M-Labs/nac3
M-Labs
/
nac3
8
0
Fork 10
Code Issues 100 Pull Requests 20 Releases Wiki Activity

Compare commits

base: M-Labs:master
Branches Tags
M-Labs:master
M-Labs:ndstrides-14-end
M-Labs:ndstrides-13-tests
M-Labs:ndstrides-12-builtins
M-Labs:ndstrides-11-matmul
M-Labs:ndstrides-10-ops
M-Labs:ndstrides-9-subassign
M-Labs:ndstrides-8-transpose
M-Labs:ndstrides-7-broadcasting
M-Labs:ndstrides-6-reshape
M-Labs:ndstrides-5-miscfuncs
M-Labs:ndstrides-4-nparray
M-Labs:ndstrides-3-indexing
M-Labs:ndstrides-2-basic
M-Labs:ndstrides-1-model
M-Labs:ndstrides
M-Labs:refactor_composer
M-Labs:misc/update-pyo3
M-Labs:misc/impl-tracert
M-Labs:misc/mold
M-Labs:ndstrides-1-model-lifetime
M-Labs:ndstrides-enhance-irrt
M-Labs:ndstrides-intro
M-Labs:issue-313
M-Labs:refactor-primstore
M-Labs:numpy-anyall
M-Labs:enhance/cargo-test-standalone-demos
M-Labs:issue-136
M-Labs:rpc-obj-as-param
M-Labs:iterators
M-Labs:escape-analysis
M-Labs:refactor_anto
lyken:to-models-tmptmp
lyken:to-models-save
lyken:ndstrides-wip
lyken:nds3
lyken:ndstrides-wip-9
lyken:ndstrides-wip-transpose
lyken:ndarray-strides-6
lyken:ndarray-strides-5
lyken:ndarray-strides-wip-2
lyken:ndarray-strides
lyken:ndarray-strides-wip
lyken:master
lyken:misc/impl-tracert
lyken:misc/mold
lyken:misc/update-pyo3
lyken:issue-313
lyken:refactor-primstore
lyken:numpy-anyall
lyken:enhance/cargo-test-standalone-demos
lyken:issue-136
lyken:rpc-obj-as-param
lyken:iterators
lyken:escape-analysis
lyken:refactor_anto
..
compare: lyken:ndstrides-wip-9
Branches Tags
lyken:to-models-tmptmp
lyken:to-models-save
lyken:ndstrides-wip
lyken:nds3
lyken:ndstrides-wip-9
lyken:ndstrides-wip-transpose
lyken:ndarray-strides-6
lyken:ndarray-strides-5
lyken:ndarray-strides-wip-2
lyken:ndarray-strides
lyken:ndarray-strides-wip
lyken:master
lyken:misc/impl-tracert
lyken:misc/mold
lyken:misc/update-pyo3
lyken:issue-313
lyken:refactor-primstore
lyken:numpy-anyall
lyken:enhance/cargo-test-standalone-demos
lyken:issue-136
lyken:rpc-obj-as-param
lyken:iterators
lyken:escape-analysis
lyken:refactor_anto
M-Labs:master
M-Labs:ndstrides-14-end
M-Labs:ndstrides-13-tests
M-Labs:ndstrides-12-builtins
M-Labs:ndstrides-11-matmul
M-Labs:ndstrides-10-ops
M-Labs:ndstrides-9-subassign
M-Labs:ndstrides-8-transpose
M-Labs:ndstrides-7-broadcasting
M-Labs:ndstrides-6-reshape
M-Labs:ndstrides-5-miscfuncs
M-Labs:ndstrides-4-nparray
M-Labs:ndstrides-3-indexing
M-Labs:ndstrides-2-basic
M-Labs:ndstrides-1-model
M-Labs:ndstrides
M-Labs:refactor_composer
M-Labs:misc/update-pyo3
M-Labs:misc/impl-tracert
M-Labs:misc/mold
M-Labs:ndstrides-1-model-lifetime
M-Labs:ndstrides-enhance-irrt
M-Labs:ndstrides-intro
M-Labs:issue-313
M-Labs:refactor-primstore
M-Labs:numpy-anyall
M-Labs:enhance/cargo-test-standalone-demos
M-Labs:issue-136
M-Labs:rpc-obj-as-param
M-Labs:iterators
M-Labs:escape-analysis
M-Labs:refactor_anto

