndims++;
}
else if (indices[i].type == ND_INDEX_TYPE_ELLIPSIS)
{
num_ellipsis++;
if (num_ellipsis > 1)
{
raise_exception(SizeT, EXN_INDEX_ERROR, "an index can only have a single ellipsis ('...')", NO_PARAM,
NO_PARAM, NO_PARAM);
}
}
else
{
__builtin_unreachable();
}
}
debug_assert_eq(SizeT, expected_dst_ndims, dst_ndarray->ndims);
if (src_ndarray->ndims - num_indexed < 0)
{
raise_exception(SizeT, EXN_INDEX_ERROR,
"too many indices for array: array is {0}-dimensional, "
"but {1} were indexed",
src_ndarray->ndims, num_indices, NO_PARAM);
}
dst_ndarray->data = src_ndarray->data;
dst_ndarray->itemsize = src_ndarray->itemsize;
// Reference code: https://github.com/wadetb/tinynumpy/blob/0d23d22e07062ffab2afa287374c7b366eebdda1/tinynumpy/tinynumpy.py#L652
SizeT src_axis = 0;
SizeT dst_axis = 0;
for (int32_t i = 0; i < num_indices; i++)
{
const NDIndex *index = &indices[i];
if (index->type == ND_INDEX_TYPE_SINGLE_ELEMENT)
{
SizeT input = (SizeT) * ((int32_t *)index->data);
SizeT k = slice::resolve_index_in_length(src_ndarray->shape[src_axis], input);
if (k == -1)
{
raise_exception(SizeT, EXN_INDEX_ERROR,
"index {0} is out of bounds for axis {1} "
"with size {2}",
input, src_axis, src_ndarray->shape[src_axis]);
}
dst_ndarray->data += k * src_ndarray->strides[src_axis];
src_axis++;
}
else if (index->type == ND_INDEX_TYPE_SLICE)
{
Slice<int32_t> *slice = (Slice<int32_t> *)index->data;
Range<int32_t> range = slice->indices_checked<SizeT>(src_ndarray->shape[src_axis]);
dst_ndarray->data += (SizeT)range.start * src_ndarray->strides[src_axis];
dst_ndarray->strides[dst_axis] = ((SizeT)range.step) * src_ndarray->strides[src_axis];
dst_ndarray->shape[dst_axis] = (SizeT)range.len<SizeT>();
dst_axis++;
src_axis++;
}
else if (index->type == ND_INDEX_TYPE_NEWAXIS)
{
dst_ndarray->strides[dst_axis] = 0;
dst_ndarray->shape[dst_axis] = 1;
dst_axis++;
}
else if (index->type == ND_INDEX_TYPE_ELLIPSIS)
{
// The number of ':' entries this '...' implies.
SizeT ellipsis_size = src_ndarray->ndims - num_indexed;
for (SizeT j = 0; j < ellipsis_size; j++)
{
dst_ndarray->strides[dst_axis] = src_ndarray->strides[src_axis];
dst_ndarray->shape[dst_axis] = src_ndarray->shape[src_axis];
dst_axis++;
src_axis++;
}
}
else
{
__builtin_unreachable();
}
}
for (; dst_axis < dst_ndarray->ndims; dst_axis++, src_axis++)
{
dst_ndarray->shape[dst_axis] = src_ndarray->shape[src_axis];
dst_ndarray->strides[dst_axis] = src_ndarray->strides[src_axis];
}
debug_assert_eq(SizeT, src_ndarray->ndims, src_axis);
debug_assert_eq(SizeT, dst_ndarray->ndims, dst_axis);
}
} // namespace indexing
} // namespace ndarray
} // namespace
extern "C"
{
using namespace ndarray::indexing;
void __nac3_ndarray_index(int32_t num_indices, NDIndex *indices, NDArray<int32_t> *src_ndarray,
NDArray<int32_t> *dst_ndarray)
{
index(num_indices, indices, src_ndarray, dst_ndarray);
}
void __nac3_ndarray_index64(int64_t num_indices, NDIndex *indices, NDArray<int64_t> *src_ndarray,
NDArray<int64_t> *dst_ndarray)
{
index(num_indices, indices, src_ndarray, dst_ndarray);
}
}

View File

@ -0,0 +1,142 @@
#pragma once
#include <irrt/int_types.hpp>
#include <irrt/ndarray/def.hpp>
namespace
{
/**
* @brief Helper struct to enumerate through all indices under a shape.
*
* i.e., If `shape` is `[3, 2]`, by repeating `next()`, then you get:
* - `[0, 0]`
* - `[0, 1]`
* - `[1, 0]`
* - `[1, 1]`
* - `[2, 0]`
* - `[2, 1]`
* - end.
*
* Interesting cases:
* - If ndims == 0, there is one enumeration.
* - If shape contains zeroes, there are no enumerations.
*/
template <typename SizeT> struct NDIter
{
SizeT ndims;
SizeT *shape;
SizeT *strides;
/**
* @brief The current indices.
*
* Must be allocated by the caller.
*/
SizeT *indices;
/**
* @brief The nth (0-based) index of the current indices.
*/
SizeT nth;
/**
* @brief Pointer to the current element.
*/
uint8_t *element;
/**
* @brief The product of shape.
*/
SizeT size;
// TODO:: There is something called backstrides to speedup iteration.
// See https://ajcr.net/stride-guide-part-1/, and https://docs.scipy.org/doc/numpy-1.13.0/reference/c-api.types-and-structures.html#c.PyArrayIterObject.PyArrayIterObject.backstrides.
// Maybe LLVM is clever and knows how to optimize.
void initialize(SizeT ndims, SizeT *shape, SizeT *strides, uint8_t *element, SizeT *indices)
{
this->ndims = ndims;
this->shape = shape;
this->strides = strides;
this->indices = indices;
this->element = element;
// Compute size and backstrides
this->size = 1;
for (SizeT i = 0; i < ndims; i++)
{
this->size *= shape[i];
}
for (SizeT axis = 0; axis < ndims; axis++)
indices[axis] = 0;
nth = 0;
}
void initialize_by_ndarray(NDArray<SizeT> *ndarray, SizeT *indices)
{
this->initialize(ndarray->ndims, ndarray->shape, ndarray->strides, ndarray->data, indices);
}
bool has_next()
{
return nth < size;
}
void next()
{
for (SizeT i = 0; i < ndims; i++)
{
SizeT axis = ndims - i - 1;
indices[axis]++;
if (indices[axis] >= shape[axis])
{
indices[axis] = 0;
// TODO: Can be optimized with backstrides.
element -= strides[axis] * (shape[axis] - 1);
}
else
{
element += strides[axis];
break;
}
}
nth++;
}
};
} // namespace
extern "C"
{
void __nac3_nditer_initialize(NDIter<int32_t> *iter, NDArray<int32_t> *ndarray, int32_t *indices)
{
iter->initialize_by_ndarray(ndarray, indices);
}
void __nac3_nditer_initialize64(NDIter<int64_t> *iter, NDArray<int64_t> *ndarray, int64_t *indices)
{
iter->initialize_by_ndarray(ndarray, indices);
}
bool __nac3_nditer_has_next(NDIter<int32_t> *iter)
{
return iter->has_next();
}
bool __nac3_nditer_has_next64(NDIter<int64_t> *iter)
{
return iter->has_next();
}
void __nac3_nditer_next(NDIter<int32_t> *iter)
{
iter->next();
}
void __nac3_nditer_next64(NDIter<int64_t> *iter)
{
iter->next();
}
}

View File

@ -0,0 +1,92 @@
#pragma once
#include <irrt/debug.hpp>
#include <irrt/exception.hpp>
#include <irrt/int_types.hpp>
#include <irrt/ndarray/basic.hpp>
#include <irrt/ndarray/broadcast.hpp>
#include <irrt/ndarray/iter.hpp>
// NOTE: Everything would be much easier and elegant if einsum is implemented.
namespace
{
namespace ndarray
{
namespace matmul
{
/**
* @brief Perform the broadcast in `np.einsum("...ij,...jk->...ik", a, b)`.
*
* Example:
* Suppose `a_shape == [1, 97, 4, 2]`
* and `b_shape == [99, 98, 1, 2, 5]`,
*
* ...then `new_a_shape == [99, 98, 97, 4, 2]`,
* `new_b_shape == [99, 98, 97, 2, 5]`,
* and `dst_shape == [99, 98, 97, 4, 5]`.
* ^^^^^^^^^^ ^^^^
* (broadcasted) (4x2 @ 2x5 => 4x5)
*
* @param a_ndims Length of `a_shape`.
* @param a_shape Shape of `a`.
* @param b_ndims Length of `b_shape`.
* @param b_shape Shape of `b`.
* @param final_ndims Should be equal to `max(a_ndims, b_ndims)`. This is the length of `new_a_shape`,
* `new_b_shape`, and `dst_shape` - the number of dimensions after broadcasting.
*/
template <typename SizeT>
void calculate_shapes(SizeT a_ndims, SizeT *a_shape, SizeT b_ndims, SizeT *b_shape, SizeT final_ndims,
SizeT *new_a_shape, SizeT *new_b_shape, SizeT *dst_shape)
{
debug_assert(SizeT, a_ndims >= 2);
debug_assert(SizeT, b_ndims >= 2);
debug_assert_eq(SizeT, max(a_ndims, b_ndims), final_ndims);
// Check that a and b are compatible for matmul
if (a_shape[a_ndims - 1] != b_shape[b_ndims - 2])
{
// This is a custom error message. Different from NumPy.
raise_exception(SizeT, EXN_VALUE_ERROR, "Cannot multiply LHS (shape ?x{0}) with RHS (shape {1}x?})",
a_shape[a_ndims - 1], b_shape[b_ndims - 2], NO_PARAM);
}
const SizeT num_entries = 2;
ShapeEntry<SizeT> entries[num_entries] = {{.ndims = a_ndims - 2, .shape = a_shape},
{.ndims = b_ndims - 2, .shape = b_shape}};
// TODO: Optimize this
ndarray::broadcast::broadcast_shapes<SizeT>(num_entries, entries, final_ndims - 2, new_a_shape);
ndarray::broadcast::broadcast_shapes<SizeT>(num_entries, entries, final_ndims - 2, new_b_shape);
ndarray::broadcast::broadcast_shapes<SizeT>(num_entries, entries, final_ndims - 2, dst_shape);
new_a_shape[final_ndims - 2] = a_shape[a_ndims - 2];
new_a_shape[final_ndims - 1] = a_shape[a_ndims - 1];
new_b_shape[final_ndims - 2] = b_shape[b_ndims - 2];
new_b_shape[final_ndims - 1] = b_shape[b_ndims - 1];
dst_shape[final_ndims - 2] = a_shape[a_ndims - 2];
dst_shape[final_ndims - 1] = b_shape[b_ndims - 1];
}
} // namespace matmul
} // namespace ndarray
} // namespace
extern "C"
{
using namespace ndarray::matmul;
void __nac3_ndarray_matmul_calculate_shapes(int32_t a_ndims, int32_t *a_shape, int32_t b_ndims, int32_t *b_shape,
int32_t final_ndims, int32_t *new_a_shape, int32_t *new_b_shape,
int32_t *dst_shape)
{
calculate_shapes(a_ndims, a_shape, b_ndims, b_shape, final_ndims, new_a_shape, new_b_shape, dst_shape);
}
void __nac3_ndarray_matmul_calculate_shapes64(int64_t a_ndims, int64_t *a_shape, int64_t b_ndims, int64_t *b_shape,
int64_t final_ndims, int64_t *new_a_shape, int64_t *new_b_shape,
int64_t *dst_shape)
{
calculate_shapes(a_ndims, a_shape, b_ndims, b_shape, final_ndims, new_a_shape, new_b_shape, dst_shape);
}
}

View File

@ -0,0 +1,125 @@
#pragma once
#include <irrt/int_types.hpp>
#include <irrt/ndarray/def.hpp>
namespace
{
namespace ndarray
{
namespace reshape
{
/**
* @brief Perform assertions on and resolve unknown dimensions in `new_shape` in `np.reshape(<ndarray>, new_shape)`
*
* If `new_shape` indeed contains unknown dimensions (specified with `-1`, just like numpy), `new_shape` will be
* modified to contain the resolved dimension.
*
* To perform assertions on and resolve unknown dimensions in `new_shape`, we don't need the actual
* `<ndarray>` object itself, but only the `.size` of the `<ndarray>`.
*
* @param size The `.size` of `<ndarray>`
* @param new_ndims Number of elements in `new_shape`
* @param new_shape Target shape to reshape to
*/
template <typename SizeT> void resolve_and_check_new_shape(SizeT size, SizeT new_ndims, SizeT *new_shape)
{
// Is there a -1 in `new_shape`?
bool neg1_exists = false;
// Location of -1, only initialized if `neg1_exists` is true
SizeT neg1_axis_i;
// The computed ndarray size of `new_shape`
SizeT new_size = 1;
for (SizeT axis_i = 0; axis_i < new_ndims; axis_i++)
{
SizeT dim = new_shape[axis_i];
if (dim < 0)
{
if (dim == -1)
{
if (neg1_exists)
{
// Multiple `-1` found. Throw an error.
raise_exception(SizeT, EXN_VALUE_ERROR, "can only specify one unknown dimension", NO_PARAM,
NO_PARAM, NO_PARAM);
}
else
{
neg1_exists = true;
neg1_axis_i = axis_i;
}
}
else
{
// TODO: What? In `np.reshape` any negative dimensions is
// treated like its `-1`.
//
// Try running `np.zeros((3, 4)).reshape((-999, 2))`
//
// It is not documented by numpy.
// Throw an error for now...
raise_exception(SizeT, EXN_VALUE_ERROR, "Found non -1 negative dimension {0} on axis {1}", dim, axis_i,
NO_PARAM);
}
}
else
{
new_size *= dim;
}
}
bool can_reshape;
if (neg1_exists)
{
// Let `x` be the unknown dimension
// Solve `x * <new_size> = <size>`
if (new_size == 0 && size == 0)
{
// `x` has infinitely many solutions
can_reshape = false;
}
else if (new_size == 0 && size != 0)
{
// `x` has no solutions
can_reshape = false;
}
else if (size % new_size != 0)
{
// `x` has no integer solutions
can_reshape = false;
}
else
{
can_reshape = true;
new_shape[neg1_axis_i] = size / new_size; // Resolve dimension
}
}
else
{
can_reshape = (new_size == size);
}
if (!can_reshape)
{
raise_exception(SizeT, EXN_VALUE_ERROR, "cannot reshape array of size {0} into given shape", size, NO_PARAM,
NO_PARAM);
}
}
} // namespace reshape
} // namespace ndarray
} // namespace
extern "C"
{
void __nac3_ndarray_reshape_resolve_and_check_new_shape(int32_t size, int32_t new_ndims, int32_t *new_shape)
{
ndarray::reshape::resolve_and_check_new_shape(size, new_ndims, new_shape);
}
void __nac3_ndarray_reshape_resolve_and_check_new_shape64(int64_t size, int64_t new_ndims, int64_t *new_shape)
{
ndarray::reshape::resolve_and_check_new_shape(size, new_ndims, new_shape);
}
}

View File

@ -0,0 +1,155 @@
#pragma once
#include <irrt/int_types.hpp>
#include <irrt/ndarray/def.hpp>
#include <irrt/slice.hpp>
/*
* Notes on `np.transpose(<array>, <axes>)`
*
* TODO: `axes`, if specified, can actually contain negative indices,
* but it is not documented in numpy.
*
* Supporting it for now.
*/
namespace
{
namespace ndarray
{
namespace transpose
{
/**
* @brief Do assertions on `<axes>` in `np.transpose(<array>, <axes>)`.
*
* Note that `np.transpose`'s `<axe>` argument is optional. If the argument
* is specified but the user, use this function to do assertions on it.
*
* @param ndims The number of dimensions of `<array>`
* @param num_axes Number of elements in `<axes>` as specified by the user.
* This should be equal to `ndims`. If not, a "ValueError: axes don't match array" is thrown.
* @param axes The user specified `<axes>`.
*/
template <typename SizeT> void assert_transpose_axes(SizeT ndims, SizeT num_axes, const SizeT *axes)
{
if (ndims != num_axes)
{
raise_exception(SizeT, EXN_VALUE_ERROR, "axes don't match array", NO_PARAM, NO_PARAM, NO_PARAM);
}
// TODO: Optimize this
bool *axe_specified = (bool *)__builtin_alloca(sizeof(bool) * ndims);
for (SizeT i = 0; i < ndims; i++)
axe_specified[i] = false;
for (SizeT i = 0; i < ndims; i++)
{
SizeT axis = slice::resolve_index_in_length(ndims, axes[i]);
if (axis == -1)
{
// TODO: numpy actually throws a `numpy.exceptions.AxisError`
raise_exception(SizeT, EXN_VALUE_ERROR, "axis {0} is out of bounds for array of dimension {1}", axis, ndims,
NO_PARAM);
}
if (axe_specified[axis])
{
raise_exception(SizeT, EXN_VALUE_ERROR, "repeated axis in transpose", NO_PARAM, NO_PARAM, NO_PARAM);
}
axe_specified[axis] = true;
}
}
/**
* @brief Create a transpose view of `src_ndarray` and perform proper assertions.
*
* This function is very similar to doing `dst_ndarray = np.transpose(src_ndarray, <axes>)`.
* If `<axes>` is supposed to be `None`, caller can pass in a `nullptr` to `<axes>`.
*
* The transpose view created is returned by modifying `dst_ndarray`.
*
* The caller is responsible for setting up `dst_ndarray` before calling this function.
* Here is what this function expects from `dst_ndarray` when called:
* - `dst_ndarray->data` does not have to be initialized.
* - `dst_ndarray->itemsize` does not have to be initialized.
* - `dst_ndarray->ndims` must be initialized, must be equal to `src_ndarray->ndims`.
* - `dst_ndarray->shape` must be allocated, through it can contain uninitialized values.
* - `dst_ndarray->strides` must be allocated, through it can contain uninitialized values.
* When this function call ends:
* - `dst_ndarray->data` is set to `src_ndarray->data` (`dst_ndarray` is just a view to `src_ndarray`)
* - `dst_ndarray->itemsize` is set to `src_ndarray->itemsize`
* - `dst_ndarray->ndims` is unchanged
* - `dst_ndarray->shape` is updated according to how `np.transpose` works
* - `dst_ndarray->strides` is updated according to how `np.transpose` works
*
* @param src_ndarray The NDArray to build a transpose view on
* @param dst_ndarray The resulting NDArray after transpose. Further details in the comments above,
* @param num_axes Number of elements in axes. Unused if `axes` is nullptr.
* @param axes Axes permutation. Set it to `nullptr` if `<axes>` is `None`.
*/
template <typename SizeT>
void transpose(const NDArray<SizeT> *src_ndarray, NDArray<SizeT> *dst_ndarray, SizeT num_axes, const SizeT *axes)
{
debug_assert_eq(SizeT, src_ndarray->ndims, dst_ndarray->ndims);
const auto ndims = src_ndarray->ndims;
if (axes != nullptr)
assert_transpose_axes(ndims, num_axes, axes);
dst_ndarray->data = src_ndarray->data;
dst_ndarray->itemsize = src_ndarray->itemsize;
// Check out https://ajcr.net/stride-guide-part-2/ to see how `np.transpose` works behind the scenes.
if (axes == nullptr)
{
// `np.transpose(<array>, axes=None)`
/*
* Minor note: `np.transpose(<array>, axes=None)` is equivalent to
* `np.transpose(<array>, axes=[N-1, N-2, ..., 0])` - basically it
* is reversing the order of strides and shape.
*
* This is a fast implementation to handle this special (but very common) case.
*/
for (SizeT axis = 0; axis < ndims; axis++)
{
dst_ndarray->shape[axis] = src_ndarray->shape[ndims - axis - 1];
dst_ndarray->strides[axis] = src_ndarray->strides[ndims - axis - 1];
}
}
else
{
// `np.transpose(<array>, <axes>)`
// Permute strides and shape according to `axes`, while resolving negative indices in `axes`
for (SizeT axis = 0; axis < ndims; axis++)
{
// `i` cannot be OUT_OF_BOUNDS because of assertions
SizeT i = slice::resolve_index_in_length(ndims, axes[axis]);
dst_ndarray->shape[axis] = src_ndarray->shape[i];
dst_ndarray->strides[axis] = src_ndarray->strides[i];
}
}
}
} // namespace transpose
} // namespace ndarray
} // namespace
extern "C"
{
using namespace ndarray::transpose;
void __nac3_ndarray_transpose(const NDArray<int32_t> *src_ndarray, NDArray<int32_t> *dst_ndarray, int32_t num_axes,
const int32_t *axes)
{
transpose(src_ndarray, dst_ndarray, num_axes, axes);
}
void __nac3_ndarray_transpose64(const NDArray<int64_t> *src_ndarray, NDArray<int64_t> *dst_ndarray,
int64_t num_axes, const int64_t *axes)
{
transpose(src_ndarray, dst_ndarray, num_axes, axes);
}
}

View File

@ -0,0 +1,215 @@
#pragma once
#include <irrt/int_types.hpp>
#include <irrt/math_util.hpp>
// The type of an index or a value describing the length of a range/slice is always `int32_t`.
using SliceIndex = 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;
}
} // namespace
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);
}
} // extern "C"

View File

@ -0,0 +1,224 @@
#pragma once
#include <irrt/debug.hpp>
#include <irrt/exception.hpp>
#include <irrt/int_types.hpp>
#include <irrt/math_util.hpp>
namespace
{
namespace range
{
template <typename T> T len(T start, T stop, T step)
{
// Reference: https://github.com/python/cpython/blob/9dbd12375561a393eaec4b21ee4ac568a407cdb0/Objects/rangeobject.c#L933
if (step > 0 && start < stop)
return 1 + (stop - 1 - start) / step;
else if (step < 0 && start > stop)
return 1 + (start - 1 - stop) / (-step);
else
return 0;
}
} // namespace range
/**
* @brief A Python range.
*/
template <typename T> struct Range
{
T start;
T stop;
T step;
/**
* @brief Calculate the `len()` of this range.
*/
template <typename SizeT> T len()
{
debug_assert(SizeT, step != 0);
return range::len(start, stop, step);
}
};
namespace slice
{
/**
* @brief Resolve a slice index under a given length like Python indexing.
*
* In Python, if you have a `list` of length 100, `list[-1]` resolves to
* `list[99]`, so `resolve_index_in_length_clamped(100, -1)` returns `99`.
*
* If `length` is 0, 0 is returned for any value of `index`.
*
* If `index` is out of bounds, clamps the returned value between `0` and
* `length - 1` (inclusive).
*
*/
template <typename T> T resolve_index_in_length_clamped(T length, T index)
{
if (index < 0)
{
return max<T>(length + index, 0);
}
else
{
return min<T>(length, index);
}
}
/**
* @brief Like `resolve_index_in_length_clamped`, but returns `-1` if `index` is out of bounds.
*/
template <typename T> T resolve_index_in_length(T length, T index)
{
T resolved = index < 0 ? length + index : index;
if (0 <= resolved && resolved < length)
{
return resolved;
}
else
{
return -1;
}
}
/**
* @brief Resolve a slice as a range.
*
* In Python, this would be `range(*slice(start, stop, step).indices(length))`.
*/
template <typename T>
void indices(bool start_defined, T start, bool stop_defined, T stop, bool step_defined, T step, T length,
T *range_start, T *range_stop, T *range_step)
{
// Reference: https://github.com/python/cpython/blob/main/Objects/sliceobject.c#L388
*range_step = step_defined ? step : 1;
bool step_is_negative = *range_step < 0;
T lower, upper;
if (step_is_negative)
{
lower = -1;
upper = length - 1;
}
else
{
lower = 0;
upper = length;
}
if (start_defined)
{
*range_start = start < 0 ? max(lower, start + length) : min(upper, start);
}
else
{
*range_start = step_is_negative ? upper : lower;
}
if (stop_defined)
{
*range_stop = stop < 0 ? max(lower, stop + length) : min(upper, stop);
}
else
{
*range_stop = step_is_negative ? lower : upper;
}
}
} // namespace slice
/**
* @brief A Python-like slice with **unresolved** indices.
*/
template <typename T> struct Slice
{
bool start_defined;
T start;
bool stop_defined;
T stop;
bool step_defined;
T step;
Slice()
{
this->reset();
}
void reset()
{
this->start_defined = false;
this->stop_defined = false;
this->step_defined = false;
}
void set_start(T start)
{
this->start_defined = true;
this->start = start;
}
void set_stop(T stop)
{
this->stop_defined = true;
this->stop = stop;
}
void set_step(T step)
{
this->step_defined = true;
this->step = step;
}
/**
* @brief Resolve this slice as a range.
*
* In Python, this would be `range(*slice(start, stop, step).indices(length))`.
*/
template <typename SizeT> Range<T> indices(T length)
{
// Reference: https://github.com/python/cpython/blob/main/Objects/sliceobject.c#L388
debug_assert(SizeT, length >= 0);
Range<T> result;
slice::indices(start_defined, start, stop_defined, stop, step_defined, step, length, &result.start,
&result.stop, &result.step);
return result;
}
/**
* @brief Like `.indices()` but with assertions.
*/
template <typename SizeT> Range<T> indices_checked(T length)
{
// TODO: Switch to `SizeT length`
if (length < 0)
{
raise_exception(SizeT, EXN_VALUE_ERROR, "length should not be negative, got {0}", length, NO_PARAM,
NO_PARAM);
}
if (this->step_defined && this->step == 0)
{
raise_exception(SizeT, EXN_VALUE_ERROR, "slice step cannot be zero", NO_PARAM, NO_PARAM, NO_PARAM);
}
return this->indices<SizeT>(length);
}
};
} // namespace
extern "C"
{
int32_t __nac3_range_len_i32(int32_t start, int32_t stop, int32_t step)
{
range::len<int32_t>(start, stop, step);
}
int32_t __nac3_range_len_i3264(int32_t start, int32_t stop, int32_t step)
{
range::len<int64_t>(start, stop, step);
}
}

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@ -25,6 +25,7 @@ pub struct ConcreteFuncArg {
pub name: StrRef,
pub ty: ConcreteType,
pub default_value: Option<SymbolValue>,
pub is_vararg: bool,
}
#[derive(Clone, Debug)]
@ -46,6 +47,7 @@ pub enum ConcreteTypeEnum {
TPrimitive(Primitive),
TTuple {
ty: Vec<ConcreteType>,
is_vararg_ctx: bool,
},
TObj {
obj_id: DefinitionId,
@ -102,8 +104,16 @@ impl ConcreteTypeStore {
.iter()
.map(|arg| ConcreteFuncArg {
name: arg.name,
ty: self.from_unifier_type(unifier, primitives, arg.ty, cache),
ty: if arg.is_vararg {
let tuple_ty = unifier
.add_ty(TypeEnum::TTuple { ty: vec![arg.ty], is_vararg_ctx: true });
self.from_unifier_type(unifier, primitives, tuple_ty, cache)
} else {
self.from_unifier_type(unifier, primitives, arg.ty, cache)
},
default_value: arg.default_value.clone(),
is_vararg: arg.is_vararg,
})
.collect(),
ret: self.from_unifier_type(unifier, primitives, signature.ret, cache),
@ -158,11 +168,12 @@ impl ConcreteTypeStore {
cache.insert(ty, None);
let ty_enum = unifier.get_ty(ty);
let result = match &*ty_enum {
TypeEnum::TTuple { ty } => ConcreteTypeEnum::TTuple {
TypeEnum::TTuple { ty, is_vararg_ctx } => ConcreteTypeEnum::TTuple {
ty: ty
.iter()
.map(|t| self.from_unifier_type(unifier, primitives, *t, cache))
.collect(),
is_vararg_ctx: *is_vararg_ctx,
},
TypeEnum::TObj { obj_id, fields, params } => ConcreteTypeEnum::TObj {
obj_id: *obj_id,
@ -248,11 +259,12 @@ impl ConcreteTypeStore {
*cache.get_mut(&cty).unwrap() = Some(ty);
return ty;
}
ConcreteTypeEnum::TTuple { ty } => TypeEnum::TTuple {
ConcreteTypeEnum::TTuple { ty, is_vararg_ctx } => TypeEnum::TTuple {
ty: ty
.iter()
.map(|cty| self.to_unifier_type(unifier, primitives, *cty, cache))
.collect(),
is_vararg_ctx: *is_vararg_ctx,
},
ConcreteTypeEnum::TVirtual { ty } => {
TypeEnum::TVirtual { ty: self.to_unifier_type(unifier, primitives, *ty, cache) }
@ -277,6 +289,7 @@ impl ConcreteTypeStore {
name: arg.name,
ty: self.to_unifier_type(unifier, primitives, arg.ty, cache),
default_value: arg.default_value.clone(),
is_vararg: false,
})
.collect(),
ret: self.to_unifier_type(unifier, primitives, *ret, cache),

File diff suppressed because it is too large Load Diff

View File

@ -13,8 +13,8 @@ use crate::codegen::CodeGenContext;
/// * `$extern_fn:literal`: Name of underlying extern function
///
/// Optional Arguments:
/// * `$(,$attributes:literal)*)`: Attributes linked with the extern function
/// The default attributes are "mustprogress", "nofree", "nounwind", "willreturn", and "writeonly"
/// * `$(,$attributes:literal)*)`: Attributes linked with the extern function.
/// The default attributes are "mustprogress", "nofree", "nounwind", "willreturn", and "writeonly".
/// These will be used unless other attributes are specified
/// * `$(,$args:ident)*`: Operands of the extern function
/// The data type of these operands will be set to `FloatValue`
@ -130,3 +130,62 @@ pub fn call_ldexp<'ctx>(
.map(Either::unwrap_left)
.unwrap()
}
/// Macro to generate `np_linalg` and `sp_linalg` functions
/// The function takes as input `NDArray` and returns ()
///
/// Arguments:
/// * `$fn_name:ident`: The identifier of the rust function to be generated
/// * `$extern_fn:literal`: Name of underlying extern function
/// * (2/3/4): Number of `NDArray` that function takes as input
///
/// Note:
/// The operands and resulting `NDArray` are both passed as input to the funcion
/// It is the responsibility of caller to ensure that output `NDArray` is properly allocated on stack
/// The function changes the content of the output `NDArray` in-place
macro_rules! generate_linalg_extern_fn {
($fn_name:ident, $extern_fn:literal, 2) => {
generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2);
};
($fn_name:ident, $extern_fn:literal, 3) => {
generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2, mat3);
};
($fn_name:ident, $extern_fn:literal, 4) => {
generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2, mat3, mat4);
};
($fn_name:ident, $extern_fn:literal $(,$input_matrix:ident)*) => {
#[doc = concat!("Invokes the linalg `", stringify!($extern_fn), " function." )]
pub fn $fn_name<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>
$(,$input_matrix: BasicValueEnum<'ctx>)*,
name: Option<&str>,
){
const FN_NAME: &str = $extern_fn;
let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
let fn_type = ctx.ctx.void_type().fn_type(&[$($input_matrix.get_type().into()),*], false);
let func = ctx.module.add_function(FN_NAME, fn_type, None);
for attr in ["mustprogress", "nofree", "nounwind", "willreturn", "writeonly"] {
func.add_attribute(
AttributeLoc::Function,
ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
);
}
func
});
ctx.builder.build_call(extern_fn, &[$($input_matrix.into(),)*], name.unwrap_or_default()).unwrap();
}
};
}
generate_linalg_extern_fn!(call_np_linalg_cholesky, "np_linalg_cholesky", 2);
generate_linalg_extern_fn!(call_np_linalg_qr, "np_linalg_qr", 3);
generate_linalg_extern_fn!(call_np_linalg_svd, "np_linalg_svd", 4);
generate_linalg_extern_fn!(call_np_linalg_inv, "np_linalg_inv", 2);
generate_linalg_extern_fn!(call_np_linalg_pinv, "np_linalg_pinv", 2);
generate_linalg_extern_fn!(call_np_linalg_matrix_power, "np_linalg_matrix_power", 3);
generate_linalg_extern_fn!(call_np_linalg_det, "np_linalg_det", 2);
generate_linalg_extern_fn!(call_sp_linalg_lu, "sp_linalg_lu", 3);
generate_linalg_extern_fn!(call_sp_linalg_schur, "sp_linalg_schur", 3);
generate_linalg_extern_fn!(call_sp_linalg_hessenberg, "sp_linalg_hessenberg", 3);

View File

@ -57,6 +57,7 @@ pub trait CodeGenerator {
/// - fun: Function signature, definition ID and the substitution key.
/// - params: Function parameters. Note that this does not include the object even if the
/// 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>(
@ -123,11 +124,45 @@ pub trait CodeGenerator {
ctx: &mut CodeGenContext<'ctx, '_>,
target: &Expr<Option<Type>>,
value: ValueEnum<'ctx>,
value_ty: Type,
) -> Result<(), String>
where
Self: Sized,
{
gen_assign(self, ctx, target, value)
gen_assign(self, ctx, target, value, value_ty)
}
/// Generate code for an assignment expression where LHS is a `"target_list"`.
///
/// See <https://docs.python.org/3/reference/simple_stmts.html#assignment-statements>.
fn gen_assign_target_list<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
targets: &Vec<Expr<Option<Type>>>,
value: ValueEnum<'ctx>,
value_ty: Type,
) -> Result<(), String>
where
Self: Sized,
{
gen_assign_target_list(self, ctx, targets, value, value_ty)
}
/// Generate code for an item assignment.
///
/// i.e., `target[key] = value`
fn gen_setitem<'ctx>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
target: &Expr<Option<Type>>,
key: &Expr<Option<Type>>,
value: ValueEnum<'ctx>,
value_ty: Type,
) -> Result<(), String>
where
Self: Sized,
{
gen_setitem(self, ctx, target, key, value, value_ty)
}
/// Generate code for a while expression.

View File

@ -1,414 +0,0 @@
using int8_t = _BitInt(8);
using uint8_t = unsigned _BitInt(8);
using int32_t = _BitInt(32);
using uint32_t = unsigned _BitInt(32);
using int64_t = _BitInt(64);
using uint64_t = unsigned _BitInt(64);
// NDArray indices are always `uint32_t`.
using NDIndex = uint32_t;
// The type of an index or a value describing the length of a range/slice is always `int32_t`.
using SliceIndex = int32_t;
namespace {
template <typename T>
const T& max(const T& a, const T& b) {
return a > b ? a : b;
}
template <typename T>
const T& min(const T& a, const 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>
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>
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>
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>
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>
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>
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];
}
}
} // namespace
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);
}
} // extern "C"