36 Commits
master ... ndstrides-

Author SHA1 Message Date
lyken 2ab7b299b8 core/ndstrides: refactor numpy indexing 2024-07-31 09:53:15 +08:00
lyken 86b0d31290 core/ndstrides: pub ScalarOrNDArray::to_basic_value_enum 2024-07-31 09:53:15 +08:00
lyken 6369db94ab core/codegen: gen_assign to take in value_ty 2024-07-31 09:53:15 +08:00
lyken 3d8240259c core/typecheck: Inferencer allow heterogenerous assignemnt 2024-07-31 09:53:15 +08:00
lyken e4f6adb1ec core/ndstrides: add numpy broadcasting utils 2024-07-31 09:53:15 +08:00
lyken eb295cf7e4 core/ndstrides: implement numpy broadcasting IRRT 2024-07-31 09:53:15 +08:00
lyken 7501a086d0 core/irrt: print_value add bool 2024-07-31 09:53:15 +08:00
lyken fb54d5d112 core/ndstrides: add TODO in np_reshape 2024-07-31 09:53:15 +08:00
lyken 3dc4b17310 core/ndstrides: introduce NDArrayObject & refactor reshape 2024-07-31 09:53:15 +08:00
lyken 7436513b64 core/model: add util.rs & gen_model_memcpy 2024-07-31 09:53:15 +08:00
lyken 7e056b9747 core/ndstrides: fix alloca_ndarray comment 2024-07-31 09:53:15 +08:00
lyken ac7cc15d90 core/ndstrides: remove unnecessary Result<_, String> 2024-07-31 09:53:15 +08:00
lyken 28e6f23034 core/ndstrides: rewrite and fix np_reshape() bug 
Data content should be copied and strides should be updated after
negative indices are resolved.
 2024-07-31 09:53:15 +08:00
lyken dfb8bf9748 core/ndstrides: fix and rewrite is_c_contiguous 2024-07-31 09:53:15 +08:00
lyken d5880b119a core/ndstrides: move functions to numpy_new/util.rs 2024-07-31 09:53:15 +08:00
lyken 2747869a45 core/ndstrides: implement general ndarray reshaping 2024-07-31 09:53:15 +08:00
lyken bd5cb14d0d core/ndstrides: implement general ndarray basic indexing 2024-07-31 09:53:15 +08:00
lyken 4b14609342 core/ndstrides: implement IRRT slice 
Needed by ndarray indexing
 2024-07-31 09:53:15 +08:00
lyken 2211c4d852 core/ndstrides: implement gen_foreach_ndarray_elements & np_{empty,ndarray,zeros,ones,full} 2024-07-31 09:53:15 +08:00
lyken 5b9ac9b09c core/ndstrides: implement ndarray len() 2024-07-31 09:53:15 +08:00
lyken 02e3ddfce6 core: make get_llvm_type return new NDArray with strides 
NOTE: All old numpy functions are now impossible to run, until NDArray
with strides is fully implemented.
 2024-07-31 09:53:15 +08:00
lyken 8ae9a4294b core/ndstrides: add basic ndarray IRRT functions 2024-07-31 09:53:15 +08:00
lyken e5fe86cc93 core/ndstrides: add ArrayWriter & make_shape_writer 2024-07-31 09:53:15 +08:00
lyken fd3d02bff0 core/ndstrides: add NDArray with strides definition 2024-07-31 09:53:15 +08:00
lyken 7502b14d55 core/irrt: add ErrorContext 2024-07-31 09:53:15 +08:00
lyken 5b7588df75 core/model: add and use CSlice and Exception 2024-07-31 09:53:15 +08:00
lyken 0477e2acfa core/irrt: comment arrays_match() 2024-07-31 09:53:15 +08:00
lyken bf0dcf325e core/irrt: add cstr_utils 2024-07-31 09:53:15 +08:00
lyken c772fdb83a core/model: introduce codegen/model 2024-07-31 09:53:15 +08:00
lyken c1369ea5bd core/irrt: introduce irrt testing 
`cargo test -F test` would compile `nac3core/irrt/irrt_test.cpp`
targetted to the host machine (it gets to use `std`) and run the
test executable.
 2024-07-31 09:52:43 +08:00
lyken ef28138291 core/irrt: split irrt.cpp into headers 
To scale IRRT implementations
 2024-07-31 09:52:43 +08:00
lyken 984843a46a core/irrt: build.rs capture IR defined constants 2024-07-31 09:52:43 +08:00
lyken c5626e4947 core/irrt: build.rs capture IR defined types 2024-07-31 09:52:43 +08:00
lyken e4ba5e6411 core/irrt: reformat 2024-07-31 09:52:43 +08:00
lyken 31d0fdd818 core: add .clang-format 2024-07-31 09:52:43 +08:00
lyken 3f0e7e28b8 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.
 2024-07-31 09:52:43 +08:00
143 changed files with 8081 additions and 8921 deletions
Show Stats Expand all files Collapse all files

33
.clang-format
View File

@ -1,32 +1,3 @@
BasedOnStyle: LLVM
Language: Cpp
Standard: Cpp11
AccessModifierOffset: -1
AlignEscapedNewlines: Left
AlwaysBreakAfterReturnType: None
AlwaysBreakTemplateDeclarations: Yes
AllowAllParametersOfDeclarationOnNextLine: false
AllowShortFunctionsOnASingleLine: Inline
BinPackParameters: false
BreakBeforeBinaryOperators: NonAssignment
BreakBeforeTernaryOperators: true
BreakConstructorInitializers: AfterColon
BreakInheritanceList: AfterColon
ColumnLimit: 120
ConstructorInitializerAllOnOneLineOrOnePerLine: true
ContinuationIndentWidth: 4
DerivePointerAlignment: false
IndentCaseLabels: true
IndentPPDirectives: None
BasedOnStyle: Google
IndentWidth: 4
MaxEmptyLinesToKeep: 1
PointerAlignment: Left
ReflowComments: true
SortIncludes: false
SortUsingDeclarations: true
SpaceAfterTemplateKeyword: false
SpacesBeforeTrailingComments: 2
TabWidth: 4
UseTab: Never
ReflowComments: false

1
.gitignore vendored
View File

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

8
.pre-commit-config.yaml
View File

@ -8,17 +8,17 @@ repos:
hooks:
- id: nac3-cargo-fmt
name: nac3 cargo format
entry: nix
entry: cargo
language: system
types: [file, rust]
pass_filenames: false
description: Runs cargo fmt on the codebase.
args: [develop, -c, cargo, fmt, --all]
args: [fmt]
- id: nac3-cargo-clippy
name: nac3 cargo clippy
entry: nix
entry: cargo
language: system
types: [file, rust]
pass_filenames: false
description: Runs cargo clippy on the codebase.
args: [develop, -c, cargo, clippy, --tests]
args: [clippy, --tests]

423
Cargo.lock generated
View File

@ -26,9 +26,9 @@ dependencies = [
[[package]]
name = "anstream"
version = "0.6.17"
version = "0.6.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "23a1e53f0f5d86382dafe1cf314783b2044280f406e7e1506368220ad11b1338"
checksum = "64e15c1ab1f89faffbf04a634d5e1962e9074f2741eef6d97f3c4e322426d526"
dependencies = [
"anstyle",
"anstyle-parse",
@ -41,67 +41,67 @@ dependencies = [
[[package]]
name = "anstyle"
version = "1.0.9"
version = "1.0.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8365de52b16c035ff4fcafe0092ba9390540e3e352870ac09933bebcaa2c8c56"
checksum = "1bec1de6f59aedf83baf9ff929c98f2ad654b97c9510f4e70cf6f661d49fd5b1"
[[package]]
name = "anstyle-parse"
version = "0.2.6"
version = "0.2.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3b2d16507662817a6a20a9ea92df6652ee4f94f914589377d69f3b21bc5798a9"
checksum = "eb47de1e80c2b463c735db5b217a0ddc39d612e7ac9e2e96a5aed1f57616c1cb"
dependencies = [
"utf8parse",
]
[[package]]
name = "anstyle-query"
version = "1.1.2"
version = "1.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "79947af37f4177cfead1110013d678905c37501914fba0efea834c3fe9a8d60c"
checksum = "6d36fc52c7f6c869915e99412912f22093507da8d9e942ceaf66fe4b7c14422a"
dependencies = [
"windows-sys 0.59.0",
"windows-sys",
]
[[package]]
name = "anstyle-wincon"
version = "3.0.6"
version = "3.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2109dbce0e72be3ec00bed26e6a7479ca384ad226efdd66db8fa2e3a38c83125"
checksum = "5bf74e1b6e971609db8ca7a9ce79fd5768ab6ae46441c572e46cf596f59e57f8"
dependencies = [
"anstyle",
"windows-sys 0.59.0",
"windows-sys",
]
[[package]]
name = "ascii-canvas"
version = "4.0.0"
version = "3.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ef1e3e699d84ab1b0911a1010c5c106aa34ae89aeac103be5ce0c3859db1e891"
checksum = "8824ecca2e851cec16968d54a01dd372ef8f95b244fb84b84e70128be347c3c6"
dependencies = [
"term",
]
[[package]]
name = "autocfg"
version = "1.4.0"
version = "1.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ace50bade8e6234aa140d9a2f552bbee1db4d353f69b8217bc503490fc1a9f26"
checksum = "0c4b4d0bd25bd0b74681c0ad21497610ce1b7c91b1022cd21c80c6fbdd9476b0"
[[package]]
name = "bit-set"
version = "0.8.0"
version = "0.5.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "08807e080ed7f9d5433fa9b275196cfc35414f66a0c79d864dc51a0d825231a3"
checksum = "0700ddab506f33b20a03b13996eccd309a48e5ff77d0d95926aa0210fb4e95f1"
dependencies = [
"bit-vec",
]
[[package]]
name = "bit-vec"
version = "0.8.0"
version = "0.6.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5e764a1d40d510daf35e07be9eb06e75770908c27d411ee6c92109c9840eaaf7"
checksum = "349f9b6a179ed607305526ca489b34ad0a41aed5f7980fa90eb03160b69598fb"
[[package]]
name = "bitflags"
@ -109,15 +109,6 @@ version = "2.6.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b048fb63fd8b5923fc5aa7b340d8e156aec7ec02f0c78fa8a6ddc2613f6f71de"
[[package]]
name = "block-buffer"
version = "0.10.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3078c7629b62d3f0439517fa394996acacc5cbc91c5a20d8c658e77abd503a71"
dependencies = [
"generic-array",
]
[[package]]
name = "byteorder"
version = "1.5.0"
@ -126,12 +117,9 @@ checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
[[package]]
name = "cc"
version = "1.1.31"
version = "1.1.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c2e7962b54006dcfcc61cb72735f4d89bb97061dd6a7ed882ec6b8ee53714c6f"
dependencies = [
"shlex",
]
checksum = "2aba8f4e9906c7ce3c73463f62a7f0c65183ada1a2d47e397cc8810827f9694f"
[[package]]
name = "cfg-if"
@ -141,9 +129,9 @@ checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
[[package]]
name = "clap"
version = "4.5.20"
version = "4.5.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b97f376d85a664d5837dbae44bf546e6477a679ff6610010f17276f686d867e8"
checksum = "35723e6a11662c2afb578bcf0b88bf6ea8e21282a953428f240574fcc3a2b5b3"
dependencies = [
"clap_builder",
"clap_derive",
@ -151,9 +139,9 @@ dependencies = [
[[package]]
name = "clap_builder"
version = "4.5.20"
version = "4.5.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "19bc80abd44e4bed93ca373a0704ccbd1b710dc5749406201bb018272808dc54"
checksum = "49eb96cbfa7cfa35017b7cd548c75b14c3118c98b423041d70562665e07fb0fa"
dependencies = [
"anstream",
"anstyle",
@ -163,14 +151,14 @@ dependencies = [
[[package]]
name = "clap_derive"
version = "4.5.18"
version = "4.5.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4ac6a0c7b1a9e9a5186361f67dfa1b88213572f427fb9ab038efb2bd8c582dab"
checksum = "5d029b67f89d30bbb547c89fd5161293c0aec155fc691d7924b64550662db93e"
dependencies = [
"heck 0.5.0",
"proc-macro2",
"quote",
"syn 2.0.85",
"syn 2.0.72",
]
[[package]]
@ -181,9 +169,9 @@ checksum = "1462739cb27611015575c0c11df5df7601141071f07518d56fcc1be504cbec97"