View File

@ -1,21 +1,21 @@
use crate::typecheck::typedef::Type;
use crate::{symbol_resolver::SymbolResolver, typecheck::typedef::Type};
use super::{
classes::{
ArrayLikeIndexer, ArrayLikeValue, ArraySliceValue, ListValue, NDArrayValue,
TypedArrayLikeAdapter, UntypedArrayLikeAccessor,
model::*,
object::{
list::List,
ndarray::{broadcast::ShapeEntry, indexing::NDIndex, nditer::NDIter, NDArray},
},
llvm_intrinsics, CodeGenContext, CodeGenerator,
CodeGenContext, CodeGenerator,
};
use crate::codegen::classes::TypedArrayLikeAccessor;
use crate::codegen::stmt::gen_for_callback_incrementing;
use function::CallFunction;
use inkwell::{
attributes::{Attribute, AttributeLoc},
context::Context,
memory_buffer::MemoryBuffer,
module::Module,
types::{BasicTypeEnum, IntType},
values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue},
types::BasicTypeEnum,
values::{BasicValue, BasicValueEnum, CallSiteValue, FloatValue, IntValue},
AddressSpace, IntPredicate,
};
use itertools::Either;
@ -326,12 +326,11 @@ pub fn list_slice_assignment<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ty: BasicTypeEnum<'ctx>,
dest_arr: ListValue<'ctx>,
dest_arr: Instance<'ctx, Ptr<Struct<List<Any<'ctx>>>>>,
dest_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
src_arr: ListValue<'ctx>,
src_arr: Instance<'ctx, Ptr<Struct<List<Any<'ctx>>>>>,
src_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
) {
let size_ty = generator.get_size_type(ctx.ctx);
let int8_ptr = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
let int32 = ctx.ctx.i32_type();
let (fun_symbol, elem_ptr_type) = ("__nac3_list_slice_assign_var_size", int8_ptr);
@ -357,16 +356,14 @@ pub fn list_slice_assignment<'ctx, G: CodeGenerator + ?Sized>(
let zero = int32.const_zero();
let one = int32.const_int(1, false);
let dest_arr_ptr = dest_arr.data().base_ptr(ctx, generator);
let dest_arr_ptr =
ctx.builder.build_pointer_cast(dest_arr_ptr, elem_ptr_type, "dest_arr_ptr_cast").unwrap();
let dest_len = dest_arr.load_size(ctx, Some("dest.len"));
let dest_len = ctx.builder.build_int_truncate_or_bit_cast(dest_len, int32, "srclen32").unwrap();
let src_arr_ptr = src_arr.data().base_ptr(ctx, generator);
dest_arr.get(generator, ctx, |f| f.items).pointer_cast(generator, ctx, Int(Byte)).value;
let src_arr_ptr =
ctx.builder.build_pointer_cast(src_arr_ptr, elem_ptr_type, "src_arr_ptr_cast").unwrap();
let src_len = src_arr.load_size(ctx, Some("src.len"));
let src_len = ctx.builder.build_int_truncate_or_bit_cast(src_len, int32, "srclen32").unwrap();
src_arr.get(generator, ctx, |f| f.items).pointer_cast(generator, ctx, Int(Byte)).value;
let dest_len = dest_arr.get(generator, ctx, |f| f.len).truncate(generator, ctx, Int32).value;
let src_len = src_arr.get(generator, ctx, |f| f.len).truncate(generator, ctx, Int32).value;
// index in bound and positive should be done
// assert if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest), and
@ -462,8 +459,10 @@ pub fn list_slice_assignment<'ctx, G: CodeGenerator + ?Sized>(
let cont_bb = ctx.ctx.append_basic_block(current, "cont");
ctx.builder.build_conditional_branch(need_update, update_bb, cont_bb).unwrap();
ctx.builder.position_at_end(update_bb);
let new_len = ctx.builder.build_int_z_extend_or_bit_cast(new_len, size_ty, "new_len").unwrap();
dest_arr.store_size(ctx, generator, new_len);
let new_len = Int(SizeT).z_extend_or_bit_cast(generator, ctx, new_len);
dest_arr.set(ctx, |f| f.len, new_len);
ctx.builder.build_unconditional_branch(cont_bb).unwrap();
ctx.builder.position_at_end(cont_bb);
}
@ -563,367 +562,339 @@ pub fn call_j0<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> Flo
.unwrap()
}
/// Generates a call to `__nac3_ndarray_calc_size`. Returns an [`IntValue`] representing the
/// calculated total size.
///
/// * `dims` - An [`ArrayLikeIndexer`] containing the size of each dimension.
/// * `range` - The dimension index to begin and end (exclusively) calculating the dimensions for,
/// or [`None`] if starting from the first dimension and ending at the last dimension respectively.
pub fn call_ndarray_calc_size<'ctx, G, Dims>(
generator: &G,
ctx: &CodeGenContext<'ctx, '_>,
dims: &Dims,
(begin, end): (Option<IntValue<'ctx>>, Option<IntValue<'ctx>>),
) -> IntValue<'ctx>
where
G: CodeGenerator + ?Sized,
Dims: ArrayLikeIndexer<'ctx>,
{
let llvm_usize = generator.get_size_type(ctx.ctx);
let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
let ndarray_calc_size_fn_name = match llvm_usize.get_bit_width() {
32 => "__nac3_ndarray_calc_size",
64 => "__nac3_ndarray_calc_size64",
bw => unreachable!("Unsupported size type bit width: {}", bw),
};
let ndarray_calc_size_fn_t = llvm_usize.fn_type(
&[llvm_pusize.into(), llvm_usize.into(), llvm_usize.into(), llvm_usize.into()],
false,
);
let ndarray_calc_size_fn =
ctx.module.get_function(ndarray_calc_size_fn_name).unwrap_or_else(|| {
ctx.module.add_function(ndarray_calc_size_fn_name, ndarray_calc_size_fn_t, None)
});
let begin = begin.unwrap_or_else(|| llvm_usize.const_zero());
let end = end.unwrap_or_else(|| dims.size(ctx, generator));
ctx.builder
.build_call(
ndarray_calc_size_fn,
&[
dims.base_ptr(ctx, generator).into(),
dims.size(ctx, generator).into(),
begin.into(),
end.into(),
],
"",
)
.map(CallSiteValue::try_as_basic_value)
.map(|v| v.map_left(BasicValueEnum::into_int_value))
.map(Either::unwrap_left)
.unwrap()
// When [`TypeContext::size_type`] is 32-bits, the function name is "{fn_name}".
// When [`TypeContext::size_type`] is 64-bits, the function name is "{fn_name}64".
#[must_use]
pub fn get_sizet_dependent_function_name<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &CodeGenContext<'_, '_>,
name: &str,
) -> String {
let mut name = name.to_owned();
match generator.get_size_type(ctx.ctx).get_bit_width() {
32 => {}
64 => name.push_str("64"),
bit_width => {
panic!("Unsupported int type bit width {bit_width}, must be either 32-bits or 64-bits")
}
}
name
}
/// Generates a call to `__nac3_ndarray_calc_nd_indices`. Returns a [`TypeArrayLikeAdpater`]
/// containing `i32` indices of the flattened index.
///
/// * `index` - The index to compute the multidimensional index for.
/// * `ndarray` - LLVM pointer to the `NDArray`. This value must be the LLVM representation of an
/// `NDArray`.
pub fn call_ndarray_calc_nd_indices<'ctx, G: CodeGenerator + ?Sized>(
generator: &G,
ctx: &mut CodeGenContext<'ctx, '_>,
index: IntValue<'ctx>,
ndarray: NDArrayValue<'ctx>,
) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
let llvm_void = ctx.ctx.void_type();
let llvm_i32 = ctx.ctx.i32_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
/// Initialize all global `EXN_*` exception IDs in IRRT with the [`SymbolResolver`].
pub fn setup_irrt_exceptions<'ctx>(
ctx: &'ctx Context,
module: &Module<'ctx>,
symbol_resolver: &dyn SymbolResolver,
) {
let exn_id_type = ctx.i32_type();
let ndarray_calc_nd_indices_fn_name = match llvm_usize.get_bit_width() {
32 => "__nac3_ndarray_calc_nd_indices",
64 => "__nac3_ndarray_calc_nd_indices64",
bw => unreachable!("Unsupported size type bit width: {}", bw),
};
let ndarray_calc_nd_indices_fn =
ctx.module.get_function(ndarray_calc_nd_indices_fn_name).unwrap_or_else(|| {
let fn_type = llvm_void.fn_type(
&[llvm_usize.into(), llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into()],
false,
);
let errors = &[
("EXN_INDEX_ERROR", "0:IndexError"),
("EXN_VALUE_ERROR", "0:ValueError"),
("EXN_ASSERTION_ERROR", "0:AssertionError"),
("EXN_TYPE_ERROR", "0:TypeError"),
];
ctx.module.add_function(ndarray_calc_nd_indices_fn_name, fn_type, None)
for (irrt_name, symbol_name) in errors {
let exn_id = symbol_resolver.get_string_id(symbol_name);
let exn_id = exn_id_type.const_int(exn_id as u64, false).as_basic_value_enum();
let global = module.get_global(irrt_name).unwrap_or_else(|| {
panic!("Exception symbol name '{irrt_name}' should exist in the IRRT LLVM module")
});
let ndarray_num_dims = ndarray.load_ndims(ctx);
let ndarray_dims = ndarray.dim_sizes();
let indices = ctx.builder.build_array_alloca(llvm_i32, ndarray_num_dims, "").unwrap();
ctx.builder
.build_call(
ndarray_calc_nd_indices_fn,
&[
index.into(),
ndarray_dims.base_ptr(ctx, generator).into(),
ndarray_num_dims.into(),
indices.into(),
],
"",
)
.unwrap();
TypedArrayLikeAdapter::from(
ArraySliceValue::from_ptr_val(indices, ndarray_num_dims, None),
Box::new(|_, v| v.into_int_value()),
Box::new(|_, v| v.into()),
)
global.set_initializer(&exn_id);
}
}
fn call_ndarray_flatten_index_impl<'ctx, G, Indices>(
generator: &G,
ctx: &CodeGenContext<'ctx, '_>,
ndarray: NDArrayValue<'ctx>,
indices: &Indices,
) -> IntValue<'ctx>
where
G: CodeGenerator + ?Sized,
Indices: ArrayLikeIndexer<'ctx>,
{
let llvm_i32 = ctx.ctx.i32_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
debug_assert_eq!(
IntType::try_from(indices.element_type(ctx, generator))
.map(IntType::get_bit_width)
.unwrap_or_default(),
llvm_i32.get_bit_width(),
"Expected i32 value for argument `indices` to `call_ndarray_flatten_index_impl`"
);
debug_assert_eq!(
indices.size(ctx, generator).get_type().get_bit_width(),
llvm_usize.get_bit_width(),
"Expected usize integer value for argument `indices_size` to `call_ndarray_flatten_index_impl`"
);
let ndarray_flatten_index_fn_name = match llvm_usize.get_bit_width() {
32 => "__nac3_ndarray_flatten_index",
64 => "__nac3_ndarray_flatten_index64",
bw => unreachable!("Unsupported size type bit width: {}", bw),
};
let ndarray_flatten_index_fn =
ctx.module.get_function(ndarray_flatten_index_fn_name).unwrap_or_else(|| {
let fn_type = llvm_usize.fn_type(
&[llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into(), llvm_usize.into()],
false,
);
ctx.module.add_function(ndarray_flatten_index_fn_name, fn_type, None)
});
let ndarray_num_dims = ndarray.load_ndims(ctx);
let ndarray_dims = ndarray.dim_sizes();
let index = ctx
.builder
.build_call(
ndarray_flatten_index_fn,
&[
ndarray_dims.base_ptr(ctx, generator).into(),
ndarray_num_dims.into(),
indices.base_ptr(ctx, generator).into(),
indices.size(ctx, generator).into(),
],
"",
)
.map(CallSiteValue::try_as_basic_value)
.map(|v| v.map_left(BasicValueEnum::into_int_value))
.map(Either::unwrap_left)
.unwrap();
index
}
/// Generates a call to `__nac3_ndarray_flatten_index`. Returns the flattened index for the
/// multidimensional index.
///
/// * `ndarray` - LLVM pointer to the `NDArray`. This value must be the LLVM representation of an
/// `NDArray`.
/// * `indices` - The multidimensional index to compute the flattened index for.
pub fn call_ndarray_flatten_index<'ctx, G, Index>(
pub fn call_nac3_range_len_i32<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: NDArrayValue<'ctx>,
indices: &Index,
) -> IntValue<'ctx>
where
G: CodeGenerator + ?Sized,
Index: ArrayLikeIndexer<'ctx>,
{
call_ndarray_flatten_index_impl(generator, ctx, ndarray, indices)
start: Instance<'ctx, Int<Int32>>,
stop: Instance<'ctx, Int<Int32>>,
step: Instance<'ctx, Int<Int32>>,
) -> Instance<'ctx, Int<Int32>> {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_range_len_i32");
CallFunction::begin(generator, ctx, &name)
.arg(start)
.arg(stop)
.arg(step)
.returning_auto("range_len")
}
/// Generates a call to `__nac3_ndarray_calc_broadcast`. Returns a tuple containing the number of
/// dimension and size of each dimension of the resultant `ndarray`.
pub fn call_ndarray_calc_broadcast<'ctx, G: CodeGenerator + ?Sized>(
pub fn call_nac3_ndarray_util_assert_shape_no_negative<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
lhs: NDArrayValue<'ctx>,
rhs: NDArrayValue<'ctx>,
) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
let llvm_usize = generator.get_size_type(ctx.ctx);
let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
32 => "__nac3_ndarray_calc_broadcast",
64 => "__nac3_ndarray_calc_broadcast64",
bw => unreachable!("Unsupported size type bit width: {}", bw),
};
let ndarray_calc_broadcast_fn =
ctx.module.get_function(ndarray_calc_broadcast_fn_name).unwrap_or_else(|| {
let fn_type = llvm_usize.fn_type(
&[
llvm_pusize.into(),
llvm_usize.into(),
llvm_pusize.into(),
llvm_usize.into(),
llvm_pusize.into(),
],
false,
);
ctx.module.add_function(ndarray_calc_broadcast_fn_name, fn_type, None)
});
let lhs_ndims = lhs.load_ndims(ctx);
let rhs_ndims = rhs.load_ndims(ctx);
let min_ndims = llvm_intrinsics::call_int_umin(ctx, lhs_ndims, rhs_ndims, None);
gen_for_callback_incrementing(
ndims: Instance<'ctx, Int<SizeT>>,
shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let name = get_sizet_dependent_function_name(
generator,
ctx,
llvm_usize.const_zero(),
(min_ndims, false),
|generator, ctx, _, idx| {
let idx = ctx.builder.build_int_sub(min_ndims, idx, "").unwrap();
let (lhs_dim_sz, rhs_dim_sz) = unsafe {
(
lhs.dim_sizes().get_typed_unchecked(ctx, generator, &idx, None),
rhs.dim_sizes().get_typed_unchecked(ctx, generator, &idx, None),
)
};
let llvm_usize_const_one = llvm_usize.const_int(1, false);
let lhs_eqz = ctx
.builder
.build_int_compare(IntPredicate::EQ, lhs_dim_sz, llvm_usize_const_one, "")
.unwrap();
let rhs_eqz = ctx
.builder
.build_int_compare(IntPredicate::EQ, rhs_dim_sz, llvm_usize_const_one, "")
.unwrap();
let lhs_or_rhs_eqz = ctx.builder.build_or(lhs_eqz, rhs_eqz, "").unwrap();
let lhs_eq_rhs = ctx
.builder
.build_int_compare(IntPredicate::EQ, lhs_dim_sz, rhs_dim_sz, "")
.unwrap();
let is_compatible = ctx.builder.build_or(lhs_or_rhs_eqz, lhs_eq_rhs, "").unwrap();
ctx.make_assert(
generator,
is_compatible,
"0:ValueError",
"operands could not be broadcast together",
[None, None, None],
ctx.current_loc,
"__nac3_ndarray_util_assert_shape_no_negative",
);
Ok(())
},
llvm_usize.const_int(1, false),
)
.unwrap();
let max_ndims = llvm_intrinsics::call_int_umax(ctx, lhs_ndims, rhs_ndims, None);
let lhs_dims = lhs.dim_sizes().base_ptr(ctx, generator);
let lhs_ndims = lhs.load_ndims(ctx);
let rhs_dims = rhs.dim_sizes().base_ptr(ctx, generator);
let rhs_ndims = rhs.load_ndims(ctx);
let out_dims = ctx.builder.build_array_alloca(llvm_usize, max_ndims, "").unwrap();
let out_dims = ArraySliceValue::from_ptr_val(out_dims, max_ndims, None);
ctx.builder
.build_call(
ndarray_calc_broadcast_fn,
&[
lhs_dims.into(),
lhs_ndims.into(),
rhs_dims.into(),
rhs_ndims.into(),
out_dims.base_ptr(ctx, generator).into(),
],
"",
)
.unwrap();
TypedArrayLikeAdapter::from(
out_dims,
Box::new(|_, v| v.into_int_value()),
Box::new(|_, v| v.into()),
)
CallFunction::begin(generator, ctx, &name).arg(ndims).arg(shape).returning_void();
}
/// Generates a call to `__nac3_ndarray_calc_broadcast_idx`. Returns an [`ArrayAllocaValue`]
/// containing the indices used for accessing `array` corresponding to the index of the broadcasted
/// array `broadcast_idx`.
pub fn call_ndarray_calc_broadcast_index<
'ctx,
G: CodeGenerator + ?Sized,
BroadcastIdx: UntypedArrayLikeAccessor<'ctx>,
>(
pub fn call_nac3_ndarray_util_assert_output_shape_same<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
array: NDArrayValue<'ctx>,
broadcast_idx: &BroadcastIdx,
) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
let llvm_i32 = ctx.ctx.i32_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
32 => "__nac3_ndarray_calc_broadcast_idx",
64 => "__nac3_ndarray_calc_broadcast_idx64",
bw => unreachable!("Unsupported size type bit width: {}", bw),
};
let ndarray_calc_broadcast_fn =
ctx.module.get_function(ndarray_calc_broadcast_fn_name).unwrap_or_else(|| {
let fn_type = llvm_usize.fn_type(
&[llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into(), llvm_pi32.into()],
false,
ndarray_ndims: Instance<'ctx, Int<SizeT>>,
ndarray_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
output_ndims: Instance<'ctx, Int<SizeT>>,
output_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let name = get_sizet_dependent_function_name(
generator,
ctx,
"__nac3_ndarray_util_assert_output_shape_same",
);
ctx.module.add_function(ndarray_calc_broadcast_fn_name, fn_type, None)
});
let broadcast_size = broadcast_idx.size(ctx, generator);
let out_idx = ctx.builder.build_array_alloca(llvm_i32, broadcast_size, "").unwrap();
let array_dims = array.dim_sizes().base_ptr(ctx, generator);
let array_ndims = array.load_ndims(ctx);
let broadcast_idx_ptr = unsafe {
broadcast_idx.ptr_offset_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
};
ctx.builder
.build_call(
ndarray_calc_broadcast_fn,
&[array_dims.into(), array_ndims.into(), broadcast_idx_ptr.into(), out_idx.into()],
"",
)
.unwrap();
TypedArrayLikeAdapter::from(
ArraySliceValue::from_ptr_val(out_idx, broadcast_size, None),
Box::new(|_, v| v.into_int_value()),
Box::new(|_, v| v.into()),
)
CallFunction::begin(generator, ctx, &name)
.arg(ndarray_ndims)
.arg(ndarray_shape)
.arg(output_ndims)
.arg(output_shape)
.returning_void();
}
pub fn call_nac3_ndarray_size<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) -> Instance<'ctx, Int<SizeT>> {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_size");
CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("size")
}
pub fn call_nac3_ndarray_nbytes<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) -> Instance<'ctx, Int<SizeT>> {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_nbytes");
CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("nbytes")
}
pub fn call_nac3_ndarray_len<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) -> Instance<'ctx, Int<SizeT>> {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_len");
CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("len")
}
pub fn call_nac3_ndarray_is_c_contiguous<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) -> Instance<'ctx, Int<Bool>> {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_is_c_contiguous");
CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("is_c_contiguous")
}
pub fn call_nac3_ndarray_get_nth_pelement<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
index: Instance<'ctx, Int<SizeT>>,
) -> Instance<'ctx, Ptr<Int<Byte>>> {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_get_nth_pelement");
CallFunction::begin(generator, ctx, &name).arg(ndarray).arg(index).returning_auto("pelement")
}
pub fn call_nac3_ndarray_get_pelement_by_indices<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
indices: Instance<'ctx, Ptr<Int<SizeT>>>,
) -> Instance<'ctx, Ptr<Int<Byte>>> {
let name =
get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_get_pelement_by_indices");
CallFunction::begin(generator, ctx, &name).arg(ndarray).arg(indices).returning_auto("pelement")
}
pub fn call_nac3_ndarray_set_strides_by_shape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) {
let name =
get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_set_strides_by_shape");
CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_void();
}
pub fn call_nac3_ndarray_copy_data<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_copy_data");
CallFunction::begin(generator, ctx, &name).arg(src_ndarray).arg(dst_ndarray).returning_void();
}
pub fn call_nac3_nditer_initialize<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
indices: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_initialize");
CallFunction::begin(generator, ctx, &name).arg(iter).arg(ndarray).arg(indices).returning_void();
}
pub fn call_nac3_nditer_has_next<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
) -> Instance<'ctx, Int<Bool>> {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_has_next");
CallFunction::begin(generator, ctx, &name).arg(iter).returning_auto("has_next")
}
pub fn call_nac3_nditer_next<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
) {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_next");
CallFunction::begin(generator, ctx, &name).arg(iter).returning_void();
}
pub fn call_nac3_ndarray_index<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
num_indices: Instance<'ctx, Int<SizeT>>,
indices: Instance<'ctx, Ptr<Struct<NDIndex>>>,
src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_index");
CallFunction::begin(generator, ctx, &name)
.arg(num_indices)
.arg(indices)
.arg(src_ndarray)
.arg(dst_ndarray)
.returning_void();
}
pub fn call_nac3_ndarray_array_set_and_validate_list_shape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
list: Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>>,
ndims: Instance<'ctx, Int<SizeT>>,
shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let name = get_sizet_dependent_function_name(
generator,
ctx,
"__nac3_ndarray_array_set_and_validate_list_shape",
);
CallFunction::begin(generator, ctx, &name).arg(list).arg(ndims).arg(shape).returning_void();
}
pub fn call_nac3_ndarray_array_write_list_to_array<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
list: Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>>,
ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) {
let name = get_sizet_dependent_function_name(
generator,
ctx,
"__nac3_ndarray_array_write_list_to_array",
);
CallFunction::begin(generator, ctx, &name).arg(list).arg(ndarray).returning_void();
}
pub fn call_nac3_ndarray_reshape_resolve_and_check_new_shape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
size: Instance<'ctx, Int<SizeT>>,
new_ndims: Instance<'ctx, Int<SizeT>>,
new_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let name = get_sizet_dependent_function_name(
generator,
ctx,
"__nac3_ndarray_reshape_resolve_and_check_new_shape",
);
CallFunction::begin(generator, ctx, &name)
.arg(size)
.arg(new_ndims)
.arg(new_shape)
.returning_void();
}
pub fn call_nac3_ndarray_broadcast_to<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
) {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_broadcast_to");
CallFunction::begin(generator, ctx, &name).arg(src_ndarray).arg(dst_ndarray).returning_void();
}
pub fn call_nac3_ndarray_broadcast_shapes<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
num_shape_entries: Instance<'ctx, Int<SizeT>>,
shape_entries: Instance<'ctx, Ptr<Struct<ShapeEntry>>>,
dst_ndims: Instance<'ctx, Int<SizeT>>,
dst_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_broadcast_shapes");
CallFunction::begin(generator, ctx, &name)
.arg(num_shape_entries)
.arg(shape_entries)
.arg(dst_ndims)
.arg(dst_shape)
.returning_void();
}
pub fn call_nac3_ndarray_transpose<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
num_axes: Instance<'ctx, Int<SizeT>>,
axes: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_transpose");
CallFunction::begin(generator, ctx, &name)
.arg(src_ndarray)
.arg(dst_ndarray)
.arg(num_axes)
.arg(axes)
.returning_void();
}
#[allow(clippy::too_many_arguments)]
pub fn call_nac3_ndarray_matmul_calculate_shapes<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
a_ndims: Instance<'ctx, Int<SizeT>>,
a_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
b_ndims: Instance<'ctx, Int<SizeT>>,
b_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
final_ndims: Instance<'ctx, Int<SizeT>>,
new_a_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
new_b_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
dst_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let name =
get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_matmul_calculate_shapes");
CallFunction::begin(generator, ctx, &name)
.arg(a_ndims)
.arg(a_shape)
.arg(b_ndims)
.arg(b_shape)
.arg(final_ndims)
.arg(new_a_shape)
.arg(new_b_shape)
.arg(dst_shape)
.returning_void();
}

View File

@ -35,6 +35,40 @@ fn get_float_intrinsic_repr(ctx: &Context, ft: FloatType) -> &'static str {
unreachable!()
}
/// Invokes the [`llvm.va_start`](https://llvm.org/docs/LangRef.html#llvm-va-start-intrinsic)
/// intrinsic.
pub fn call_va_start<'ctx>(ctx: &CodeGenContext<'ctx, '_>, arglist: PointerValue<'ctx>) {
const FN_NAME: &str = "llvm.va_start";
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, &[arglist.into()], "").unwrap();
}
/// Invokes the [`llvm.va_start`](https://llvm.org/docs/LangRef.html#llvm-va-start-intrinsic)
/// intrinsic.
pub fn call_va_end<'ctx>(ctx: &CodeGenContext<'ctx, '_>, arglist: PointerValue<'ctx>) {
const FN_NAME: &str = "llvm.va_end";
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, &[arglist.into()], "").unwrap();
}
/// Invokes the [`llvm.stacksave`](https://llvm.org/docs/LangRef.html#llvm-stacksave-intrinsic)
/// intrinsic.
pub fn call_stacksave<'ctx>(
@ -171,8 +205,9 @@ pub fn call_memcpy_generic<'ctx>(
/// * `$ctx:ident`: Reference to the current Code Generation Context
/// * `$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
/// * `$map_fn:ident`: Mapping function to be applied on `BasicValue` (`BasicValue` -> Function Return Type).
/// Use `BasicValueEnum::into_int_value` for Integer return type and
/// `BasicValueEnum::into_float_value` for Float return type
/// * `$llvm_ty:ident`: Type of first operand
/// * `,($val:ident)*`: Comma separated list of operands
macro_rules! generate_llvm_intrinsic_fn_body {
@ -188,7 +223,7 @@ macro_rules! generate_llvm_intrinsic_fn_body {
/// Arguments:
/// * `float/int`: Indicates the return and argument type of the function
/// * `$fn_name:ident`: The identifier of the rust function to be generated
/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function
/// * `$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

View File

@ -1,7 +1,6 @@
use crate::{
codegen::classes::{ListType, NDArrayType, ProxyType, RangeType},
symbol_resolver::{StaticValue, SymbolResolver},
toplevel::{helper::PrimDef, numpy::unpack_ndarray_var_tys, TopLevelContext, TopLevelDef},
toplevel::{helper::PrimDef, TopLevelContext, TopLevelDef},
typecheck::{
type_inferencer::{CodeLocation, PrimitiveStore},
typedef::{CallId, FuncArg, Type, TypeEnum, Unifier},
@ -24,7 +23,9 @@ use inkwell::{
AddressSpace, IntPredicate, OptimizationLevel,
};
use itertools::Itertools;
use model::*;
use nac3parser::ast::{Location, Stmt, StrRef};
use object::{list::List, ndarray::NDArray, range::Range};
use parking_lot::{Condvar, Mutex};
use std::collections::{HashMap, HashSet};
use std::sync::{
@ -41,7 +42,9 @@ pub mod extern_fns;
mod generator;
pub mod irrt;
pub mod llvm_intrinsics;
pub mod model;
pub mod numpy;
pub mod object;
pub mod stmt;
#[cfg(test)]
@ -68,6 +71,16 @@ pub struct CodeGenLLVMOptions {
pub target: CodeGenTargetMachineOptions,
}
impl CodeGenLLVMOptions {
/// Creates a [`TargetMachine`] using the target options specified by this struct.
///
/// See [`Target::create_target_machine`].
#[must_use]
pub fn create_target_machine(&self) -> Option<TargetMachine> {
self.target.create_target_machine(self.opt_level)
}
}
/// Additional options for code generation for the target machine.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct CodeGenTargetMachineOptions {
@ -338,6 +351,10 @@ impl WorkerRegistry {
let mut builder = context.create_builder();
let mut module = context.create_module(generator.get_name());
let target_machine = self.llvm_options.create_target_machine().unwrap();
module.set_data_layout(&target_machine.get_target_data().get_data_layout());
module.set_triple(&target_machine.get_triple());
module.add_basic_value_flag(
"Debug Info Version",
inkwell::module::FlagBehavior::Warning,
@ -361,6 +378,10 @@ impl WorkerRegistry {
errors.insert(e);
// create a new empty module just to continue codegen and collect errors
module = context.create_module(&format!("{}_recover", generator.get_name()));
let target_machine = self.llvm_options.create_target_machine().unwrap();
module.set_data_layout(&target_machine.get_target_data().get_data_layout());
module.set_triple(&target_machine.get_triple());
}
}
*self.task_count.lock() -= 1;
@ -426,7 +447,7 @@ pub struct CodeGenTask {
fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
ctx: &'ctx Context,
module: &Module<'ctx>,
generator: &mut G,
generator: &G,
unifier: &mut Unifier,
top_level: &TopLevelContext,
type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
@ -466,17 +487,12 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
type_cache,
*params.iter().next().unwrap().1,
);
ListType::new(generator, ctx, element_type).as_base_type().into()
let item = Any(element_type);
Ptr(Struct(List { item })).get_type(generator, ctx).as_basic_type_enum()
}
TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
let (dtype, _) = unpack_ndarray_var_tys(unifier, ty);
let element_type = get_llvm_type(
ctx, module, generator, unifier, top_level, type_cache, dtype,
);
NDArrayType::new(generator, ctx, element_type).as_base_type().into()
Ptr(Struct(NDArray)).get_type(generator, ctx).as_basic_type_enum()
}
_ => unreachable!(
@ -520,8 +536,10 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
};
return ty;
}
TTuple { ty } => {
TTuple { ty, is_vararg_ctx } => {
// a struct with fields in the order present in the tuple
assert!(!is_vararg_ctx, "Tuples in vararg context must be instantiated with the correct number of arguments before calling get_llvm_type");
let fields = ty
.iter()
.map(|ty| {
@ -551,7 +569,7 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
fn get_llvm_abi_type<'ctx, G: CodeGenerator + ?Sized>(
ctx: &'ctx Context,
module: &Module<'ctx>,
generator: &mut G,
generator: &G,
unifier: &mut Unifier,
top_level: &TopLevelContext,
type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
@ -560,11 +578,11 @@ fn get_llvm_abi_type<'ctx, G: CodeGenerator + ?Sized>(
) -> BasicTypeEnum<'ctx> {
// If the type is used in the definition of a function, return `i1` instead of `i8` for ABI
// consistency.
return if unifier.unioned(ty, primitives.bool) {
if unifier.unioned(ty, primitives.bool) {
ctx.bool_type().into()
} else {
get_llvm_type(ctx, module, generator, unifier, top_level, type_cache, ty)
};
}
}
/// Whether `sret` is needed for a return value with type `ty`.
@ -589,6 +607,40 @@ fn need_sret(ty: BasicTypeEnum) -> bool {
need_sret_impl(ty, true)
}
/// Returns the [`BasicTypeEnum`] representing a `va_list` struct for variadic arguments.
fn get_llvm_valist_type<'ctx>(ctx: &'ctx Context, triple: &TargetTriple) -> BasicTypeEnum<'ctx> {
let triple = TargetMachine::normalize_triple(triple);
let triple = triple.as_str().to_str().unwrap();
let arch = triple.split('-').next().unwrap();
let llvm_pi8 = ctx.i8_type().ptr_type(AddressSpace::default());
// Referenced from parseArch() in llvm/lib/Support/Triple.cpp
match arch {
"i386" | "i486" | "i586" | "i686" | "riscv32" => {
ctx.i8_type().ptr_type(AddressSpace::default()).into()
}
"amd64" | "x86_64" | "x86_64h" => {
let llvm_i32 = ctx.i32_type();
let va_list_tag = ctx.opaque_struct_type("struct.__va_list_tag");
va_list_tag.set_body(
&[llvm_i32.into(), llvm_i32.into(), llvm_pi8.into(), llvm_pi8.into()],
false,
);
va_list_tag.into()
}
"armv7" => {
let va_list = ctx.opaque_struct_type("struct.__va_list");
va_list.set_body(&[llvm_pi8.into()], false);
va_list.into()
}
triple => {
todo!("Unsupported platform for varargs: {triple}")
}
}
}
/// Implementation for generating LLVM IR for a function.
pub fn gen_func_impl<
'ctx,
@ -668,7 +720,7 @@ pub fn gen_func_impl<
Some(t) => t.as_basic_type_enum(),
}
}),
(primitives.range, RangeType::new(context).as_base_type().into()),
(primitives.range, Ptr(Range::<Int32>::default()).get_type(generator, context).into()),
(primitives.exception, {
let name = "Exception";
if let Some(t) = module.get_struct_type(name) {
@ -700,6 +752,7 @@ pub fn gen_func_impl<
name: arg.name,
ty: task.store.to_unifier_type(&mut unifier, &primitives, arg.ty, &mut cache),
default_value: arg.default_value.clone(),
is_vararg: arg.is_vararg,
})
.collect_vec(),
task.store.to_unifier_type(&mut unifier, &primitives, *ret, &mut cache),
@ -722,7 +775,10 @@ pub fn gen_func_impl<
let has_sret = ret_type.map_or(false, |ty| need_sret(ty));
let mut params = args
.iter()
.filter(|arg| !arg.is_vararg)
.map(|arg| {
debug_assert!(!arg.is_vararg);
get_llvm_abi_type(
context,
&module,
@ -741,9 +797,12 @@ pub fn gen_func_impl<
params.insert(0, ret_type.unwrap().ptr_type(AddressSpace::default()).into());
}
debug_assert!(matches!(args.iter().filter(|arg| arg.is_vararg).count(), 0..=1));
let vararg_arg = args.iter().find(|arg| arg.is_vararg);
let fn_type = match ret_type {
Some(ret_type) if !has_sret => ret_type.fn_type(&params, false),
_ => context.void_type().fn_type(&params, false),
Some(ret_type) if !has_sret => ret_type.fn_type(&params, vararg_arg.is_some()),
_ => context.void_type().fn_type(&params, vararg_arg.is_some()),
};
let symbol = &task.symbol_name;
@ -773,7 +832,9 @@ pub fn gen_func_impl<
let mut var_assignment = HashMap::new();
let offset = u32::from(has_sret);
for (n, arg) in args.iter().enumerate() {
// Store non-vararg argument values into local variables
for (n, arg) in args.iter().enumerate().filter(|(_, arg)| !arg.is_vararg) {
let param = fn_val.get_nth_param((n as u32) + offset).unwrap();
let local_type = get_llvm_type(
context,
@ -806,6 +867,8 @@ pub fn gen_func_impl<
var_assignment.insert(arg.name, (alloca, None, 0));
}
// TODO: Save vararg parameters as list
let return_buffer = if has_sret {
Some(fn_val.get_nth_param(0).unwrap().into_pointer_value())
} else {
@ -1028,3 +1091,9 @@ fn gen_in_range_check<'ctx>(
ctx.builder.build_int_compare(IntPredicate::SLT, lo, hi, "cmp").unwrap()
}
/// Returns the internal name for the `va_count` argument, used to indicate the number of arguments
/// passed to the variadic function.
fn get_va_count_arg_name(arg_name: StrRef) -> StrRef {
format!("__{}_va_count", &arg_name).into()
}

View File

@ -0,0 +1,42 @@
use inkwell::{
context::Context,
types::{BasicType, BasicTypeEnum},
values::BasicValueEnum,
};
use crate::codegen::CodeGenerator;
use super::*;
/// A [`Model`] of any [`BasicTypeEnum`].
///
/// Use this when you cannot know the type beforehand or cannot be abstracted with [`Model`].
#[derive(Debug, Clone, Copy)]
pub struct Any<'ctx>(pub BasicTypeEnum<'ctx>);
impl<'ctx> Model<'ctx> for Any<'ctx> {
type Value = BasicValueEnum<'ctx>;
type Type = BasicTypeEnum<'ctx>;
fn get_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
_ctx: &'ctx Context,
) -> Self::Type {
self.0
}
fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
&self,
_generator: &mut G,
_ctx: &'ctx Context,
ty: T,
) -> Result<(), ModelError> {
let ty = ty.as_basic_type_enum();
if ty == self.0 {
Ok(())
} else {
Err(ModelError(format!("Expecting {}, but got {}", self.0, ty)))
}
}
}

View File

@ -0,0 +1,140 @@
use std::fmt;
use inkwell::{
context::Context,
types::{ArrayType, BasicType, BasicTypeEnum},
values::{ArrayValue, IntValue},
};
use crate::codegen::{CodeGenContext, CodeGenerator};
use super::*;
pub trait LenKind: fmt::Debug + Clone + Copy {
fn get_length(&self) -> u32;
}
/// A statically known length.
#[derive(Debug, Clone, Copy, Default)]
pub struct Len<const N: u32>;
/// A dynamically known length.
#[derive(Debug, Clone, Copy)]
pub struct AnyLen(pub u32);
impl<const N: u32> LenKind for Len<N> {
fn get_length(&self) -> u32 {
N
}
}
impl LenKind for AnyLen {
fn get_length(&self) -> u32 {
self.0
}
}
/// A Model for an [`ArrayType`].
#[derive(Debug, Clone, Copy, Default)]
pub struct Array<Len, Item> {
/// Length of this array.
pub len: Len,
/// [`Model`] of an array item.
pub item: Item,
}
impl<'ctx, Len: LenKind, Item: Model<'ctx>> Model<'ctx> for Array<Len, Item> {
type Value = ArrayValue<'ctx>;
type Type = ArrayType<'ctx>;
fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
self.item.get_type(generator, ctx).array_type(self.len.get_length())
}
fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
ty: T,
) -> Result<(), ModelError> {
let ty = ty.as_basic_type_enum();
let BasicTypeEnum::ArrayType(ty) = ty else {
return Err(ModelError(format!("Expecting ArrayType, but got {ty:?}")));
};
if ty.len() != self.len.get_length() {
return Err(ModelError(format!(
"Expecting ArrayType with size {}, but got an ArrayType with size {}",
ty.len(),
self.len.get_length()
)));
}
self.item
.check_type(generator, ctx, ty.get_element_type())
.map_err(|err| err.under_context("an ArrayType"))?;
Ok(())
}
}
impl<'ctx, Len: LenKind, Item: Model<'ctx>> Instance<'ctx, Ptr<Array<Len, Item>>> {
/// Get the pointer to the `i`-th (0-based) array element.
pub fn gep(
&self,
ctx: &CodeGenContext<'ctx, '_>,
i: IntValue<'ctx>,
) -> Instance<'ctx, Ptr<Item>> {
let zero = ctx.ctx.i32_type().const_zero();
let ptr = unsafe { ctx.builder.build_in_bounds_gep(self.value, &[zero, i], "").unwrap() };
Ptr(self.model.0.item).believe_value(ptr)
}
/// Like `gep` but `i` is a constant.
pub fn gep_const(&self, ctx: &CodeGenContext<'ctx, '_>, i: u64) -> Instance<'ctx, Ptr<Item>> {
assert!(
i < u64::from(self.model.0.len.get_length()),
"Index {i} is out of bounds. Array length = {}",
self.model.0.len.get_length()
);
let i = ctx.ctx.i32_type().const_int(i, false);
self.gep(ctx, i)
}
/// Convenience function equivalent to `.gep(...).load(...)`.
pub fn get<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
i: IntValue<'ctx>,
) -> Instance<'ctx, Item> {
self.gep(ctx, i).load(generator, ctx)
}
/// Like `get` but `i` is a constant.
pub fn get_const<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
i: u64,
) -> Instance<'ctx, Item> {
self.gep_const(ctx, i).load(generator, ctx)
}
/// Convenience function equivalent to `.gep(...).store(...)`.
pub fn set(
&self,
ctx: &CodeGenContext<'ctx, '_>,
i: IntValue<'ctx>,
value: Instance<'ctx, Item>,
) {
self.gep(ctx, i).store(ctx, value);
}
/// Like `set` but `i` is a constant.
pub fn set_const(&self, ctx: &CodeGenContext<'ctx, '_>, i: u64, value: Instance<'ctx, Item>) {
self.gep_const(ctx, i).store(ctx, value);
}
}

View File

@ -0,0 +1,156 @@
use std::fmt;
use inkwell::{context::Context, types::*, values::*};
use itertools::Itertools;
use super::*;
use crate::codegen::{CodeGenContext, CodeGenerator};
#[derive(Debug, Clone)]
pub struct ModelError(pub String);
impl ModelError {
pub(super) fn under_context(mut self, context: &str) -> Self {
self.0.push_str(" ... in ");
self.0.push_str(context);
self
}
}
pub trait Model<'ctx>: fmt::Debug + Clone + Copy {
type Value: BasicValue<'ctx> + TryFrom<BasicValueEnum<'ctx>>;
type Type: BasicType<'ctx>;
/// Return the [`BasicType`] of this model.
#[must_use]
fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type;
fn sizeof<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
) -> IntValue<'ctx> {
self.get_type(generator, ctx).size_of().unwrap()
}
/// Check if a [`BasicType`] is the same type of this model.
fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
ty: T,
) -> Result<(), ModelError>;
/// Create an instance from a value with [`Instance::model`] being this model.
///
/// Caller must make sure the type of `value` and the type of this `model` are equivalent.
#[must_use]
fn believe_value(&self, value: Self::Value) -> Instance<'ctx, Self> {
Instance { model: *self, value }
}
/// Check if a [`BasicValue`]'s type is equivalent to the type of this model.
/// Wrap it into an [`Instance`] if it is.
fn check_value<V: BasicValue<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
value: V,
) -> Result<Instance<'ctx, Self>, ModelError> {
let value = value.as_basic_value_enum();
self.check_type(generator, ctx, value.get_type())
.map_err(|err| err.under_context(format!("the value {value:?}").as_str()))?;
let Ok(value) = Self::Value::try_from(value) else {
unreachable!("check_type() has bad implementation")
};
Ok(self.believe_value(value))
}
// Allocate a value on the stack and return its pointer.
fn alloca<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Ptr<Self>> {
let p = ctx.builder.build_alloca(self.get_type(generator, ctx.ctx), "").unwrap();
Ptr(*self).believe_value(p)
}
// Allocate an array on the stack and return its pointer.
fn array_alloca<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
len: IntValue<'ctx>,
) -> Instance<'ctx, Ptr<Self>> {
let p = ctx.builder.build_array_alloca(self.get_type(generator, ctx.ctx), len, "").unwrap();
Ptr(*self).believe_value(p)
}
fn var_alloca<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
name: Option<&str>,
) -> Result<Instance<'ctx, Ptr<Self>>, String> {
let ty = self.get_type(generator, ctx.ctx).as_basic_type_enum();
let p = generator.gen_var_alloc(ctx, ty, name)?;
Ok(Ptr(*self).believe_value(p))
}
fn array_var_alloca<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
len: IntValue<'ctx>,
name: Option<&'ctx str>,
) -> Result<Instance<'ctx, Ptr<Self>>, String> {
// TODO: Remove ArraySliceValue
let ty = self.get_type(generator, ctx.ctx).as_basic_type_enum();
let p = generator.gen_array_var_alloc(ctx, ty, len, name)?;
Ok(Ptr(*self).believe_value(PointerValue::from(p)))
}
fn const_array<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
values: &[Instance<'ctx, Self>],
) -> Instance<'ctx, Array<AnyLen, Self>> {
macro_rules! make {
($t:expr, $into_value:expr) => {
$t.const_array(
&values
.iter()
.map(|x| $into_value(x.value.as_basic_value_enum()))
.collect_vec(),
)
};
}
let value = match self.get_type(generator, ctx).as_basic_type_enum() {
BasicTypeEnum::ArrayType(t) => make!(t, BasicValueEnum::into_array_value),
BasicTypeEnum::IntType(t) => make!(t, BasicValueEnum::into_int_value),
BasicTypeEnum::FloatType(t) => make!(t, BasicValueEnum::into_float_value),
BasicTypeEnum::PointerType(t) => make!(t, BasicValueEnum::into_pointer_value),
BasicTypeEnum::StructType(t) => make!(t, BasicValueEnum::into_struct_value),
BasicTypeEnum::VectorType(t) => make!(t, BasicValueEnum::into_vector_value),
};
Array { len: AnyLen(values.len() as u32), item: *self }
.check_value(generator, ctx, value)
.unwrap()
}
}
#[derive(Debug, Clone, Copy)]
pub struct Instance<'ctx, M: Model<'ctx>> {
/// The model of this instance.
pub model: M,
/// The value of this instance.
///
/// Caller must make sure the type of `value` and the type of this `model` are equivalent,
/// down to having the same [`IntType::get_bit_width`] in case of [`IntType`] for example.
pub value: M::Value,
}

View File

@ -0,0 +1,87 @@
use std::fmt;
use inkwell::{context::Context, types::FloatType, values::FloatValue};
use crate::codegen::CodeGenerator;
use super::*;
pub trait FloatKind<'ctx>: fmt::Debug + Clone + Copy {
fn get_float_type<G: CodeGenerator + ?Sized>(
&self,
generator: &G,
ctx: &'ctx Context,
) -> FloatType<'ctx>;
}
#[derive(Debug, Clone, Copy, Default)]
pub struct Float32;
#[derive(Debug, Clone, Copy, Default)]
pub struct Float64;
impl<'ctx> FloatKind<'ctx> for Float32 {
fn get_float_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
ctx: &'ctx Context,
) -> FloatType<'ctx> {
ctx.f32_type()
}
}
impl<'ctx> FloatKind<'ctx> for Float64 {
fn get_float_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
ctx: &'ctx Context,
) -> FloatType<'ctx> {
ctx.f64_type()
}
}
#[derive(Debug, Clone, Copy)]
pub struct AnyFloat<'ctx>(FloatType<'ctx>);
impl<'ctx> FloatKind<'ctx> for AnyFloat<'ctx> {
fn get_float_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
_ctx: &'ctx Context,
) -> FloatType<'ctx> {
self.0
}
}
#[derive(Debug, Clone, Copy, Default)]
pub struct Float<N>(pub N);
impl<'ctx, N: FloatKind<'ctx>> Model<'ctx> for Float<N> {
type Value = FloatValue<'ctx>;
type Type = FloatType<'ctx>;
fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
self.0.get_float_type(generator, ctx)
}
fn check_type<T: inkwell::types::BasicType<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
ty: T,
) -> Result<(), ModelError> {
let ty = ty.as_basic_type_enum();
let Ok(ty) = FloatType::try_from(ty) else {
return Err(ModelError(format!("Expecting FloatType, but got {ty:?}")));
};
let exp_ty = self.0.get_float_type(generator, ctx);
// TODO: Inkwell does not have get_bit_width for FloatType?
// TODO: Quick hack for now, but does this actually work?
if ty != exp_ty {
return Err(ModelError(format!("Expecting {exp_ty:?}, but got {ty:?}")));
}
Ok(())
}
}