[[package]]
name = "colorchoice"
version = "1.0.3"
version = "1.0.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5b63caa9aa9397e2d9480a9b13673856c78d8ac123288526c37d7839f2a86990"
checksum = "d3fd119d74b830634cea2a0f58bbd0d54540518a14397557951e79340abc28c0"
[[package]]
name = "console"
@ -194,16 +182,7 @@ dependencies = [
"encode_unicode",
"lazy_static",
"libc",
"windows-sys 0.52.0",
]
[[package]]
name = "cpufeatures"
version = "0.2.14"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "608697df725056feaccfa42cffdaeeec3fccc4ffc38358ecd19b243e716a78e0"
dependencies = [
"libc",
"windows-sys",
]
[[package]]
@ -263,23 +242,30 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "22ec99545bb0ed0ea7bb9b8e1e9122ea386ff8a48c0922e43f36d45ab09e0e80"
[[package]]
name = "crypto-common"
version = "0.1.6"
name = "crunchy"
version = "0.2.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1bfb12502f3fc46cca1bb51ac28df9d618d813cdc3d2f25b9fe775a34af26bb3"
checksum = "7a81dae078cea95a014a339291cec439d2f232ebe854a9d672b796c6afafa9b7"
[[package]]
name = "dirs-next"
version = "2.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b98cf8ebf19c3d1b223e151f99a4f9f0690dca41414773390fc824184ac833e1"
dependencies = [
"generic-array",
"typenum",
"cfg-if",
"dirs-sys-next",
]
[[package]]
name = "digest"
version = "0.10.7"
name = "dirs-sys-next"
version = "0.1.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9ed9a281f7bc9b7576e61468ba615a66a5c8cfdff42420a70aa82701a3b1e292"
checksum = "4ebda144c4fe02d1f7ea1a7d9641b6fc6b580adcfa024ae48797ecdeb6825b4d"
dependencies = [
"block-buffer",
"crypto-common",
"libc",
"redox_users",
"winapi",
]
[[package]]
@ -316,14 +302,14 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "534c5cf6194dfab3db3242765c03bbe257cf92f22b38f6bc0c58d59108a820ba"
dependencies = [
"libc",
"windows-sys 0.52.0",
"windows-sys",
]
[[package]]
name = "fastrand"
version = "2.1.1"
version = "2.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e8c02a5121d4ea3eb16a80748c74f5549a5665e4c21333c6098f283870fbdea6"
checksum = "9fc0510504f03c51ada170672ac806f1f105a88aa97a5281117e1ddc3368e51a"
[[package]]
name = "fixedbitset"
@ -340,16 +326,6 @@ dependencies = [
"byteorder",
]
[[package]]
name = "generic-array"
version = "0.14.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "85649ca51fd72272d7821adaf274ad91c288277713d9c18820d8499a7ff69e9a"
dependencies = [
"typenum",
"version_check",
]
[[package]]
name = "getopts"
version = "0.2.21"
@ -385,12 +361,6 @@ dependencies = [
"ahash",
]
[[package]]
name = "hashbrown"
version = "0.15.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1e087f84d4f86bf4b218b927129862374b72199ae7d8657835f1e89000eea4fb"
[[package]]
name = "heck"
version = "0.4.1"
@ -403,15 +373,6 @@ version = "0.5.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
[[package]]
name = "home"
version = "0.5.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e3d1354bf6b7235cb4a0576c2619fd4ed18183f689b12b006a0ee7329eeff9a5"
dependencies = [
"windows-sys 0.52.0",
]
[[package]]
name = "indexmap"
version = "1.9.3"
@ -424,12 +385,12 @@ dependencies = [
[[package]]
name = "indexmap"
version = "2.6.0"
version = "2.2.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "707907fe3c25f5424cce2cb7e1cbcafee6bdbe735ca90ef77c29e84591e5b9da"
checksum = "168fb715dda47215e360912c096649d23d58bf392ac62f73919e831745e40f26"
dependencies = [
"equivalent",
"hashbrown 0.15.0",
"hashbrown 0.14.5",
]
[[package]]
@ -440,9 +401,9 @@ checksum = "b248f5224d1d606005e02c97f5aa4e88eeb230488bcc03bc9ca4d7991399f2b5"
[[package]]
name = "inkwell"
version = "0.5.0"
version = "0.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "40fb405537710d51f6bdbc8471365ddd4cd6d3a3c3ad6e0c8291691031ba94b2"
checksum = "b597a7b2cdf279aeef6d7149071e35e4bc87c2cf05a5b7f2d731300bffe587ea"
dependencies = [
"either",
"inkwell_internals",
@ -454,13 +415,13 @@ dependencies = [
[[package]]
name = "inkwell_internals"
version = "0.10.0"
version = "0.9.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9dd28cfd4cfba665d47d31c08a6ba637eed16770abca2eccbbc3ca831fef1e44"
checksum = "4fa4d8d74483041a882adaa9a29f633253a66dde85055f0495c121620ac484b2"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.85",
"syn 2.0.72",
]
[[package]]
@ -483,6 +444,15 @@ version = "1.70.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7943c866cc5cd64cbc25b2e01621d07fa8eb2a1a23160ee81ce38704e97b8ecf"
[[package]]
name = "itertools"
version = "0.11.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b1c173a5686ce8bfa551b3563d0c2170bf24ca44da99c7ca4bfdab5418c3fe57"
dependencies = [
"either",
]
[[package]]
name = "itertools"
version = "0.13.0"