View File

@ -0,0 +1,106 @@
use inkwell::{
attributes::{Attribute, AttributeLoc},
types::{BasicMetadataTypeEnum, BasicType, FunctionType},
values::{AnyValue, BasicMetadataValueEnum, BasicValue, BasicValueEnum, CallSiteValue},
};
use itertools::Itertools;
use crate::codegen::{CodeGenContext, CodeGenerator};
use super::*;
#[derive(Debug, Clone, Copy)]
struct Arg<'ctx> {
ty: BasicMetadataTypeEnum<'ctx>,
val: BasicMetadataValueEnum<'ctx>,
}
/// A structure to construct & call an LLVM function.
///
/// This is a helper to reduce IRRT Inkwell function call boilerplate
// TODO: Remove the lifetimes somehow? There are 4 of them.
pub struct CallFunction<'ctx, 'a, 'b, 'c, 'd, G: CodeGenerator + ?Sized> {
generator: &'d mut G,
ctx: &'b CodeGenContext<'ctx, 'a>,
/// Function name
name: &'c str,
/// Call arguments
args: Vec<Arg<'ctx>>,
/// LLVM function Attributes
attrs: Vec<&'static str>,
}
impl<'ctx, 'a, 'b, 'c, 'd, G: CodeGenerator + ?Sized> CallFunction<'ctx, 'a, 'b, 'c, 'd, G> {
pub fn begin(generator: &'d mut G, ctx: &'b CodeGenContext<'ctx, 'a>, name: &'c str) -> Self {
CallFunction { generator, ctx, name, args: Vec::new(), attrs: Vec::new() }
}
/// Push a list of LLVM function attributes to the function declaration.
#[must_use]
pub fn attrs(mut self, attrs: Vec<&'static str>) -> Self {
self.attrs = attrs;
self
}
/// Push a call argument to the function call.
#[allow(clippy::needless_pass_by_value)]
#[must_use]
pub fn arg<M: Model<'ctx>>(mut self, arg: Instance<'ctx, M>) -> Self {
let arg = Arg {
ty: arg.model.get_type(self.generator, self.ctx.ctx).as_basic_type_enum().into(),
val: arg.value.as_basic_value_enum().into(),
};
self.args.push(arg);
self
}
/// Call the function and expect the function to return a value of type of `return_model`.
#[must_use]
pub fn returning<M: Model<'ctx>>(self, name: &str, return_model: M) -> Instance<'ctx, M> {
let ret_ty = return_model.get_type(self.generator, self.ctx.ctx);
let ret = self.get_function(|tys| ret_ty.fn_type(tys, false), name);
let ret = BasicValueEnum::try_from(ret.as_any_value_enum()).unwrap(); // Must work
let ret = return_model.check_value(self.generator, self.ctx.ctx, ret).unwrap(); // Must work
ret
}
/// Like [`CallFunction::returning_`] but `return_model` is automatically inferred.
#[must_use]
pub fn returning_auto<M: Model<'ctx> + Default>(self, name: &str) -> Instance<'ctx, M> {
self.returning(name, M::default())
}
/// Call the function and expect the function to return a void-type.
pub fn returning_void(self) {
let ret_ty = self.ctx.ctx.void_type();
let _ = self.get_function(|tys| ret_ty.fn_type(tys, false), "");
}
fn get_function<F>(&self, make_fn_type: F, return_value_name: &str) -> CallSiteValue<'ctx>
where
F: FnOnce(&[BasicMetadataTypeEnum<'ctx>]) -> FunctionType<'ctx>,
{
// Get the LLVM function.
let func = self.ctx.module.get_function(self.name).unwrap_or_else(|| {
// Declare the function if it doesn't exist.
let tys = self.args.iter().map(|arg| arg.ty).collect_vec();
let func_type = make_fn_type(&tys);
let func = self.ctx.module.add_function(self.name, func_type, None);
for attr in &self.attrs {
func.add_attribute(
AttributeLoc::Function,
self.ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
);
}
func
});
let vals = self.args.iter().map(|arg| arg.val).collect_vec();
self.ctx.builder.build_call(func, &vals, return_value_name).unwrap()
}
}

View File

@ -0,0 +1,349 @@
use std::fmt;
use inkwell::{
context::Context,
types::{BasicType, IntType},
values::IntValue,
IntPredicate,
};
use crate::codegen::{CodeGenContext, CodeGenerator};
use super::*;
pub trait IntKind<'ctx>: fmt::Debug + Clone + Copy {
fn get_int_type<G: CodeGenerator + ?Sized>(
&self,
generator: &G,
ctx: &'ctx Context,
) -> IntType<'ctx>;
}
#[derive(Debug, Clone, Copy, Default)]
pub struct Bool;
#[derive(Debug, Clone, Copy, Default)]
pub struct Byte;
#[derive(Debug, Clone, Copy, Default)]
pub struct Int32;
#[derive(Debug, Clone, Copy, Default)]
pub struct Int64;
#[derive(Debug, Clone, Copy, Default)]
pub struct SizeT;
impl<'ctx> IntKind<'ctx> for Bool {
fn get_int_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
ctx: &'ctx Context,
) -> IntType<'ctx> {
ctx.bool_type()
}
}
impl<'ctx> IntKind<'ctx> for Byte {
fn get_int_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
ctx: &'ctx Context,
) -> IntType<'ctx> {
ctx.i8_type()
}
}
impl<'ctx> IntKind<'ctx> for Int32 {
fn get_int_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
ctx: &'ctx Context,
) -> IntType<'ctx> {
ctx.i32_type()
}
}
impl<'ctx> IntKind<'ctx> for Int64 {
fn get_int_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
ctx: &'ctx Context,
) -> IntType<'ctx> {
ctx.i64_type()
}
}
impl<'ctx> IntKind<'ctx> for SizeT {
fn get_int_type<G: CodeGenerator + ?Sized>(
&self,
generator: &G,
ctx: &'ctx Context,
) -> IntType<'ctx> {
generator.get_size_type(ctx)
}
}
#[derive(Debug, Clone, Copy)]
pub struct AnyInt<'ctx>(pub IntType<'ctx>);
impl<'ctx> IntKind<'ctx> for AnyInt<'ctx> {
fn get_int_type<G: CodeGenerator + ?Sized>(
&self,
_generator: &G,
_ctx: &'ctx Context,
) -> IntType<'ctx> {
self.0
}
}
#[derive(Debug, Clone, Copy, Default)]
pub struct Int<N>(pub N);
impl<'ctx, N: IntKind<'ctx>> Model<'ctx> for Int<N> {
type Value = IntValue<'ctx>;
type Type = IntType<'ctx>;
fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
self.0.get_int_type(generator, ctx)
}
fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
ty: T,
) -> Result<(), ModelError> {
let ty = ty.as_basic_type_enum();
let Ok(ty) = IntType::try_from(ty) else {
return Err(ModelError(format!("Expecting IntType, but got {ty:?}")));
};
let exp_ty = self.0.get_int_type(generator, ctx);
if ty.get_bit_width() != exp_ty.get_bit_width() {
return Err(ModelError(format!(
"Expecting IntType to have {} bit(s), but got {} bit(s)",
exp_ty.get_bit_width(),
ty.get_bit_width()
)));
}
Ok(())
}
}
impl<'ctx, N: IntKind<'ctx>> Int<N> {
pub fn const_int<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
value: u64,
) -> Instance<'ctx, Self> {
let value = self.get_type(generator, ctx).const_int(value, false);
self.believe_value(value)
}
pub fn const_0<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
) -> Instance<'ctx, Self> {
let value = self.get_type(generator, ctx).const_zero();
self.believe_value(value)
}
pub fn const_1<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
) -> Instance<'ctx, Self> {
self.const_int(generator, ctx, 1)
}
pub fn const_all_ones<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
) -> Instance<'ctx, Self> {
let value = self.get_type(generator, ctx).const_all_ones();
self.believe_value(value)
}
pub fn s_extend_or_bit_cast<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
) -> Instance<'ctx, Self> {
let value = ctx
.builder
.build_int_s_extend_or_bit_cast(value, self.get_type(generator, ctx.ctx), "")
.unwrap();
self.believe_value(value)
}
pub fn s_extend<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
) -> Instance<'ctx, Self> {
let value =
ctx.builder.build_int_s_extend(value, self.get_type(generator, ctx.ctx), "").unwrap();
self.believe_value(value)
}
pub fn z_extend_or_bit_cast<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
) -> Instance<'ctx, Self> {
let value = ctx
.builder
.build_int_z_extend_or_bit_cast(value, self.get_type(generator, ctx.ctx), "")
.unwrap();
self.believe_value(value)
}
pub fn z_extend<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
) -> Instance<'ctx, Self> {
let value =
ctx.builder.build_int_z_extend(value, self.get_type(generator, ctx.ctx), "").unwrap();
self.believe_value(value)
}
pub fn truncate_or_bit_cast<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
) -> Instance<'ctx, Self> {
let value = ctx
.builder
.build_int_truncate_or_bit_cast(value, self.get_type(generator, ctx.ctx), "")
.unwrap();
self.believe_value(value)
}
pub fn truncate<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
) -> Instance<'ctx, Self> {
let value =
ctx.builder.build_int_truncate(value, self.get_type(generator, ctx.ctx), "").unwrap();
self.believe_value(value)
}
}
impl Int<Bool> {
#[must_use]
pub fn const_false<'ctx, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
) -> Instance<'ctx, Self> {
self.const_int(generator, ctx, 0)
}
#[must_use]
pub fn const_true<'ctx, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
) -> Instance<'ctx, Self> {
self.const_int(generator, ctx, 1)
}
}
impl<'ctx, N: IntKind<'ctx>> Instance<'ctx, Int<N>> {
pub fn s_extend_or_bit_cast<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
to_int_kind: NewN,
) -> Instance<'ctx, Int<NewN>> {
Int(to_int_kind).s_extend_or_bit_cast(generator, ctx, self.value)
}
pub fn s_extend<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
to_int_kind: NewN,
) -> Instance<'ctx, Int<NewN>> {
Int(to_int_kind).s_extend(generator, ctx, self.value)
}
pub fn z_extend_or_bit_cast<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
to_int_kind: NewN,
) -> Instance<'ctx, Int<NewN>> {
Int(to_int_kind).z_extend_or_bit_cast(generator, ctx, self.value)
}
pub fn z_extend<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
to_int_kind: NewN,
) -> Instance<'ctx, Int<NewN>> {
Int(to_int_kind).z_extend(generator, ctx, self.value)
}
pub fn truncate_or_bit_cast<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
to_int_kind: NewN,
) -> Instance<'ctx, Int<NewN>> {
Int(to_int_kind).truncate_or_bit_cast(generator, ctx, self.value)
}
pub fn truncate<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
to_int_kind: NewN,
) -> Instance<'ctx, Int<NewN>> {
Int(to_int_kind).truncate(generator, ctx, self.value)
}
#[must_use]
pub fn add(&self, ctx: &CodeGenContext<'ctx, '_>, other: Self) -> Self {
let value = ctx.builder.build_int_add(self.value, other.value, "").unwrap();
self.model.believe_value(value)
}
#[must_use]
pub fn sub(&self, ctx: &CodeGenContext<'ctx, '_>, other: Self) -> Self {
let value = ctx.builder.build_int_sub(self.value, other.value, "").unwrap();
self.model.believe_value(value)
}
#[must_use]
pub fn mul(&self, ctx: &CodeGenContext<'ctx, '_>, other: Self) -> Self {
let value = ctx.builder.build_int_mul(self.value, other.value, "").unwrap();
self.model.believe_value(value)
}
pub fn compare(
&self,
ctx: &CodeGenContext<'ctx, '_>,
op: IntPredicate,
other: Self,
) -> Instance<'ctx, Int<Bool>> {
let value = ctx.builder.build_int_compare(op, self.value, other.value, "").unwrap();
Int(Bool).believe_value(value)
}
#[must_use]
pub fn not(&self, ctx: &CodeGenContext<'ctx, '_>) -> Self {
let value = ctx.builder.build_not(self.value, "").unwrap();
self.model.believe_value(value)
}
}

View File

@ -0,0 +1,17 @@
mod any;
mod array;
mod core;
mod float;
pub mod function;
mod int;
mod ptr;
mod structure;
pub mod util;
pub use any::*;
pub use array::*;
pub use core::*;
pub use float::*;
pub use int::*;
pub use ptr::*;
pub use structure::*;

View File

@ -0,0 +1,192 @@
use inkwell::{
context::Context,
types::{BasicType, BasicTypeEnum, PointerType},
values::{IntValue, PointerValue},
AddressSpace,
};
use crate::codegen::{llvm_intrinsics::call_memcpy_generic, CodeGenContext, CodeGenerator};
use super::*;
#[derive(Debug, Clone, Copy, Default)]
pub struct Ptr<Item>(pub Item);
impl<'ctx, Item: Model<'ctx>> Model<'ctx> for Ptr<Item> {
type Value = PointerValue<'ctx>;
type Type = PointerType<'ctx>;
fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
self.0.get_type(generator, ctx).ptr_type(AddressSpace::default())
}
fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
ty: T,
) -> Result<(), ModelError> {
let ty = ty.as_basic_type_enum();
let Ok(ty) = PointerType::try_from(ty) else {
return Err(ModelError(format!("Expecting PointerType, but got {ty:?}")));
};
let elem_ty = ty.get_element_type();
let Ok(elem_ty) = BasicTypeEnum::try_from(elem_ty) else {
return Err(ModelError(format!(
"Expecting pointer element type to be a BasicTypeEnum, but got {elem_ty:?}"
)));
};
// TODO: inkwell `get_element_type()` will be deprecated.
// Remove the check for `get_element_type()` when the time comes.
self.0
.check_type(generator, ctx, elem_ty)
.map_err(|err| err.under_context("a PointerType"))?;
Ok(())
}
}
impl<'ctx, Element: Model<'ctx>> Ptr<Element> {
/// Return a ***constant*** nullptr.
pub fn nullptr<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
) -> Instance<'ctx, Ptr<Element>> {
let ptr = self.get_type(generator, ctx).const_null();
self.believe_value(ptr)
}
/// Cast a pointer into this model with [`inkwell::builder::Builder::build_pointer_cast`]
pub fn pointer_cast<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
ptr: PointerValue<'ctx>,
) -> Instance<'ctx, Ptr<Element>> {
let t = self.get_type(generator, ctx.ctx);
let ptr = ctx.builder.build_pointer_cast(ptr, t, "").unwrap();
self.believe_value(ptr)
}
}
impl<'ctx, Item: Model<'ctx>> Instance<'ctx, Ptr<Item>> {
/// Offset the pointer by [`inkwell::builder::Builder::build_in_bounds_gep`].
#[must_use]
pub fn offset(
&self,
ctx: &CodeGenContext<'ctx, '_>,
offset: IntValue<'ctx>,
) -> Instance<'ctx, Ptr<Item>> {
let p = unsafe { ctx.builder.build_in_bounds_gep(self.value, &[offset], "").unwrap() };
self.model.believe_value(p)
}
/// Offset the pointer by [`inkwell::builder::Builder::build_in_bounds_gep`] by a constant offset.
#[must_use]
pub fn offset_const(
&self,
ctx: &CodeGenContext<'ctx, '_>,
offset: u64,
) -> Instance<'ctx, Ptr<Item>> {
let offset = ctx.ctx.i32_type().const_int(offset, false);
self.offset(ctx, offset)
}
pub fn set_index(
&self,
ctx: &CodeGenContext<'ctx, '_>,
index: IntValue<'ctx>,
value: Instance<'ctx, Item>,
) {
self.offset(ctx, index).store(ctx, value);
}
pub fn set_index_const(
&self,
ctx: &CodeGenContext<'ctx, '_>,
index: u64,
value: Instance<'ctx, Item>,
) {
self.offset_const(ctx, index).store(ctx, value);
}
pub fn get_index<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
index: IntValue<'ctx>,
) -> Instance<'ctx, Item> {
self.offset(ctx, index).load(generator, ctx)
}
pub fn get_index_const<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
index: u64,
) -> Instance<'ctx, Item> {
self.offset_const(ctx, index).load(generator, ctx)
}
/// Load the value with [`inkwell::builder::Builder::build_load`].
pub fn load<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Item> {
let value = ctx.builder.build_load(self.value, "").unwrap();
self.model.0.check_value(generator, ctx.ctx, value).unwrap() // If unwrap() panics, there is a logic error.
}
/// Store a value with [`inkwell::builder::Builder::build_store`].
pub fn store(&self, ctx: &CodeGenContext<'ctx, '_>, value: Instance<'ctx, Item>) {
ctx.builder.build_store(self.value, value.value).unwrap();
}
/// Return a casted pointer of element type `NewElement` with [`inkwell::builder::Builder::build_pointer_cast`].
pub fn pointer_cast<NewItem: Model<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
new_item: NewItem,
) -> Instance<'ctx, Ptr<NewItem>> {
Ptr(new_item).pointer_cast(generator, ctx, self.value)
}
/// Check if the pointer is null with [`inkwell::builder::Builder::build_is_null`].
pub fn is_null(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Int<Bool>> {
let value = ctx.builder.build_is_null(self.value, "").unwrap();
Int(Bool).believe_value(value)
}
/// Check if the pointer is not null with [`inkwell::builder::Builder::build_is_not_null`].
pub fn is_not_null(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Int<Bool>> {
let value = ctx.builder.build_is_not_null(self.value, "").unwrap();
Int(Bool).believe_value(value)
}
pub fn copy_from<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
source: Self,
num_items: IntValue<'ctx>,
) {
let llvm_usize = generator.get_size_type(ctx.ctx);
// Force extend `num_items` and `itemsize` so their types would match.
let itemsize = self.model.sizeof(generator, ctx.ctx);
let itemsize =
ctx.builder.build_int_z_extend_or_bit_cast(itemsize, llvm_usize, "").unwrap();
let num_items =
ctx.builder.build_int_z_extend_or_bit_cast(num_items, llvm_usize, "").unwrap();
let totalsize = ctx.builder.build_int_mul(itemsize, num_items, "totalsize").unwrap();
let is_volatile = ctx.ctx.bool_type().const_zero(); // is_volatile = false
call_memcpy_generic(ctx, self.value, source.value, totalsize, is_volatile);
}
}

View File

@ -0,0 +1,261 @@
use std::fmt;
use inkwell::{
context::Context,
types::{BasicType, BasicTypeEnum, StructType},
values::{BasicValueEnum, StructValue},
};
use crate::codegen::{CodeGenContext, CodeGenerator};
use super::*;
/// A traveral that traverses a Rust `struct` that is used to declare an LLVM's struct's field types.
pub trait FieldTraversal<'ctx> {
/// Output type of [`FieldTraversal::add`].
type Out<M>;
/// Traverse through the type of a declared field and do something with it.
///
/// * `name` - The cosmetic name of the LLVM field. Used for debugging.
/// * `model` - The [`Model`] representing the LLVM type of this field.
fn add<M: Model<'ctx>>(&mut self, name: &'static str, model: M) -> Self::Out<M>;
/// Like [`FieldTraversal::add`] but [`Model`] is automatically inferred from its [`Default`] trait.
fn add_auto<M: Model<'ctx> + Default>(&mut self, name: &'static str) -> Self::Out<M> {
self.add(name, M::default())
}
}
/// Descriptor of an LLVM struct field.
#[derive(Debug, Clone, Copy)]
pub struct GepField<M> {
/// The GEP index of this field. This is the index to use with `build_gep`.
pub gep_index: u64,
/// The cosmetic name of this field.
pub name: &'static str,
/// The [`Model`] of this field's type.
pub model: M,
}
/// A traversal to get the GEP index of fields.
pub struct GepFieldTraversal {
/// The current GEP index.
gep_index_counter: u64,
}
impl<'ctx> FieldTraversal<'ctx> for GepFieldTraversal {
type Out<M> = GepField<M>;
fn add<M: Model<'ctx>>(&mut self, name: &'static str, model: M) -> Self::Out<M> {
let gep_index = self.gep_index_counter;
self.gep_index_counter += 1;
Self::Out { gep_index, name, model }
}
}
/// A traversal to collect the field types of a struct.
///
/// This is used to collect the field types for [`Context::struct_type`].
struct TypeFieldTraversal<'ctx, 'a, G: CodeGenerator + ?Sized> {
generator: &'a G,
ctx: &'ctx Context,
/// The collected field types so far, in order.
field_types: Vec<BasicTypeEnum<'ctx>>,
}
impl<'ctx, 'a, G: CodeGenerator + ?Sized> FieldTraversal<'ctx> for TypeFieldTraversal<'ctx, 'a, G> {
type Out<M> = (); // Checking types return nothing.
fn add<M: Model<'ctx>>(&mut self, _name: &'static str, model: M) -> Self::Out<M> {
let t = model.get_type(self.generator, self.ctx).as_basic_type_enum();
self.field_types.push(t);
}
}
/// A traversal to check the types of a field for debug assertions.
struct CheckTypeFieldTraversal<'ctx, 'a, G: CodeGenerator + ?Sized> {
generator: &'a mut G,
ctx: &'ctx Context,
/// The current GEP index, so we can tell the index of the field we are checking
/// and report the GEP index.
index: u32,
/// The [`StructType`] to check.
scrutinee: StructType<'ctx>,
/// A list of collected errors so far.
errors: Vec<ModelError>,
}
impl<'ctx, 'a, G: CodeGenerator + ?Sized> FieldTraversal<'ctx>
for CheckTypeFieldTraversal<'ctx, 'a, G>
{
type Out<M> = (); // Checking types return nothing.
fn add<M: Model<'ctx>>(&mut self, name: &'static str, model: M) -> Self::Out<M> {
let i = self.index;
self.index += 1;
if let Some(t) = self.scrutinee.get_field_type_at_index(i) {
if let Err(err) = model.check_type(self.generator, self.ctx, t) {
self.errors.push(err.under_context(format!("field #{i} '{name}'").as_str()));
}
} // Otherwise, it will be caught by Struct's `check_type`.
}
}
pub trait StructKind<'ctx>: fmt::Debug + Clone + Copy {
type Fields<F: FieldTraversal<'ctx>>;
/// Traverse map through all fields of this [`StructKind`].
fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F>;
/// Get a convenience structure to get a struct field's GEP index through its corresponding Rust field.
fn fields(&self) -> Self::Fields<GepFieldTraversal> {
self.traverse_fields(&mut GepFieldTraversal { gep_index_counter: 0 })
}
/// Get the LLVM [`StructType`] of this [`StructKind`].
fn get_struct_type<G: CodeGenerator + ?Sized>(
&self,
generator: &G,
ctx: &'ctx Context,
) -> StructType<'ctx> {
let mut traversal = TypeFieldTraversal { generator, ctx, field_types: Vec::new() };
self.traverse_fields(&mut traversal);
ctx.struct_type(&traversal.field_types, false)
}
}
#[derive(Debug, Clone, Copy, Default)]
pub struct Struct<S>(pub S);
impl<'ctx, S: StructKind<'ctx>> Struct<S> {
/// Create a constant struct value.
///
/// This function also validates `fields` and panic when there is something wrong.
pub fn const_struct<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
fields: &[BasicValueEnum<'ctx>],
) -> Instance<'ctx, Self> {
// NOTE: There *could* have been a functor `F<M> = Instance<'ctx, M>` for `S::Fields<F>`
// to create a more user-friendly interface, but Rust's type system is not sophisticated enough
// and if you try doing that Rust would force you put lifetimes everywhere.
let val = ctx.const_struct(fields, false);
self.check_value(generator, ctx, val).unwrap()
}
}
impl<'ctx, S: StructKind<'ctx>> Model<'ctx> for Struct<S> {
type Value = StructValue<'ctx>;
type Type = StructType<'ctx>;
fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
self.0.get_struct_type(generator, ctx)
}
fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
ty: T,
) -> Result<(), ModelError> {
let ty = ty.as_basic_type_enum();
let Ok(ty) = StructType::try_from(ty) else {
return Err(ModelError(format!("Expecting StructType, but got {ty:?}")));
};
let mut traversal =
CheckTypeFieldTraversal { generator, ctx, index: 0, errors: Vec::new(), scrutinee: ty };
self.0.traverse_fields(&mut traversal);
let exp_num_fields = traversal.index;
let got_num_fields = u32::try_from(ty.get_field_types().len()).unwrap();
if exp_num_fields != got_num_fields {
return Err(ModelError(format!(
"Expecting StructType with {exp_num_fields} field(s), but got {got_num_fields}"
)));
}
if !traversal.errors.is_empty() {
return Err(traversal.errors[0].clone()); // TODO: Return other errors as well
}
Ok(())
}
}
impl<'ctx, S: StructKind<'ctx>> Instance<'ctx, Struct<S>> {
/// Get a field with [`StructValue::get_field_at_index`].
pub fn get_field<G: CodeGenerator + ?Sized, M, GetField>(
&self,
generator: &mut G,
ctx: &'ctx Context,
get_field: GetField,
) -> Instance<'ctx, M>
where
M: Model<'ctx>,
GetField: FnOnce(S::Fields<GepFieldTraversal>) -> GepField<M>,
{
let field = get_field(self.model.0.fields());
let val = self.value.get_field_at_index(field.gep_index as u32).unwrap();
field.model.check_value(generator, ctx, val).unwrap()
}
}
impl<'ctx, S: StructKind<'ctx>> Instance<'ctx, Ptr<Struct<S>>> {
/// Get a pointer to a field with [`Builder::build_in_bounds_gep`].
pub fn gep<M, GetField>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
get_field: GetField,
) -> Instance<'ctx, Ptr<M>>
where
M: Model<'ctx>,
GetField: FnOnce(S::Fields<GepFieldTraversal>) -> GepField<M>,
{
let field = get_field(self.model.0 .0.fields());
let llvm_i32 = ctx.ctx.i32_type(); // i64 would segfault
let ptr = unsafe {
ctx.builder
.build_in_bounds_gep(
self.value,
&[llvm_i32.const_zero(), llvm_i32.const_int(field.gep_index, false)],
field.name,
)
.unwrap()
};
Ptr(field.model).believe_value(ptr)
}
/// Convenience function equivalent to `.gep(...).load(...)`.
pub fn get<M, GetField, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
get_field: GetField,
) -> Instance<'ctx, M>
where
M: Model<'ctx>,
GetField: FnOnce(S::Fields<GepFieldTraversal>) -> GepField<M>,
{
self.gep(ctx, get_field).load(generator, ctx)
}
/// Convenience function equivalent to `.gep(...).store(...)`.
pub fn set<M, GetField>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
get_field: GetField,
value: Instance<'ctx, M>,
) where
M: Model<'ctx>,
GetField: FnOnce(S::Fields<GepFieldTraversal>) -> GepField<M>,
{
self.gep(ctx, get_field).store(ctx, value);
}
}

View File

@ -0,0 +1,40 @@
use crate::codegen::{
stmt::{gen_for_callback_incrementing, BreakContinueHooks},
CodeGenContext, CodeGenerator,
};
use super::*;
/// Like [`gen_for_callback_incrementing`] with [`Model`] abstractions.
pub fn gen_for_model<'ctx, 'a, G, F, N>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
start: Instance<'ctx, Int<N>>,
stop: Instance<'ctx, Int<N>>,
step: Instance<'ctx, Int<N>>,
body: F,
) -> Result<(), String>
where
G: CodeGenerator + ?Sized,
F: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
BreakContinueHooks<'ctx>,
Instance<'ctx, Int<N>>,
) -> Result<(), String>,
N: IntKind<'ctx> + Default,
{
let int_model = Int(N::default());
gen_for_callback_incrementing(
generator,
ctx,
None,
start.value,
(stop.value, false),
|g, ctx, hooks, i| {
let i = int_model.believe_value(i);
body(g, ctx, hooks, i)
},
step.value,
)
}

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,12 @@
use inkwell::values::BasicValueEnum;
use crate::typecheck::typedef::Type;
/// An NAC3 LLVM Python object.
#[derive(Debug, Clone, Copy)]
pub struct AnyObject<'ctx> {
/// Typechecker type of the object.
pub ty: Type,
/// LLVM value of the object.
pub value: BasicValueEnum<'ctx>,
}

View File

@ -0,0 +1,150 @@
use inkwell::types::BasicType;
use crate::{
codegen::{model::*, CodeGenContext, CodeGenerator},
typecheck::typedef::{iter_type_vars, Type, TypeEnum},
};
use super::any::AnyObject;
/// Fields of [`List`]
pub struct ListFields<'ctx, F: FieldTraversal<'ctx>, Item: Model<'ctx>> {
/// Array pointer to content
pub items: F::Out<Ptr<Item>>,
/// Number of items in the array
pub len: F::Out<Int<SizeT>>,
}
/// A list in NAC3.
#[derive(Debug, Clone, Copy, Default)]
pub struct List<Item> {
/// Model of the list items
pub item: Item,
}
impl<'ctx, Item: Model<'ctx>> StructKind<'ctx> for List<Item> {
type Fields<F: FieldTraversal<'ctx>> = ListFields<'ctx, F, Item>;
fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
Self::Fields {
items: traversal.add("items", Ptr(self.item)),
len: traversal.add_auto("len"),
}
}
}
/// A NAC3 Python List object.
#[derive(Debug, Clone, Copy)]
pub struct ListObject<'ctx> {
/// Typechecker type of the list items. Could be [`TypeEnum::TVar`] if unresolved (like
/// in the case of empty lists and the typechecker does not have enough hints).
pub item_type: Type,
pub instance: Instance<'ctx, Ptr<Struct<List<Any<'ctx>>>>>,
}
impl<'ctx> ListObject<'ctx> {
/// Create a [`ListObject`] from an LLVM value and its typechecker [`Type`].
pub fn from_object<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
object: AnyObject<'ctx>,
) -> Self {
assert!(matches!(
&*ctx.unifier.get_ty(object.ty),
TypeEnum::TObj { obj_id, .. } if *obj_id == ctx.primitives.list.obj_id(&ctx.unifier).unwrap()
));
// Check typechecker type and extract `item_type`
let item_type = match &*ctx.unifier.get_ty(object.ty) {
TypeEnum::TObj { obj_id, params, .. }
if *obj_id == ctx.primitives.list.obj_id(&ctx.unifier).unwrap() =>
{
iter_type_vars(params).next().unwrap().ty // Extract `item_type`
}
_ => {
panic!("Expecting type to be a list, but got {}", ctx.unifier.stringify(object.ty))
}
};
// If `item_type` is unresolved, the list's ptr will default to `size_t*`
// as a placeholder because there are no opaque pointers in LLVM 14.
let item_type_llvm = if let TypeEnum::TVar { .. } = &*ctx.unifier.get_ty(item_type) {
generator.get_size_type(ctx.ctx).as_basic_type_enum()
} else {
ctx.get_llvm_type(generator, item_type)
};
let plist = Ptr(Struct(List { item: Any(item_type_llvm) }));
// Create object
let value = plist.check_value(generator, ctx.ctx, object.value).unwrap();
ListObject { item_type, instance: value }
}
/// Get the `len()` of this list.
pub fn len<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<SizeT>> {
self.instance.get(generator, ctx, |f| f.len)
}
/// Get the `items` field as an opaque pointer.
pub fn get_opaque_items_ptr<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Ptr<Int<Byte>>> {
self.instance.get(generator, ctx, |f| f.items).pointer_cast(generator, ctx, Int(Byte))
}
/// Get the value of this [`ListObject`] as a list with opaque items.
///
/// This function allocates on the stack to create the list, but the
/// reference to the `items` is preserved.
pub fn get_opaque_list_ptr<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>> {
let opaque_list = Struct(List { item: Int(Byte) }).alloca(generator, ctx);
// Copy items pointer
let items = self.get_opaque_items_ptr(generator, ctx);
opaque_list.set(ctx, |f| f.items, items);
// Copy len
let len = self.instance.get(generator, ctx, |f| f.len);
opaque_list.set(ctx, |f| f.len, len);
opaque_list
}
/// Allocate a list on the stack given its `item_type` and `len`.
///
/// The returned list's content will be:
/// - `items`: allocated with an array of length `len` with uninitialized values.
/// - `len`: set to `len`.
pub fn alloca<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
item_type: Type,
len: Instance<'ctx, Int<SizeT>>,
) -> Self {
// If `item_type` is unresolved, the list's ptr will default to `size_t*`
// as a placeholder because there are no opaque pointers in LLVM 14.
let item_type_llvm = if let TypeEnum::TVar { .. } = &*ctx.unifier.get_ty(item_type) {
generator.get_size_type(ctx.ctx).as_basic_type_enum()
} else {
ctx.get_llvm_type(generator, item_type)
};
let items = Any(item_type_llvm).array_alloca(generator, ctx, len.value);
let instance = Struct(List { item: Any(item_type_llvm) }).alloca(generator, ctx);
instance.set(ctx, |f| f.items, items);
instance.set(ctx, |f| f.len, len);
ListObject { item_type, instance }
}
}

View File

@ -0,0 +1,5 @@
pub mod any;
pub mod list;
pub mod ndarray;
pub mod range;
pub mod tuple;

View File

@ -0,0 +1,178 @@
use super::NDArrayObject;
use crate::{
codegen::{
irrt::{
call_nac3_ndarray_array_set_and_validate_list_shape,
call_nac3_ndarray_array_write_list_to_array,
},
model::*,
object::{any::AnyObject, list::ListObject},
stmt::gen_if_else_expr_callback,
CodeGenContext, CodeGenerator,
},
toplevel::helper::{arraylike_flatten_element_type, arraylike_get_ndims},
typecheck::typedef::{Type, TypeEnum},
};
fn get_list_object_dtype_and_ndims<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
list: ListObject<'ctx>,
) -> (Type, u64) {
let dtype = arraylike_flatten_element_type(&mut ctx.unifier, list.item_type);
let ndims = arraylike_get_ndims(&mut ctx.unifier, list.item_type);
let ndims = ndims + 1; // To count `list` itself.
(dtype, ndims)
}
impl<'ctx> NDArrayObject<'ctx> {
fn make_np_array_list_copy_impl<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
list: ListObject<'ctx>,
) -> Self {
let (dtype, ndims_int) = get_list_object_dtype_and_ndims(ctx, list);
let list_value = list.get_opaque_list_ptr(generator, ctx);
// Validate `list` has a consistent shape.
// Raise an exception if `list` is something abnormal like `[[1, 2], [3]]`.
// If `list` has a consistent shape, deduce the shape and write it to `shape`.
let ndims = Int(SizeT).const_int(generator, ctx.ctx, ndims_int);
let shape = Int(SizeT).array_alloca(generator, ctx, ndims.value);
call_nac3_ndarray_array_set_and_validate_list_shape(
generator, ctx, list_value, ndims, shape,
);
let ndarray = NDArrayObject::alloca(generator, ctx, dtype, ndims_int);
ndarray.copy_shape_from_array(generator, ctx, shape);
ndarray.create_data(generator, ctx);
// Copy all contents from the list.
call_nac3_ndarray_array_write_list_to_array(generator, ctx, list_value, ndarray.instance);
ndarray
}
fn make_np_array_list_try_no_copy_impl<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
list: ListObject<'ctx>,
) -> Self {
// np_array without copying is only possible `list` is not nested.
//
// If `list` is `list[T]`, we can create an ndarray with `data` set
// to the array pointer of `list`.
//
// If `list` is `list[list[T]]` or worse, copy.
let (dtype, ndims) = get_list_object_dtype_and_ndims(ctx, list);
if ndims == 1 {
// `list` is not nested
let ndarray = NDArrayObject::alloca(generator, ctx, dtype, 1);
// Set data
let data = list.get_opaque_items_ptr(generator, ctx);
ndarray.instance.set(ctx, |f| f.data, data);
// ndarray->shape[0] = list->len;
let shape = ndarray.instance.get(generator, ctx, |f| f.shape);
let list_len = list.instance.get(generator, ctx, |f| f.len);
shape.set_index_const(ctx, 0, list_len);
// Set strides, the `data` is contiguous
ndarray.set_strides_contiguous(generator, ctx);
ndarray
} else {
// `list` is nested, copy
NDArrayObject::make_np_array_list_copy_impl(generator, ctx, list)
}
}
fn make_np_array_list_impl<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
list: ListObject<'ctx>,
copy: Instance<'ctx, Int<Bool>>,
) -> Self {
let (dtype, ndims) = get_list_object_dtype_and_ndims(ctx, list);
let ndarray = gen_if_else_expr_callback(
generator,
ctx,
|_generator, _ctx| Ok(copy.value),
|generator, ctx| {
let ndarray = NDArrayObject::make_np_array_list_copy_impl(generator, ctx, list);
Ok(Some(ndarray.instance.value))
},
|generator, ctx| {
let ndarray =
NDArrayObject::make_np_array_list_try_no_copy_impl(generator, ctx, list);
Ok(Some(ndarray.instance.value))
},
)
.unwrap()
.unwrap();
NDArrayObject::from_value_and_unpacked_types(generator, ctx, ndarray, dtype, ndims)
}
pub fn make_np_array_ndarray_impl<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: NDArrayObject<'ctx>,
copy: Instance<'ctx, Int<Bool>>,
) -> Self {
let ndarray_val = gen_if_else_expr_callback(
generator,
ctx,
|_generator, _ctx| Ok(copy.value),
|generator, ctx| {
let ndarray = ndarray.make_copy(generator, ctx); // Force copy
Ok(Some(ndarray.instance.value))
},
|_generator, _ctx| {
// No need to copy. Return `ndarray` itself.
Ok(Some(ndarray.instance.value))
},
)
.unwrap()
.unwrap();
NDArrayObject::from_value_and_unpacked_types(
generator,
ctx,
ndarray_val,
ndarray.dtype,
ndarray.ndims,
)
}
/// Create a new ndarray like `np.array()`.
///
/// NOTE: The `ndmin` argument is not here. You may want to
/// do [`NDArrayObject::atleast_nd`] to achieve that.
pub fn make_np_array<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
object: AnyObject<'ctx>,
copy: Instance<'ctx, Int<Bool>>,
) -> Self {
match &*ctx.unifier.get_ty(object.ty) {
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.list.obj_id(&ctx.unifier).unwrap() =>
{
let list = ListObject::from_object(generator, ctx, object);
NDArrayObject::make_np_array_list_impl(generator, ctx, list, copy)
}
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.ndarray.obj_id(&ctx.unifier).unwrap() =>
{
let ndarray = NDArrayObject::from_object(generator, ctx, object);
NDArrayObject::make_np_array_ndarray_impl(generator, ctx, ndarray, copy)
}
_ => panic!("Unrecognized object type: {}", ctx.unifier.stringify(object.ty)), // Typechecker ensures this
}
}
}

View File

@ -0,0 +1,135 @@
use itertools::Itertools;
use crate::codegen::{
irrt::{call_nac3_ndarray_broadcast_shapes, call_nac3_ndarray_broadcast_to},
model::*,
CodeGenContext, CodeGenerator,
};
use super::NDArrayObject;
/// Fields of [`ShapeEntry`]
pub struct ShapeEntryFields<'ctx, F: FieldTraversal<'ctx>> {
pub ndims: F::Out<Int<SizeT>>,
pub shape: F::Out<Ptr<Int<SizeT>>>,
}
/// An IRRT structure used in broadcasting.
#[derive(Debug, Clone, Copy, Default)]
pub struct ShapeEntry;
impl<'ctx> StructKind<'ctx> for ShapeEntry {
type Fields<F: FieldTraversal<'ctx>> = ShapeEntryFields<'ctx, F>;
fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
Self::Fields { ndims: traversal.add_auto("ndims"), shape: traversal.add_auto("shape") }
}
}
impl<'ctx> NDArrayObject<'ctx> {
/// Create a broadcast view on this ndarray with a target shape.
///
/// The input shape will be checked to make sure that it contains no negative values.
///
/// * `target_ndims` - The ndims type after broadcasting to the given shape.
/// The caller has to figure this out for this function.
/// * `target_shape` - An array pointer pointing to the target shape.
#[must_use]
pub fn broadcast_to<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
target_ndims: u64,
target_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) -> Self {
let broadcast_ndarray = NDArrayObject::alloca(generator, ctx, self.dtype, target_ndims);
broadcast_ndarray.copy_shape_from_array(generator, ctx, target_shape);
call_nac3_ndarray_broadcast_to(generator, ctx, self.instance, broadcast_ndarray.instance);
broadcast_ndarray
}
}
/// A result produced by [`broadcast_all_ndarrays`]
#[derive(Debug, Clone)]
pub struct BroadcastAllResult<'ctx> {
/// The statically known `ndims` of the broadcast result.
pub ndims: u64,
/// The broadcasting shape.
pub shape: Instance<'ctx, Ptr<Int<SizeT>>>,
/// Broadcasted views on the inputs.
///
/// All of them will have `shape` [`BroadcastAllResult::shape`] and
/// `ndims` [`BroadcastAllResult::ndims`]. The length of the vector
/// is the same as the input.
pub ndarrays: Vec<NDArrayObject<'ctx>>,
}
/// Helper function to call `call_nac3_ndarray_broadcast_shapes`
fn broadcast_shapes<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
in_shape_entries: &[(Instance<'ctx, Ptr<Int<SizeT>>>, u64)], // (shape, shape's length/ndims)
broadcast_ndims: u64,
broadcast_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
// Prepare input shape entries to be passed to `call_nac3_ndarray_broadcast_shapes`.
let num_shape_entries =
Int(SizeT).const_int(generator, ctx.ctx, u64::try_from(in_shape_entries.len()).unwrap());
let shape_entries = Struct(ShapeEntry).array_alloca(generator, ctx, num_shape_entries.value);
for (i, (in_shape, in_ndims)) in in_shape_entries.iter().enumerate() {
let pshape_entry = shape_entries.offset_const(ctx, i as u64);
let in_ndims = Int(SizeT).const_int(generator, ctx.ctx, *in_ndims);
pshape_entry.set(ctx, |f| f.ndims, in_ndims);
pshape_entry.set(ctx, |f| f.shape, *in_shape);
}
let broadcast_ndims = Int(SizeT).const_int(generator, ctx.ctx, broadcast_ndims);
call_nac3_ndarray_broadcast_shapes(
generator,
ctx,
num_shape_entries,
shape_entries,
broadcast_ndims,
broadcast_shape,
);
}
impl<'ctx> NDArrayObject<'ctx> {
/// Broadcast all ndarrays according to `np.broadcast()` and return a [`BroadcastAllResult`]
/// containing all the information of the result of the broadcast operation.
pub fn broadcast<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarrays: &[Self],
) -> BroadcastAllResult<'ctx> {
assert!(!ndarrays.is_empty());
// Infer the broadcast output ndims.
let broadcast_ndims_int = ndarrays.iter().map(|ndarray| ndarray.ndims).max().unwrap();
let broadcast_ndims = Int(SizeT).const_int(generator, ctx.ctx, broadcast_ndims_int);
let broadcast_shape = Int(SizeT).array_alloca(generator, ctx, broadcast_ndims.value);
let shape_entries = ndarrays
.iter()
.map(|ndarray| (ndarray.instance.get(generator, ctx, |f| f.shape), ndarray.ndims))
.collect_vec();
broadcast_shapes(generator, ctx, &shape_entries, broadcast_ndims_int, broadcast_shape);
// Broadcast all the inputs to shape `dst_shape`.
let broadcast_ndarrays: Vec<_> = ndarrays
.iter()
.map(|ndarray| {
ndarray.broadcast_to(generator, ctx, broadcast_ndims_int, broadcast_shape)
})
.collect_vec();
BroadcastAllResult {
ndims: broadcast_ndims_int,
shape: broadcast_shape,
ndarrays: broadcast_ndarrays,
}
}
}