@ -498,45 +468,35 @@ version = "1.0.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "49f1f14873335454500d59611f1cf4a4b0f786f9ac11f4312a78e4cf2566695b"
[[package]]
name = "keccak"
version = "0.1.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ecc2af9a1119c51f12a14607e783cb977bde58bc069ff0c3da1095e635d70654"
dependencies = [
"cpufeatures",
]
[[package]]
name = "lalrpop"
version = "0.22.0"
version = "0.20.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "06093b57658c723a21da679530e061a8c25340fa5a6f98e313b542268c7e2a1f"
checksum = "55cb077ad656299f160924eb2912aa147d7339ea7d69e1b5517326fdcec3c1ca"
dependencies = [
"ascii-canvas",
"bit-set",
"ena",
"itertools",
"itertools 0.11.0",
"lalrpop-util",
"petgraph",
"pico-args",
"regex",
"regex-syntax",
"sha3",
"string_cache",
"term",
"tiny-keccak",
"unicode-xid",
"walkdir",
]
[[package]]
name = "lalrpop-util"
version = "0.22.0"
version = "0.20.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "feee752d43abd0f4807a921958ab4131f692a44d4d599733d4419c5d586176ce"
checksum = "507460a910eb7b32ee961886ff48539633b788a36b65692b95f225b844c82553"
dependencies = [
"regex-automata",
"rustversion",
]
[[package]]
@ -547,9 +507,9 @@ checksum = "bbd2bcb4c963f2ddae06a2efc7e9f3591312473c50c6685e1f298068316e66fe"
[[package]]
name = "libc"
version = "0.2.161"
version = "0.2.155"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8e9489c2807c139ffd9c1794f4af0ebe86a828db53ecdc7fea2111d0fed085d1"
checksum = "97b3888a4aecf77e811145cadf6eef5901f4782c53886191b2f693f24761847c"
[[package]]
name = "libloading"
@ -561,6 +521,16 @@ dependencies = [
"windows-targets",
]
[[package]]
name = "libredox"
version = "0.1.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c0ff37bd590ca25063e35af745c343cb7a0271906fb7b37e4813e8f79f00268d"
dependencies = [
"bitflags",
"libc",
]
[[package]]
name = "linked-hash-map"
version = "0.5.6"
@ -621,9 +591,11 @@ dependencies = [
name = "nac3artiq"
version = "0.1.0"
dependencies = [
"itertools",
"inkwell",
"itertools 0.13.0",
"nac3core",
"nac3ld",
"nac3parser",
"parking_lot",
"pyo3",
"tempfile",
@ -634,6 +606,7 @@ name = "nac3ast"
version = "0.1.0"
dependencies = [
"fxhash",
"lazy_static",
"parking_lot",
"string-interner",
]
@ -643,11 +616,11 @@ name = "nac3core"
version = "0.1.0"
dependencies = [
"crossbeam",
"indexmap 2.6.0",
"indexmap 2.2.6",
"indoc",
"inkwell",
"insta",
"itertools",
"itertools 0.13.0",
"nac3parser",
"parking_lot",
"rayon",
@ -685,7 +658,9 @@ name = "nac3standalone"
version = "0.1.0"
dependencies = [
"clap",
"inkwell",
"nac3core",
"nac3parser",
"parking_lot",
]
@ -697,9 +672,9 @@ checksum = "650eef8c711430f1a879fdd01d4745a7deea475becfb90269c06775983bbf086"
[[package]]
name = "once_cell"
version = "1.20.2"
version = "1.19.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1261fe7e33c73b354eab43b1273a57c8f967d0391e80353e51f764ac02cf6775"
checksum = "3fdb12b2476b595f9358c5161aa467c2438859caa136dec86c26fdd2efe17b92"
[[package]]
name = "parking_lot"
@ -731,7 +706,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b4c5cc86750666a3ed20bdaf5ca2a0344f9c67674cae0515bec2da16fbaa47db"
dependencies = [
"fixedbitset",
"indexmap 2.6.0",
"indexmap 2.2.6",
]
[[package]]
@ -774,7 +749,7 @@ dependencies = [
"phf_shared 0.11.2",
"proc-macro2",
"quote",
"syn 2.0.85",
"syn 2.0.72",
]
[[package]]
@ -803,18 +778,15 @@ checksum = "5be167a7af36ee22fe3115051bc51f6e6c7054c9348e28deb4f49bd6f705a315"
[[package]]
name = "portable-atomic"
version = "1.9.0"
version = "1.7.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cc9c68a3f6da06753e9335d63e27f6b9754dd1920d941135b7ea8224f141adb2"
checksum = "da544ee218f0d287a911e9c99a39a8c9bc8fcad3cb8db5959940044ecfc67265"
[[package]]
name = "ppv-lite86"
version = "0.2.20"
version = "0.2.17"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "77957b295656769bb8ad2b6a6b09d897d94f05c41b069aede1fcdaa675eaea04"
dependencies = [
"zerocopy",
]
checksum = "5b40af805b3121feab8a3c29f04d8ad262fa8e0561883e7653e024ae4479e6de"
[[package]]
name = "precomputed-hash"
@ -824,9 +796,9 @@ checksum = "925383efa346730478fb4838dbe9137d2a47675ad789c546d150a6e1dd4ab31c"
[[package]]
name = "proc-macro2"
version = "1.0.89"
version = "1.0.86"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f139b0662de085916d1fb67d2b4169d1addddda1919e696f3252b740b629986e"
checksum = "5e719e8df665df0d1c8fbfd238015744736151d4445ec0836b8e628aae103b77"
dependencies = [
"unicode-ident",
]
@ -878,7 +850,7 @@ dependencies = [
"proc-macro2",
"pyo3-macros-backend",
"quote",
"syn 2.0.85",
"syn 2.0.72",
]
[[package]]
@ -891,14 +863,14 @@ dependencies = [
"proc-macro2",
"pyo3-build-config",
"quote",
"syn 2.0.85",
"syn 2.0.72",
]
[[package]]
name = "quote"
version = "1.0.37"
version = "1.0.36"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b5b9d34b8991d19d98081b46eacdd8eb58c6f2b201139f7c5f643cc155a633af"
checksum = "0fa76aaf39101c457836aec0ce2316dbdc3ab723cdda1c6bd4e6ad4208acaca7"
dependencies = [
"proc-macro2",
]
@ -955,18 +927,29 @@ dependencies = [
[[package]]
name = "redox_syscall"
version = "0.5.7"
version = "0.5.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9b6dfecf2c74bce2466cabf93f6664d6998a69eb21e39f4207930065b27b771f"
checksum = "2a908a6e00f1fdd0dfd9c0eb08ce85126f6d8bbda50017e74bc4a4b7d4a926a4"
dependencies = [
"bitflags",
]
[[package]]
name = "regex"
version = "1.11.1"
name = "redox_users"
version = "0.4.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b544ef1b4eac5dc2db33ea63606ae9ffcfac26c1416a2806ae0bf5f56b201191"
checksum = "bd283d9651eeda4b2a83a43c1c91b266c40fd76ecd39a50a8c630ae69dc72891"
dependencies = [
"getrandom",
"libredox",
"thiserror",
]
[[package]]
name = "regex"
version = "1.10.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b91213439dad192326a0d7c6ee3955910425f441d7038e0d6933b0aec5c4517f"
dependencies = [
"aho-corasick",
"memchr",
@ -976,9 +959,9 @@ dependencies = [
[[package]]
name = "regex-automata"
version = "0.4.8"
version = "0.4.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "368758f23274712b504848e9d5a6f010445cc8b87a7cdb4d7cbee666c1288da3"
checksum = "38caf58cc5ef2fed281f89292ef23f6365465ed9a41b7a7754eb4e26496c92df"
dependencies = [
"aho-corasick",
"memchr",
@ -987,9 +970,9 @@ dependencies = [
[[package]]
name = "regex-syntax"
version = "0.8.5"
version = "0.8.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2b15c43186be67a4fd63bee50d0303afffcef381492ebe2c5d87f324e1b8815c"
checksum = "7a66a03ae7c801facd77a29370b4faec201768915ac14a721ba36f20bc9c209b"
[[package]]
name = "runkernel"
@ -1000,22 +983,22 @@ dependencies = [
[[package]]
name = "rustix"
version = "0.38.38"
version = "0.38.34"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "aa260229e6538e52293eeb577aabd09945a09d6d9cc0fc550ed7529056c2e32a"
checksum = "70dc5ec042f7a43c4a73241207cecc9873a06d45debb38b329f8541d85c2730f"
dependencies = [
"bitflags",
"errno",
"libc",
"linux-raw-sys",
"windows-sys 0.52.0",
"windows-sys",
]
[[package]]
name = "rustversion"
version = "1.0.18"
version = "1.0.17"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0e819f2bc632f285be6d7cd36e25940d45b2391dd6d9b939e79de557f7014248"
checksum = "955d28af4278de8121b7ebeb796b6a45735dc01436d898801014aced2773a3d6"
[[package]]
name = "ryu"
@ -1046,32 +1029,31 @@ checksum = "61697e0a1c7e512e84a621326239844a24d8207b4669b41bc18b32ea5cbf988b"
[[package]]
name = "serde"
version = "1.0.214"
version = "1.0.204"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f55c3193aca71c12ad7890f1785d2b73e1b9f63a0bbc353c08ef26fe03fc56b5"
checksum = "bc76f558e0cbb2a839d37354c575f1dc3fdc6546b5be373ba43d95f231bf7c12"
dependencies = [
"serde_derive",
]
[[package]]
name = "serde_derive"
version = "1.0.214"
version = "1.0.204"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "de523f781f095e28fa605cdce0f8307e451cc0fd14e2eb4cd2e98a355b147766"
checksum = "e0cd7e117be63d3c3678776753929474f3b04a43a080c744d6b0ae2a8c28e222"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.85",
"syn 2.0.72",
]
[[package]]
name = "serde_json"
version = "1.0.132"
version = "1.0.120"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d726bfaff4b320266d395898905d0eba0345aae23b54aee3a737e260fd46db03"
checksum = "4e0d21c9a8cae1235ad58a00c11cb40d4b1e5c784f1ef2c537876ed6ffd8b7c5"
dependencies = [
"itoa",
"memchr",
"ryu",
"serde",
]
@ -1088,22 +1070,6 @@ dependencies = [
"yaml-rust",
]
[[package]]
name = "sha3"
version = "0.10.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "75872d278a8f37ef87fa0ddbda7802605cb18344497949862c0d4dcb291eba60"
dependencies = [
"digest",
"keccak",
]
[[package]]
name = "shlex"
version = "1.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0fda2ff0d084019ba4d7c6f371c95d8fd75ce3524c3cb8fb653a3023f6323e64"
[[package]]
name = "similar"
version = "2.6.0"
@ -1168,7 +1134,7 @@ dependencies = [
"proc-macro2",
"quote",
"rustversion",
"syn 2.0.85",
"syn 2.0.72",
]
[[package]]
@ -1184,9 +1150,9 @@ dependencies = [
[[package]]
name = "syn"
version = "2.0.85"
version = "2.0.72"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5023162dfcd14ef8f32034d8bcd4cc5ddc61ef7a247c024a33e24e1f24d21b56"
checksum = "dc4b9b9bf2add8093d3f2c0204471e951b2285580335de42f9d2534f3ae7a8af"
dependencies = [
"proc-macro2",
"quote",
@ -1195,31 +1161,31 @@ dependencies = [
[[package]]
name = "target-lexicon"
version = "0.12.16"
version = "0.12.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "61c41af27dd6d1e27b1b16b489db798443478cef1f06a660c96db617ba5de3b1"
checksum = "4873307b7c257eddcb50c9bedf158eb669578359fb28428bef438fec8e6ba7c2"
[[package]]