View File

@ -0,0 +1,134 @@
use crate::{
codegen::{model::*, CodeGenContext, CodeGenerator},
typecheck::typedef::Type,
};
use super::NDArrayObject;
/// Fields of [`ContiguousNDArray`]
pub struct ContiguousNDArrayFields<'ctx, F: FieldTraversal<'ctx>, Item: Model<'ctx>> {
pub ndims: F::Out<Int<SizeT>>,
pub shape: F::Out<Ptr<Int<SizeT>>>,
pub data: F::Out<Ptr<Item>>,
}
/// An ndarray without strides and non-opaque `data` field in NAC3.
#[derive(Debug, Clone, Copy)]
pub struct ContiguousNDArray<M> {
/// [`Model`] of the items.
pub item: M,
}
impl<'ctx, Item: Model<'ctx>> StructKind<'ctx> for ContiguousNDArray<Item> {
type Fields<F: FieldTraversal<'ctx>> = ContiguousNDArrayFields<'ctx, F, Item>;
fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
Self::Fields {
ndims: traversal.add_auto("ndims"),
shape: traversal.add_auto("shape"),
data: traversal.add("data", Ptr(self.item)),
}
}
}
impl<'ctx> NDArrayObject<'ctx> {
/// Create a [`ContiguousNDArray`] from the contents of this ndarray.
///
/// This function may or may not be expensive depending on if this ndarray has contiguous data.
///
/// If this ndarray is not C-contiguous, this function will allocate memory on the stack for the `data` field of
/// the returned [`ContiguousNDArray`] and copy contents of this ndarray to there.
///
/// If this ndarray is C-contiguous, contents of this ndarray will not be copied. The created [`ContiguousNDArray`]
/// will share memory with this ndarray.
///
/// The `item_model` sets the [`Model`] of the returned [`ContiguousNDArray`]'s `Item` model for type-safety, and
/// should match the `ctx.get_llvm_type()` of this ndarray's `dtype`. Otherwise this function panics. Use model [`Any`]
/// if you don't care/cannot know the [`Model`] in advance.
pub fn make_contiguous_ndarray<G: CodeGenerator + ?Sized, Item: Model<'ctx>>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
item_model: Item,
) -> Instance<'ctx, Ptr<Struct<ContiguousNDArray<Item>>>> {
// Sanity check on `self.dtype` and `item_model`.
let dtype_llvm = ctx.get_llvm_type(generator, self.dtype);
item_model.check_type(generator, ctx.ctx, dtype_llvm).unwrap();
let cdarray_model = Struct(ContiguousNDArray { item: item_model });
let current_bb = ctx.builder.get_insert_block().unwrap();
let then_bb = ctx.ctx.insert_basic_block_after(current_bb, "then_bb");
let else_bb = ctx.ctx.insert_basic_block_after(then_bb, "else_bb");
let end_bb = ctx.ctx.insert_basic_block_after(else_bb, "end_bb");
// Allocate and setup the resulting [`ContiguousNDArray`].
let result = cdarray_model.alloca(generator, ctx);
// Set ndims and shape.
let ndims = self.ndims_llvm(generator, ctx.ctx);
result.set(ctx, |f| f.ndims, ndims);
let shape = self.instance.get(generator, ctx, |f| f.shape);
result.set(ctx, |f| f.shape, shape);
let is_contiguous = self.is_c_contiguous(generator, ctx);
ctx.builder.build_conditional_branch(is_contiguous.value, then_bb, else_bb).unwrap();
// Inserting into then_bb; This ndarray is contiguous.
ctx.builder.position_at_end(then_bb);
let data = self.instance.get(generator, ctx, |f| f.data);
let data = data.pointer_cast(generator, ctx, item_model);
result.set(ctx, |f| f.data, data);
ctx.builder.build_unconditional_branch(end_bb).unwrap();
// Inserting into else_bb; This ndarray is not contiguous. Do a full-copy on `data`.
// `make_copy` produces an ndarray with contiguous `data`.
ctx.builder.position_at_end(else_bb);
let copied_ndarray = self.make_copy(generator, ctx);
let data = copied_ndarray.instance.get(generator, ctx, |f| f.data);
let data = data.pointer_cast(generator, ctx, item_model);
result.set(ctx, |f| f.data, data);
ctx.builder.build_unconditional_branch(end_bb).unwrap();
// Reposition to end_bb for continuation
ctx.builder.position_at_end(end_bb);
result
}
/// Create an [`NDArrayObject`] from a [`ContiguousNDArray`].
///
/// The operation is super cheap. The newly created [`NDArrayObject`] will share the
/// same memory as the [`ContiguousNDArray`].
///
/// `ndims` has to be provided as [`NDArrayObject`] requires a statically known `ndims` value, despite
/// the fact that the information should be contained within the [`ContiguousNDArray`].
pub fn from_contiguous_ndarray<G: CodeGenerator + ?Sized, Item: Model<'ctx>>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
carray: Instance<'ctx, Ptr<Struct<ContiguousNDArray<Item>>>>,
dtype: Type,
ndims: u64,
) -> Self {
// Sanity check on `dtype` and `contiguous_array`'s `Item` model.
let dtype_llvm = ctx.get_llvm_type(generator, dtype);
carray.model.0 .0.item.check_type(generator, ctx.ctx, dtype_llvm).unwrap();
// TODO: Debug assert `ndims == carray.ndims` to catch bugs.
// Allocate the resulting ndarray.
let ndarray = NDArrayObject::alloca(generator, ctx, dtype, ndims);
// Copy shape and update strides
let shape = carray.get(generator, ctx, |f| f.shape);
ndarray.copy_shape_from_array(generator, ctx, shape);
ndarray.set_strides_contiguous(generator, ctx);
// Share data
let data = carray.get(generator, ctx, |f| f.data).pointer_cast(generator, ctx, Int(Byte));
ndarray.instance.set(ctx, |f| f.data, data);
ndarray
}
}

View File

@ -0,0 +1,176 @@
use inkwell::{values::BasicValueEnum, IntPredicate};
use crate::{
codegen::{
irrt::call_nac3_ndarray_util_assert_shape_no_negative, model::*, CodeGenContext,
CodeGenerator,
},
typecheck::typedef::Type,
};
use super::NDArrayObject;
/// Get the zero value in `np.zeros()` of a `dtype`.
fn ndarray_zero_value<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
) -> BasicValueEnum<'ctx> {
if [ctx.primitives.int32, ctx.primitives.uint32]
.iter()
.any(|ty| ctx.unifier.unioned(dtype, *ty))
{
ctx.ctx.i32_type().const_zero().into()
} else if [ctx.primitives.int64, ctx.primitives.uint64]
.iter()
.any(|ty| ctx.unifier.unioned(dtype, *ty))
{
ctx.ctx.i64_type().const_zero().into()
} else if ctx.unifier.unioned(dtype, ctx.primitives.float) {
ctx.ctx.f64_type().const_zero().into()
} else if ctx.unifier.unioned(dtype, ctx.primitives.bool) {
ctx.ctx.bool_type().const_zero().into()
} else if ctx.unifier.unioned(dtype, ctx.primitives.str) {
ctx.gen_string(generator, "").into()
} else {
panic!("unrecognized dtype: {}", ctx.unifier.stringify(dtype));
}
}
/// Get the one value in `np.ones()` of a `dtype`.
fn ndarray_one_value<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
) -> BasicValueEnum<'ctx> {
if [ctx.primitives.int32, ctx.primitives.uint32]
.iter()
.any(|ty| ctx.unifier.unioned(dtype, *ty))
{
let is_signed = ctx.unifier.unioned(dtype, ctx.primitives.int32);
ctx.ctx.i32_type().const_int(1, is_signed).into()
} else if [ctx.primitives.int64, ctx.primitives.uint64]
.iter()
.any(|ty| ctx.unifier.unioned(dtype, *ty))
{
let is_signed = ctx.unifier.unioned(dtype, ctx.primitives.int64);
ctx.ctx.i64_type().const_int(1, is_signed).into()
} else if ctx.unifier.unioned(dtype, ctx.primitives.float) {
ctx.ctx.f64_type().const_float(1.0).into()
} else if ctx.unifier.unioned(dtype, ctx.primitives.bool) {
ctx.ctx.bool_type().const_int(1, false).into()
} else if ctx.unifier.unioned(dtype, ctx.primitives.str) {
ctx.gen_string(generator, "1").into()
} else {
panic!("unrecognized dtype: {}", ctx.unifier.stringify(dtype));
}
}
impl<'ctx> NDArrayObject<'ctx> {
/// Create an ndarray like `np.empty`.
pub fn make_np_empty<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
ndims: u64,
shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) -> Self {
// Validate `shape`
let ndims_llvm = Int(SizeT).const_int(generator, ctx.ctx, ndims);
call_nac3_ndarray_util_assert_shape_no_negative(generator, ctx, ndims_llvm, shape);
let ndarray = NDArrayObject::alloca(generator, ctx, dtype, ndims);
ndarray.copy_shape_from_array(generator, ctx, shape);
ndarray.create_data(generator, ctx);
ndarray
}
/// Create an ndarray like `np.full`.
pub fn make_np_full<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
ndims: u64,
shape: Instance<'ctx, Ptr<Int<SizeT>>>,
fill_value: BasicValueEnum<'ctx>,
) -> Self {
let ndarray = NDArrayObject::make_np_empty(generator, ctx, dtype, ndims, shape);
ndarray.fill(generator, ctx, fill_value);
ndarray
}
/// Create an ndarray like `np.zero`.
pub fn make_np_zeros<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
ndims: u64,
shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) -> Self {
let fill_value = ndarray_zero_value(generator, ctx, dtype);
NDArrayObject::make_np_full(generator, ctx, dtype, ndims, shape, fill_value)
}
/// Create an ndarray like `np.ones`.
pub fn make_np_ones<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
ndims: u64,
shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) -> Self {
let fill_value = ndarray_one_value(generator, ctx, dtype);
NDArrayObject::make_np_full(generator, ctx, dtype, ndims, shape, fill_value)
}
/// Create an ndarray like `np.eye`.
pub fn make_np_eye<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
nrows: Instance<'ctx, Int<SizeT>>,
ncols: Instance<'ctx, Int<SizeT>>,
offset: Instance<'ctx, Int<SizeT>>,
) -> Self {
let ndzero = ndarray_zero_value(generator, ctx, dtype);
let ndone = ndarray_one_value(generator, ctx, dtype);
let ndarray = NDArrayObject::alloca_dynamic_shape(generator, ctx, dtype, &[nrows, ncols]);
// Create data and make the matrix like look np.eye()
ndarray.create_data(generator, ctx);
ndarray
.foreach(generator, ctx, |generator, ctx, _hooks, nditer| {
// NOTE: rows and cols can never be zero here, since this ndarray's `np.size` would be zero
// and this loop would not execute.
// Load up `row_i` and `col_i` from indices.
let row_i = nditer.get_indices().get_index_const(generator, ctx, 0);
let col_i = nditer.get_indices().get_index_const(generator, ctx, 1);
let be_one = row_i.add(ctx, offset).compare(ctx, IntPredicate::EQ, col_i);
let value = ctx.builder.build_select(be_one.value, ndone, ndzero, "value").unwrap();
let p = nditer.get_pointer(generator, ctx);
ctx.builder.build_store(p, value).unwrap();
Ok(())
})
.unwrap();
ndarray
}
/// Create an ndarray like `np.identity`.
pub fn make_np_identity<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
size: Instance<'ctx, Int<SizeT>>,
) -> Self {
// Convenient implementation
let offset = Int(SizeT).const_0(generator, ctx.ctx);
NDArrayObject::make_np_eye(generator, ctx, dtype, size, size, offset)
}
}

View File

@ -0,0 +1,295 @@
use crate::codegen::{irrt::call_nac3_ndarray_index, model::*, CodeGenContext, CodeGenerator};
use super::NDArrayObject;
pub type NDIndexType = Byte;
/// Fields of [`NDIndex`]
#[derive(Debug, Clone, Copy)]
pub struct NDIndexFields<'ctx, F: FieldTraversal<'ctx>> {
pub type_: F::Out<Int<NDIndexType>>, // Defined to be uint8_t in IRRT
pub data: F::Out<Ptr<Int<Byte>>>,
}
/// An IRRT representation of an ndarray subscript index.
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
pub struct NDIndex;
impl<'ctx> StructKind<'ctx> for NDIndex {
type Fields<F: FieldTraversal<'ctx>> = NDIndexFields<'ctx, F>;
fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
Self::Fields { type_: traversal.add_auto("type"), data: traversal.add_auto("data") }
}
}
/// Fields of [`Slice`]
#[derive(Debug, Clone)]
pub struct SliceFields<'ctx, F: FieldTraversal<'ctx>, N: IntKind<'ctx>> {
pub start_defined: F::Out<Int<Bool>>,
pub start: F::Out<Int<N>>,
pub stop_defined: F::Out<Int<Bool>>,
pub stop: F::Out<Int<N>>,
pub step_defined: F::Out<Int<Bool>>,
pub step: F::Out<Int<N>>,
}
/// An IRRT representation of an (unresolved) slice.
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
pub struct Slice<N> {
int_kind: N,
}
impl<'ctx, N: IntKind<'ctx>> StructKind<'ctx> for Slice<N> {
type Fields<F: FieldTraversal<'ctx>> = SliceFields<'ctx, F, N>;
fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
Self::Fields {
start_defined: traversal.add_auto("start_defined"),
start: traversal.add("start", Int(self.int_kind)),
stop_defined: traversal.add_auto("stop_defined"),
stop: traversal.add("stop", Int(self.int_kind)),
step_defined: traversal.add_auto("step_defined"),
step: traversal.add("step", Int(self.int_kind)),
}
}
}
/// A convenience structure to prepare a [`Slice`].
#[derive(Debug, Clone)]
pub struct RustSlice<'ctx, N: IntKind<'ctx>> {
pub start: Option<Instance<'ctx, Int<N>>>,
pub stop: Option<Instance<'ctx, Int<N>>>,
pub step: Option<Instance<'ctx, Int<N>>>,
}
impl<'ctx, N: IntKind<'ctx>> RustSlice<'ctx, N> {
/// Write the contents to an LLVM [`Slice`].
pub fn write_to_slice<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
dst_slice_ptr: Instance<'ctx, Ptr<Struct<Slice<N>>>>,
) {
let false_ = Int(Bool).const_false(generator, ctx.ctx);
let true_ = Int(Bool).const_true(generator, ctx.ctx);
match self.start {
Some(start) => {
dst_slice_ptr.gep(ctx, |f| f.start_defined).store(ctx, true_);
dst_slice_ptr.gep(ctx, |f| f.start).store(ctx, start);
}
None => dst_slice_ptr.gep(ctx, |f| f.start_defined).store(ctx, false_),
}
match self.stop {
Some(stop) => {
dst_slice_ptr.gep(ctx, |f| f.stop_defined).store(ctx, true_);
dst_slice_ptr.gep(ctx, |f| f.stop).store(ctx, stop);
}
None => dst_slice_ptr.gep(ctx, |f| f.stop_defined).store(ctx, false_),
}
match self.step {
Some(step) => {
dst_slice_ptr.gep(ctx, |f| f.step_defined).store(ctx, true_);
dst_slice_ptr.gep(ctx, |f| f.step).store(ctx, step);
}
None => dst_slice_ptr.gep(ctx, |f| f.step_defined).store(ctx, false_),
}
}
}
// A convenience enum to prepare an [`NDIndex`].
#[derive(Debug, Clone)]
pub enum RustNDIndex<'ctx> {
SingleElement(Instance<'ctx, Int<Int32>>), // TODO: To be SizeT
Slice(RustSlice<'ctx, Int32>),
NewAxis,
Ellipsis,
}
impl<'ctx> RustNDIndex<'ctx> {
/// Get the value to set `NDIndex::type` for this variant.
fn get_type_id(&self) -> u64 {
// Defined in IRRT, must be in sync
match self {
RustNDIndex::SingleElement(_) => 0,
RustNDIndex::Slice(_) => 1,
RustNDIndex::NewAxis => 2,
RustNDIndex::Ellipsis => 3,
}
}
/// Write the contents to an LLVM [`NDIndex`].
fn write_to_ndindex<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
dst_ndindex_ptr: Instance<'ctx, Ptr<Struct<NDIndex>>>,
) {
// Set `dst_ndindex_ptr->type`
dst_ndindex_ptr.gep(ctx, |f| f.type_).store(
ctx,
Int(NDIndexType::default()).const_int(generator, ctx.ctx, self.get_type_id()),
);
// Set `dst_ndindex_ptr->data`
match self {
RustNDIndex::SingleElement(in_index) => {
let index_ptr = Int(Int32).alloca(generator, ctx);
index_ptr.store(ctx, *in_index);
dst_ndindex_ptr
.gep(ctx, |f| f.data)
.store(ctx, index_ptr.pointer_cast(generator, ctx, Int(Byte)));
}
RustNDIndex::Slice(in_rust_slice) => {
let user_slice_ptr = Struct(Slice { int_kind: Int32 }).alloca(generator, ctx);
in_rust_slice.write_to_slice(generator, ctx, user_slice_ptr);
dst_ndindex_ptr
.gep(ctx, |f| f.data)
.store(ctx, user_slice_ptr.pointer_cast(generator, ctx, Int(Byte)));
}
RustNDIndex::NewAxis | RustNDIndex::Ellipsis => {}
}
}
/// Allocate an array of `NDIndex`es on the stack and return its stack pointer.
pub fn alloca_ndindices<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
in_ndindices: &[RustNDIndex<'ctx>],
) -> (Instance<'ctx, Int<SizeT>>, Instance<'ctx, Ptr<Struct<NDIndex>>>) {
let ndindex_model = Struct(NDIndex);
let num_ndindices = Int(SizeT).const_int(generator, ctx.ctx, in_ndindices.len() as u64);
let ndindices = ndindex_model.array_alloca(generator, ctx, num_ndindices.value);
for (i, in_ndindex) in in_ndindices.iter().enumerate() {
let pndindex = ndindices.offset_const(ctx, i as u64);
in_ndindex.write_to_ndindex(generator, ctx, pndindex);
}
(num_ndindices, ndindices)
}
}
impl<'ctx> NDArrayObject<'ctx> {
/// Get the ndims [`Type`] after indexing with a given slice.
#[must_use]
pub fn deduce_ndims_after_indexing_with(&self, indices: &[RustNDIndex<'ctx>]) -> u64 {
let mut ndims = self.ndims;
for index in indices {
match index {
RustNDIndex::SingleElement(_) => {
ndims -= 1; // Single elements decrements ndims
}
RustNDIndex::NewAxis => {
ndims += 1; // `np.newaxis` / `none` adds a new axis
}
RustNDIndex::Ellipsis | RustNDIndex::Slice(_) => {}
}
}
ndims
}
/// Index into the ndarray, and return a newly-allocated view on this ndarray.
///
/// This function behaves like NumPy's ndarray indexing, but if the indices index
/// into a single element, an unsized ndarray is returned.
#[must_use]
pub fn index<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
indices: &[RustNDIndex<'ctx>],
) -> Self {
let dst_ndims = self.deduce_ndims_after_indexing_with(indices);
let dst_ndarray = NDArrayObject::alloca(generator, ctx, self.dtype, dst_ndims);
let (num_indices, indices) = RustNDIndex::alloca_ndindices(generator, ctx, indices);
call_nac3_ndarray_index(
generator,
ctx,
num_indices,
indices,
self.instance,
dst_ndarray.instance,
);
dst_ndarray
}
}
pub mod util {
use itertools::Itertools;
use nac3parser::ast::{Expr, ExprKind};
use crate::{
codegen::{expr::gen_slice, model::*, CodeGenContext, CodeGenerator},
typecheck::typedef::Type,
};
use super::{RustNDIndex, RustSlice};
/// Generate LLVM code to transform an ndarray subscript expression to
/// its list of [`RustNDIndex`]
///
/// i.e.,
/// ```python
/// my_ndarray[::3, 1, :2:]
/// ^^^^^^^^^^^ Then these into a three `RustNDIndex`es
/// ```
pub fn gen_ndarray_subscript_ndindices<'ctx, G: CodeGenerator>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
subscript: &Expr<Option<Type>>,
) -> Result<Vec<RustNDIndex<'ctx>>, String> {
// TODO: Support https://numpy.org/doc/stable/user/basics.indexing.html#dimensional-indexing-tools
// Annoying notes about `slice`
// - `my_array[5]`
// - slice is a `Constant`
// - `my_array[:5]`
// - slice is a `Slice`
// - `my_array[:]`
// - slice is a `Slice`, but lower upper step would all be `Option::None`
// - `my_array[:, :]`
// - slice is now a `Tuple` of two `Slice`-s
//
// In summary:
// - when there is a comma "," within [], `slice` will be a `Tuple` of the entries.
// - when there is not comma "," within [] (i.e., just a single entry), `slice` will be that entry itself.
//
// So we first "flatten" out the slice expression
let index_exprs = match &subscript.node {
ExprKind::Tuple { elts, .. } => elts.iter().collect_vec(),
_ => vec![subscript],
};
// Process all index expressions
let mut rust_ndindices: Vec<RustNDIndex> = Vec::with_capacity(index_exprs.len()); // Not using iterators here because `?` is used here.
for index_expr in index_exprs {
// NOTE: Currently nac3core's slices do not have an object representation,
// so the code/implementation looks awkward - we have to do pattern matching on the expression
let ndindex = if let ExprKind::Slice { lower, upper, step } = &index_expr.node {
// Handle slices
let (lower, upper, step) = gen_slice(generator, ctx, lower, upper, step)?;
RustNDIndex::Slice(RustSlice { start: lower, stop: upper, step })
} else {
// Treat and handle everything else as a single element index.
let index = generator.gen_expr(ctx, index_expr)?.unwrap().to_basic_value_enum(
ctx,
generator,
ctx.primitives.int32, // Must be int32, this checks for illegal values
)?;
let index = Int(Int32).check_value(generator, ctx.ctx, index).unwrap();
RustNDIndex::SingleElement(index)
};
rust_ndindices.push(ndindex);
}
Ok(rust_ndindices)
}
}

View File

@ -0,0 +1,220 @@
use inkwell::values::BasicValueEnum;
use itertools::Itertools;
use crate::{
codegen::{
object::ndarray::{AnyObject, NDArrayObject},
stmt::gen_for_callback,
CodeGenContext, CodeGenerator,
},
typecheck::typedef::Type,
};
use super::{nditer::NDIterHandle, NDArrayOut, ScalarOrNDArray};
impl<'ctx> NDArrayObject<'ctx> {
/// Generate LLVM IR to broadcast `ndarray`s together, and starmap through them with `mapping` elementwise.
///
/// `mapping` is an LLVM IR generator. The input of `mapping` is the list of elements when iterating through
/// the input `ndarrays` after broadcasting. The output of `mapping` is the result of the elementwise operation.
///
/// `out` specifies whether the result should be a new ndarray or to be written an existing ndarray.
pub fn broadcast_starmap<'a, G, MappingFn>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
ndarrays: &[Self],
out: NDArrayOut<'ctx>,
mapping: MappingFn,
) -> Result<Self, String>
where
G: CodeGenerator + ?Sized,
MappingFn: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
&[BasicValueEnum<'ctx>],
) -> Result<BasicValueEnum<'ctx>, String>,
{
// Broadcast inputs
let broadcast_result = NDArrayObject::broadcast(generator, ctx, ndarrays);
let out_ndarray = match out {
NDArrayOut::NewNDArray { dtype } => {
// Create a new ndarray based on the broadcast shape.
let result_ndarray =
NDArrayObject::alloca(generator, ctx, dtype, broadcast_result.ndims);
result_ndarray.copy_shape_from_array(generator, ctx, broadcast_result.shape);
result_ndarray.create_data(generator, ctx);
result_ndarray
}
NDArrayOut::WriteToNDArray { ndarray: result_ndarray } => {
// Use an existing ndarray.
// Check that its shape is compatible with the broadcast shape.
result_ndarray.assert_can_be_written_by_out(
generator,
ctx,
broadcast_result.ndims,
broadcast_result.shape,
);
result_ndarray
}
};
// Map element-wise and store results into `mapped_ndarray`.
let nditer = NDIterHandle::new(generator, ctx, out_ndarray);
gen_for_callback(
generator,
ctx,
Some("broadcast_starmap"),
|generator, ctx| {
// Create NDIters for all broadcasted input ndarrays.
let other_nditers = broadcast_result
.ndarrays
.iter()
.map(|ndarray| NDIterHandle::new(generator, ctx, *ndarray))
.collect_vec();
Ok((nditer, other_nditers))
},
|generator, ctx, (out_nditer, _in_nditers)| {
// We can simply use `out_nditer`'s `has_next()`.
// `in_nditers`' `has_next()`s should return the same value.
Ok(out_nditer.has_next(generator, ctx).value)
},
|generator, ctx, _hooks, (out_nditer, in_nditers)| {
// Get all the scalars from the broadcasted input ndarrays, pass them to `mapping`,
// and write to `out_ndarray`.
let in_scalars = in_nditers
.iter()
.map(|nditer| nditer.get_scalar(generator, ctx).value)
.collect_vec();
let result = mapping(generator, ctx, &in_scalars)?;
let p = out_nditer.get_pointer(generator, ctx);
ctx.builder.build_store(p, result).unwrap();
Ok(())
},
|generator, ctx, (out_nditer, in_nditers)| {
// Advance all iterators
out_nditer.next(generator, ctx);
in_nditers.iter().for_each(|nditer| nditer.next(generator, ctx));
Ok(())
},
)?;
Ok(out_ndarray)
}
/// Map through this ndarray with an elementwise function.
pub fn map<'a, G, Mapping>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
out: NDArrayOut<'ctx>,
mapping: Mapping,
) -> Result<Self, String>
where
G: CodeGenerator + ?Sized,
Mapping: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
BasicValueEnum<'ctx>,
) -> Result<BasicValueEnum<'ctx>, String>,
{
NDArrayObject::broadcast_starmap(
generator,
ctx,
&[*self],
out,
|generator, ctx, scalars| mapping(generator, ctx, scalars[0]),
)
}
}
impl<'ctx> ScalarOrNDArray<'ctx> {
/// Starmap through a list of inputs using `mapping`, where an input could be an ndarray, a scalar.
///
/// This function is very helpful when implementing NumPy functions that takes on either scalars or ndarrays or a mix of them
/// as their inputs and produces either an ndarray with broadcast, or a scalar if all its inputs are all scalars.
///
/// For example ,this function can be used to implement `np.add`, which has the following behaviors:
/// - `np.add(3, 4) = 7` # (scalar, scalar) -> scalar
/// - `np.add(3, np.array([4, 5, 6]))` # (scalar, ndarray) -> ndarray; the first `scalar` is converted into an ndarray and broadcasted.
/// - `np.add(np.array([[1], [2], [3]]), np.array([[4, 5, 6]]))` # (ndarray, ndarray) -> ndarray; there is broadcasting.
///
/// ## Details:
///
/// If `inputs` are all [`ScalarOrNDArray::Scalar`], the output will be a [`ScalarOrNDArray::Scalar`] with type `ret_dtype`.
///
/// Otherwise (if there are any [`ScalarOrNDArray::NDArray`] in `inputs`), all inputs will be 'as-ndarray'-ed into ndarrays,
/// then all inputs (now all ndarrays) will be passed to [`NDArrayObject::broadcasting_starmap`] and **create** a new ndarray
/// with dtype `ret_dtype`.
pub fn broadcasting_starmap<'a, G, MappingFn>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
inputs: &[ScalarOrNDArray<'ctx>],
ret_dtype: Type,
mapping: MappingFn,
) -> Result<ScalarOrNDArray<'ctx>, String>
where
G: CodeGenerator + ?Sized,
MappingFn: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
&[BasicValueEnum<'ctx>],
) -> Result<BasicValueEnum<'ctx>, String>,
{
// Check if all inputs are Scalars
let all_scalars: Option<Vec<_>> = inputs.iter().map(AnyObject::try_from).try_collect().ok();
if let Some(scalars) = all_scalars {
let scalars = scalars.iter().map(|scalar| scalar.value).collect_vec();
let value = mapping(generator, ctx, &scalars)?;
Ok(ScalarOrNDArray::Scalar(AnyObject { ty: ret_dtype, value }))
} else {
// Promote all input to ndarrays and map through them.
let inputs = inputs.iter().map(|input| input.to_ndarray(generator, ctx)).collect_vec();
let ndarray = NDArrayObject::broadcast_starmap(
generator,
ctx,
&inputs,
NDArrayOut::NewNDArray { dtype: ret_dtype },
mapping,
)?;
Ok(ScalarOrNDArray::NDArray(ndarray))
}
}
/// Map through this [`ScalarOrNDArray`] with an elementwise function.
///
/// If this is a scalar, `mapping` will directly act on the scalar. This function will return a [`ScalarOrNDArray::Scalar`] of that result.
///
/// If this is an ndarray, `mapping` will be applied to the elements of the ndarray. A new ndarray of the results will be created and
/// returned as a [`ScalarOrNDArray::NDArray`].
pub fn map<'a, G, Mapping>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
ret_dtype: Type,
mapping: Mapping,
) -> Result<ScalarOrNDArray<'ctx>, String>
where
G: CodeGenerator + ?Sized,
Mapping: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
BasicValueEnum<'ctx>,
) -> Result<BasicValueEnum<'ctx>, String>,
{
ScalarOrNDArray::broadcasting_starmap(
generator,
ctx,
&[*self],
ret_dtype,
|generator, ctx, scalars| mapping(generator, ctx, scalars[0]),
)
}
}

View File

@ -0,0 +1,207 @@
use std::cmp::max;
use nac3parser::ast::Operator;
use util::gen_for_model;
use crate::{
codegen::{
expr::gen_binop_expr_with_values, irrt::call_nac3_ndarray_matmul_calculate_shapes,
model::*, object::ndarray::indexing::RustNDIndex, CodeGenContext, CodeGenerator,
},
typecheck::{magic_methods::Binop, typedef::Type},
};
use super::{NDArrayObject, NDArrayOut};
/// Perform `np.einsum("...ij,...jk->...ik", in_a, in_b)`.
///
/// `dst_dtype` defines the dtype of the returned ndarray.
fn matmul_at_least_2d<'ctx, G: CodeGenerator>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dst_dtype: Type,
in_a: NDArrayObject<'ctx>,
in_b: NDArrayObject<'ctx>,
) -> NDArrayObject<'ctx> {
assert!(in_a.ndims >= 2);
assert!(in_b.ndims >= 2);
// Deduce ndims of the result of matmul.
let ndims_int = max(in_a.ndims, in_b.ndims);
let ndims = Int(SizeT).const_int(generator, ctx.ctx, ndims_int);
let num_0 = Int(SizeT).const_int(generator, ctx.ctx, 0);
let num_1 = Int(SizeT).const_int(generator, ctx.ctx, 1);
// Broadcasts `in_a.shape[:-2]` and `in_b.shape[:-2]` together and allocate the
// destination ndarray to store the result of matmul.
let (a, b, dst) = {
let in_a_ndims = in_a.ndims_llvm(generator, ctx.ctx);
let in_a_shape = in_a.instance.get(generator, ctx, |f| f.shape);
let in_b_ndims = in_b.ndims_llvm(generator, ctx.ctx);
let in_b_shape = in_b.instance.get(generator, ctx, |f| f.shape);
let a_shape = Int(SizeT).array_alloca(generator, ctx, ndims.value);
let b_shape = Int(SizeT).array_alloca(generator, ctx, ndims.value);
let dst_shape = Int(SizeT).array_alloca(generator, ctx, ndims.value);
// Matmul dimension compatibility is checked here.
call_nac3_ndarray_matmul_calculate_shapes(
generator, ctx, in_a_ndims, in_a_shape, in_b_ndims, in_b_shape, ndims, a_shape,
b_shape, dst_shape,
);
let a = in_a.broadcast_to(generator, ctx, ndims_int, a_shape);
let b = in_b.broadcast_to(generator, ctx, ndims_int, b_shape);
let dst = NDArrayObject::alloca(generator, ctx, dst_dtype, ndims_int);
dst.copy_shape_from_array(generator, ctx, dst_shape);
dst.create_data(generator, ctx);
(a, b, dst)
};
let len =
a.instance.get(generator, ctx, |f| f.shape).get_index_const(generator, ctx, ndims_int - 1);
let at_row = ndims_int - 2;
let at_col = ndims_int - 1;
let dst_dtype_llvm = ctx.get_llvm_type(generator, dst_dtype);
let dst_zero = dst_dtype_llvm.const_zero();
dst.foreach(generator, ctx, |generator, ctx, _, hdl| {
let pdst_ij = hdl.get_pointer(generator, ctx);
ctx.builder.build_store(pdst_ij, dst_zero).unwrap();
let indices = hdl.get_indices();
let i = indices.get_index_const(generator, ctx, at_row);
let j = indices.get_index_const(generator, ctx, at_col);
gen_for_model(generator, ctx, num_0, len, num_1, |generator, ctx, _, k| {
// `indices` is modified to index into `a` and `b`, and restored.
indices.set_index_const(ctx, at_row, i);
indices.set_index_const(ctx, at_col, k);
let a_ik = a.get_scalar_by_indices(generator, ctx, indices);
indices.set_index_const(ctx, at_row, k);
indices.set_index_const(ctx, at_col, j);
let b_kj = b.get_scalar_by_indices(generator, ctx, indices);
// Restore `indices`.
indices.set_index_const(ctx, at_row, i);
indices.set_index_const(ctx, at_col, j);
// x = a_[...]ik * b_[...]kj
let x = gen_binop_expr_with_values(
generator,
ctx,
(&Some(a.dtype), a_ik.value),
Binop::normal(Operator::Mult),
(&Some(b.dtype), b_kj.value),
ctx.current_loc,
)?
.unwrap()
.to_basic_value_enum(ctx, generator, dst_dtype)?;
// dst_[...]ij += x
let dst_ij = ctx.builder.build_load(pdst_ij, "").unwrap();
let dst_ij = gen_binop_expr_with_values(
generator,
ctx,
(&Some(dst_dtype), dst_ij),
Binop::normal(Operator::Add),
(&Some(dst_dtype), x),
ctx.current_loc,
)?
.unwrap()
.to_basic_value_enum(ctx, generator, dst_dtype)?;
ctx.builder.build_store(pdst_ij, dst_ij).unwrap();
Ok(())
})
})
.unwrap();
dst
}
impl<'ctx> NDArrayObject<'ctx> {
/// Perform `np.matmul` according to the rules in
/// <https://numpy.org/doc/stable/reference/generated/numpy.matmul.html>.
///
/// This function always return an [`NDArrayObject`]. You may want to use [`NDArrayObject::split_unsized`]
/// to handle when the output could be a scalar.
///
/// `dst_dtype` defines the dtype of the returned ndarray.
pub fn matmul<G: CodeGenerator>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
a: Self,
b: Self,
out: NDArrayOut<'ctx>,
) -> Self {
// Sanity check, but type inference should prevent this.
assert!(a.ndims > 0 && b.ndims > 0, "np.matmul disallows scalar input");
/*
If both arguments are 2-D they are multiplied like conventional matrices.
If either argument is N-D, N > 2, it is treated as a stack of matrices residing in the last two indices and broadcast accordingly.
If the first argument is 1-D, it is promoted to a matrix by prepending a 1 to its dimensions. After matrix multiplication the prepended 1 is removed.
If the second argument is 1-D, it is promoted to a matrix by appending a 1 to its dimensions. After matrix multiplication the appended 1 is removed.
*/
let new_a = if a.ndims == 1 {
// Prepend 1 to its dimensions
a.index(generator, ctx, &[RustNDIndex::NewAxis, RustNDIndex::Ellipsis])
} else {
a
};
let new_b = if b.ndims == 1 {
// Append 1 to its dimensions
b.index(generator, ctx, &[RustNDIndex::Ellipsis, RustNDIndex::NewAxis])
} else {
b
};
// NOTE: `result` will always be a newly allocated ndarray.
// Current implementation cannot do in-place matrix muliplication.
let mut result = matmul_at_least_2d(generator, ctx, out.get_dtype(), new_a, new_b);
// Postprocessing on the result to remove prepended/appended axes.
let mut postindices = vec![];
let zero = Int(Int32).const_0(generator, ctx.ctx);
if a.ndims == 1 {
// Remove the prepended 1
postindices.push(RustNDIndex::SingleElement(zero));
}
if b.ndims == 1 {
// Remove the appended 1
postindices.push(RustNDIndex::Ellipsis);
postindices.push(RustNDIndex::SingleElement(zero));
}
if !postindices.is_empty() {
result = result.index(generator, ctx, &postindices);
}
match out {
NDArrayOut::NewNDArray { .. } => result,
NDArrayOut::WriteToNDArray { ndarray: out_ndarray } => {
let result_shape = result.instance.get(generator, ctx, |f| f.shape);
out_ndarray.assert_can_be_written_by_out(
generator,
ctx,
result.ndims,
result_shape,
);
out_ndarray.copy_data_from(generator, ctx, result);
out_ndarray
}
}
}
}

View File

@ -0,0 +1,671 @@
pub mod array;
pub mod broadcast;
pub mod contiguous;
pub mod factory;
pub mod indexing;
pub mod map;
pub mod matmul;
pub mod nditer;
pub mod shape_util;
pub mod view;
use inkwell::{
context::Context,
types::BasicType,
values::{BasicValue, BasicValueEnum, PointerValue},
AddressSpace,
};
use crate::{
codegen::{
irrt::{
call_nac3_ndarray_copy_data, call_nac3_ndarray_get_nth_pelement,
call_nac3_ndarray_get_pelement_by_indices, call_nac3_ndarray_is_c_contiguous,
call_nac3_ndarray_len, call_nac3_ndarray_nbytes,
call_nac3_ndarray_set_strides_by_shape, call_nac3_ndarray_size,
call_nac3_ndarray_util_assert_output_shape_same,
},
model::*,
CodeGenContext, CodeGenerator,
},
toplevel::{
helper::{create_ndims, extract_ndims},
numpy::{make_ndarray_ty, unpack_ndarray_var_tys},
},
typecheck::typedef::{Type, TypeEnum},
};
use super::{any::AnyObject, tuple::TupleObject};
/// Fields of [`NDArray`]
pub struct NDArrayFields<'ctx, F: FieldTraversal<'ctx>> {
pub data: F::Out<Ptr<Int<Byte>>>,
pub itemsize: F::Out<Int<SizeT>>,
pub ndims: F::Out<Int<SizeT>>,
pub shape: F::Out<Ptr<Int<SizeT>>>,
pub strides: F::Out<Ptr<Int<SizeT>>>,
}
/// A strided ndarray in NAC3.
///
/// See IRRT implementation for details about its fields.
#[derive(Debug, Clone, Copy, Default)]
pub struct NDArray;
impl<'ctx> StructKind<'ctx> for NDArray {
type Fields<F: FieldTraversal<'ctx>> = NDArrayFields<'ctx, F>;
fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
Self::Fields {
data: traversal.add_auto("data"),
itemsize: traversal.add_auto("itemsize"),
ndims: traversal.add_auto("ndims"),
shape: traversal.add_auto("shape"),
strides: traversal.add_auto("strides"),
}
}
}
/// A NAC3 Python ndarray object.
#[derive(Debug, Clone, Copy)]
pub struct NDArrayObject<'ctx> {
pub dtype: Type,
pub ndims: u64,
pub instance: Instance<'ctx, Ptr<Struct<NDArray>>>,
}
impl<'ctx> NDArrayObject<'ctx> {
/// Attempt to convert an [`AnyObject`] into an [`NDArrayObject`].
pub fn from_object<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
object: AnyObject<'ctx>,
) -> NDArrayObject<'ctx> {
let (dtype, ndims) = unpack_ndarray_var_tys(&mut ctx.unifier, object.ty);
let ndims = extract_ndims(&ctx.unifier, ndims);
Self::from_value_and_unpacked_types(generator, ctx, object.value, dtype, ndims)
}
/// Like [`NDArrayObject::from_object`] but you directly supply the ndarray's
/// `dtype` and `ndims`.
pub fn from_value_and_unpacked_types<V: BasicValue<'ctx>, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
value: V,
dtype: Type,
ndims: u64,
) -> Self {
let value = Ptr(Struct(NDArray)).check_value(generator, ctx.ctx, value).unwrap();
NDArrayObject { dtype, ndims, instance: value }
}
/// Get this ndarray's `ndims` as an LLVM constant.
pub fn ndims_llvm<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &'ctx Context,
) -> Instance<'ctx, Int<SizeT>> {
Int(SizeT).const_int(generator, ctx, self.ndims)
}
/// Get the typechecker ndarray type of this [`NDArrayObject`].
pub fn get_type(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> Type {
let ndims = create_ndims(&mut ctx.unifier, self.ndims);
make_ndarray_ty(&mut ctx.unifier, &ctx.primitives, Some(self.dtype), Some(ndims))
}
/// Forget that this is an ndarray and convert into an [`AnyObject`].
pub fn to_any(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> AnyObject<'ctx> {
let ty = self.get_type(ctx);
AnyObject { value: self.instance.value.as_basic_value_enum(), ty }
}
/// Allocate an ndarray on the stack given its `ndims` and `dtype`.
///
/// `shape` and `strides` will be automatically allocated on the stack.
//e
/// The returned ndarray's content will be:
/// - `data`: set to `nullptr`.
/// - `itemsize`: set to the `sizeof()` of `dtype`.
/// - `ndims`: set to the value of `ndims`.
/// - `shape`: allocated with an array of length `ndims` with uninitialized values.
/// - `strides`: allocated with an array of length `ndims` with uninitialized values.
pub fn alloca<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
ndims: u64,
) -> Self {
let ndarray = Struct(NDArray).alloca(generator, ctx);
let data = Ptr(Int(Byte)).nullptr(generator, ctx.ctx);
ndarray.set(ctx, |f| f.data, data);
let itemsize = ctx.get_llvm_type(generator, dtype).size_of().unwrap();
let itemsize = Int(SizeT).s_extend_or_bit_cast(generator, ctx, itemsize);
ndarray.set(ctx, |f| f.itemsize, itemsize);
let ndims_val = Int(SizeT).const_int(generator, ctx.ctx, ndims);
ndarray.set(ctx, |f| f.ndims, ndims_val);
let shape = Int(SizeT).array_alloca(generator, ctx, ndims_val.value);
ndarray.set(ctx, |f| f.shape, shape);
let strides = Int(SizeT).array_alloca(generator, ctx, ndims_val.value);
ndarray.set(ctx, |f| f.strides, strides);
NDArrayObject { dtype, ndims, instance: ndarray }
}
/// Convenience function. Allocate an [`NDArrayObject`] with a statically known shape.
///
/// The returned [`NDArrayObject`]'s `data` and `strides` are uninitialized.
pub fn alloca_constant_shape<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
shape: &[u64],
) -> Self {
let ndarray = NDArrayObject::alloca(generator, ctx, dtype, shape.len() as u64);
// Write shape
let dst_shape = ndarray.instance.get(generator, ctx, |f| f.shape);
for (i, dim) in shape.iter().enumerate() {
let dim = Int(SizeT).const_int(generator, ctx.ctx, *dim);
dst_shape.offset_const(ctx, i as u64).store(ctx, dim);
}
ndarray
}
/// Convenience function. Allocate an [`NDArrayObject`] with a dynamically known shape.
///
/// The returned [`NDArrayObject`]'s `data` and `strides` are uninitialized.
pub fn alloca_dynamic_shape<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
dtype: Type,
shape: &[Instance<'ctx, Int<SizeT>>],
) -> Self {
let ndarray = NDArrayObject::alloca(generator, ctx, dtype, shape.len() as u64);
// Write shape
let dst_shape = ndarray.instance.get(generator, ctx, |f| f.shape);
for (i, dim) in shape.iter().enumerate() {
dst_shape.offset_const(ctx, i as u64).store(ctx, *dim);
}
ndarray
}
/// Initialize an ndarray's `data` by allocating a buffer on the stack.
/// The allocated data buffer is considered to be *owned* by the ndarray.
///
/// `strides` of the ndarray will also be updated with `set_strides_by_shape`.
///
/// `shape` and `itemsize` of the ndarray ***must*** be initialized first.
pub fn create_data<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) {
let nbytes = self.nbytes(generator, ctx);
let data = Int(Byte).array_alloca(generator, ctx, nbytes.value);
self.instance.set(ctx, |f| f.data, data);
self.set_strides_contiguous(generator, ctx);
}
/// Copy shape dimensions from an array.
pub fn copy_shape_from_array<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let num_items = self.ndims_llvm(generator, ctx.ctx).value;
self.instance.get(generator, ctx, |f| f.shape).copy_from(generator, ctx, shape, num_items);
}
/// Copy shape dimensions from an ndarray.
/// Panics if `ndims` mismatches.
pub fn copy_shape_from_ndarray<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_ndarray: NDArrayObject<'ctx>,
) {
assert_eq!(self.ndims, src_ndarray.ndims);
let src_shape = src_ndarray.instance.get(generator, ctx, |f| f.shape);
self.copy_shape_from_array(generator, ctx, src_shape);
}
/// Copy strides dimensions from an array.
pub fn copy_strides_from_array<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
strides: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let num_items = self.ndims_llvm(generator, ctx.ctx).value;
self.instance
.get(generator, ctx, |f| f.strides)
.copy_from(generator, ctx, strides, num_items);
}
/// Copy strides dimensions from an ndarray.
/// Panics if `ndims` mismatches.
pub fn copy_strides_from_ndarray<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_ndarray: NDArrayObject<'ctx>,
) {
assert_eq!(self.ndims, src_ndarray.ndims);
let src_strides = src_ndarray.instance.get(generator, ctx, |f| f.strides);
self.copy_strides_from_array(generator, ctx, src_strides);
}
/// Get the `np.size()` of this ndarray.
pub fn size<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<SizeT>> {
call_nac3_ndarray_size(generator, ctx, self.instance)
}
/// Get the `ndarray.nbytes` of this ndarray.
pub fn nbytes<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<SizeT>> {
call_nac3_ndarray_nbytes(generator, ctx, self.instance)
}
/// Get the `len()` of this ndarray.
pub fn len<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<SizeT>> {
call_nac3_ndarray_len(generator, ctx, self.instance)
}
/// Check if this ndarray is C-contiguous.
///
/// See NumPy's `flags["C_CONTIGUOUS"]`: <https://numpy.org/doc/stable/reference/generated/numpy.ndarray.flags.html#numpy.ndarray.flags>
pub fn is_c_contiguous<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<Bool>> {
call_nac3_ndarray_is_c_contiguous(generator, ctx, self.instance)
}
/// Get the pointer to the n-th (0-based) element.
///
/// The returned pointer has the element type of the LLVM type of this ndarray's `dtype`.
pub fn get_nth_pelement<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
nth: Instance<'ctx, Int<SizeT>>,
) -> PointerValue<'ctx> {
let elem_ty = ctx.get_llvm_type(generator, self.dtype);
let p = call_nac3_ndarray_get_nth_pelement(generator, ctx, self.instance, nth);
ctx.builder
.build_pointer_cast(p.value, elem_ty.ptr_type(AddressSpace::default()), "")
.unwrap()
}
/// Get the n-th (0-based) scalar.
pub fn get_nth_scalar<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
nth: Instance<'ctx, Int<SizeT>>,
) -> AnyObject<'ctx> {
let ptr = self.get_nth_pelement(generator, ctx, nth);
let value = ctx.builder.build_load(ptr, "").unwrap();
AnyObject { ty: self.dtype, value }
}
/// Get the pointer to the element indexed by `indices`.
///
/// The returned pointer has the element type of the LLVM type of this ndarray's `dtype`.
pub fn get_pelement_by_indices<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
indices: Instance<'ctx, Ptr<Int<SizeT>>>,
) -> PointerValue<'ctx> {
let elem_ty = ctx.get_llvm_type(generator, self.dtype);
let p = call_nac3_ndarray_get_pelement_by_indices(generator, ctx, self.instance, indices);
ctx.builder
.build_pointer_cast(p.value, elem_ty.ptr_type(AddressSpace::default()), "")
.unwrap()
}
/// Get the scalar indexed by `indices`.
pub fn get_scalar_by_indices<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
indices: Instance<'ctx, Ptr<Int<SizeT>>>,
) -> AnyObject<'ctx> {
let ptr = self.get_pelement_by_indices(generator, ctx, indices);
let value = ctx.builder.build_load(ptr, "").unwrap();
AnyObject { ty: self.dtype, value }
}
/// Call [`call_nac3_ndarray_set_strides_by_shape`] on this ndarray to update `strides`.
///
/// Update the ndarray's strides to make the ndarray contiguous.
pub fn set_strides_contiguous<G: CodeGenerator + ?Sized>(
self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) {
call_nac3_ndarray_set_strides_by_shape(generator, ctx, self.instance);
}
/// Clone/Copy this ndarray - Allocate a new ndarray with the same shape as this ndarray and copy the contents over.
///
/// The new ndarray will own its data and will be C-contiguous.
#[must_use]
pub fn make_copy<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Self {
let clone = NDArrayObject::alloca(generator, ctx, self.dtype, self.ndims);
let shape = self.instance.gep(ctx, |f| f.shape).load(generator, ctx);
clone.copy_shape_from_array(generator, ctx, shape);
clone.create_data(generator, ctx);
clone.copy_data_from(generator, ctx, *self);
clone
}
/// Copy data from another ndarray.
///
/// This ndarray and `src` is that their `np.size()` should be the same. Their shapes
/// do not matter. The copying order is determined by how their flattened views look.
///
/// Panics if the `dtype`s of ndarrays are different.
pub fn copy_data_from<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src: NDArrayObject<'ctx>,
) {
assert!(ctx.unifier.unioned(self.dtype, src.dtype), "self and src dtype should match");
call_nac3_ndarray_copy_data(generator, ctx, src.instance, self.instance);
}
/// Returns true if this ndarray is unsized - `ndims == 0` and only contains a scalar.
#[must_use]
pub fn is_unsized(&self) -> bool {
self.ndims == 0
}
/// If this ndarray is unsized, return its sole value as an [`AnyObject`].
/// Otherwise, do nothing and return the ndarray itself.
pub fn split_unsized<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> ScalarOrNDArray<'ctx> {
if self.is_unsized() {
// NOTE: `np.size(self) == 0` here is never possible.
let zero = Int(SizeT).const_0(generator, ctx.ctx);
let value = self.get_nth_scalar(generator, ctx, zero).value;
ScalarOrNDArray::Scalar(AnyObject { ty: self.dtype, value })
} else {
ScalarOrNDArray::NDArray(*self)
}
}
/// Fill the ndarray with a scalar.
///
/// `fill_value` must have the same LLVM type as the `dtype` of this ndarray.
pub fn fill<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
value: BasicValueEnum<'ctx>,
) {
self.foreach(generator, ctx, |generator, ctx, _hooks, nditer| {
let p = nditer.get_pointer(generator, ctx);
ctx.builder.build_store(p, value).unwrap();
Ok(())
})
.unwrap();
}
/// Create the shape tuple of this ndarray like `np.shape(<ndarray>)`.
///
/// The returned integers in the tuple are in int32.
pub fn make_shape_tuple<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> TupleObject<'ctx> {
// TODO: Return a tuple of SizeT
let mut objects = Vec::with_capacity(self.ndims as usize);
for i in 0..self.ndims {
let dim = self
.instance
.get(generator, ctx, |f| f.shape)
.get_index_const(generator, ctx, i)
.truncate(generator, ctx, Int32);
objects.push(AnyObject {
ty: ctx.primitives.int32,
value: dim.value.as_basic_value_enum(),
});
}
TupleObject::from_objects(generator, ctx, objects)
}
/// Create the strides tuple of this ndarray like `np.strides(<ndarray>)`.
///
/// The returned integers in the tuple are in int32.
pub fn make_strides_tuple<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> TupleObject<'ctx> {
// TODO: Return a tuple of SizeT.
let mut objects = Vec::with_capacity(self.ndims as usize);
for i in 0..self.ndims {
let dim = self
.instance
.get(generator, ctx, |f| f.strides)
.get_index_const(generator, ctx, i)
.truncate(generator, ctx, Int32);
objects.push(AnyObject {
ty: ctx.primitives.int32,
value: dim.value.as_basic_value_enum(),
});
}
TupleObject::from_objects(generator, ctx, objects)
}
/// Create an unsized ndarray to contain `object`.
pub fn make_unsized<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
object: AnyObject<'ctx>,
) -> NDArrayObject<'ctx> {
// We have to put the value on the stack to get a data pointer.
let data = ctx.builder.build_alloca(object.value.get_type(), "make_unsized").unwrap();
ctx.builder.build_store(data, object.value).unwrap();
let data = Ptr(Int(Byte)).pointer_cast(generator, ctx, data);
let ndarray = NDArrayObject::alloca(generator, ctx, object.ty, 0);
ndarray.instance.set(ctx, |f| f.data, data);
ndarray
}
/// Check if this `NDArray` can be used as an `out` ndarray for an operation.
///
/// Raise an exception if the shapes do not match.
pub fn assert_can_be_written_by_out<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
out_ndims: u64,
out_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) {
let ndarray_ndims = self.ndims_llvm(generator, ctx.ctx);
let ndarray_shape = self.instance.get(generator, ctx, |f| f.shape);
let output_ndims = Int(SizeT).const_int(generator, ctx.ctx, out_ndims);
let output_shape = out_shape;
call_nac3_ndarray_util_assert_output_shape_same(
generator,
ctx,
ndarray_ndims,
ndarray_shape,
output_ndims,
output_shape,
);
}
}
/// A convenience enum for implementing functions that acts on scalars or ndarrays or both.
#[derive(Debug, Clone, Copy)]
pub enum ScalarOrNDArray<'ctx> {
Scalar(AnyObject<'ctx>),
NDArray(NDArrayObject<'ctx>),
}
impl<'ctx> TryFrom<&ScalarOrNDArray<'ctx>> for AnyObject<'ctx> {
type Error = ();
fn try_from(value: &ScalarOrNDArray<'ctx>) -> Result<Self, Self::Error> {
match value {
ScalarOrNDArray::Scalar(scalar) => Ok(*scalar),
ScalarOrNDArray::NDArray(_ndarray) => Err(()),
}
}
}
impl<'ctx> TryFrom<&ScalarOrNDArray<'ctx>> for NDArrayObject<'ctx> {
type Error = ();
fn try_from(value: &ScalarOrNDArray<'ctx>) -> Result<Self, Self::Error> {
match value {
ScalarOrNDArray::Scalar(_scalar) => Err(()),
ScalarOrNDArray::NDArray(ndarray) => Ok(*ndarray),
}
}
}
impl<'ctx> ScalarOrNDArray<'ctx> {
/// Split on `object` either into a scalar or an ndarray.
///
/// If `object` is an ndarray, [`ScalarOrNDArray::NDArray`].
///
/// For everything else, it is wrapped with [`ScalarOrNDArray::Scalar`].
pub fn split_object<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
object: AnyObject<'ctx>,
) -> ScalarOrNDArray<'ctx> {
match &*ctx.unifier.get_ty(object.ty) {
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.ndarray.obj_id(&ctx.unifier).unwrap() =>
{
let ndarray = NDArrayObject::from_object(generator, ctx, object);
ScalarOrNDArray::NDArray(ndarray)
}
_ => ScalarOrNDArray::Scalar(object),
}
}
/// Get the underlying [`BasicValueEnum<'ctx>`] of this [`ScalarOrNDArray`].
#[must_use]
pub fn to_basic_value_enum(self) -> BasicValueEnum<'ctx> {
match self {
ScalarOrNDArray::Scalar(scalar) => scalar.value,
ScalarOrNDArray::NDArray(ndarray) => ndarray.instance.value.as_basic_value_enum(),
}
}
/// Convert this [`ScalarOrNDArray`] to an ndarray - behaves like `np.asarray`.
/// - If this is an ndarray, the ndarray is returned.
/// - If this is a scalar, this function returns new ndarray created with [`NDArrayObject::make_unsized`].
pub fn to_ndarray<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> NDArrayObject<'ctx> {
match self {
ScalarOrNDArray::NDArray(ndarray) => *ndarray,
ScalarOrNDArray::Scalar(scalar) => NDArrayObject::make_unsized(generator, ctx, *scalar),
}
}
/// Get the dtype of the ndarray created if this were called with [`ScalarOrNDArray::to_ndarray`].
#[must_use]
pub fn get_dtype(&self) -> Type {
match self {
ScalarOrNDArray::NDArray(ndarray) => ndarray.dtype,
ScalarOrNDArray::Scalar(scalar) => scalar.ty,
}
}
}
/// An helper enum specifying how a function should produce its output.
///
/// Many functions in NumPy has an optional `out` parameter (e.g., `matmul`). If `out` is specified
/// with an ndarray, the result of a function will be written to `out`. If `out` is not specified, a function will
/// create a new ndarray and store the result in it.
#[derive(Debug, Clone, Copy)]
pub enum NDArrayOut<'ctx> {
/// Tell a function should create a new ndarray with the expected element type `dtype`.
NewNDArray { dtype: Type },
/// Tell a function to write the result to `ndarray`.
WriteToNDArray { ndarray: NDArrayObject<'ctx> },
}
impl<'ctx> NDArrayOut<'ctx> {
/// Get the dtype of this output.
#[must_use]
pub fn get_dtype(&self) -> Type {
match self {
NDArrayOut::NewNDArray { dtype } => *dtype,
NDArrayOut::WriteToNDArray { ndarray } => ndarray.dtype,
}
}
}
/// A version of [`call_nac3_ndarray_set_strides_by_shape`] in Rust.
///
/// This function is used generating strides for globally defined contiguous ndarrays.
#[must_use]
pub fn make_contiguous_strides(itemsize: u64, ndims: u64, shape: &[u64]) -> Vec<u64> {
let mut strides = Vec::with_capacity(ndims as usize);
let mut stride_product = 1u64;
for i in 0..ndims {
let axis = ndims - i - 1;
strides[axis as usize] = stride_product * itemsize;
stride_product *= shape[axis as usize];
}
strides
}

View File

@ -0,0 +1,168 @@
use inkwell::{types::BasicType, values::PointerValue, AddressSpace};
use crate::codegen::{
irrt::{call_nac3_nditer_has_next, call_nac3_nditer_initialize, call_nac3_nditer_next},
model::*,
object::any::AnyObject,
stmt::{gen_for_callback, BreakContinueHooks},
CodeGenContext, CodeGenerator,
};
use super::NDArrayObject;
/// Fields of [`NDIter`]
pub struct NDIterFields<'ctx, F: FieldTraversal<'ctx>> {
pub ndims: F::Out<Int<SizeT>>,
pub shape: F::Out<Ptr<Int<SizeT>>>,
pub strides: F::Out<Ptr<Int<SizeT>>>,
pub indices: F::Out<Ptr<Int<SizeT>>>,
pub nth: F::Out<Int<SizeT>>,
pub element: F::Out<Ptr<Int<Byte>>>,
pub size: F::Out<Int<SizeT>>,
}
/// An IRRT helper structure used to iterate through an ndarray.
#[derive(Debug, Clone, Copy, Default)]
pub struct NDIter;
impl<'ctx> StructKind<'ctx> for NDIter {
type Fields<F: FieldTraversal<'ctx>> = NDIterFields<'ctx, F>;
fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
Self::Fields {
ndims: traversal.add_auto("ndims"),
shape: traversal.add_auto("shape"),
strides: traversal.add_auto("strides"),
indices: traversal.add_auto("indices"),
nth: traversal.add_auto("nth"),
element: traversal.add_auto("element"),
size: traversal.add_auto("size"),
}
}
}
/// A helper structure containing extra details of an [`NDIter`].
#[derive(Debug, Clone)]
pub struct NDIterHandle<'ctx> {
instance: Instance<'ctx, Ptr<Struct<NDIter>>>,
/// The ndarray this [`NDIter`] to iterating over.
ndarray: NDArrayObject<'ctx>,
/// The current indices of [`NDIter`].
indices: Instance<'ctx, Ptr<Int<SizeT>>>,
}
impl<'ctx> NDIterHandle<'ctx> {
/// Allocate an [`NDIter`] that iterates through an ndarray.
pub fn new<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray: NDArrayObject<'ctx>,
) -> Self {
let nditer = Struct(NDIter).alloca(generator, ctx);
let ndims = ndarray.ndims_llvm(generator, ctx.ctx);
// The caller has the responsibility to allocate 'indices' for `NDIter`.
let indices = Int(SizeT).array_alloca(generator, ctx, ndims.value);
call_nac3_nditer_initialize(generator, ctx, nditer, ndarray.instance, indices);
NDIterHandle { ndarray, instance: nditer, indices }
}
#[must_use]
pub fn has_next<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<Bool>> {
call_nac3_nditer_has_next(generator, ctx, self.instance)
}
pub fn next<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) {
call_nac3_nditer_next(generator, ctx, self.instance);
}
/// Get pointer to the current element.
#[must_use]
pub fn get_pointer<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> PointerValue<'ctx> {
let elem_ty = ctx.get_llvm_type(generator, self.ndarray.dtype);
let p = self.instance.get(generator, ctx, |f| f.element);
ctx.builder
.build_pointer_cast(p.value, elem_ty.ptr_type(AddressSpace::default()), "element")
.unwrap()
}
/// Get the value of the current element.
#[must_use]
pub fn get_scalar<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> AnyObject<'ctx> {
let p = self.get_pointer(generator, ctx);
let value = ctx.builder.build_load(p, "value").unwrap();
AnyObject { ty: self.ndarray.dtype, value }
}
/// Get the index of the current element.
#[must_use]
pub fn get_index<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<SizeT>> {
self.instance.get(generator, ctx, |f| f.nth)
}
/// Get the indices of the current element.
#[must_use]
pub fn get_indices(&self) -> Instance<'ctx, Ptr<Int<SizeT>>> {
self.indices
}
}
impl<'ctx> NDArrayObject<'ctx> {
/// Iterate through every element in the ndarray.
///
/// `body` also access to [`BreakContinueHooks`] to short-circuit.
pub fn foreach<'a, G, F>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
body: F,
) -> Result<(), String>
where
G: CodeGenerator + ?Sized,
F: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
BreakContinueHooks<'ctx>,
NDIterHandle<'ctx>,
) -> Result<(), String>,
{
gen_for_callback(
generator,
ctx,
Some("ndarray_foreach"),
|generator, ctx| Ok(NDIterHandle::new(generator, ctx, *self)),
|generator, ctx, nditer| Ok(nditer.has_next(generator, ctx).value),
|generator, ctx, hooks, nditer| body(generator, ctx, hooks, nditer),
|generator, ctx, nditer| {
nditer.next(generator, ctx);
Ok(())
},
)
}
}

View File

@ -0,0 +1,105 @@
use util::gen_for_model;
use crate::{
codegen::{
model::*,
object::{any::AnyObject, list::ListObject, tuple::TupleObject},
CodeGenContext, CodeGenerator,
},
typecheck::typedef::TypeEnum,
};
/// Parse a NumPy-like "int sequence" input and return the int sequence as an array and its length.
///
/// * `sequence` - The `sequence` parameter.
/// * `sequence_ty` - The typechecker type of `sequence`
///
/// The `sequence` argument type may only be one of the following:
/// 1. A list of `int32`; e.g., `np.empty([600, 800, 3])`
/// 2. A tuple of `int32`; e.g., `np.empty((600, 800, 3))`
/// 3. A scalar `int32`; e.g., `np.empty(3)`, this is functionally equivalent to `np.empty([3])`
///
/// All `int32` values will be sign-extended to `SizeT`.
pub fn parse_numpy_int_sequence<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
input_sequence: AnyObject<'ctx>,
) -> (Instance<'ctx, Int<SizeT>>, Instance<'ctx, Ptr<Int<SizeT>>>) {
let zero = Int(SizeT).const_0(generator, ctx.ctx);
let one = Int(SizeT).const_1(generator, ctx.ctx);
// The result `list` to return.
match &*ctx.unifier.get_ty(input_sequence.ty) {
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.list.obj_id(&ctx.unifier).unwrap() =>
{
// 1. A list of `int32`; e.g., `np.empty([600, 800, 3])`
// Check `input_sequence`
let input_sequence = ListObject::from_object(generator, ctx, input_sequence);
let len = input_sequence.instance.get(generator, ctx, |f| f.len);
let result = Int(SizeT).array_alloca(generator, ctx, len.value);
// Load all the `int32`s from the input_sequence, cast them to `SizeT`, and store them into `result`
gen_for_model(generator, ctx, zero, len, one, |generator, ctx, _hooks, i| {
// Load the i-th int32 in the input sequence
let int = input_sequence
.instance
.get(generator, ctx, |f| f.items)
.get_index(generator, ctx, i.value)
.value
.into_int_value();
// Cast to SizeT
let int = Int(SizeT).s_extend_or_bit_cast(generator, ctx, int);
// Store
result.set_index(ctx, i.value, int);
Ok(())
})
.unwrap();
(len, result)
}
TypeEnum::TTuple { .. } => {
// 2. A tuple of ints; e.g., `np.empty((600, 800, 3))`
let input_sequence = TupleObject::from_object(ctx, input_sequence);
let len = input_sequence.len(generator, ctx);
let result = Int(SizeT).array_alloca(generator, ctx, len.value);
for i in 0..input_sequence.num_elements() {
// Get the i-th element off of the tuple and load it into `result`.
let int = input_sequence.index(ctx, i).value.into_int_value();
let int = Int(SizeT).s_extend_or_bit_cast(generator, ctx, int);
result.set_index_const(ctx, i as u64, int);
}
(len, result)
}
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.int32.obj_id(&ctx.unifier).unwrap() =>
{
// 3. A scalar int; e.g., `np.empty(3)`, this is functionally equivalent to `np.empty([3])`
let input_int = input_sequence.value.into_int_value();
let len = Int(SizeT).const_1(generator, ctx.ctx);
let result = Int(SizeT).array_alloca(generator, ctx, len.value);
let int = Int(SizeT).s_extend_or_bit_cast(generator, ctx, input_int);
// Storing into result[0]
result.store(ctx, int);
(len, result)
}
_ => panic!(
"encountered unknown sequence type: {}",
ctx.unifier.stringify(input_sequence.ty)
),
}
}

View File

@ -0,0 +1,119 @@
use crate::codegen::{
irrt::{call_nac3_ndarray_reshape_resolve_and_check_new_shape, call_nac3_ndarray_transpose},
model::*,
CodeGenContext, CodeGenerator,
};
use super::{indexing::RustNDIndex, NDArrayObject};
impl<'ctx> NDArrayObject<'ctx> {
/// Make sure the ndarray is at least `ndmin`-dimensional.
///
/// If this ndarray's `ndims` is less than `ndmin`, a view is created on this with 1s prepended to the shape.
/// If this ndarray's `ndims` is not less than `ndmin`, this function does nothing and return this ndarray.
#[must_use]
pub fn atleast_nd<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndmin: u64,
) -> Self {
if self.ndims < ndmin {
// return this_ndarray[np.newaxis, np.newaxis, and more, ...]
let mut indices = vec![];
for _ in self.ndims..ndmin {
indices.push(RustNDIndex::NewAxis);
}
indices.push(RustNDIndex::Ellipsis);
self.index(generator, ctx, &indices)
} else {
*self
}
}
/// Create a reshaped view on this ndarray like `np.reshape()`.
///
/// If there is a `-1` in `new_shape`, it will be resolved; `new_shape` would **NOT** be modified as a result.
///
/// If reshape without copying is impossible, this function will allocate a new ndarray and copy contents.
///
/// * `new_ndims` - The number of dimensions of `new_shape` as a [`Type`].
/// * `new_shape` - The target shape to do `np.reshape()`.
#[must_use]
pub fn reshape_or_copy<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
new_ndims: u64,
new_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
) -> Self {
// TODO: The current criterion for whether to do a full copy or not is by checking `is_c_contiguous`,
// but this is not optimal - there are cases when the ndarray is not contiguous but could be reshaped
// without copying data. Look into how numpy does it.
let current_bb = ctx.builder.get_insert_block().unwrap();
let then_bb = ctx.ctx.insert_basic_block_after(current_bb, "then_bb");
let else_bb = ctx.ctx.insert_basic_block_after(then_bb, "else_bb");
let end_bb = ctx.ctx.insert_basic_block_after(else_bb, "end_bb");
let dst_ndarray = NDArrayObject::alloca(generator, ctx, self.dtype, new_ndims);
dst_ndarray.copy_shape_from_array(generator, ctx, new_shape);
// Reolsve negative indices
let size = self.size(generator, ctx);
let dst_ndims = dst_ndarray.ndims_llvm(generator, ctx.ctx);
let dst_shape = dst_ndarray.instance.get(generator, ctx, |f| f.shape);
call_nac3_ndarray_reshape_resolve_and_check_new_shape(
generator, ctx, size, dst_ndims, dst_shape,
);
let is_c_contiguous = self.is_c_contiguous(generator, ctx);
ctx.builder.build_conditional_branch(is_c_contiguous.value, then_bb, else_bb).unwrap();
// Inserting into then_bb: reshape is possible without copying
ctx.builder.position_at_end(then_bb);
dst_ndarray.set_strides_contiguous(generator, ctx);
dst_ndarray.instance.set(ctx, |f| f.data, self.instance.get(generator, ctx, |f| f.data));
ctx.builder.build_unconditional_branch(end_bb).unwrap();
// Inserting into else_bb: reshape is impossible without copying
ctx.builder.position_at_end(else_bb);
dst_ndarray.create_data(generator, ctx);
dst_ndarray.copy_data_from(generator, ctx, *self);
ctx.builder.build_unconditional_branch(end_bb).unwrap();
// Reposition for continuation
ctx.builder.position_at_end(end_bb);
dst_ndarray
}
/// Create a transposed view on this ndarray like `np.transpose(<ndarray>, <axes> = None)`.
/// * `axes` - If specified, should be an array of the permutation (negative indices are **allowed**).
#[must_use]
pub fn transpose<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
axes: Option<Instance<'ctx, Ptr<Int<SizeT>>>>,
) -> Self {
// Define models
let transposed_ndarray = NDArrayObject::alloca(generator, ctx, self.dtype, self.ndims);
let num_axes = self.ndims_llvm(generator, ctx.ctx);
// `axes = nullptr` if `axes` is unspecified.
let axes = axes.unwrap_or_else(|| Ptr(Int(SizeT)).nullptr(generator, ctx.ctx));
call_nac3_ndarray_transpose(
generator,
ctx,
self.instance,
transposed_ndarray.instance,
num_axes,
axes,
);
transposed_ndarray
}
}

View File

@ -0,0 +1,70 @@
use crate::codegen::{irrt::call_nac3_range_len_i32, model::*, CodeGenContext, CodeGenerator};
use super::any::AnyObject;
/// A range in NAC3.
pub type Range<N> = Array<Len<3>, Int<N>>;
impl<'ctx, N: IntKind<'ctx>> Instance<'ctx, Ptr<Range<N>>> {
/// Get GEP to `range.start`.
pub fn start(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Ptr<Int<N>>> {
self.gep_const(ctx, 0)
}
/// Get GEP to `range.stop`.
pub fn stop(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Ptr<Int<N>>> {
self.gep_const(ctx, 1)
}
/// Get GEP to `range.step`.
pub fn step(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Ptr<Int<N>>> {
self.gep_const(ctx, 2)
}
/// Convenience function to get the load the `(start, stop, step)` of this range.
#[allow(clippy::type_complexity)]
pub fn destructure<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &CodeGenContext<'ctx, '_>,
) -> (Instance<'ctx, Int<N>>, Instance<'ctx, Int<N>>, Instance<'ctx, Int<N>>) {
let start = self.start(ctx).load(generator, ctx);
let stop = self.stop(ctx).load(generator, ctx);
let step = self.step(ctx).load(generator, ctx);
(start, stop, step)
}
}
// TODO: `RangeObject` in the future will have range32, range64
/// A NAC3 Python range object.
#[derive(Debug, Clone, Copy)]
pub struct RangeObject<'ctx> {
pub instance: Instance<'ctx, Ptr<Range<Int32>>>,
}
impl<'ctx> RangeObject<'ctx> {
/// Attempt to convert an [`AnyObject`] into a [`RangeObject`].
pub fn from_object<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
object: AnyObject<'ctx>,
) -> RangeObject<'ctx> {
assert!(ctx.unifier.unioned(object.ty, ctx.primitives.range));
let instance = Ptr(Range::default()).check_value(generator, ctx.ctx, object.value).unwrap();
RangeObject { instance }
}
/// Get the `len()` of this range.
pub fn len<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<Int32>> {
let start = self.instance.start(ctx).load(generator, ctx);
let stop = self.instance.stop(ctx).load(generator, ctx);
let step = self.instance.step(ctx).load(generator, ctx);
call_nac3_range_len_i32(generator, ctx, start, stop, step)
}
}

View File

@ -0,0 +1,101 @@
use inkwell::values::StructValue;
use itertools::Itertools;
use crate::{
codegen::{model::*, CodeGenContext, CodeGenerator},
typecheck::typedef::{Type, TypeEnum},
};
use super::any::AnyObject;
/// A NAC3 tuple object.
///
/// NOTE: This struct has no copy trait.
#[derive(Debug, Clone)]
pub struct TupleObject<'ctx> {
/// The type of the tuple.
pub tys: Vec<Type>,
/// The underlying LLVM struct value of this tuple.
pub value: StructValue<'ctx>,
}
impl<'ctx> TupleObject<'ctx> {
pub fn from_object(ctx: &mut CodeGenContext<'ctx, '_>, object: AnyObject<'ctx>) -> Self {
// TODO: Keep `is_vararg_ctx` from TTuple?
assert!(matches!(&*ctx.unifier.get_ty(object.ty), TypeEnum::TTuple { .. }));
// Sanity check on object type.
let TypeEnum::TTuple { ty: tys, .. } = &*ctx.unifier.get_ty(object.ty) else {
panic!(
"Expected type to be a TypeEnum::TTuple, got {}",
ctx.unifier.stringify(object.ty)
);
};
// Check number of fields
let value = object.value.into_struct_value();
let value_num_fields = value.get_type().count_fields() as usize;
assert!(
value_num_fields == tys.len(),
"Tuple type has {} item(s), but the LLVM struct value has {} field(s)",
tys.len(),
value_num_fields
);
TupleObject { tys: tys.clone(), value }
}
/// Convenience function. Create a [`TupleObject`] from an iterator of objects.
pub fn from_objects<I, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
objects: I,
) -> Self
where
I: IntoIterator<Item = AnyObject<'ctx>>,
{
let (values, tys): (Vec<_>, Vec<_>) =
objects.into_iter().map(|object| (object.value, object.ty)).unzip();
let llvm_tys = tys.iter().map(|ty| ctx.get_llvm_type(generator, *ty)).collect_vec();
let llvm_tuple_ty = ctx.ctx.struct_type(&llvm_tys, false);
let pllvm_tuple = ctx.builder.build_alloca(llvm_tuple_ty, "tuple").unwrap();
for (i, val) in values.into_iter().enumerate() {
let pval = ctx.builder.build_struct_gep(pllvm_tuple, i as u32, "value").unwrap();
ctx.builder.build_store(pval, val).unwrap();
}
let value = ctx.builder.build_load(pllvm_tuple, "").unwrap().into_struct_value();
TupleObject { tys, value }
}
#[must_use]
pub fn num_elements(&self) -> usize {
self.tys.len()
}
/// Get the `len()` of this tuple.
#[must_use]
pub fn len<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Instance<'ctx, Int<SizeT>> {
Int(SizeT).const_int(generator, ctx.ctx, self.num_elements() as u64)
}
/// Get the `i`-th (0-based) object in this tuple.
pub fn index(&self, ctx: &mut CodeGenContext<'ctx, '_>, i: usize) -> AnyObject<'ctx> {
assert!(
i < self.num_elements(),
"Tuple object with length {} have index {i}",
self.num_elements()
);
let value = ctx.builder.build_extract_value(self.value, i as u32, "tuple[{i}]").unwrap();
let ty = self.tys[i];
AnyObject { ty, value }
}
}