name = "tempfile"
version = "3.13.0"
version = "3.10.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f0f2c9fc62d0beef6951ccffd757e241266a2c833136efbe35af6cd2567dca5b"
checksum = "85b77fafb263dd9d05cbeac119526425676db3784113aa9295c88498cbf8bff1"
dependencies = [
"cfg-if",
"fastrand",
"once_cell",
"rustix",
"windows-sys 0.59.0",
"windows-sys",
]
[[package]]
name = "term"
version = "1.0.0"
version = "0.7.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4df4175de05129f31b80458c6df371a15e7fc3fd367272e6bf938e5c351c7ea0"
checksum = "c59df8ac95d96ff9bede18eb7300b0fda5e5d8d90960e76f8e14ae765eedbf1f"
dependencies = [
"home",
"windows-sys 0.52.0",
"dirs-next",
"rustversion",
"winapi",
]
[[package]]
@ -1237,29 +1203,32 @@ dependencies = [
[[package]]
name = "thiserror"
version = "1.0.65"
version = "1.0.63"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5d11abd9594d9b38965ef50805c5e469ca9cc6f197f883f717e0269a3057b3d5"
checksum = "c0342370b38b6a11b6cc11d6a805569958d54cfa061a29969c3b5ce2ea405724"
dependencies = [
"thiserror-impl",
]
[[package]]
name = "thiserror-impl"
version = "1.0.65"
version = "1.0.63"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ae71770322cbd277e69d762a16c444af02aa0575ac0d174f0b9562d3b37f8602"
checksum = "a4558b58466b9ad7ca0f102865eccc95938dca1a74a856f2b57b6629050da261"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.85",
"syn 2.0.72",
]
[[package]]
name = "typenum"
version = "1.17.0"
name = "tiny-keccak"
version = "2.0.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "42ff0bf0c66b8238c6f3b578df37d0b7848e55df8577b3f74f92a69acceeb825"
checksum = "2c9d3793400a45f954c52e73d068316d76b6f4e36977e3fcebb13a2721e80237"
dependencies = [
"crunchy",
]
[[package]]
name = "unic-char-property"
@ -1315,27 +1284,27 @@ dependencies = [
[[package]]
name = "unicode-ident"
version = "1.0.13"
version = "1.0.12"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e91b56cd4cadaeb79bbf1a5645f6b4f8dc5bde8834ad5894a8db35fda9efa1fe"
checksum = "3354b9ac3fae1ff6755cb6db53683adb661634f67557942dea4facebec0fee4b"
[[package]]
name = "unicode-width"
version = "0.1.14"
version = "0.1.13"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7dd6e30e90baa6f72411720665d41d89b9a3d039dc45b8faea1ddd07f617f6af"
checksum = "0336d538f7abc86d282a4189614dfaa90810dfc2c6f6427eaf88e16311dd225d"
[[package]]
name = "unicode-xid"
version = "0.2.6"
version = "0.2.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"
checksum = "f962df74c8c05a667b5ee8bcf162993134c104e96440b663c8daa176dc772d8c"
[[package]]
name = "unicode_names2"
version = "1.3.0"
version = "1.2.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d1673eca9782c84de5f81b82e4109dcfb3611c8ba0d52930ec4a9478f547b2dd"
checksum = "addeebf294df7922a1164f729fb27ebbbcea99cc32b3bf08afab62757f707677"
dependencies = [
"phf",
"unicode_names2_generator",
@ -1343,9 +1312,9 @@ dependencies = [
[[package]]
name = "unicode_names2_generator"
version = "1.3.0"
version = "1.2.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b91e5b84611016120197efd7dc93ef76774f4e084cd73c9fb3ea4a86c570c56e"
checksum = "f444b8bba042fe3c1251ffaca35c603f2dc2ccc08d595c65a8c4f76f3e8426c0"
dependencies = [
"getopts",
"log",
@ -1388,14 +1357,36 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9c8d87e72b64a3b4db28d11ce29237c246188f4f51057d65a7eab63b7987e423"
[[package]]
name = "winapi-util"
version = "0.1.9"
name = "winapi"
version = "0.3.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cf221c93e13a30d793f7645a0e7762c55d169dbb0a49671918a2319d289b10bb"
checksum = "5c839a674fcd7a98952e593242ea400abe93992746761e38641405d28b00f419"
dependencies = [
"windows-sys 0.59.0",
"winapi-i686-pc-windows-gnu",
"winapi-x86_64-pc-windows-gnu",
]
[[package]]
name = "winapi-i686-pc-windows-gnu"
version = "0.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ac3b87c63620426dd9b991e5ce0329eff545bccbbb34f3be09ff6fb6ab51b7b6"
[[package]]
name = "winapi-util"
version = "0.1.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4d4cc384e1e73b93bafa6fb4f1df8c41695c8a91cf9c4c64358067d15a7b6c6b"
dependencies = [
"windows-sys",
]
[[package]]
name = "winapi-x86_64-pc-windows-gnu"
version = "0.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"
[[package]]
name = "windows-sys"
version = "0.52.0"
@ -1405,15 +1396,6 @@ 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"
@ -1493,7 +1475,6 @@ version = "0.7.35"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1b9b4fd18abc82b8136838da5d50bae7bdea537c574d8dc1a34ed098d6c166f0"
dependencies = [
"byteorder",
"zerocopy-derive",
]
@ -1505,5 +1486,5 @@ checksum = "fa4f8080344d4671fb4e831a13ad1e68092748387dfc4f55e356242fae12ce3e"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.85",
"syn 2.0.72",
]