View File

@ -1,19 +1,22 @@
use super::{
super::symbol_resolver::ValueEnum,
expr::destructure_range,
irrt::{handle_slice_indices, list_slice_assignment},
object::{
any::AnyObject,
list::ListObject,
ndarray::{
indexing::util::gen_ndarray_subscript_ndindices, NDArrayObject, ScalarOrNDArray,
},
range::RangeObject,
},
CodeGenContext, CodeGenerator,
};
use crate::{
codegen::{
classes::{ArrayLikeIndexer, ArraySliceValue, ListValue, RangeValue},
expr::gen_binop_expr,
gen_in_range_check,
},
toplevel::{helper::PrimDef, numpy::unpack_ndarray_var_tys, DefinitionId, TopLevelDef},
codegen::{classes::ArraySliceValue, expr::gen_binop_expr, gen_in_range_check, model::*},
toplevel::{DefinitionId, TopLevelDef},
typecheck::{
magic_methods::Binop,
typedef::{FunSignature, Type, TypeEnum},
typedef::{iter_type_vars, FunSignature, Type, TypeEnum},
},
};
use inkwell::{
@ -23,10 +26,10 @@ use inkwell::{
values::{BasicValue, BasicValueEnum, FunctionValue, IntValue, PointerValue},
IntPredicate,
};
use itertools::{izip, Itertools};
use nac3parser::ast::{
Constant, ExcepthandlerKind, Expr, ExprKind, Location, Stmt, StmtKind, StrRef,
};
use std::convert::TryFrom;
/// See [`CodeGenerator::gen_var_alloc`].
pub fn gen_var<'ctx>(
@ -97,8 +100,6 @@ pub fn gen_store_target<'ctx, G: CodeGenerator>(
pattern: &Expr<Option<Type>>,
name: Option<&str>,
) -> Result<Option<PointerValue<'ctx>>, String> {
let llvm_usize = generator.get_size_type(ctx.ctx);
// very similar to gen_expr, but we don't do an extra load at the end
// and we flatten nested tuples
Ok(Some(match &pattern.node {
@ -137,65 +138,6 @@ pub fn gen_store_target<'ctx, G: CodeGenerator>(
}
.unwrap()
}
ExprKind::Subscript { value, slice, .. } => {
match ctx.unifier.get_ty_immutable(value.custom.unwrap()).as_ref() {
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::List.id() => {
let v = generator
.gen_expr(ctx, value)?
.unwrap()
.to_basic_value_enum(ctx, generator, value.custom.unwrap())?
.into_pointer_value();
let v = ListValue::from_ptr_val(v, llvm_usize, None);
let len = v.load_size(ctx, Some("len"));
let raw_index = generator
.gen_expr(ctx, slice)?
.unwrap()
.to_basic_value_enum(ctx, generator, slice.custom.unwrap())?
.into_int_value();
let raw_index = ctx
.builder
.build_int_s_extend(raw_index, generator.get_size_type(ctx.ctx), "sext")
.unwrap();
// handle negative index
let is_negative = ctx
.builder
.build_int_compare(
IntPredicate::SLT,
raw_index,
generator.get_size_type(ctx.ctx).const_zero(),
"is_neg",
)
.unwrap();
let adjusted = ctx.builder.build_int_add(raw_index, len, "adjusted").unwrap();
let index = ctx
.builder
.build_select(is_negative, adjusted, raw_index, "index")
.map(BasicValueEnum::into_int_value)
.unwrap();
// unsigned less than is enough, because negative index after adjustment is
// bigger than the length (for unsigned cmp)
let bound_check = ctx
.builder
.build_int_compare(IntPredicate::ULT, index, len, "inbound")
.unwrap();
ctx.make_assert(
generator,
bound_check,
"0:IndexError",
"index {0} out of bounds 0:{1}",
[Some(raw_index), Some(len), None],
slice.location,
);
v.data().ptr_offset(ctx, generator, &index, name)
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
todo!()
}
_ => unreachable!(),
}
}
_ => unreachable!(),
}))
}
@ -206,70 +148,20 @@ pub fn gen_assign<'ctx, G: CodeGenerator>(
ctx: &mut CodeGenContext<'ctx, '_>,
target: &Expr<Option<Type>>,
value: ValueEnum<'ctx>,
value_ty: Type,
) -> Result<(), String> {
let llvm_usize = generator.get_size_type(ctx.ctx);
// See https://docs.python.org/3/reference/simple_stmts.html#assignment-statements.
match &target.node {
ExprKind::Tuple { elts, .. } => {
let BasicValueEnum::StructValue(v) =
value.to_basic_value_enum(ctx, generator, target.custom.unwrap())?
else {
unreachable!()
};
for (i, elt) in elts.iter().enumerate() {
let v = ctx
.builder
.build_extract_value(v, u32::try_from(i).unwrap(), "struct_elem")
.unwrap();
generator.gen_assign(ctx, elt, v.into())?;
ExprKind::Subscript { value: target, slice: key, .. } => {
// Handle "slicing" or "subscription"
generator.gen_setitem(ctx, target, key, value, value_ty)?;
}
}
ExprKind::Subscript { value: ls, slice, .. }
if matches!(&slice.node, ExprKind::Slice { .. }) =>
{
let ExprKind::Slice { lower, upper, step } = &slice.node else { unreachable!() };
let ls = generator
.gen_expr(ctx, ls)?
.unwrap()
.to_basic_value_enum(ctx, generator, ls.custom.unwrap())?
.into_pointer_value();
let ls = ListValue::from_ptr_val(ls, llvm_usize, None);
let Some((start, end, step)) =
handle_slice_indices(lower, upper, step, ctx, generator, ls.load_size(ctx, None))?
else {
return Ok(());
};
let value = value
.to_basic_value_enum(ctx, generator, target.custom.unwrap())?
.into_pointer_value();
let value = ListValue::from_ptr_val(value, llvm_usize, None);
let ty = match &*ctx.unifier.get_ty_immutable(target.custom.unwrap()) {
TypeEnum::TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
*params.iter().next().unwrap().1
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
unpack_ndarray_var_tys(&mut ctx.unifier, target.custom.unwrap()).0
}
_ => unreachable!(),
};
let ty = ctx.get_llvm_type(generator, ty);
let Some(src_ind) = handle_slice_indices(
&None,
&None,
&None,
ctx,
generator,
value.load_size(ctx, None),
)?
else {
return Ok(());
};
list_slice_assignment(generator, ctx, ty, ls, (start, end, step), value, src_ind);
ExprKind::Tuple { elts, .. } | ExprKind::List { elts, .. } => {
// Fold on `"[" [target_list] "]"` and `"(" [target_list] ")"`
generator.gen_assign_target_list(ctx, elts, value, value_ty)?;
}
_ => {
// Handle attribute and direct variable assignments.
let name = if let ExprKind::Name { id, .. } = &target.node {
format!("{id}.addr")
} else {
@ -293,6 +185,269 @@ pub fn gen_assign<'ctx, G: CodeGenerator>(
Ok(())
}
/// See [`CodeGenerator::gen_assign_target_list`].
pub fn gen_assign_target_list<'ctx, G: CodeGenerator>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
targets: &Vec<Expr<Option<Type>>>,
value: ValueEnum<'ctx>,
value_ty: Type,
) -> Result<(), String> {
// Deconstruct the tuple `value`
let BasicValueEnum::StructValue(tuple) = value.to_basic_value_enum(ctx, generator, value_ty)?
else {
unreachable!()
};
// NOTE: Currently, RHS's type is forced to be a Tuple by the type inferencer.
let TypeEnum::TTuple { ty: tuple_tys, .. } = &*ctx.unifier.get_ty(value_ty) else {
unreachable!();
};
assert_eq!(tuple.get_type().count_fields() as usize, tuple_tys.len());
let tuple = (0..tuple.get_type().count_fields())
.map(|i| ctx.builder.build_extract_value(tuple, i, "item").unwrap())
.collect_vec();
// Find the starred target if it exists.
let mut starred_target_index: Option<usize> = None; // Index of the "starred" target. If it exists, there may only be one.
for (i, target) in targets.iter().enumerate() {
if matches!(target.node, ExprKind::Starred { .. }) {
assert!(starred_target_index.is_none()); // The typechecker ensures this
starred_target_index = Some(i);
}
}
if let Some(starred_target_index) = starred_target_index {
assert!(tuple_tys.len() >= targets.len() - 1); // The typechecker ensures this
let a = starred_target_index; // Number of RHS values before the starred target
let b = tuple_tys.len() - (targets.len() - 1 - starred_target_index); // Number of RHS values after the starred target
// Thus `tuple[a..b]` is assigned to the starred target.
// Handle assignment before the starred target
for (target, val, val_ty) in
izip!(&targets[..starred_target_index], &tuple[..a], &tuple_tys[..a])
{
generator.gen_assign(ctx, target, ValueEnum::Dynamic(*val), *val_ty)?;
}
// Handle assignment to the starred target
if let ExprKind::Starred { value: target, .. } = &targets[starred_target_index].node {
let vals = &tuple[a..b];
let val_tys = &tuple_tys[a..b];
// Create a sub-tuple from `value` for the starred target.
let sub_tuple_ty = ctx
.ctx
.struct_type(&vals.iter().map(BasicValueEnum::get_type).collect_vec(), false);
let psub_tuple_val =
ctx.builder.build_alloca(sub_tuple_ty, "starred_target_value_ptr").unwrap();
for (i, val) in vals.iter().enumerate() {
let pitem = ctx
.builder
.build_struct_gep(psub_tuple_val, i as u32, "starred_target_value_item")
.unwrap();
ctx.builder.build_store(pitem, *val).unwrap();
}
let sub_tuple_val =
ctx.builder.build_load(psub_tuple_val, "starred_target_value").unwrap();
// Create the typechecker type of the sub-tuple
let sub_tuple_ty =
ctx.unifier.add_ty(TypeEnum::TTuple { ty: val_tys.to_vec(), is_vararg_ctx: false });
// Now assign with that sub-tuple to the starred target.
generator.gen_assign(ctx, target, ValueEnum::Dynamic(sub_tuple_val), sub_tuple_ty)?;
} else {
unreachable!() // The typechecker ensures this
}
// Handle assignment after the starred target
for (target, val, val_ty) in
izip!(&targets[starred_target_index + 1..], &tuple[b..], &tuple_tys[b..])
{
generator.gen_assign(ctx, target, ValueEnum::Dynamic(*val), *val_ty)?;
}
} else {
assert_eq!(tuple_tys.len(), targets.len()); // The typechecker ensures this
for (target, val, val_ty) in izip!(targets, tuple, tuple_tys) {
generator.gen_assign(ctx, target, ValueEnum::Dynamic(val), *val_ty)?;
}
}
Ok(())
}
/// See [`CodeGenerator::gen_setitem`].
pub fn gen_setitem<'ctx, G: CodeGenerator>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
target: &Expr<Option<Type>>,
key: &Expr<Option<Type>>,
value: ValueEnum<'ctx>,
value_ty: Type,
) -> Result<(), String> {
let target_ty = target.custom.unwrap();
let key_ty = key.custom.unwrap();
match &*ctx.unifier.get_ty(target_ty) {
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.list.obj_id(&ctx.unifier).unwrap() =>
{
// Handle list item assignment
let target = generator
.gen_expr(ctx, target)?
.unwrap()
.to_basic_value_enum(ctx, generator, target_ty)?;
let target = AnyObject { ty: target_ty, value: target };
let target = ListObject::from_object(generator, ctx, target);
let target_len = target.instance.get(generator, ctx, |f| f.len);
let target_item_type_llvm = ctx.get_llvm_type(generator, target.item_type);
if let ExprKind::Slice { .. } = &key.node {
// Handle assigning to a slice
let ExprKind::Slice { lower, upper, step } = &key.node else { unreachable!() };
let Some((start, end, step)) =
handle_slice_indices(lower, upper, step, ctx, generator, target_len.value)?
else {
return Ok(());
};
let value = value.to_basic_value_enum(ctx, generator, value_ty)?;
let value = AnyObject { ty: value_ty, value };
let value = ListObject::from_object(generator, ctx, value);
let value_len = value.instance.get(generator, ctx, |f| f.len);
let Some(src_ind) =
handle_slice_indices(&None, &None, &None, ctx, generator, value_len.value)?
else {
return Ok(());
};
list_slice_assignment(
generator,
ctx,
target_item_type_llvm,
target.instance,
(start, end, step),
value.instance,
src_ind,
);
} else {
// Handle assigning to an index
let value = value.to_basic_value_enum(ctx, generator, value_ty)?;
let value =
Any(target_item_type_llvm).check_value(generator, ctx.ctx, value).unwrap();
let index = generator
.gen_expr(ctx, key)?
.unwrap()
.to_basic_value_enum(ctx, generator, key_ty)?
.into_int_value();
let index = ctx
.builder
.build_int_s_extend(index, generator.get_size_type(ctx.ctx), "sext")
.unwrap();
// handle negative index
let is_negative = ctx
.builder
.build_int_compare(
IntPredicate::SLT,
index,
generator.get_size_type(ctx.ctx).const_zero(),
"is_neg",
)
.unwrap();
let adjusted =
ctx.builder.build_int_add(index, target_len.value, "adjusted").unwrap();
let index = ctx
.builder
.build_select(is_negative, adjusted, index, "index")
.map(BasicValueEnum::into_int_value)
.unwrap();
// unsigned less than is enough, because negative index after adjustment is
// bigger than the length (for unsigned cmp)
let bound_check = ctx
.builder
.build_int_compare(IntPredicate::ULT, index, target_len.value, "inbound")
.unwrap();
ctx.make_assert(
generator,
bound_check,
"0:IndexError",
"index {0} out of bounds 0:{1}",
[Some(index), Some(target_len.value), None],
key.location,
);
// Write value to index on list
target
.instance
.get(generator, ctx, |f| f.items)
.offset(ctx, index)
.store(ctx, value);
}
}
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.ndarray.obj_id(&ctx.unifier).unwrap() =>
{
// Handle NDArray item assignment
// Process target
let target = generator
.gen_expr(ctx, target)?
.unwrap()
.to_basic_value_enum(ctx, generator, target_ty)?;
let target = AnyObject { value: target, ty: target_ty };
// Process key
let key = gen_ndarray_subscript_ndindices(generator, ctx, key)?;
// Process value
let value = value.to_basic_value_enum(ctx, generator, value_ty)?;
let value = AnyObject { value, ty: value_ty };
/*
Reference code:
```python
target = target[key]
value = np.asarray(value)
shape = np.broadcast_shape((target, value))
target = np.broadcast_to(target, shape)
value = np.broadcast_to(value, shape)
...and finally copy 1-1 from value to target.
```
*/
let target = NDArrayObject::from_object(generator, ctx, target);
let target = target.index(generator, ctx, &key);
let value =
ScalarOrNDArray::split_object(generator, ctx, value).to_ndarray(generator, ctx);
let broadcast_result = NDArrayObject::broadcast(generator, ctx, &[target, value]);
let target = broadcast_result.ndarrays[0];
let value = broadcast_result.ndarrays[1];
target.copy_data_from(generator, ctx, value);
}
_ => {
panic!("encountered unknown target type: {}", ctx.unifier.stringify(target_ty));
}
}
Ok(())
}
/// See [`CodeGenerator::gen_for`].
pub fn gen_for<G: CodeGenerator>(
generator: &mut G,
@ -315,9 +470,6 @@ pub fn gen_for<G: CodeGenerator>(
let orelse_bb =
if orelse.is_empty() { cont_bb } else { ctx.ctx.append_basic_block(current, "for.orelse") };
// Whether the iterable is a range() expression
let is_iterable_range_expr = ctx.unifier.unioned(iter.custom.unwrap(), ctx.primitives.range);
// The BB containing the increment expression
let incr_bb = ctx.ctx.append_basic_block(current, "for.incr");
// The BB containing the loop condition check
@ -326,27 +478,41 @@ pub fn gen_for<G: CodeGenerator>(
// store loop bb information and restore it later
let loop_bb = ctx.loop_target.replace((incr_bb, cont_bb));
let iter_ty = iter.custom.unwrap();
let iter_val = if let Some(v) = generator.gen_expr(ctx, iter)? {
v.to_basic_value_enum(ctx, generator, iter.custom.unwrap())?
v.to_basic_value_enum(ctx, generator, iter_ty)?
} else {
return Ok(());
};
if is_iterable_range_expr {
let iter_val = RangeValue::from_ptr_val(iter_val.into_pointer_value(), Some("range"));
match &*ctx.unifier.get_ty(iter_ty) {
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.range.obj_id(&ctx.unifier).unwrap() =>
{
let range = AnyObject { value: iter_val, ty: iter_ty };
let range = RangeObject::from_object(generator, ctx, range);
let (start, stop, step) = range.instance.destructure(generator, ctx);
let start = start.value;
let stop = stop.value;
let step = step.value;
// Internal variable for loop; Cannot be assigned
let i = generator.gen_var_alloc(ctx, int32.into(), Some("for.i.addr"))?;
// Variable declared in "target" expression of the loop; Can be reassigned *or* shadowed
let Some(target_i) = generator.gen_store_target(ctx, target, Some("for.target.addr"))?
let Some(target_i) =
generator.gen_store_target(ctx, target, Some("for.target.addr"))?
else {
unreachable!()
};
let (start, stop, step) = destructure_range(ctx, iter_val);
ctx.builder.build_store(i, start).unwrap();
// Check "If step is zero, ValueError is raised."
let rangenez =
ctx.builder.build_int_compare(IntPredicate::NE, step, int32.const_zero(), "").unwrap();
let rangenez = ctx
.builder
.build_int_compare(IntPredicate::NE, step, int32.const_zero(), "")
.unwrap();
ctx.make_assert(
generator,
rangenez,
@ -363,7 +529,10 @@ pub fn gen_for<G: CodeGenerator>(
.build_conditional_branch(
gen_in_range_check(
ctx,
ctx.builder.build_load(i, "").map(BasicValueEnum::into_int_value).unwrap(),
ctx.builder
.build_load(i, "")
.map(BasicValueEnum::into_int_value)
.unwrap(),
stop,
step,
),
@ -393,7 +562,10 @@ pub fn gen_for<G: CodeGenerator>(
)
.unwrap();
generator.gen_block(ctx, body.iter())?;
} else {
}
TypeEnum::TObj { obj_id, params: list_params, .. }
if *obj_id == ctx.primitives.list.obj_id(&ctx.unifier).unwrap() =>
{
let index_addr = generator.gen_var_alloc(ctx, size_t.into(), Some("for.index.addr"))?;
ctx.builder.build_store(index_addr, size_t.const_zero()).unwrap();
let len = ctx
@ -431,9 +603,14 @@ pub fn gen_for<G: CodeGenerator>(
.map(BasicValueEnum::into_int_value)
.unwrap();
let val = ctx.build_gep_and_load(arr_ptr, &[index], Some("val"));
generator.gen_assign(ctx, target, val.into())?;
let val_ty = iter_type_vars(list_params).next().unwrap().ty;
generator.gen_assign(ctx, target, val.into(), val_ty)?;
generator.gen_block(ctx, body.iter())?;
}
_ => {
panic!("unsupported for loop iterator type: {}", ctx.unifier.stringify(iter_ty));
}
}
for (k, (_, _, counter)) in &var_assignment {
let (_, static_val, counter2) = ctx.var_assignment.get_mut(k).unwrap();
@ -494,6 +671,7 @@ pub struct BreakContinueHooks<'ctx> {
pub fn gen_for_callback<'ctx, 'a, G, I, InitFn, CondFn, BodyFn, UpdateFn>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
label: Option<&str>,
init: InitFn,
cond: CondFn,
body: BodyFn,
@ -504,18 +682,24 @@ where
I: Clone,
InitFn: FnOnce(&mut G, &mut CodeGenContext<'ctx, 'a>) -> Result<I, String>,
CondFn: FnOnce(&mut G, &mut CodeGenContext<'ctx, 'a>, I) -> Result<IntValue<'ctx>, String>,
BodyFn:
FnOnce(&mut G, &mut CodeGenContext<'ctx, 'a>, BreakContinueHooks, I) -> Result<(), String>,
BodyFn: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
BreakContinueHooks<'ctx>,
I,
) -> Result<(), String>,
UpdateFn: FnOnce(&mut G, &mut CodeGenContext<'ctx, 'a>, I) -> Result<(), String>,
{
let label = label.unwrap_or("for");
let current_bb = ctx.builder.get_insert_block().unwrap();
let init_bb = ctx.ctx.insert_basic_block_after(current_bb, "for.init");
let init_bb = ctx.ctx.insert_basic_block_after(current_bb, &format!("{label}.init"));
// The BB containing the loop condition check
let cond_bb = ctx.ctx.insert_basic_block_after(init_bb, "for.cond");
let body_bb = ctx.ctx.insert_basic_block_after(cond_bb, "for.body");
let cond_bb = ctx.ctx.insert_basic_block_after(init_bb, &format!("{label}.cond"));
let body_bb = ctx.ctx.insert_basic_block_after(cond_bb, &format!("{label}.body"));
// The BB containing the increment expression
let update_bb = ctx.ctx.insert_basic_block_after(body_bb, "for.update");
let cont_bb = ctx.ctx.insert_basic_block_after(update_bb, "for.end");
let update_bb = ctx.ctx.insert_basic_block_after(body_bb, &format!("{label}.update"));
let cont_bb = ctx.ctx.insert_basic_block_after(update_bb, &format!("{label}.end"));
// store loop bb information and restore it later
let loop_bb = ctx.loop_target.replace((update_bb, cont_bb));
@ -572,6 +756,7 @@ where
pub fn gen_for_callback_incrementing<'ctx, 'a, G, BodyFn>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
label: Option<&str>,
init_val: IntValue<'ctx>,
max_val: (IntValue<'ctx>, bool),
body: BodyFn,
@ -582,7 +767,7 @@ where
BodyFn: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
BreakContinueHooks,
BreakContinueHooks<'ctx>,
IntValue<'ctx>,
) -> Result<(), String>,
{
@ -591,6 +776,7 @@ where
gen_for_callback(
generator,
ctx,
label,
|generator, ctx| {
let i_addr = generator.gen_var_alloc(ctx, init_val_t.into(), None)?;
ctx.builder.build_store(i_addr, init_val).unwrap();
@ -642,9 +828,11 @@ where
/// - `step_fn`: A lambda of IR statements that retrieves the `step` value of the `range`-like
/// iterable. This value will be extended to the size of `start`.
/// - `body_fn`: A lambda of IR statements within the loop body.
#[allow(clippy::too_many_arguments)]
pub fn gen_for_range_callback<'ctx, 'a, G, StartFn, StopFn, StepFn, BodyFn>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
label: Option<&str>,
is_unsigned: bool,
start_fn: StartFn,
(stop_fn, stop_inclusive): (StopFn, bool),
@ -656,13 +844,19 @@ where
StartFn: Fn(&mut G, &mut CodeGenContext<'ctx, 'a>) -> Result<IntValue<'ctx>, String>,
StopFn: Fn(&mut G, &mut CodeGenContext<'ctx, 'a>) -> Result<IntValue<'ctx>, String>,
StepFn: Fn(&mut G, &mut CodeGenContext<'ctx, 'a>) -> Result<IntValue<'ctx>, String>,
BodyFn: FnOnce(&mut G, &mut CodeGenContext<'ctx, 'a>, IntValue<'ctx>) -> Result<(), String>,
BodyFn: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
BreakContinueHooks<'ctx>,
IntValue<'ctx>,
) -> Result<(), String>,
{
let init_val_t = start_fn(generator, ctx).map(IntValue::get_type).unwrap();
gen_for_callback(
generator,
ctx,
label,
|generator, ctx| {
let i_addr = generator.gen_var_alloc(ctx, init_val_t.into(), None)?;
@ -720,10 +914,10 @@ where
Ok(cond)
},
|generator, ctx, _, (i_addr, _)| {
|generator, ctx, hooks, (i_addr, _)| {
let i = ctx.builder.build_load(i_addr, "").map(BasicValueEnum::into_int_value).unwrap();
body_fn(generator, ctx, i)
body_fn(generator, ctx, hooks, i)
},
|generator, ctx, (i_addr, _)| {
let i = ctx.builder.build_load(i_addr, "").map(BasicValueEnum::into_int_value).unwrap();
@ -1575,14 +1769,14 @@ pub fn gen_stmt<G: CodeGenerator>(
}
StmtKind::AnnAssign { target, value, .. } => {
if let Some(value) = value {
let Some(value) = generator.gen_expr(ctx, value)? else { return Ok(()) };
generator.gen_assign(ctx, target, value)?;
let Some(value_enum) = generator.gen_expr(ctx, value)? else { return Ok(()) };
generator.gen_assign(ctx, target, value_enum, value.custom.unwrap())?;
}
}
StmtKind::Assign { targets, value, .. } => {
let Some(value) = generator.gen_expr(ctx, value)? else { return Ok(()) };
let Some(value_enum) = generator.gen_expr(ctx, value)? else { return Ok(()) };
for target in targets {
generator.gen_assign(ctx, target, value.clone())?;
generator.gen_assign(ctx, target, value_enum.clone(), value.custom.unwrap())?;
}
}
StmtKind::Continue { .. } => {
@ -1596,15 +1790,16 @@ pub fn gen_stmt<G: CodeGenerator>(
StmtKind::For { .. } => generator.gen_for(ctx, stmt)?,
StmtKind::With { .. } => generator.gen_with(ctx, stmt)?,
StmtKind::AugAssign { target, op, value, .. } => {
let value = gen_binop_expr(
let value_enum = gen_binop_expr(
generator,
ctx,
target,
Binop::aug_assign(*op),
value,
stmt.location,
)?;
generator.gen_assign(ctx, target, value.unwrap())?;
)?
.unwrap();
generator.gen_assign(ctx, target, value_enum, value.custom.unwrap())?;
}
StmtKind::Try { .. } => gen_try(generator, ctx, stmt)?,
StmtKind::Raise { exc, .. } => {
@ -1633,11 +1828,11 @@ pub fn gen_stmt<G: CodeGenerator>(
return Ok(());
}
}
None => ctx.gen_string(generator, ""),
None => ctx.gen_string(generator, "").into(),
};
ctx.make_assert_impl(
generator,
test.into_int_value(),
generator.bool_to_i1(ctx, test.into_int_value()),
"0:AssertionError",
err_msg,
[None, None, None],

View File

@ -1,9 +1,7 @@
use crate::{
codegen::{
classes::{ListType, NDArrayType, ProxyType, RangeType},
concrete_type::ConcreteTypeStore,
CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask,
CodeGenerator, DefaultCodeGenerator, WithCall, WorkerRegistry,
concrete_type::ConcreteTypeStore, CodeGenContext, CodeGenLLVMOptions,
CodeGenTargetMachineOptions, CodeGenTask, DefaultCodeGenerator, WithCall, WorkerRegistry,
},
symbol_resolver::{SymbolResolver, ValueEnum},
toplevel::{
@ -94,7 +92,7 @@ fn test_primitives() {
"};
let statements = parse_program(source, FileName::default()).unwrap();
let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
let composer = TopLevelComposer::new(Vec::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());
@ -109,8 +107,18 @@ fn test_primitives() {
let threads = vec![DefaultCodeGenerator::new("test".into(), 32).into()];
let signature = FunSignature {
args: vec![
FuncArg { name: "a".into(), ty: primitives.int32, default_value: None },
FuncArg { name: "b".into(), ty: primitives.int32, default_value: None },
FuncArg {
name: "a".into(),
ty: primitives.int32,
default_value: None,
is_vararg: false,
},
FuncArg {
name: "b".into(),
ty: primitives.int32,
default_value: None,
is_vararg: false,
},
],
ret: primitives.int32,
vars: VarMap::new(),
@ -189,6 +197,8 @@ fn test_primitives() {
let expected = indoc! {"
; ModuleID = 'test'
source_filename = \"test\"
target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"
target triple = \"x86_64-unknown-linux-gnu\"
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
define i32 @testing(i32 %0, i32 %1) local_unnamed_addr #0 !dbg !4 {
@ -246,14 +256,19 @@ fn test_simple_call() {
"};
let statements_2 = parse_program(source_2, FileName::default()).unwrap();
let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
let composer = TopLevelComposer::new(Vec::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());
unifier.top_level = Some(top_level.clone());
let signature = FunSignature {
args: vec![FuncArg { name: "a".into(), ty: primitives.int32, default_value: None }],
args: vec![FuncArg {
name: "a".into(),
ty: primitives.int32,
default_value: None,
is_vararg: false,
}],
ret: primitives.int32,
vars: VarMap::new(),
};
@ -368,6 +383,8 @@ fn test_simple_call() {
let expected = indoc! {"
; ModuleID = 'test'
source_filename = \"test\"
target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"
target triple = \"x86_64-unknown-linux-gnu\"
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
define i32 @testing(i32 %0) local_unnamed_addr #0 !dbg !5 {
@ -417,35 +434,3 @@ fn test_simple_call() {
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

@ -78,14 +78,14 @@ impl SymbolValue {
}
Constant::Tuple(t) => {
let expected_ty = unifier.get_ty(expected_ty);
let TypeEnum::TTuple { ty } = expected_ty.as_ref() else {
let TypeEnum::TTuple { ty, is_vararg_ctx } = expected_ty.as_ref() else {
return Err(format!(
"Expected {:?}, but got Tuple",
expected_ty.get_type_name()
));
};
assert_eq!(ty.len(), t.len());
assert!(*is_vararg_ctx || ty.len() == t.len());
let elems = t
.iter()
@ -155,7 +155,7 @@ impl SymbolValue {
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 })
unifier.add_ty(TypeEnum::TTuple { ty: vs_tys, is_vararg_ctx: false })
}
SymbolValue::OptionSome(_) | SymbolValue::OptionNone => primitives.option,
}
@ -482,7 +482,7 @@ pub fn parse_type_annotation<T>(
parse_type_annotation(resolver, top_level_defs, unifier, primitives, elt)
})
.collect::<Result<Vec<_>, _>>()?;
Ok(unifier.add_ty(TypeEnum::TTuple { ty }))
Ok(unifier.add_ty(TypeEnum::TTuple { ty, is_vararg_ctx: false }))
} else {
Err(HashSet::from(["Expected multiple elements for tuple".into()]))
}

View File

@ -1,6 +1,6 @@
use std::iter::once;
use helper::{debug_assert_prim_is_allowed, make_exception_fields, PrimDefDetails};
use helper::{debug_assert_prim_is_allowed, extract_ndims, make_exception_fields, PrimDefDetails};
use indexmap::IndexMap;
use inkwell::{
attributes::{Attribute, AttributeLoc},
@ -9,15 +9,19 @@ use inkwell::{
IntPredicate,
};
use itertools::Either;
use numpy::unpack_ndarray_var_tys;
use strum::IntoEnumIterator;
use crate::{
codegen::{
builtin_fns,
classes::{ArrayLikeValue, NDArrayValue, ProxyValue, RangeValue, TypedArrayLikeAccessor},
expr::destructure_range,
irrt::*,
model::*,
numpy::*,
object::{
any::AnyObject,
ndarray::{shape_util::parse_numpy_int_sequence, NDArrayObject},
range::RangeObject,
},
stmt::exn_constructor,
},
symbol_resolver::SymbolValue,
@ -45,10 +49,26 @@ pub fn get_exn_constructor(
name: "msg".into(),
ty: string,
default_value: Some(SymbolValue::Str(String::new())),
is_vararg: false,
},
FuncArg {
name: "param0".into(),
ty: int64,
default_value: Some(SymbolValue::I64(0)),
is_vararg: false,
},
FuncArg {
name: "param1".into(),
ty: int64,
default_value: Some(SymbolValue::I64(0)),
is_vararg: false,
},
FuncArg {
name: "param2".into(),
ty: int64,
default_value: Some(SymbolValue::I64(0)),
is_vararg: false,
},
FuncArg { name: "param0".into(), ty: int64, default_value: Some(SymbolValue::I64(0)) },
FuncArg { name: "param1".into(), ty: int64, default_value: Some(SymbolValue::I64(0)) },
FuncArg { name: "param2".into(), ty: int64, default_value: Some(SymbolValue::I64(0)) },
];
let exn_type = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(class_id),
@ -114,7 +134,12 @@ fn create_fn_by_codegen(
signature: unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: param_ty
.iter()
.map(|p| FuncArg { name: p.1.into(), ty: p.0, default_value: None })
.map(|p| FuncArg {
name: p.1.into(),
ty: p.0,
default_value: None,
is_vararg: false,
})
.collect(),
ret: ret_ty,
vars: var_map.clone(),
@ -346,8 +371,8 @@ impl<'a> BuiltinBuilder<'a> {
let (is_some_ty, unwrap_ty, option_tvar) =
if let TypeEnum::TObj { fields, params, .. } = unifier.get_ty(option).as_ref() {
(
*fields.get(&PrimDef::OptionIsSome.simple_name().into()).unwrap(),
*fields.get(&PrimDef::OptionUnwrap.simple_name().into()).unwrap(),
*fields.get(&PrimDef::FunOptionIsSome.simple_name().into()).unwrap(),
*fields.get(&PrimDef::FunOptionUnwrap.simple_name().into()).unwrap(),
iter_type_vars(params).next().unwrap(),
)
} else {
@ -362,9 +387,9 @@ impl<'a> BuiltinBuilder<'a> {
let ndarray_dtype_tvar = iter_type_vars(ndarray_params).next().unwrap();
let ndarray_ndims_tvar = iter_type_vars(ndarray_params).nth(1).unwrap();
let ndarray_copy_ty =
*ndarray_fields.get(&PrimDef::NDArrayCopy.simple_name().into()).unwrap();
*ndarray_fields.get(&PrimDef::FunNDArrayCopy.simple_name().into()).unwrap();
let ndarray_fill_ty =
*ndarray_fields.get(&PrimDef::NDArrayFill.simple_name().into()).unwrap();
*ndarray_fields.get(&PrimDef::FunNDArrayFill.simple_name().into()).unwrap();
let num_ty = unifier.get_fresh_var_with_range(
&[int32, int64, float, boolean, uint32, uint64],
@ -464,14 +489,14 @@ impl<'a> BuiltinBuilder<'a> {
PrimDef::Exception => self.build_exception_class_related(prim),
PrimDef::Option
| PrimDef::OptionIsSome
| PrimDef::OptionIsNone
| PrimDef::OptionUnwrap
| PrimDef::FunOptionIsSome
| PrimDef::FunOptionIsNone
| PrimDef::FunOptionUnwrap
| PrimDef::FunSome => self.build_option_class_related(prim),
PrimDef::List => self.build_list_class_related(prim),
PrimDef::NDArray | PrimDef::NDArrayCopy | PrimDef::NDArrayFill => {
PrimDef::NDArray | PrimDef::FunNDArrayCopy | PrimDef::FunNDArrayFill => {
self.build_ndarray_class_related(prim)
}
@ -492,6 +517,14 @@ impl<'a> BuiltinBuilder<'a> {
| PrimDef::FunNpEye
| PrimDef::FunNpIdentity => self.build_ndarray_other_factory_function(prim),
PrimDef::FunNpSize | PrimDef::FunNpShape | PrimDef::FunNpStrides => {
self.build_ndarray_property_getter_function(prim)
}
PrimDef::FunNpBroadcastTo | PrimDef::FunNpTranspose | PrimDef::FunNpReshape => {
self.build_ndarray_view_function(prim)
}
PrimDef::FunStr => self.build_str_function(),
PrimDef::FunFloor | PrimDef::FunFloor64 | PrimDef::FunCeil | PrimDef::FunCeil64 => {
@ -510,7 +543,9 @@ impl<'a> BuiltinBuilder<'a> {
PrimDef::FunMin | PrimDef::FunMax => self.build_min_max_function(prim),
PrimDef::FunNpMin | PrimDef::FunNpMax => self.build_np_min_max_function(prim),
PrimDef::FunNpArgmin | PrimDef::FunNpArgmax | PrimDef::FunNpMin | PrimDef::FunNpMax => {
self.build_np_max_min_function(prim)
}
PrimDef::FunNpMinimum | PrimDef::FunNpMaximum => {
self.build_np_minimum_maximum_function(prim)
@ -554,6 +589,18 @@ impl<'a> BuiltinBuilder<'a> {
| PrimDef::FunNpLdExp
| PrimDef::FunNpHypot
| PrimDef::FunNpNextAfter => self.build_np_2ary_function(prim),
PrimDef::FunNpDot
| PrimDef::FunNpLinalgCholesky
| PrimDef::FunNpLinalgQr
| PrimDef::FunNpLinalgSvd
| PrimDef::FunNpLinalgInv
| PrimDef::FunNpLinalgPinv
| PrimDef::FunNpLinalgMatrixPower
| PrimDef::FunNpLinalgDet
| PrimDef::FunSpLinalgLu
| PrimDef::FunSpLinalgSchur
| PrimDef::FunSpLinalgHessenberg => self.build_linalg_methods(prim),
};
if cfg!(debug_assertions) {
@ -562,7 +609,7 @@ impl<'a> BuiltinBuilder<'a> {
match (&tld, prim.details()) {
(
TopLevelDef::Class { name, object_id, .. },
PrimDefDetails::PrimClass { name: exp_name },
PrimDefDetails::PrimClass { name: exp_name, .. },
) => {
let exp_object_id = prim.id();
assert_eq!(name, &exp_name.into());
@ -611,17 +658,24 @@ impl<'a> BuiltinBuilder<'a> {
let make_ctor_signature = |unifier: &mut Unifier| {
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![
FuncArg { name: "start".into(), ty: int32, default_value: None },
FuncArg {
name: "start".into(),
ty: int32,
default_value: None,
is_vararg: false,
},
FuncArg {
name: "stop".into(),
ty: int32,
// placeholder
default_value: Some(SymbolValue::I32(0)),
is_vararg: false,
},
FuncArg {
name: "step".into(),
ty: int32,
default_value: Some(SymbolValue::I32(1)),
is_vararg: false,
},
],
ret: range,
@ -663,9 +717,10 @@ impl<'a> BuiltinBuilder<'a> {
codegen_callback: Some(Arc::new(GenCall::new(Box::new(
|ctx, obj, _, args, generator| {
let (zelf_ty, zelf) = obj.unwrap();
let zelf =
zelf.to_basic_value_enum(ctx, generator, zelf_ty)?.into_pointer_value();
let zelf = RangeValue::from_ptr_val(zelf, Some("range"));
let zelf = zelf.to_basic_value_enum(ctx, generator, zelf_ty)?;
let zelf = AnyObject { ty: zelf_ty, value: zelf };
let zelf = RangeObject::from_object(generator, ctx, zelf);
let mut start = None;
let mut stop = None;
@ -748,11 +803,14 @@ impl<'a> BuiltinBuilder<'a> {
});
let start = start.unwrap_or_else(|| int32.const_zero());
zelf.store_start(ctx, start);
zelf.store_end(ctx, stop);
zelf.store_step(ctx, step);
let start = Int(Int32).believe_value(start);
let stop = Int(Int32).believe_value(stop);
let step = Int(Int32).believe_value(step);
zelf.instance.start(ctx).store(ctx, start);
zelf.instance.stop(ctx).store(ctx, stop);
zelf.instance.step(ctx).store(ctx, step);
Ok(Some(zelf.as_base_value().into()))
Ok(Some(zelf.instance.value.as_basic_value_enum()))
},
)))),
loc: None,
@ -792,9 +850,9 @@ impl<'a> BuiltinBuilder<'a> {
prim,
&[
PrimDef::Option,
PrimDef::OptionIsSome,
PrimDef::OptionIsNone,
PrimDef::OptionUnwrap,
PrimDef::FunOptionIsSome,
PrimDef::FunOptionIsNone,
PrimDef::FunOptionUnwrap,
PrimDef::FunSome,
],
);
@ -807,9 +865,9 @@ impl<'a> BuiltinBuilder<'a> {
fields: Vec::default(),
attributes: Vec::default(),
methods: vec![
Self::create_method(PrimDef::OptionIsSome, self.is_some_ty.0),
Self::create_method(PrimDef::OptionIsNone, self.is_some_ty.0),
Self::create_method(PrimDef::OptionUnwrap, self.unwrap_ty.0),
Self::create_method(PrimDef::FunOptionIsSome, self.is_some_ty.0),
Self::create_method(PrimDef::FunOptionIsNone, self.is_some_ty.0),
Self::create_method(PrimDef::FunOptionUnwrap, self.unwrap_ty.0),
],
ancestors: vec![TypeAnnotation::CustomClass {
id: prim.id(),
@ -820,7 +878,7 @@ impl<'a> BuiltinBuilder<'a> {
loc: None,
},
PrimDef::OptionUnwrap => TopLevelDef::Function {
PrimDef::FunOptionUnwrap => TopLevelDef::Function {
name: prim.name().into(),
simple_name: prim.simple_name().into(),
signature: self.unwrap_ty.0,
@ -834,7 +892,7 @@ impl<'a> BuiltinBuilder<'a> {
loc: None,
},
PrimDef::OptionIsNone | PrimDef::OptionIsSome => TopLevelDef::Function {
PrimDef::FunOptionIsNone | PrimDef::FunOptionIsSome => TopLevelDef::Function {
name: prim.name().to_string(),
simple_name: prim.simple_name().into(),
signature: self.is_some_ty.0,
@ -855,10 +913,10 @@ impl<'a> BuiltinBuilder<'a> {
};
let returned_int = match prim {
PrimDef::OptionIsNone => {
PrimDef::FunOptionIsNone => {
ctx.builder.build_is_null(ptr, prim.simple_name())
}
PrimDef::OptionIsSome => {
PrimDef::FunOptionIsSome => {
ctx.builder.build_is_not_null(ptr, prim.simple_name())
}
_ => unreachable!(),
@ -877,6 +935,7 @@ impl<'a> BuiltinBuilder<'a> {
name: "n".into(),
ty: self.option_tvar.ty,
default_value: None,
is_vararg: false,
}],
ret: self.primitives.option,
vars: into_var_map([self.option_tvar]),
@ -931,7 +990,7 @@ impl<'a> BuiltinBuilder<'a> {
fn build_ndarray_class_related(&self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(
prim,
&[PrimDef::NDArray, PrimDef::NDArrayCopy, PrimDef::NDArrayFill],
&[PrimDef::NDArray, PrimDef::FunNDArrayCopy, PrimDef::FunNDArrayFill],
);
match prim {
@ -942,8 +1001,8 @@ impl<'a> BuiltinBuilder<'a> {
fields: Vec::default(),
attributes: Vec::default(),
methods: vec![
Self::create_method(PrimDef::NDArrayCopy, self.ndarray_copy_ty.0),
Self::create_method(PrimDef::NDArrayFill, self.ndarray_fill_ty.0),
Self::create_method(PrimDef::FunNDArrayCopy, self.ndarray_copy_ty.0),
Self::create_method(PrimDef::FunNDArrayFill, self.ndarray_fill_ty.0),
],
ancestors: Vec::default(),
constructor: None,
@ -951,7 +1010,7 @@ impl<'a> BuiltinBuilder<'a> {
loc: None,
},
PrimDef::NDArrayCopy => TopLevelDef::Function {
PrimDef::FunNDArrayCopy => TopLevelDef::Function {
name: prim.name().into(),
simple_name: prim.simple_name().into(),
signature: self.ndarray_copy_ty.0,
@ -968,7 +1027,7 @@ impl<'a> BuiltinBuilder<'a> {
loc: None,
},
PrimDef::NDArrayFill => TopLevelDef::Function {
PrimDef::FunNDArrayFill => TopLevelDef::Function {
name: prim.name().into(),
simple_name: prim.simple_name().into(),
signature: self.ndarray_fill_ty.0,
@ -1011,6 +1070,7 @@ impl<'a> BuiltinBuilder<'a> {
name: "n".into(),
ty: self.num_or_ndarray_ty.ty,
default_value: None,
is_vararg: false,
}],
ret: self.num_or_ndarray_ty.ty,
vars: self.num_or_ndarray_var_map.clone(),
@ -1230,16 +1290,23 @@ impl<'a> BuiltinBuilder<'a> {
simple_name: prim.simple_name().into(),
signature: self.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![
FuncArg { name: "object".into(), ty: tv.ty, default_value: None },
FuncArg {
name: "object".into(),
ty: tv.ty,
default_value: None,
is_vararg: false,
},
FuncArg {
name: "copy".into(),
ty: bool,
default_value: Some(SymbolValue::Bool(true)),
is_vararg: false,
},
FuncArg {
name: "ndmin".into(),
ty: int32,
default_value: Some(SymbolValue::U32(0)),
is_vararg: false,
},
],
ret: ndarray,
@ -1281,17 +1348,24 @@ impl<'a> BuiltinBuilder<'a> {
simple_name: prim.simple_name().into(),
signature: self.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![
FuncArg { name: "N".into(), ty: int32, default_value: None },
FuncArg {
name: "N".into(),
ty: int32,
default_value: None,
is_vararg: false,
},
// TODO(Derppening): Default values current do not work?
FuncArg {
name: "M".into(),
ty: int32,
default_value: Some(SymbolValue::OptionNone),
is_vararg: false,
},
FuncArg {
name: "k".into(),
ty: int32,
default_value: Some(SymbolValue::I32(0)),
is_vararg: false,
},
],
ret: self.ndarray_float_2d,
@ -1325,6 +1399,169 @@ impl<'a> BuiltinBuilder<'a> {
}
}
fn build_ndarray_property_getter_function(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(
prim,
&[PrimDef::FunNpSize, PrimDef::FunNpShape, PrimDef::FunNpStrides],
);
let mut var_map = self.num_var_map.clone();
var_map.insert(self.ndarray_dtype_tvar.id, self.ndarray_dtype_tvar.ty);
let in_ndarray_ty = self.unifier.get_fresh_var_with_range(
&[self.primitives.ndarray],
Some("T".into()),
None,
);
match prim {
PrimDef::FunNpSize => create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
self.primitives.int32,
&[(in_ndarray_ty.ty, "a")],
Box::new(|ctx, obj, fun, args, generator| {
assert!(obj.is_none());
assert_eq!(args.len(), 1);
let ndarray_ty = fun.0.args[0].ty;
let ndarray =
args[0].1.clone().to_basic_value_enum(ctx, generator, ndarray_ty)?;
let ndarray = AnyObject { ty: ndarray_ty, value: ndarray };
let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
let size = ndarray.size(generator, ctx).truncate(generator, ctx, Int32);
Ok(Some(size.value.as_basic_value_enum()))
}),
),
PrimDef::FunNpShape | PrimDef::FunNpStrides => {
// The fnuction signatures of `np_shape` an `np_size` are the same.
// The return type is a tuple of variable length depending on the ndims of the input ndarray.
let ret_ty = self.unifier.get_dummy_var().ty; // Handled by special folding
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
ret_ty,
&[(in_ndarray_ty.ty, "a")],
Box::new(move |ctx, obj, fun, args, generator| {
assert!(obj.is_none());
assert_eq!(args.len(), 1);
let ndarray_ty = fun.0.args[0].ty;
let ndarray =
args[0].1.clone().to_basic_value_enum(ctx, generator, ndarray_ty)?;
let ndarray = AnyObject { ty: ndarray_ty, value: ndarray };
let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
let result_tuple = match prim {
PrimDef::FunNpShape => ndarray.make_shape_tuple(generator, ctx),
PrimDef::FunNpStrides => ndarray.make_strides_tuple(generator, ctx),
_ => unreachable!(),
};
Ok(Some(result_tuple.value.as_basic_value_enum()))
}),
)
}
_ => unreachable!(),
}
}
/// Build np/sp functions that take as input `NDArray` only
fn build_ndarray_view_function(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(
prim,
&[PrimDef::FunNpBroadcastTo, PrimDef::FunNpTranspose, PrimDef::FunNpReshape],
);
let in_ndarray_ty = self.unifier.get_fresh_var_with_range(
&[self.primitives.ndarray],
Some("T".into()),
None,
);
match prim {
PrimDef::FunNpTranspose => {
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
in_ndarray_ty.ty,
&[(in_ndarray_ty.ty, "x")],
Box::new(move |ctx, _, fun, args, generator| {
let arg_ty = fun.0.args[0].ty;
let arg_val =
args[0].1.clone().to_basic_value_enum(ctx, generator, arg_ty)?;
let arg = AnyObject { ty: arg_ty, value: arg_val };
let ndarray = NDArrayObject::from_object(generator, ctx, arg);
let ndarray = ndarray.transpose(generator, ctx, None); // TODO: Add axes argument
Ok(Some(ndarray.instance.value.as_basic_value_enum()))
}),
)
}
// NOTE: on `ndarray_factory_fn_shape_arg_tvar` and
// the `param_ty` for `create_fn_by_codegen`.
//
// Similar to `build_ndarray_from_shape_factory_function` we delegate the responsibility of typechecking
// to [`typecheck::type_inferencer::Inferencer::fold_numpy_function_call_shape_argument`],
// and use a dummy [`TypeVar`] `ndarray_factory_fn_shape_arg_tvar` as a placeholder for `param_ty`.
PrimDef::FunNpBroadcastTo | PrimDef::FunNpReshape => {
let ret_ty = self.unifier.get_dummy_var().ty; // Handled by special holding
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
ret_ty,
&[
(in_ndarray_ty.ty, "x"),
(self.ndarray_factory_fn_shape_arg_tvar.ty, "shape"), // Handled by special folding
],
Box::new(move |ctx, _, fun, args, generator| {
let ndarray_ty = fun.0.args[0].ty;
let ndarray_val =
args[0].1.clone().to_basic_value_enum(ctx, generator, ndarray_ty)?;
let shape_ty = fun.0.args[1].ty;
let shape_val =
args[1].1.clone().to_basic_value_enum(ctx, generator, shape_ty)?;
let ndarray = AnyObject { value: ndarray_val, ty: ndarray_ty };
let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
let shape = AnyObject { value: shape_val, ty: shape_ty };
let (_, shape) = parse_numpy_int_sequence(generator, ctx, shape);
// The ndims after reshaping is gotten from the return type of the call.
let (_, ndims) = unpack_ndarray_var_tys(&mut ctx.unifier, fun.0.ret);
let ndims = extract_ndims(&ctx.unifier, ndims);
let new_ndarray = match prim {
PrimDef::FunNpBroadcastTo => {
ndarray.broadcast_to(generator, ctx, ndims, shape)
}
PrimDef::FunNpReshape => {
ndarray.reshape_or_copy(generator, ctx, ndims, shape)
}
_ => unreachable!(),
};
Ok(Some(new_ndarray.instance.value.as_basic_value_enum()))
}),
)
}
_ => unreachable!(),
}
}
/// Build the `str()` function.
fn build_str_function(&mut self) -> TopLevelDef {
let prim = PrimDef::FunStr;
@ -1335,7 +1572,12 @@ impl<'a> BuiltinBuilder<'a> {
name: prim.name().into(),
simple_name: prim.simple_name().into(),
signature: self.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![FuncArg { name: "s".into(), ty: str, default_value: None }],
args: vec![FuncArg {
name: "s".into(),
ty: str,
default_value: None,
is_vararg: false,
}],
ret: str,
vars: VarMap::default(),
})),
@ -1399,31 +1641,21 @@ impl<'a> BuiltinBuilder<'a> {
fn build_len_function(&mut self) -> TopLevelDef {
let prim = PrimDef::FunLen;
let PrimitiveStore { uint64, int32, .. } = *self.primitives;
// Type handled in [`Inferencer::try_fold_special_call`]
let arg_tvar = self.unifier.get_dummy_var();
let tvar = self.unifier.get_fresh_var(Some("L".into()), None);
let list = self
.unifier
.subst(
self.primitives.list,
&into_var_map([TypeVar { id: self.list_tvar.id, ty: tvar.ty }]),
)
.unwrap();
let ndims = self.unifier.get_fresh_const_generic_var(uint64, Some("N".into()), None);
let ndarray = make_ndarray_ty(self.unifier, self.primitives, Some(tvar.ty), Some(ndims.ty));
let arg_ty = self.unifier.get_fresh_var_with_range(
&[list, ndarray, self.primitives.range],
Some("I".into()),
None,
);
TopLevelDef::Function {
name: prim.name().into(),
simple_name: prim.simple_name().into(),
signature: self.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![FuncArg { name: "ls".into(), ty: arg_ty.ty, default_value: None }],
ret: int32,
vars: into_var_map([tvar, arg_ty]),
args: vec![FuncArg {
name: "obj".into(),
ty: arg_tvar.ty,
default_value: None,
is_vararg: false,
}],
ret: self.primitives.int32,
vars: into_var_map([arg_tvar]),
})),
var_id: Vec::default(),
instance_to_symbol: HashMap::default(),
@ -1431,86 +1663,10 @@ impl<'a> BuiltinBuilder<'a> {
resolver: None,
codegen_callback: Some(Arc::new(GenCall::new(Box::new(
move |ctx, _, fun, args, generator| {
let range_ty = ctx.primitives.range;
let arg_ty = fun.0.args[0].ty;
let arg = args[0].1.clone().to_basic_value_enum(ctx, generator, arg_ty)?;
Ok(if ctx.unifier.unioned(arg_ty, range_ty) {
let arg = RangeValue::from_ptr_val(arg.into_pointer_value(), Some("range"));
let (start, end, step) = destructure_range(ctx, arg);
Some(calculate_len_for_slice_range(generator, ctx, start, end, step).into())
} else {
match &*ctx.unifier.get_ty_immutable(arg_ty) {
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::List.id() => {
let int32 = ctx.ctx.i32_type();
let zero = int32.const_zero();
let len = ctx
.build_gep_and_load(
arg.into_pointer_value(),
&[zero, int32.const_int(1, false)],
None,
)
.into_int_value();
if len.get_type().get_bit_width() == 32 {
Some(len.into())
} else {
Some(
ctx.builder
.build_int_truncate(len, int32, "len2i32")
.map(Into::into)
.unwrap(),
)
}
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
let llvm_i32 = ctx.ctx.i32_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let arg = NDArrayValue::from_ptr_val(
arg.into_pointer_value(),
llvm_usize,
None,
);
let ndims = arg.dim_sizes().size(ctx, generator);
ctx.make_assert(
generator,
ctx.builder
.build_int_compare(
IntPredicate::NE,
ndims,
llvm_usize.const_zero(),
"",
)
.unwrap(),
"0:TypeError",
&format!("{name}() of unsized object", name = prim.name()),
[None, None, None],
ctx.current_loc,
);
let len = unsafe {
arg.dim_sizes().get_typed_unchecked(
ctx,
generator,
&llvm_usize.const_zero(),
None,
)
};
if len.get_type().get_bit_width() == 32 {
Some(len.into())
} else {
Some(
ctx.builder
.build_int_truncate(len, llvm_i32, "len")
.map(Into::into)
.unwrap(),
)
}
}
_ => unreachable!(),
}
})
builtin_fns::call_len(generator, ctx, (arg_ty, arg)).map(|ret| Some(ret.into()))
},
)))),
loc: None,
@ -1526,8 +1682,18 @@ impl<'a> BuiltinBuilder<'a> {
simple_name: prim.simple_name().into(),
signature: self.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![
FuncArg { name: "m".into(), ty: self.num_ty.ty, default_value: None },
FuncArg { name: "n".into(), ty: self.num_ty.ty, default_value: None },
FuncArg {
name: "m".into(),
ty: self.num_ty.ty,
default_value: None,
is_vararg: false,
},
FuncArg {
name: "n".into(),
ty: self.num_ty.ty,
default_value: None,
is_vararg: false,
},
],
ret: self.num_ty.ty,
vars: self.num_var_map.clone(),
@ -1555,10 +1721,19 @@ impl<'a> BuiltinBuilder<'a> {
}
}
/// Build the functions `np_min()` and `np_max()`.
fn build_np_min_max_function(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(prim, &[PrimDef::FunNpMin, PrimDef::FunNpMax]);
/// Build the functions `np_max()`, `np_min()`, `np_argmax()` and `np_argmin()`
/// Calls `call_numpy_max_min` with the function name
fn build_np_max_min_function(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(
prim,
&[PrimDef::FunNpArgmin, PrimDef::FunNpArgmax, PrimDef::FunNpMin, PrimDef::FunNpMax],
);
let (var_map, ret_ty) = match prim {
PrimDef::FunNpArgmax | PrimDef::FunNpArgmin => {
(self.num_or_ndarray_var_map.clone(), self.primitives.int64)
}
PrimDef::FunNpMax | PrimDef::FunNpMin => {
let ret_ty = self.unifier.get_fresh_var(Some("R".into()), None);
let var_map = self
.num_or_ndarray_var_map
@ -1566,28 +1741,25 @@ impl<'a> BuiltinBuilder<'a> {
.into_iter()
.chain(once((ret_ty.id, ret_ty.ty)))
.collect::<IndexMap<_, _>>();
(var_map, ret_ty.ty)
}
_ => unreachable!(),
};
create_fn_by_codegen(
self.unifier,
&var_map,
prim.name(),
ret_ty.ty,
&[(self.float_or_ndarray_ty.ty, "a")],
ret_ty,
&[(self.num_or_ndarray_ty.ty, "a")],
Box::new(move |ctx, _, fun, args, generator| {
let a_ty = fun.0.args[0].ty;
let a = args[0].1.clone().to_basic_value_enum(ctx, generator, a_ty)?;
let func = match prim {
PrimDef::FunNpMin => builtin_fns::call_numpy_min,
PrimDef::FunNpMax => builtin_fns::call_numpy_max,
_ => unreachable!(),
};
Ok(Some(func(generator, ctx, (a_ty, a))?))
Ok(Some(builtin_fns::call_numpy_max_min(generator, ctx, (a_ty, a), prim.name())?))
}),
)
}
/// Build the functions `np_minimum()` and `np_maximum()`.
fn build_np_minimum_maximum_function(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(prim, &[PrimDef::FunNpMinimum, PrimDef::FunNpMaximum]);
@ -1603,7 +1775,12 @@ impl<'a> BuiltinBuilder<'a> {
signature: self.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: param_ty
.iter()
.map(|p| FuncArg { name: p.1.into(), ty: p.0, default_value: None })
.map(|p| FuncArg {
name: p.1.into(),
ty: p.0,
default_value: None,
is_vararg: false,
})
.collect(),
ret: ret_ty.ty,
vars: into_var_map([x1_ty, x2_ty, ret_ty]),
@ -1644,6 +1821,7 @@ impl<'a> BuiltinBuilder<'a> {
name: "n".into(),
ty: self.num_or_ndarray_ty.ty,
default_value: None,
is_vararg: false,
}],
ret: self.num_or_ndarray_ty.ty,
vars: self.num_or_ndarray_var_map.clone(),
@ -1832,7 +2010,12 @@ impl<'a> BuiltinBuilder<'a> {
signature: self.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: param_ty
.iter()
.map(|p| FuncArg { name: p.1.into(), ty: p.0, default_value: None })
.map(|p| FuncArg {
name: p.1.into(),
ty: p.0,
default_value: None,
is_vararg: false,
})
.collect(),
ret: ret_ty.ty,
vars: into_var_map([x1_ty, x2_ty, ret_ty]),
@ -1866,6 +2049,158 @@ impl<'a> BuiltinBuilder<'a> {
}
}
/// Build `np_linalg` and `sp_linalg` functions
///
/// The input to these functions must be floating point `NDArray`
fn build_linalg_methods(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(
prim,
&[
PrimDef::FunNpDot,
PrimDef::FunNpLinalgCholesky,
PrimDef::FunNpLinalgQr,
PrimDef::FunNpLinalgSvd,
PrimDef::FunNpLinalgInv,
PrimDef::FunNpLinalgPinv,
PrimDef::FunNpLinalgMatrixPower,
PrimDef::FunNpLinalgDet,
PrimDef::FunSpLinalgLu,
PrimDef::FunSpLinalgSchur,
PrimDef::FunSpLinalgHessenberg,
],
);
match prim {
PrimDef::FunNpDot => create_fn_by_codegen(
self.unifier,
&self.num_or_ndarray_var_map,
prim.name(),
self.num_ty.ty,
&[(self.num_or_ndarray_ty.ty, "x1"), (self.num_or_ndarray_ty.ty, "x2")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val = args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
let x2_ty = fun.0.args[1].ty;
let x2_val = args[1].1.clone().to_basic_value_enum(ctx, generator, x2_ty)?;
let result = ndarray_dot(generator, ctx, (x1_ty, x1_val), (x2_ty, x2_val))?;
Ok(Some(result))
}),
),
PrimDef::FunNpLinalgCholesky | PrimDef::FunNpLinalgInv | PrimDef::FunNpLinalgPinv => {
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
self.ndarray_float_2d,
&[(self.ndarray_float_2d, "x1")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val =
args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
let func = match prim {
PrimDef::FunNpLinalgCholesky => builtin_fns::call_np_linalg_cholesky,
PrimDef::FunNpLinalgInv => builtin_fns::call_np_linalg_inv,
PrimDef::FunNpLinalgPinv => builtin_fns::call_np_linalg_pinv,
_ => unreachable!(),
};
Ok(Some(func(generator, ctx, (x1_ty, x1_val))?))
}),
)
}
PrimDef::FunNpLinalgQr
| PrimDef::FunSpLinalgLu
| PrimDef::FunSpLinalgSchur
| PrimDef::FunSpLinalgHessenberg => {
let ret_ty = self.unifier.add_ty(TypeEnum::TTuple {
ty: vec![self.ndarray_float_2d, self.ndarray_float_2d],
is_vararg_ctx: false,
});
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
ret_ty,
&[(self.ndarray_float_2d, "x1")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val =
args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
let func = match prim {
PrimDef::FunNpLinalgQr => builtin_fns::call_np_linalg_qr,
PrimDef::FunSpLinalgLu => builtin_fns::call_sp_linalg_lu,
PrimDef::FunSpLinalgSchur => builtin_fns::call_sp_linalg_schur,
PrimDef::FunSpLinalgHessenberg => {
builtin_fns::call_sp_linalg_hessenberg
}
_ => unreachable!(),
};
Ok(Some(func(generator, ctx, (x1_ty, x1_val))?))
}),
)
}
PrimDef::FunNpLinalgSvd => {
let ret_ty = self.unifier.add_ty(TypeEnum::TTuple {
ty: vec![self.ndarray_float_2d, self.ndarray_float, self.ndarray_float_2d],
is_vararg_ctx: false,
});
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
ret_ty,
&[(self.ndarray_float_2d, "x1")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val =
args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
Ok(Some(builtin_fns::call_np_linalg_svd(generator, ctx, (x1_ty, x1_val))?))
}),
)
}
PrimDef::FunNpLinalgMatrixPower => create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
self.ndarray_float_2d,
&[(self.ndarray_float_2d, "x1"), (self.primitives.int32, "power")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val = args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
let x2_ty = fun.0.args[1].ty;
let x2_val = args[1].1.clone().to_basic_value_enum(ctx, generator, x2_ty)?;
Ok(Some(builtin_fns::call_np_linalg_matrix_power(
generator,
ctx,
(x1_ty, x1_val),
(x2_ty, x2_val),
)?))
}),
),
PrimDef::FunNpLinalgDet => create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
self.primitives.float,
&[(self.ndarray_float_2d, "x1")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val = args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
Ok(Some(builtin_fns::call_np_linalg_det(generator, ctx, (x1_ty, x1_val))?))
}),
),
_ => unreachable!(),
}
}
fn create_method(prim: PrimDef, method_ty: Type) -> (StrRef, Type, DefinitionId) {
(prim.simple_name().into(), method_ty, prim.id())
}

View File

@ -44,12 +44,27 @@ pub struct TopLevelComposer {
pub size_t: u32,
}
/// The specification for a builtin function, consisting of the function name, the function
/// signature, and a [code generation callback][`GenCall`].
pub type BuiltinFuncSpec = (StrRef, FunSignature, Arc<GenCall>);
/// A function that creates a [`BuiltinFuncSpec`] using the provided [`PrimitiveStore`] and
/// [`Unifier`].
pub type BuiltinFuncCreator = dyn Fn(&PrimitiveStore, &mut Unifier) -> BuiltinFuncSpec;
impl TopLevelComposer {
/// return a composer and things to make a "primitive" symbol resolver, so that the symbol
/// resolver can later figure out primitive type definitions when passed a primitive type name
/// resolver can later figure out primitive tye definitions when passed a primitive type name
///
/// `lateinit_builtins` are specifically for the ARTIQ module. Since the [`Unifier`] instance
/// used to create builtin functions do not persist until method compilation, any types
/// created (e.g. [`TypeEnum::TVar`]) also do not persist. Those functions should be instead put
/// in `lateinit_builtins`, where they will be instantiated with the [`Unifier`] instance used
/// for method compilation.
#[must_use]
pub fn new(
builtins: Vec<(StrRef, FunSignature, Arc<GenCall>)>,
builtins: Vec<BuiltinFuncSpec>,
lateinit_builtins: Vec<Box<BuiltinFuncCreator>>,
core_config: ComposerConfig,
size_t: u32,
) -> (Self, HashMap<StrRef, DefinitionId>, HashMap<StrRef, Type>) {
@ -119,7 +134,13 @@ impl TopLevelComposer {
}
}
for (name, sig, codegen_callback) in builtins {
// Materialize lateinit_builtins, now that the unifier is ready
let lateinit_builtins = lateinit_builtins
.into_iter()
.map(|builtin| builtin(&primitives_ty, &mut unifier))
.collect_vec();
for (name, sig, codegen_callback) in builtins.into_iter().chain(lateinit_builtins) {
let fun_sig = unifier.add_ty(TypeEnum::TFunc(sig));
builtin_ty.insert(name, fun_sig);
builtin_id.insert(name, DefinitionId(definition_ast_list.len()));
@ -766,6 +787,7 @@ impl TopLevelComposer {
let target_ty = get_type_from_type_annotation_kinds(
&temp_def_list,
unifier,
primitives,
&def,
&mut subst_list,
)?;
@ -859,7 +881,73 @@ impl TopLevelComposer {
let resolver = &**resolver;
let mut function_var_map = VarMap::new();
let arg_types = {
let vararg = args
.vararg
.as_ref()
.map(|vararg| -> Result<_, HashSet<String>> {
let vararg = vararg.as_ref();
let annotation = vararg
.node
.annotation
.as_ref()
.ok_or_else(|| {
HashSet::from([format!(
"function parameter `{}` needs type annotation at {}",
vararg.node.arg, vararg.location
)])
})?
.as_ref();
let type_annotation = parse_ast_to_type_annotation_kinds(
resolver,
temp_def_list.as_slice(),
unifier,
primitives_store,
annotation,
// NOTE: since only class need this, for function
// it should be fine to be empty map
HashMap::new(),
)?;
let type_vars_within =
get_type_var_contained_in_type_annotation(&type_annotation)
.into_iter()
.map(|x| -> Result<TypeVar, HashSet<String>> {
let TypeAnnotation::TypeVar(ty) = x else {
unreachable!("must be type var annotation kind")
};
let id = Self::get_var_id(ty, unifier)?;
Ok(TypeVar { id, ty })
})
.collect::<Result<Vec<_>, _>>()?;
for var in type_vars_within {
if let Some(prev_ty) = function_var_map.insert(var.id, var.ty) {
// if already have the type inserted, make sure they are the same thing
assert_eq!(prev_ty, var.ty);
}
}
let ty = get_type_from_type_annotation_kinds(
temp_def_list.as_ref(),
unifier,
primitives_store,
&type_annotation,
&mut None,
)?;
Ok(FuncArg {
name: vararg.node.arg,
ty,
default_value: Some(SymbolValue::Tuple(Vec::default())),
is_vararg: true,
})
})
.transpose()?;
let mut arg_types = {
// make sure no duplicate parameter
let mut defined_parameter_name: HashSet<_> = HashSet::new();
for x in &args.args {
@ -936,6 +1024,7 @@ impl TopLevelComposer {
let ty = get_type_from_type_annotation_kinds(
temp_def_list.as_ref(),
unifier,
primitives_store,
&type_annotation,
&mut None,
)?;
@ -959,11 +1048,18 @@ impl TopLevelComposer {
v
}),
},
is_vararg: false,
})
})
.collect::<Result<Vec<_>, _>>()?
};
if let Some(vararg) = vararg {
arg_types.push(vararg);
};
let arg_types = arg_types;
let return_ty = {
if let Some(returns) = returns {
let return_ty_annotation = {
@ -1002,6 +1098,7 @@ impl TopLevelComposer {
get_type_from_type_annotation_kinds(
&temp_def_list,
unifier,
primitives_store,
&return_ty_annotation,
&mut None,
)?
@ -1214,6 +1311,7 @@ impl TopLevelComposer {
})
}
},
is_vararg: false,
};
// push the dummy type and the type annotation
// into the list for later unification
@ -1622,6 +1720,7 @@ impl TopLevelComposer {
let self_type = get_type_from_type_annotation_kinds(
&def_list,
unifier,
primitives_ty,
&make_self_type_annotation(type_vars, *object_id),
&mut None,
)?;
@ -1638,21 +1737,25 @@ impl TopLevelComposer {
name: "msg".into(),
ty: string,
default_value: Some(SymbolValue::Str(String::new())),
is_vararg: false,
},
FuncArg {
name: "param0".into(),
ty: int64,
default_value: Some(SymbolValue::I64(0)),
is_vararg: false,
},
FuncArg {
name: "param1".into(),
ty: int64,
default_value: Some(SymbolValue::I64(0)),
is_vararg: false,
},
FuncArg {
name: "param2".into(),
ty: int64,
default_value: Some(SymbolValue::I64(0)),
is_vararg: false,
},
],
ret: self_type,
@ -1803,7 +1906,11 @@ impl TopLevelComposer {
let ty_ann = make_self_type_annotation(type_vars, *class_id);
let self_ty = get_type_from_type_annotation_kinds(
&def_list, unifier, &ty_ann, &mut None,
&def_list,
unifier,
primitives_ty,
&ty_ann,
&mut None,
)?;
vars.extend(type_vars.iter().map(|ty| {
let TypeEnum::TVar { id, .. } = &*unifier.get_ty(*ty) else {
@ -1858,6 +1965,7 @@ impl TopLevelComposer {
name: a.name,
ty: unifier.subst(a.ty, &subst).unwrap_or(a.ty),
default_value: a.default_value.clone(),
is_vararg: false,
})
.collect_vec()
};

View File

@ -27,17 +27,22 @@ pub enum PrimDef {
List,
NDArray,
// Member Functions
OptionIsSome,
OptionIsNone,
OptionUnwrap,
NDArrayCopy,
NDArrayFill,
FunInt32,
FunInt64,
FunUInt32,
FunUInt64,
FunFloat,
// Option methods
FunOptionIsSome,
FunOptionIsNone,
FunOptionUnwrap,
// Option-related functions
FunSome,
// NDArray methods
FunNDArrayCopy,
FunNDArrayFill,
// Range methods
FunRangeInit,
// NumPy factory functions
FunNpNDArray,
FunNpEmpty,
FunNpZeros,
@ -46,26 +51,27 @@ pub enum PrimDef {
FunNpArray,
FunNpEye,
FunNpIdentity,
FunRound,
FunRound64,
// NumPy ndarray property getters
FunNpSize,
FunNpShape,
FunNpStrides,
// NumPy ndarray view functions
FunNpBroadcastTo,
FunNpTranspose,
FunNpReshape,
// Miscellaneous NumPy & SciPy functions
FunNpRound,
FunRangeInit,
FunStr,
FunBool,
FunFloor,
FunFloor64,
FunNpFloor,
FunCeil,
FunCeil64,
FunNpCeil,
FunLen,
FunMin,
FunNpMin,
FunNpMinimum,
FunMax,
FunNpArgmin,
FunNpMax,
FunNpMaximum,
FunAbs,
FunNpArgmax,
FunNpIsNan,
FunNpIsInf,
FunNpSin,
@ -104,14 +110,43 @@ pub enum PrimDef {
FunNpHypot,
FunNpNextAfter,
// Top-Level Functions
FunSome,
// Linalg functions
FunNpDot,
FunNpLinalgCholesky,
FunNpLinalgQr,
FunNpLinalgSvd,
FunNpLinalgInv,
FunNpLinalgPinv,
FunNpLinalgMatrixPower,
FunNpLinalgDet,
FunSpLinalgLu,
FunSpLinalgSchur,
FunSpLinalgHessenberg,
// Miscellaneous Python & NAC3 functions
FunInt32,
FunInt64,
FunUInt32,
FunUInt64,
FunFloat,
FunRound,
FunRound64,
FunStr,
FunBool,
FunFloor,
FunFloor64,
FunCeil,
FunCeil64,
FunLen,
FunMin,
FunMax,
FunAbs,
}
/// Associated details of a [`PrimDef`]
pub enum PrimDefDetails {
PrimFunction { name: &'static str, simple_name: &'static str },
PrimClass { name: &'static str },
PrimClass { name: &'static str, get_ty_fn: fn(&PrimitiveStore) -> Type },
}
impl PrimDef {
@ -153,15 +188,17 @@ impl PrimDef {
#[must_use]
pub fn name(&self) -> &'static str {
match self.details() {
PrimDefDetails::PrimFunction { name, .. } | PrimDefDetails::PrimClass { name } => name,
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 class(name: &'static str, get_ty_fn: fn(&PrimitiveStore) -> Type) -> PrimDefDetails {
PrimDefDetails::PrimClass { name, get_ty_fn }
}
fn fun(name: &'static str, simple_name: Option<&'static str>) -> PrimDefDetails {
@ -169,29 +206,37 @@ impl PrimDef {
}
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),
// Classes
PrimDef::Int32 => class("int32", |primitives| primitives.int32),
PrimDef::Int64 => class("int64", |primitives| primitives.int64),
PrimDef::Float => class("float", |primitives| primitives.float),
PrimDef::Bool => class("bool", |primitives| primitives.bool),
PrimDef::None => class("none", |primitives| primitives.none),
PrimDef::Range => class("range", |primitives| primitives.range),
PrimDef::Str => class("str", |primitives| primitives.str),
PrimDef::Exception => class("Exception", |primitives| primitives.exception),
PrimDef::UInt32 => class("uint32", |primitives| primitives.uint32),
PrimDef::UInt64 => class("uint64", |primitives| primitives.uint64),
PrimDef::Option => class("Option", |primitives| primitives.option),
PrimDef::List => class("list", |primitives| primitives.list),
PrimDef::NDArray => class("ndarray", |primitives| primitives.ndarray),
// Option methods
PrimDef::FunOptionIsSome => fun("Option.is_some", Some("is_some")),
PrimDef::FunOptionIsNone => fun("Option.is_none", Some("is_none")),
PrimDef::FunOptionUnwrap => fun("Option.unwrap", Some("unwrap")),
// Option-related functions
PrimDef::FunSome => fun("Some", None),
// NDArray methods
PrimDef::FunNDArrayCopy => fun("ndarray.copy", Some("copy")),
PrimDef::FunNDArrayFill => fun("ndarray.fill", Some("fill")),
// Range methods
PrimDef::FunRangeInit => fun("range.__init__", Some("__init__")),
// NumPy factory functions
PrimDef::FunNpNDArray => fun("np_ndarray", None),
PrimDef::FunNpEmpty => fun("np_empty", None),
PrimDef::FunNpZeros => fun("np_zeros", None),
@ -200,26 +245,27 @@ impl PrimDef {
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),
// NumPy NDArray property getters,
PrimDef::FunNpSize => fun("np_size", None),
PrimDef::FunNpShape => fun("np_shape", None),
PrimDef::FunNpStrides => fun("np_strides", None),
// NumPy NDArray view functions
PrimDef::FunNpBroadcastTo => fun("np_broadcast_to", None),
PrimDef::FunNpTranspose => fun("np_transpose", None),
PrimDef::FunNpReshape => fun("np_reshape", None),
// Miscellaneous NumPy & SciPy functions
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::FunNpArgmin => fun("np_argmin", None),
PrimDef::FunNpMax => fun("np_max", None),
PrimDef::FunNpMaximum => fun("np_maximum", None),
PrimDef::FunAbs => fun("abs", None),
PrimDef::FunNpArgmax => fun("np_argmax", None),
PrimDef::FunNpIsNan => fun("np_isnan", None),
PrimDef::FunNpIsInf => fun("np_isinf", None),
PrimDef::FunNpSin => fun("np_sin", None),
@ -257,7 +303,38 @@ impl PrimDef {
PrimDef::FunNpLdExp => fun("np_ldexp", None),
PrimDef::FunNpHypot => fun("np_hypot", None),
PrimDef::FunNpNextAfter => fun("np_nextafter", None),
PrimDef::FunSome => fun("Some", None),
// Linalg functions
PrimDef::FunNpDot => fun("np_dot", None),
PrimDef::FunNpLinalgCholesky => fun("np_linalg_cholesky", None),
PrimDef::FunNpLinalgQr => fun("np_linalg_qr", None),
PrimDef::FunNpLinalgSvd => fun("np_linalg_svd", None),
PrimDef::FunNpLinalgInv => fun("np_linalg_inv", None),
PrimDef::FunNpLinalgPinv => fun("np_linalg_pinv", None),
PrimDef::FunNpLinalgMatrixPower => fun("np_linalg_matrix_power", None),
PrimDef::FunNpLinalgDet => fun("np_linalg_det", None),
PrimDef::FunSpLinalgLu => fun("sp_linalg_lu", None),
PrimDef::FunSpLinalgSchur => fun("sp_linalg_schur", None),
PrimDef::FunSpLinalgHessenberg => fun("sp_linalg_hessenberg", None),
// Miscellaneous Python & NAC3 functions
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::FunRound => fun("round", None),
PrimDef::FunRound64 => fun("round64", None),
PrimDef::FunStr => fun("str", None),
PrimDef::FunBool => fun("bool", None),
PrimDef::FunFloor => fun("floor", None),
PrimDef::FunFloor64 => fun("floor64", None),
PrimDef::FunCeil => fun("ceil", None),
PrimDef::FunCeil64 => fun("ceil64", None),
PrimDef::FunLen => fun("len", None),
PrimDef::FunMin => fun("min", None),
PrimDef::FunMax => fun("max", None),
PrimDef::FunAbs => fun("abs", None),
}
}
}
@ -408,9 +485,9 @@ impl TopLevelComposer {
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Option.id(),
fields: vec![
(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)),
(PrimDef::FunOptionIsSome.simple_name().into(), (is_some_type_fun_ty, true)),
(PrimDef::FunOptionIsNone.simple_name().into(), (is_some_type_fun_ty, true)),
(PrimDef::FunOptionUnwrap.simple_name().into(), (unwrap_fun_ty, true)),
]
.into_iter()
.collect::<HashMap<_, _>>(),
@ -444,6 +521,7 @@ impl TopLevelComposer {
name: "value".into(),
ty: ndarray_dtype_tvar.ty,
default_value: None,
is_vararg: false,
}],
ret: none,
vars: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
@ -451,8 +529,8 @@ impl TopLevelComposer {
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)),
(PrimDef::FunNDArrayCopy.simple_name().into(), (ndarray_copy_fun_ty, true)),
(PrimDef::FunNDArrayFill.simple_name().into(), (ndarray_fill_fun_ty, true)),
]),
params: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
});
@ -938,3 +1016,23 @@ pub fn arraylike_get_ndims(unifier: &mut Unifier, ty: Type) -> u64 {
_ => 0,
}
}
/// Extract an ndarray's `ndims` [type][`Type`] in `u64`. Panic if not possible.
/// The `ndims` must only contain 1 value.
#[must_use]
pub fn extract_ndims(unifier: &Unifier, ndims_ty: Type) -> u64 {
let ndims_ty_enum = unifier.get_ty_immutable(ndims_ty);
let TypeEnum::TLiteral { values, .. } = &*ndims_ty_enum else {
panic!("ndims_ty should be a TLiteral");
};
assert_eq!(values.len(), 1, "ndims_ty TLiteral should only contain 1 value");
let ndims = values[0].clone();
u64::try_from(ndims).unwrap()
}
/// Return an ndarray's `ndims` as a typechecker [`Type`] from its `u64` value.
pub fn create_ndims(unifier: &mut Unifier, ndims: u64) -> Type {
unifier.get_fresh_literal(vec![SymbolValue::U64(ndims)], None)
}

View File

@ -5,7 +5,7 @@ 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: [TypeVarId(245)]\n}\n",
"Function {\nname: \"Generic_A.fun\",\nsig: \"fn[[a:int32], V]\",\nvar_id: [TypeVarId(241)]\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

@ -7,7 +7,7 @@ expression: res_vec
"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[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",
"Class {\nname: \"B\",\nancestors: [\"B[typevar230]\", \"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: [\"typevar230\"]\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

@ -5,8 +5,8 @@ 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: [TypeVarId(247)]\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(252)]\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[v:V], none]\",\nvar_id: [TypeVarId(243)]\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(248)]\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

@ -3,7 +3,7 @@ source: nac3core/src/toplevel/test.rs
expression: res_vec
---
[
"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",
"Class {\nname: \"A\",\nancestors: [\"A[typevar229, typevar230]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[float, bool], b:B], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\")],\ntype_vars: [\"typevar229\", \"typevar230\"]\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",

View File

@ -6,12 +6,12 @@ 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: [TypeVarId(253)]\n}\n",
"Function {\nname: \"A.foo\",\nsig: \"fn[[a:T, b:V], none]\",\nvar_id: [TypeVarId(249)]\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: [TypeVarId(261)]\n}\n",
"Function {\nname: \"ff\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(257)]\n}\n",
]

View File

@ -117,7 +117,8 @@ impl SymbolResolver for Resolver {
"register"
)]
fn test_simple_register(source: Vec<&str>) {
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let mut composer =
TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
for s in source {
let ast = parse_program(s, FileName::default()).unwrap();
@ -137,7 +138,8 @@ fn test_simple_register(source: Vec<&str>) {
"register"
)]
fn test_simple_register_without_constructor(source: &str) {
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let mut composer =
TopLevelComposer::new(Vec::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, "", true).unwrap();
@ -171,7 +173,8 @@ fn test_simple_register_without_constructor(source: &str) {
"function compose"
)]
fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let mut composer =
TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
let internal_resolver = Arc::new(ResolverInternal {
id_to_def: Mutex::default(),
@ -519,7 +522,8 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
)]
fn test_analyze(source: &[&str], res: &[&str]) {
let print = false;
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let mut composer =
TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
let internal_resolver = make_internal_resolver_with_tvar(
vec![
@ -696,7 +700,8 @@ fn test_analyze(source: &[&str], res: &[&str]) {
)]
fn test_inference(source: Vec<&str>, res: &[&str]) {
let print = true;
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let mut composer =
TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
let internal_resolver = make_internal_resolver_with_tvar(
vec![

View File

@ -1,8 +1,9 @@
use super::*;
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::helper::PrimDef;
use crate::toplevel::helper::{PrimDef, PrimDefDetails};
use crate::typecheck::typedef::VarMap;
use nac3parser::ast::Constant;
use strum::IntoEnumIterator;
#[derive(Clone, Debug)]
pub enum TypeAnnotation {
@ -357,6 +358,7 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
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, HashSet<String>> {
@ -379,10 +381,43 @@ pub fn get_type_from_type_annotation_kinds(
let param_ty = params
.iter()
.map(|x| {
get_type_from_type_annotation_kinds(top_level_defs, unifier, x, subst_list)
get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
x,
subst_list,
)
})
.collect::<Result<Vec<_>, _>>()?;
let ty = if let Some(prim_def) = PrimDef::iter().find(|prim| prim.id() == *obj_id) {
// Primitive TopLevelDefs do not contain all fields that are present in their Type
// counterparts, so directly perform subst on the Type instead.
let PrimDefDetails::PrimClass { get_ty_fn, .. } = prim_def.details() else {
unreachable!()
};
let base_ty = get_ty_fn(primitives);
let params =
if let TypeEnum::TObj { params, .. } = &*unifier.get_ty_immutable(base_ty) {
params.clone()
} else {
unreachable!()
};
unifier
.subst(
get_ty_fn(primitives),
&params
.iter()
.zip(param_ty)
.map(|(obj_tv, param)| (*obj_tv.0, param))
.collect(),
)
.unwrap_or(base_ty)
} else {
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
@ -424,12 +459,15 @@ pub fn get_type_from_type_annotation_kinds(
}
}
TypeEnum::TVar { id, range, name, loc, is_const_generic: true, .. } => {
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);
let temp =
unifier.get_fresh_const_generic_var(ty, *name, *loc);
unifier.unify(temp.ty, p).is_ok()
}
};
@ -468,11 +506,16 @@ pub fn get_type_from_type_annotation_kinds(
fields: tobj_fields,
params: subst,
});
if need_subst {
if let Some(wl) = subst_list.as_mut() {
wl.push(ty);
}
}
ty
};
Ok(ty)
}
TypeAnnotation::Primitive(ty) | TypeAnnotation::TypeVar(ty) => Ok(*ty),
@ -490,6 +533,7 @@ pub fn get_type_from_type_annotation_kinds(
let ty = get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
ty.as_ref(),
subst_list,
)?;
@ -499,10 +543,16 @@ pub fn get_type_from_type_annotation_kinds(
let tys = tys
.iter()
.map(|x| {
get_type_from_type_annotation_kinds(top_level_defs, unifier, x, subst_list)
get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
x,
subst_list,
)
})
.collect::<Result<Vec<_>, _>>()?;
Ok(unifier.add_ty(TypeEnum::TTuple { ty: tys }))
Ok(unifier.add_ty(TypeEnum::TTuple { ty: tys, is_vararg_ctx: false }))
}
}
}

View File

@ -34,13 +34,18 @@ impl<'a> Inferencer<'a> {
self.should_have_value(pattern)?;
Ok(())
}
ExprKind::Tuple { elts, .. } => {
ExprKind::List { elts, .. } | ExprKind::Tuple { elts, .. } => {
for elt in elts {
self.check_pattern(elt, defined_identifiers)?;
self.should_have_value(elt)?;
}
Ok(())
}
ExprKind::Starred { value, .. } => {
self.check_pattern(value, defined_identifiers)?;
self.should_have_value(value)?;
Ok(())
}
ExprKind::Subscript { value, slice, .. } => {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
@ -207,6 +212,9 @@ impl<'a> Inferencer<'a> {
/// 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 {
if cfg!(feature = "no-escape-analysis") {
true
} else {
match &*self.unifier.get_ty_immutable(ret_ty) {
TypeEnum::TObj { .. } => [
self.primitives.int32,
@ -218,10 +226,11 @@ impl<'a> Inferencer<'a> {
]
.iter()
.any(|allowed_ty| self.unifier.unioned(ret_ty, *allowed_ty)),
TypeEnum::TTuple { ty } => ty.iter().all(|t| self.check_return_value_ty(*t)),
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(

View File

@ -1,5 +1,5 @@
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::helper::PrimDef;
use crate::toplevel::helper::{extract_ndims, PrimDef};
use crate::toplevel::numpy::{make_ndarray_ty, unpack_ndarray_var_tys};
use crate::typecheck::{
type_inferencer::*,
@ -197,6 +197,7 @@ pub fn impl_binop(
ty: other_ty,
default_value: None,
name: "other".into(),
is_vararg: false,
}],
})),
false,
@ -261,6 +262,7 @@ pub fn impl_cmpop(
ty: other_ty,
default_value: None,
name: "other".into(),
is_vararg: false,
}],
})),
false,
@ -518,36 +520,41 @@ pub fn typeof_binop(
}
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()
let (lhs_dtype, lhs_ndims) = unpack_ndarray_var_tys(unifier, lhs);
let lhs_ndims = extract_ndims(unifier, lhs_ndims);
let (rhs_dtype, rhs_ndims) = unpack_ndarray_var_tys(unifier, rhs);
let rhs_ndims = extract_ndims(unifier, rhs_ndims);
if !(unifier.unioned(lhs_dtype, primitives.float)
&& unifier.unioned(rhs_dtype, primitives.float))
{
return Err(format!(
"ndarray.__matmul__ only supports float64 operations, but LHS has type {} and RHS has type {}",
unifier.stringify(lhs),
unifier.stringify(rhs)
));
}
_ => 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()
let result_ndims = match (lhs_ndims, rhs_ndims) {
(0, _) | (_, 0) => {
return Err(
"ndarray.__matmul__ does not allow unsized ndarray input".to_string()
)
}
_ => unreachable!(),
(1, 1) => 0,
(1, _) => rhs_ndims - 1,
(_, 1) => lhs_ndims - 1,
(m, n) => max(m, n),
};
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"
))
}
if result_ndims == 0 {
// If the result is unsized, NumPy returns a scalar.
primitives.float
} else {
let result_ndims_ty =
unifier.get_fresh_literal(vec![SymbolValue::U64(result_ndims)], None);
make_ndarray_ty(unifier, primitives, Some(primitives.float), Some(result_ndims_ty))
}
}
@ -678,6 +685,7 @@ pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifie
bool: bool_t,
uint32: uint32_t,
uint64: uint64_t,
str: str_t,
list: list_t,
ndarray: ndarray_t,
..
@ -723,6 +731,9 @@ pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifie
impl_sign(unifier, store, bool_t, Some(int32_t));
impl_eq(unifier, store, bool_t, &[bool_t, ndarray_bool_t], None);
/* str ========= */
impl_cmpop(unifier, store, str_t, &[str_t], &[Cmpop::Eq, Cmpop::NotEq], Some(bool_t));
/* 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]);
@ -746,7 +757,7 @@ pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifie
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_matmul(unifier, store, ndarray_t, &[ndarray_unsized_t], None);
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);

View File

@ -183,9 +183,10 @@ impl<'a> Display for DisplayTypeError<'a> {
}
result
}
(TypeEnum::TTuple { ty: ty1 }, TypeEnum::TTuple { ty: ty2 })
if ty1.len() != ty2.len() =>
{
(
TypeEnum::TTuple { ty: ty1, is_vararg_ctx: is_vararg1 },
TypeEnum::TTuple { ty: ty2, is_vararg_ctx: is_vararg2 },
) if !is_vararg1 && !is_vararg2 && ty1.len() != ty2.len() => {
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 {t1} and {t2}")

File diff suppressed because it is too large Load Diff

View File

@ -83,7 +83,12 @@ impl TestEnvironment {
});
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 }],
args: vec![FuncArg {
name: "other".into(),
ty: int32,
default_value: None,
is_vararg: false,
}],
ret: int32,
vars: VarMap::new(),
}));
@ -224,7 +229,12 @@ impl TestEnvironment {
});
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 }],
args: vec![FuncArg {
name: "other".into(),
ty: int32,
default_value: None,
is_vararg: false,
}],
ret: int32,
vars: VarMap::new(),
}));

View File

@ -1,16 +1,4 @@
use indexmap::IndexMap;
use itertools::Itertools;
use std::cell::RefCell;
use std::collections::HashMap;
use std::fmt::{self, Display};
use std::iter::zip;
use std::rc::Rc;
use std::sync::{Arc, Mutex};
use std::{borrow::Cow, collections::HashSet};
use nac3parser::ast::{Cmpop, Location, StrRef, Unaryop};
use super::magic_methods::Binop;
use super::magic_methods::{Binop, HasOpInfo};
use super::type_error::{TypeError, TypeErrorKind};
use super::unification_table::{UnificationKey, UnificationTable};
use crate::symbol_resolver::SymbolValue;
@ -18,6 +6,16 @@ use crate::toplevel::helper::PrimDef;
use crate::toplevel::{DefinitionId, TopLevelContext, TopLevelDef};
use crate::typecheck::magic_methods::OpInfo;
use crate::typecheck::type_inferencer::PrimitiveStore;
use indexmap::IndexMap;
use itertools::{repeat_n, Itertools};
use nac3parser::ast::{Cmpop, Location, StrRef, Unaryop};
use std::cell::RefCell;
use std::collections::HashMap;
use std::fmt::{self, Display};
use std::iter::{repeat, zip};
use std::rc::Rc;
use std::sync::{Arc, Mutex};
use std::{borrow::Cow, collections::HashSet};
#[cfg(test)]
mod test;
@ -115,6 +113,7 @@ pub struct FuncArg {
pub name: StrRef,
pub ty: Type,
pub default_value: Option<SymbolValue>,
pub is_vararg: bool,
}
impl FuncArg {
@ -233,6 +232,12 @@ pub enum TypeEnum {
TTuple {
/// The types of elements present in this tuple.
ty: Vec<Type>,
/// Whether this tuple is used in a vararg context.
///
/// If `true`, `ty` must only contain one type, and the tuple is assumed to contain any
/// number of `ty`-typed values.
is_vararg_ctx: bool,
},
/// An object type.
@ -527,7 +532,7 @@ impl Unifier {
TypeEnum::TVirtual { ty } => self.get_instantiations(*ty).map(|ty| {
ty.iter().map(|&ty| self.add_ty(TypeEnum::TVirtual { ty })).collect_vec()
}),
TypeEnum::TTuple { ty } => {
TypeEnum::TTuple { ty, is_vararg_ctx } => {
let tuples = ty
.iter()
.map(|ty| self.get_instantiations(*ty).unwrap_or_else(|| vec![*ty]))
@ -537,7 +542,12 @@ impl Unifier {
None
} else {
Some(
tuples.into_iter().map(|ty| self.add_ty(TypeEnum::TTuple { ty })).collect(),
tuples
.into_iter()
.map(|ty| {
self.add_ty(TypeEnum::TTuple { ty, is_vararg_ctx: *is_vararg_ctx })
})
.collect(),
)
}
}
@ -581,7 +591,7 @@ impl Unifier {
TVar { .. } => allowed_typevars.iter().any(|b| self.unification_table.unioned(a, *b)),
TCall { .. } => false,
TVirtual { ty } => self.is_concrete(*ty, allowed_typevars),
TTuple { ty } => ty.iter().all(|ty| self.is_concrete(*ty, allowed_typevars)),
TTuple { ty, .. } => ty.iter().all(|ty| self.is_concrete(*ty, allowed_typevars)),
TObj { params: vars, .. } => {
vars.values().all(|ty| self.is_concrete(*ty, allowed_typevars))
}
@ -649,6 +659,7 @@ impl Unifier {
// Get details about the function signature/parameters.
let num_params = signature.args.len();
let is_vararg = signature.args.iter().any(|arg| arg.is_vararg);
// Force the type vars in `b` and `signature' to be up-to-date.
let b = self.instantiate_fun(b, signature);
@ -737,7 +748,7 @@ impl Unifier {
};
// Check for "too many arguments"
if num_params < posargs.len() {
if !is_vararg && num_params < posargs.len() {
let expected_min_count =
signature.args.iter().filter(|param| param.is_required()).count();
let expected_max_count = num_params;
@ -770,6 +781,19 @@ impl Unifier {
type_check_arg(param.name, param.ty, arg_ty)?;
}
if is_vararg {
debug_assert!(!signature.args.is_empty());
let vararg_args = posargs.iter().skip(signature.args.len());
let vararg_param = signature.args.last().unwrap();
for (&arg_ty, param) in zip(vararg_args, repeat(vararg_param)) {
// `param_info` for this argument would've already been marked as supplied
// during non-vararg posarg typecheck
type_check_arg(param.name, param.ty, arg_ty)?;
}
}
// Now consume all keyword arguments and typecheck them.
for (&param_name, &arg_ty) in kwargs {
// We will also use this opportunity to check if this keyword argument is "legal".
@ -959,7 +983,10 @@ impl Unifier {
self.unify_impl(x, b, false)?;
self.set_a_to_b(a, x);
}
(TVar { fields: Some(fields), range, is_const_generic: false, .. }, TTuple { ty }) => {
(
TVar { fields: Some(fields), range, is_const_generic: false, .. },
TTuple { ty, .. },
) => {
let len = i32::try_from(ty.len()).unwrap();
for (k, v) in fields {
match *k {
@ -980,8 +1007,18 @@ impl Unifier {
self.unify_impl(v.ty, ty[ind as usize], false)
.map_err(|e| e.at(v.loc))?;
}
RecordKey::Str(_) => {
return Err(TypeError::new(TypeErrorKind::NoSuchField(*k, b), v.loc))
RecordKey::Str(s) => {
let tuple_fns = [
Cmpop::Eq.op_info().method_name,
Cmpop::NotEq.op_info().method_name,
];
if !tuple_fns.into_iter().any(|op| s.to_string() == op) {
return Err(TypeError::new(
TypeErrorKind::NoSuchField(*k, b),
v.loc,
));
}
}
}
}
@ -1056,15 +1093,47 @@ impl Unifier {
self.set_a_to_b(a, b);
}
(TTuple { ty: ty1 }, TTuple { ty: ty2 }) => {
if ty1.len() != ty2.len() {
return Err(TypeError::new(TypeErrorKind::IncompatibleTypes(a, b), None));
(
TTuple { ty: ty1, is_vararg_ctx: is_vararg1 },
TTuple { ty: ty2, is_vararg_ctx: is_vararg2 },
) => {
// Rules for Tuples:
// - ty1: is_vararg && ty2: is_vararg -> ty1[0] == ty2[0]
// - ty1: is_vararg && ty2: !is_vararg -> type error (not enough info to infer the correct number of arguments)
// - ty1: !is_vararg && ty2: is_vararg -> ty1[..] == ty2[0]
// - ty1: !is_vararg && ty2: !is_vararg -> ty1.len() == ty2.len() && ty1[i] == ty2[i]
debug_assert!(!is_vararg1 || ty1.len() == 1);
debug_assert!(!is_vararg2 || ty2.len() == 1);
match (*is_vararg1, *is_vararg2) {
(true, true) => {
if self.unify_impl(ty1[0], ty2[0], false).is_err() {
return Self::incompatible_types(a, b);
}
}
(true, false) => return Self::incompatible_types(a, b),
(false, true) => {
for y in ty2 {
if self.unify_impl(ty1[0], *y, false).is_err() {
return Self::incompatible_types(a, b);
}
}
}
(false, false) => {
if ty1.len() != ty2.len() {
return Self::incompatible_types(a, b);
}
for (x, y) in ty1.iter().zip(ty2.iter()) {
if self.unify_impl(*x, *y, false).is_err() {
return Err(TypeError::new(TypeErrorKind::IncompatibleTypes(a, b), None));
return Self::incompatible_types(a, b);
}
}
}
}
self.set_a_to_b(a, b);
}
(TVar { fields: Some(map), range, .. }, TObj { obj_id, fields, params }) => {
@ -1307,11 +1376,23 @@ impl Unifier {
TypeEnum::TLiteral { values, .. } => {
format!("const({})", values.iter().map(|v| format!("{v:?}")).join(", "))
}
TypeEnum::TTuple { ty } => {
let mut fields =
ty.iter().map(|v| self.internal_stringify(*v, obj_to_name, var_to_name, notes));
TypeEnum::TTuple { ty, is_vararg_ctx } => {
if *is_vararg_ctx {
debug_assert_eq!(ty.len(), 1);
let field = self.internal_stringify(
*ty.iter().next().unwrap(),
obj_to_name,
var_to_name,
notes,
);
format!("tuple[*{field}]")
} else {
let mut fields = ty
.iter()
.map(|v| self.internal_stringify(*v, obj_to_name, var_to_name, notes));
format!("tuple[{}]", fields.join(", "))
}
}
TypeEnum::TVirtual { ty } => {
format!(
"virtual[{}]",
@ -1335,17 +1416,21 @@ impl Unifier {
.args
.iter()
.map(|arg| {
let vararg_prefix = if arg.is_vararg { "*" } else { "" };
if let Some(dv) = &arg.default_value {
format!(
"{}:{}={}",
"{}:{}{}={}",
arg.name,
vararg_prefix,
self.internal_stringify(arg.ty, obj_to_name, var_to_name, notes),
dv
)
} else {
format!(
"{}:{}",
"{}:{}{}",
arg.name,
vararg_prefix,
self.internal_stringify(arg.ty, obj_to_name, var_to_name, notes)
)
}
@ -1431,7 +1516,7 @@ impl Unifier {
match &*ty {
TypeEnum::TRigidVar { .. } | TypeEnum::TLiteral { .. } => None,
TypeEnum::TVar { id, .. } => mapping.get(id).copied(),
TypeEnum::TTuple { ty } => {
TypeEnum::TTuple { ty, is_vararg_ctx } => {
let mut new_ty = Cow::from(ty);
for (i, t) in ty.iter().enumerate() {
if let Some(t1) = self.subst_impl(*t, mapping, cache) {
@ -1439,7 +1524,10 @@ impl Unifier {
}
}
if matches!(new_ty, Cow::Owned(_)) {
Some(self.add_ty(TypeEnum::TTuple { ty: new_ty.into_owned() }))
Some(self.add_ty(TypeEnum::TTuple {
ty: new_ty.into_owned(),
is_vararg_ctx: *is_vararg_ctx,
}))
} else {
None
}
@ -1599,16 +1687,37 @@ impl Unifier {
}
}
(TVar { range, .. }, _) => self.check_var_compatibility(b, range).or(Err(())),
(TTuple { ty: ty1 }, TTuple { ty: ty2 }) if ty1.len() == ty2.len() => {
let ty: Vec<_> = zip(ty1.iter(), ty2.iter())
.map(|(a, b)| self.get_intersection(*a, *b))
.try_collect()?;
if ty.iter().any(Option::is_some) {
Ok(Some(self.add_ty(TTuple {
ty: zip(ty, ty1.iter()).map(|(a, b)| a.unwrap_or(*b)).collect(),
})))
(
TTuple { ty: ty1, is_vararg_ctx: is_vararg1 },
TTuple { ty: ty2, is_vararg_ctx: is_vararg2 },
) => {
if *is_vararg1 && *is_vararg2 {
let isect_ty = self.get_intersection(ty1[0], ty2[0])?;
Ok(isect_ty.map(|ty| self.add_ty(TTuple { ty: vec![ty], is_vararg_ctx: true })))
} else {
Ok(None)
let zip_iter: Box<dyn Iterator<Item = (&Type, &Type)>> =
match (*is_vararg1, *is_vararg2) {
(true, _) => Box::new(repeat_n(&ty1[0], ty2.len()).zip(ty2.iter())),
(_, false) => Box::new(ty1.iter().zip(repeat_n(&ty2[0], ty1.len()))),
_ => {
if ty1.len() != ty2.len() {
return Err(());
}
Box::new(ty1.iter().zip(ty2.iter()))
}
};
let ty: Vec<_> =
zip_iter.map(|(a, b)| self.get_intersection(*a, *b)).try_collect()?;
Ok(if ty.iter().any(Option::is_some) {
Some(self.add_ty(TTuple {
ty: zip(ty, ty1.iter()).map(|(a, b)| a.unwrap_or(*b)).collect(),
is_vararg_ctx: false,
}))
} else {
None
})
}
}
// TODO(Derppening): #444

View File

@ -28,7 +28,10 @@ impl Unifier {
TypeEnum::TVar { fields: Some(map1), .. },
TypeEnum::TVar { fields: Some(map2), .. },
) => self.map_eq2(map1, map2),
(TypeEnum::TTuple { ty: ty1 }, TypeEnum::TTuple { ty: ty2 }) => {
(
TypeEnum::TTuple { ty: ty1, is_vararg_ctx: false },
TypeEnum::TTuple { ty: ty2, is_vararg_ctx: false },
) => {
ty1.len() == ty2.len()
&& ty1.iter().zip(ty2.iter()).all(|(t1, t2)| self.eq(*t1, *t2))
}
@ -178,7 +181,7 @@ impl TestEnvironment {
ty.push(result.0);
s = result.1;
}
(self.unifier.add_ty(TypeEnum::TTuple { ty }), &s[1..])
(self.unifier.add_ty(TypeEnum::TTuple { ty, is_vararg_ctx: false }), &s[1..])
}
"Record" => {
let mut s = &typ[end..];
@ -608,7 +611,7 @@ fn test_instantiation() {
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 tuple = env.unifier.add_ty(TypeEnum::TTuple { ty: vec![v, v1, v2], is_vararg_ctx: false });
let v3 = env.unifier.get_fresh_var_with_range(&[tuple, t], None, None).ty;
// t = TypeVar('t')
// v = TypeVar('v', int, bool)

View File

@ -238,7 +238,7 @@ impl<'a> EH_Frame<'a> {
/// 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.
/// > 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,

View File

@ -4,6 +4,9 @@ version = "0.1.0"
authors = ["M-Labs"]
edition = "2021"
[features]
no-escape-analysis = ["nac3core/no-escape-analysis"]
[dependencies]
parking_lot = "0.12"
nac3parser = { path = "../nac3parser" }

View File

@ -3,23 +3,66 @@
set -e
if [ -z "$1" ]; then
echo "Requires at least one argument"
echo "No argument supplied"
exit 1
fi
declare -a nac3args
while [ $# -gt 1 ]; do
case "$1" in
--help)
echo "Usage: check_demo.sh [--debug] [-i686] -- [NAC3ARGS...] demo"
exit
;;
--debug)
debug=1
;;
-i686)
i686=1
;;
--)
shift
break
;;
*)
echo "Unrecognized argument \"$1\""
exit 1
;;
esac
shift
done
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
echo "### Checking $demo..."
echo ">>>>>> Running $demo with the Python interpreter"
./interpret_demo.py "$demo" > interpreted.log
if [ -n "$i686" ]; then
echo "...... Trying NAC3's 32-bit code generator output"
if [ -n "$debug" ]; then
./run_demo.sh --debug -i686 --out run_32.log -- "${nac3args[@]}" "$demo"
else
./run_demo.sh -i686 --out run_32.log -- "${nac3args[@]}" "$demo"
fi
diff -Nau interpreted.log run_32.log
fi
echo "...... Trying NAC3's 64-bit code generator output"
if [ -n "$debug" ]; then
./run_demo.sh --debug --out run_64.log -- "${nac3args[@]}" "$demo"
else
./run_demo.sh --out run_64.log -- "${nac3args[@]}" "$demo"
fi
diff -Nau interpreted.log run_64.log
echo "...... OK"
rm -f interpreted.log \
run_32.log run_64.log

View File

@ -2,6 +2,11 @@
set -e
if [ "$1" == "--help" ]; then
echo "Usage: check_demos.sh [CHECKARGS...] [--] [NAC3ARGS...]"
exit
fi
count=0
for demo in src/*.py; do
./check_demo.sh "$@" "$demo"

View File

@ -6,8 +6,6 @@
#include <stdlib.h>
#include <string.h>
#define usize size_t
double dbl_nan(void) {
return NAN;
}
@ -64,14 +62,14 @@ void output_asciiart(int32_t x) {
struct cslice {
void *data;
usize len;
size_t 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) {
for (size_t i = 0; i < slice->len; ++i) {
if (i == slice->len - 1) {
printf("%d", data[i]);
} else {
@ -85,7 +83,7 @@ void output_int32_list(struct cslice *slice) {
void output_str(struct cslice *slice) {
const char *data = (const char *) slice->data;
for (usize i = 0; i < slice->len; ++i) {
for (size_t i = 0; i < slice->len; ++i) {
putchar(data[i]);
}
}
@ -107,8 +105,25 @@ uint32_t __nac3_personality(uint32_t state, uint32_t exception_object, uint32_t
__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);
// See `struct Exception<'a>` in
// https://github.com/m-labs/artiq/blob/master/artiq/firmware/libeh/eh_artiq.rs
struct Exception {
uint32_t id;
struct cslice file;
uint32_t line;
uint32_t column;
struct cslice function;
struct cslice message;
int64_t param[3];
};
uint32_t __nac3_raise(struct Exception* e) {
printf("__nac3_raise called. Exception details:\n");
printf(" ID: %"PRIu32"\n", e->id);
printf(" Location: %*s:%"PRIu32":%"PRIu32"\n" , (int) e->file.len, (const char*) e->file.data, e->line, e->column);
printf(" Function: %*s\n" , (int) e->function.len, (const char*) e->function.data);
printf(" Message: \"%*s\"\n" , (int) e->message.len, (const char*) e->message.data);
printf(" Params: {0}=%"PRId64", {1}=%"PRId64", {2}=%"PRId64"\n", e->param[0], e->param[1], e->param[2]);
exit(101);
__builtin_unreachable();
}

View File

@ -6,6 +6,7 @@ import importlib.machinery
import math
import numpy as np
import numpy.typing as npt
import scipy as sp
import pathlib
from numpy import int32, int64, uint32, uint64
@ -167,7 +168,7 @@ def patch(module):
module.ceil64 = _ceil
module.np_ceil = np.ceil
# NumPy ndarray functions
# NumPy NDArray factory functions
module.ndarray = NDArray
module.np_ndarray = np.ndarray
module.np_empty = np.empty
@ -178,13 +179,25 @@ def patch(module):
module.np_identity = np.identity
module.np_array = np.array
# NumPy NDArray view functions
module.np_broadcast_to = np.broadcast_to
module.np_transpose = np.transpose
module.np_reshape = np.reshape
# NumPy NDArray property getters
module.np_size = np.size
module.np_shape = np.shape
module.np_strides = lambda ndarray: ndarray.strides
# 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_argmin = np.argmin
module.np_max = np.max
module.np_maximum = np.maximum
module.np_argmax = np.argmax
module.np_sin = np.sin
module.np_cos = np.cos
module.np_exp = np.exp
@ -216,7 +229,7 @@ def patch(module):
module.np_hypot = np.hypot
module.np_nextafter = np.nextafter
# SciPy Math Functions
# SciPy Math functions
module.sp_spec_erf = special.erf
module.sp_spec_erfc = special.erfc
module.sp_spec_gamma = special.gamma
@ -224,14 +237,19 @@ def patch(module):
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
# Linalg functions
module.np_dot = np.dot
module.np_linalg_cholesky = np.linalg.cholesky
module.np_linalg_qr = np.linalg.qr
module.np_linalg_svd = np.linalg.svd
module.np_linalg_inv = np.linalg.inv
module.np_linalg_pinv = np.linalg.pinv
module.np_linalg_matrix_power = np.linalg.matrix_power
module.np_linalg_det = np.linalg.det
module.sp_linalg_lu = lambda x: sp.linalg.lu(x, True)
module.sp_linalg_schur = sp.linalg.schur
module.sp_linalg_hessenberg = lambda x: sp.linalg.hessenberg(x, True)
def file_import(filename, prefix="file_import_"):
filename = pathlib.Path(filename)

114
nac3standalone/demo/linalg/Cargo.lock generated Normal file
View File

@ -0,0 +1,114 @@
# This file is automatically @generated by Cargo.
# It is not intended for manual editing.
version = 3
[[package]]
name = "approx"
version = "0.5.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cab112f0a86d568ea0e627cc1d6be74a1e9cd55214684db5561995f6dad897c6"
dependencies = [
"num-traits",
]
[[package]]
name = "autocfg"
version = "1.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0c4b4d0bd25bd0b74681c0ad21497610ce1b7c91b1022cd21c80c6fbdd9476b0"
[[package]]
name = "cslice"
version = "0.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0f8cb7306107e4b10e64994de6d3274bd08996a7c1322a27b86482392f96be0a"
[[package]]
name = "libm"
version = "0.2.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4ec2a862134d2a7d32d7983ddcdd1c4923530833c9f2ea1a44fc5fa473989058"
[[package]]
name = "linalg"
version = "0.1.0"
dependencies = [
"cslice",
"nalgebra",
]
[[package]]
name = "nalgebra"
version = "0.32.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7b5c17de023a86f59ed79891b2e5d5a94c705dbe904a5b5c9c952ea6221b03e4"
dependencies = [
"approx",
"num-complex",
"num-rational",
"num-traits",
"simba",
"typenum",
]
[[package]]
name = "num-complex"
version = "0.4.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "73f88a1307638156682bada9d7604135552957b7818057dcef22705b4d509495"
dependencies = [
"num-traits",
]
[[package]]
name = "num-integer"
version = "0.1.46"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7969661fd2958a5cb096e56c8e1ad0444ac2bbcd0061bd28660485a44879858f"
dependencies = [
"num-traits",
]
[[package]]
name = "num-rational"
version = "0.4.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f83d14da390562dca69fc84082e73e548e1ad308d24accdedd2720017cb37824"
dependencies = [
"num-integer",
"num-traits",
]
[[package]]
name = "num-traits"
version = "0.2.19"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "071dfc062690e90b734c0b2273ce72ad0ffa95f0c74596bc250dcfd960262841"
dependencies = [
"autocfg",
"libm",
]
[[package]]
name = "paste"
version = "1.0.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "57c0d7b74b563b49d38dae00a0c37d4d6de9b432382b2892f0574ddcae73fd0a"
[[package]]
name = "simba"
version = "0.8.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "061507c94fc6ab4ba1c9a0305018408e312e17c041eb63bef8aa726fa33aceae"
dependencies = [
"approx",
"num-complex",
"num-traits",
"paste",
]
[[package]]
name = "typenum"
version = "1.17.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "42ff0bf0c66b8238c6f3b578df37d0b7848e55df8577b3f74f92a69acceeb825"

View File

@ -0,0 +1,13 @@
[package]
name = "linalg"
version = "0.1.0"
edition = "2021"
[lib]
crate-type = ["staticlib"]
[dependencies]
nalgebra = {version = "0.32.6", default-features = false, features = ["libm", "alloc"]}
cslice = "0.3.0"
[workspace]

View File

@ -0,0 +1,406 @@
// Uses `nalgebra` crate to invoke `np_linalg` and `sp_linalg` functions
// When converting between `nalgebra::Matrix` and `NDArray` following considerations are necessary
//
// * Both `nalgebra::Matrix` and `NDArray` require their content to be stored in row-major order
// * `NDArray` data pointer can be directly read and converted to `nalgebra::Matrix` (row and column number must be known)
// * `nalgebra::Matrix::as_slice` returns the content of matrix in column-major order and initial data needs to be transposed before storing it in `NDArray` data pointer
use core::slice;
use nalgebra::DMatrix;
fn report_error(
error_name: &str,
fn_name: &str,
file_name: &str,
line_num: u32,
col_num: u32,
err_msg: &str,
) -> ! {
panic!(
"Exception {} from {} in {}:{}:{}, message: {}",
error_name, fn_name, file_name, line_num, col_num, err_msg
);
}
pub struct InputMatrix {
pub ndims: usize,
pub dims: *const usize,
pub data: *mut f64,
}
impl InputMatrix {
fn get_dims(&mut self) -> Vec<usize> {
let dims = unsafe { slice::from_raw_parts(self.dims, self.ndims) };
dims.to_vec()
}
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn np_linalg_cholesky(mat1: *mut InputMatrix, out: *mut InputMatrix) {
let mat1 = mat1.as_mut().unwrap();
let out = out.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "np_linalg_cholesky", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
if dim1[0] != dim1[1] {
let err_msg =
format!("last 2 dimensions of the array must be square: {0} != {1}", dim1[0], dim1[1]);
report_error("LinAlgError", "np_linalg_cholesky", file!(), line!(), column!(), &err_msg);
}
let outdim = out.get_dims();
let out_slice = unsafe { slice::from_raw_parts_mut(out.data, outdim[0] * outdim[1]) };
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let matrix1 = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
let result = matrix1.cholesky();
match result {
Some(res) => {
out_slice.copy_from_slice(res.unpack().transpose().as_slice());
}
None => {
report_error(
"LinAlgError",
"np_linalg_cholesky",
file!(),
line!(),
column!(),
"Matrix is not positive definite",
);
}
};
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn np_linalg_qr(
mat1: *mut InputMatrix,
out_q: *mut InputMatrix,
out_r: *mut InputMatrix,
) {
let mat1 = mat1.as_mut().unwrap();
let out_q = out_q.as_mut().unwrap();
let out_r = out_r.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "np_linalg_cholesky", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
let outq_dim = (*out_q).get_dims();
let outr_dim = (*out_r).get_dims();
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let out_q_slice = unsafe { slice::from_raw_parts_mut(out_q.data, outq_dim[0] * outq_dim[1]) };
let out_r_slice = unsafe { slice::from_raw_parts_mut(out_r.data, outr_dim[0] * outr_dim[1]) };
// Refer to https://github.com/dimforge/nalgebra/issues/735
let matrix1 = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
let res = matrix1.qr();
let (q, r) = res.unpack();
// Uses different algo need to match numpy
out_q_slice.copy_from_slice(q.transpose().as_slice());
out_r_slice.copy_from_slice(r.transpose().as_slice());
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn np_linalg_svd(
mat1: *mut InputMatrix,
outu: *mut InputMatrix,
outs: *mut InputMatrix,
outvh: *mut InputMatrix,
) {
let mat1 = mat1.as_mut().unwrap();
let outu = outu.as_mut().unwrap();
let outs = outs.as_mut().unwrap();
let outvh = outvh.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "np_linalg_svd", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
let outu_dim = (*outu).get_dims();
let outs_dim = (*outs).get_dims();
let outvh_dim = (*outvh).get_dims();
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let out_u_slice = unsafe { slice::from_raw_parts_mut(outu.data, outu_dim[0] * outu_dim[1]) };
let out_s_slice = unsafe { slice::from_raw_parts_mut(outs.data, outs_dim[0]) };
let out_vh_slice =
unsafe { slice::from_raw_parts_mut(outvh.data, outvh_dim[0] * outvh_dim[1]) };
let matrix = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
let result = matrix.svd(true, true);
out_u_slice.copy_from_slice(result.u.unwrap().transpose().as_slice());
out_s_slice.copy_from_slice(result.singular_values.as_slice());
out_vh_slice.copy_from_slice(result.v_t.unwrap().transpose().as_slice());
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn np_linalg_inv(mat1: *mut InputMatrix, out: *mut InputMatrix) {
let mat1 = mat1.as_mut().unwrap();
let out = out.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "np_linalg_inv", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
if dim1[0] != dim1[1] {
let err_msg =
format!("last 2 dimensions of the array must be square: {0} != {1}", dim1[0], dim1[1]);
report_error("LinAlgError", "np_linalg_inv", file!(), line!(), column!(), &err_msg);
}
let outdim = out.get_dims();
let out_slice = unsafe { slice::from_raw_parts_mut(out.data, outdim[0] * outdim[1]) };
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let matrix = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
if !matrix.is_invertible() {
report_error(
"LinAlgError",
"np_linalg_inv",
file!(),
line!(),
column!(),
"no inverse for Singular Matrix",
);
}
let inv = matrix.try_inverse().unwrap();
out_slice.copy_from_slice(inv.transpose().as_slice());
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn np_linalg_pinv(mat1: *mut InputMatrix, out: *mut InputMatrix) {
let mat1 = mat1.as_mut().unwrap();
let out = out.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "np_linalg_pinv", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
let outdim = out.get_dims();
let out_slice = unsafe { slice::from_raw_parts_mut(out.data, outdim[0] * outdim[1]) };
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let matrix = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
let svd = matrix.svd(true, true);
let inv = svd.pseudo_inverse(1e-15);
match inv {
Ok(m) => {
out_slice.copy_from_slice(m.transpose().as_slice());
}
Err(err_msg) => {
report_error("LinAlgError", "np_linalg_pinv", file!(), line!(), column!(), err_msg);
}
}
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn np_linalg_matrix_power(
mat1: *mut InputMatrix,
mat2: *mut InputMatrix,
out: *mut InputMatrix,
) {
let mat1 = mat1.as_mut().unwrap();
let mat2 = mat2.as_mut().unwrap();
let out = out.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D", mat1.ndims);
report_error("ValueError", "np_linalg_matrix_power", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
let power = unsafe { slice::from_raw_parts_mut(mat2.data, 1) };
let power = power[0];
let outdim = out.get_dims();
let out_slice = unsafe { slice::from_raw_parts_mut(out.data, outdim[0] * outdim[1]) };
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let abs_pow = power.abs();
let matrix1 = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
let mut result = matrix1.pow(abs_pow as u32);
if power < 0.0 {
if !result.is_invertible() {
report_error(
"LinAlgError",
"np_linalg_inv",
file!(),
line!(),
column!(),
"no inverse for Singular Matrix",
);
}
result = result.try_inverse().unwrap();
}
out_slice.copy_from_slice(result.transpose().as_slice());
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn np_linalg_det(mat1: *mut InputMatrix, out: *mut InputMatrix) {
let mat1 = mat1.as_mut().unwrap();
let out = out.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "np_linalg_det", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
let out_slice = unsafe { slice::from_raw_parts_mut(out.data, 1) };
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let matrix = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
if !matrix.is_square() {
let err_msg =
format!("last 2 dimensions of the array must be square: {0} != {1}", dim1[0], dim1[1]);
report_error("LinAlgError", "np_linalg_inv", file!(), line!(), column!(), &err_msg);
}
out_slice[0] = matrix.determinant();
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn sp_linalg_lu(
mat1: *mut InputMatrix,
out_l: *mut InputMatrix,
out_u: *mut InputMatrix,
) {
let mat1 = mat1.as_mut().unwrap();
let out_l = out_l.as_mut().unwrap();
let out_u = out_u.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "sp_linalg_lu", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
let outl_dim = (*out_l).get_dims();
let outu_dim = (*out_u).get_dims();
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let out_l_slice = unsafe { slice::from_raw_parts_mut(out_l.data, outl_dim[0] * outl_dim[1]) };
let out_u_slice = unsafe { slice::from_raw_parts_mut(out_u.data, outu_dim[0] * outu_dim[1]) };
let matrix = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
let (_, l, u) = matrix.lu().unpack();
out_l_slice.copy_from_slice(l.transpose().as_slice());
out_u_slice.copy_from_slice(u.transpose().as_slice());
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn sp_linalg_schur(
mat1: *mut InputMatrix,
out_t: *mut InputMatrix,
out_z: *mut InputMatrix,
) {
let mat1 = mat1.as_mut().unwrap();
let out_t = out_t.as_mut().unwrap();
let out_z = out_z.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "sp_linalg_schur", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
if dim1[0] != dim1[1] {
let err_msg =
format!("last 2 dimensions of the array must be square: {0} != {1}", dim1[0], dim1[1]);
report_error("LinAlgError", "np_linalg_schur", file!(), line!(), column!(), &err_msg);
}
let out_t_dim = (*out_t).get_dims();
let out_z_dim = (*out_z).get_dims();
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let out_t_slice = unsafe { slice::from_raw_parts_mut(out_t.data, out_t_dim[0] * out_t_dim[1]) };
let out_z_slice = unsafe { slice::from_raw_parts_mut(out_z.data, out_z_dim[0] * out_z_dim[1]) };
let matrix = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
let (z, t) = matrix.schur().unpack();
out_t_slice.copy_from_slice(t.transpose().as_slice());
out_z_slice.copy_from_slice(z.transpose().as_slice());
}
/// # Safety
///
/// `mat1` should point to a valid 2DArray of `f64` floats in row-major order
#[no_mangle]
pub unsafe extern "C" fn sp_linalg_hessenberg(
mat1: *mut InputMatrix,
out_h: *mut InputMatrix,
out_q: *mut InputMatrix,
) {
let mat1 = mat1.as_mut().unwrap();
let out_h = out_h.as_mut().unwrap();
let out_q = out_q.as_mut().unwrap();
if mat1.ndims != 2 {
let err_msg = format!("expected 2D Vector Input, but received {}D input", mat1.ndims);
report_error("ValueError", "sp_linalg_hessenberg", file!(), line!(), column!(), &err_msg);
}
let dim1 = (*mat1).get_dims();
if dim1[0] != dim1[1] {
let err_msg =
format!("last 2 dimensions of the array must be square: {} != {}", dim1[0], dim1[1]);
report_error("LinAlgError", "sp_linalg_hessenberg", file!(), line!(), column!(), &err_msg);
}
let out_h_dim = (*out_h).get_dims();
let out_q_dim = (*out_q).get_dims();
let data_slice1 = unsafe { slice::from_raw_parts_mut(mat1.data, dim1[0] * dim1[1]) };
let out_h_slice = unsafe { slice::from_raw_parts_mut(out_h.data, out_h_dim[0] * out_h_dim[1]) };
let out_q_slice = unsafe { slice::from_raw_parts_mut(out_q.data, out_q_dim[0] * out_q_dim[1]) };
let matrix = DMatrix::from_row_slice(dim1[0], dim1[1], data_slice1);
let (q, h) = matrix.hessenberg().unpack();
out_h_slice.copy_from_slice(h.transpose().as_slice());
out_q_slice.copy_from_slice(q.transpose().as_slice());
}

View File

@ -2,6 +2,9 @@
set -e
: "${DEMO_LINALG_STUB:=linalg/target/release/liblinalg.a}"
: "${DEMO_LINALG_STUB32:=linalg/target/i686-unknown-linux-gnu/release/liblinalg.a}"
if [ -z "$1" ]; then
echo "No argument supplied"
exit 1
@ -11,25 +14,26 @@ declare -a nac3args
while [ $# -ge 1 ]; do
case "$1" in
--help)
echo "Usage: run_demo.sh [--help] [--out OUTFILE] [--lli] [--debug] -- [NAC3ARGS...]"
echo "Usage: run_demo.sh [--help] [--out OUTFILE] [--debug] [-i686] -- [NAC3ARGS...] demo"
exit
;;
--out)
shift
outfile="$1"
;;
--lli)
use_lli=1
;;
--debug)
debug=1
;;
-i686)
i686=1
;;
--)
shift
break
;;
*)
break
echo "Unrecognized argument \"$1\""
exit 1
;;
esac
shift
@ -50,29 +54,19 @@ else
fi
rm -f ./*.o ./*.bc demo
if [ -z "$use_lli" ]; then
if [ -z "$i686" ]; 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
clang -o demo module.o demo.o $DEMO_LINALG_STUB -lm -Wl,--no-warn-search-mismatch
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
$nac3standalone --triple i686-unknown-linux-gnu --target-features +sse2 "${nac3args[@]}"
clang -m32 -c -std=gnu11 -Wall -Wextra -O3 -msse2 -o demo.o demo.c
clang -m32 -o demo module.o demo.o $DEMO_LINALG_STUB32 -lm -Wl,--no-warn-search-mismatch
fi
if [ -z "$outfile" ]; then
./demo
else
./demo > "$outfile"
fi

View File

@ -0,0 +1,66 @@
@extern
def output_int32(x: int32):
...
@extern
def output_bool(x: bool):
...
def example1():
x, *ys, z = (1, 2, 3, 4, 5)
output_int32(x)
output_int32(ys[0])
output_int32(ys[1])
output_int32(ys[2])
output_int32(z)
def example2():
x, y, *zs = (1, 2, 3, 4, 5)
output_int32(x)
output_int32(y)
output_int32(zs[0])
output_int32(zs[1])
output_int32(zs[2])
def example3():
*xs, y, z = (1, 2, 3, 4, 5)
output_int32(xs[0])
output_int32(xs[1])
output_int32(xs[2])
output_int32(y)
output_int32(z)
def example4():
# Example from: https://docs.python.org/3/reference/simple_stmts.html#assignment-statements
x = [0, 1]
i = 0
i, x[i] = 1, 2 # i is updated, then x[i] is updated
output_int32(i)
output_int32(x[0])
output_int32(x[1])
class A:
value: int32
def __init__(self):
self.value = 1000
def example5():
ws = [88, 7, 8]
a = A()
x, [y, *ys, a.value], ws[0], (ws[0],) = 1, (2, False, 4, 5), 99, (6,)
output_int32(x)
output_int32(y)
output_bool(ys[0])
output_int32(ys[1])
output_int32(a.value)
output_int32(ws[0])
output_int32(ws[1])
output_int32(ws[2])
def run() -> int32:
example1()
example2()
example3()
example4()
example5()
return 0

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