6
flake.lock
View File

@ -2,11 +2,11 @@
"nodes": {
"nixpkgs": {
"locked": {
"lastModified": 1727348695,
"narHash": "sha256-J+PeFKSDV+pHL7ukkfpVzCOO7mBSrrpJ3svwBFABbhI=",
"lastModified": 1721924956,
"narHash": "sha256-Sb1jlyRO+N8jBXEX9Pg9Z1Qb8Bw9QyOgLDNMEpmjZ2M=",
"owner": "NixOS",
"repo": "nixpkgs",
"rev": "1925c603f17fc89f4c8f6bf6f631a802ad85d784",
"rev": "5ad6a14c6bf098e98800b091668718c336effc95",
"type": "github"
},
"original": {

28
flake.nix
View File

@ -6,7 +6,6 @@
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 {};
@ -14,24 +13,9 @@
''
mkdir -p $out/bin
ln -s ${pkgs.llvmPackages_14.clang-unwrapped}/bin/clang $out/bin/clang-irrt
ln -s ${pkgs.llvmPackages_14.clang}/bin/clang $out/bin/clang-irrt-test
ln -s ${pkgs.llvmPackages_14.llvm.out}/bin/llvm-as $out/bin/llvm-as-irrt
'';
demo-linalg-stub = pkgs.rustPlatform.buildRustPackage {
name = "demo-linalg-stub";
src = ./nac3standalone/demo/linalg;
cargoLock = {
lockFile = ./nac3standalone/demo/linalg/Cargo.lock;
};
doCheck = false;
};
demo-linalg-stub32 = pkgs32.rustPlatform.buildRustPackage {
name = "demo-linalg-stub32";
src = ./nac3standalone/demo/linalg;
cargoLock = {
lockFile = ./nac3standalone/demo/linalg/Cargo.lock;
};
doCheck = false;
};
nac3artiq = pkgs.python3Packages.toPythonModule (
pkgs.rustPlatform.buildRustPackage rec {
name = "nac3artiq";
@ -40,6 +24,7 @@
cargoLock = {
lockFile = ./Cargo.lock;
};
cargoTestFlags = [ "--features" "test" ];
passthru.cargoLock = cargoLock;
nativeBuildInputs = [ pkgs.python3 (pkgs.wrapClangMulti pkgs.llvmPackages_14.clang) llvm-tools-irrt pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
buildInputs = [ pkgs.python3 llvm-nac3 ];
@ -49,9 +34,7 @@
echo "Checking nac3standalone demos..."
pushd nac3standalone/demo
patchShebangs .
export DEMO_LINALG_STUB=${demo-linalg-stub}/lib/liblinalg.a
export DEMO_LINALG_STUB32=${demo-linalg-stub32}/lib/liblinalg.a
./check_demos.sh -i686
./check_demos.sh
popd
echo "Running Cargo tests..."
cargoCheckHook
@ -181,11 +164,6 @@
pre-commit
rustfmt
];
shellHook =
''
export DEMO_LINALG_STUB=${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";

9
nac3artiq/Cargo.toml
View File

@ -12,10 +12,15 @@ crate-type = ["cdylib"]
itertools = "0.13"
pyo3 = { version = "0.21", features = ["extension-module", "gil-refs"] }
parking_lot = "0.12"
tempfile = "3.13"
tempfile = "3.10"
nac3parser = { path = "../nac3parser" }
nac3core = { path = "../nac3core" }
nac3ld = { path = "../nac3ld" }
[dependencies.inkwell]
version = "0.4"
default-features = false
features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
[features]
init-llvm-profile = []
no-escape-analysis = ["nac3core/no-escape-analysis"]

13
nac3artiq/demo/min_artiq.py
View File

@ -112,15 +112,10 @@ def extern(function):
register_function(function)
return function
def rpc(arg=None, flags={}):
"""Decorates a function or method to be executed on the host interpreter."""
if arg is None:
def inner_decorator(function):
return rpc(function, flags)
return inner_decorator
register_function(arg)
return arg
def rpc(function):
"""Decorates a function declaration defined by the core device runtime."""
register_function(function)
return function
def kernel(function_or_method):
"""Decorates a function or method to be executed on the core device."""

26
nac3artiq/demo/str_abi.py
View File

@ -1,26 +0,0 @@
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()

1067
nac3artiq/src/codegen.rs
View File

File diff suppressed because it is too large Load Diff

258
nac3artiq/src/lib.rs
View File

@ -16,65 +16,64 @@
clippy::wildcard_imports
)]
use std::{
collections::{HashMap, HashSet},
fs,
io::Write,
process::Command,
rc::Rc,
sync::Arc,
};
use std::collections::{HashMap, HashSet};
use std::fs;
use std::io::Write;
use std::process::Command;
use std::rc::Rc;
use std::sync::Arc;
use itertools::Itertools;
use parking_lot::{Mutex, RwLock};
use pyo3::{
create_exception, exceptions,
prelude::*,
types::{PyBytes, PyDict, PySet},
use inkwell::{
context::Context,
memory_buffer::MemoryBuffer,
module::{Linkage, Module},
passes::PassBuilderOptions,
support::is_multithreaded,
targets::*,
OptimizationLevel,
};
use tempfile::{self, TempDir};
use itertools::Itertools;
use nac3core::codegen::{gen_func_impl, CodeGenLLVMOptions, CodeGenTargetMachineOptions};
use nac3core::toplevel::builtins::get_exn_constructor;
use nac3core::typecheck::typedef::{TypeEnum, Unifier, VarMap};
use nac3parser::{
ast::{ExprKind, Stmt, StmtKind, StrRef},
parser::parse_program,
};
use pyo3::create_exception;
use pyo3::prelude::*;
use pyo3::{exceptions, types::PyBytes, types::PyDict, types::PySet};
use parking_lot::{Mutex, RwLock};
use nac3core::{
codegen::{
concrete_type::ConcreteTypeStore, gen_func_impl, irrt::load_irrt, CodeGenLLVMOptions,
CodeGenTargetMachineOptions, CodeGenTask, WithCall, WorkerRegistry,
},
inkwell::{
context::Context,
memory_buffer::MemoryBuffer,
module::{Linkage, Module},
passes::PassBuilderOptions,
support::is_multithreaded,
targets::*,
OptimizationLevel,
},
nac3parser::{
ast::{Constant, ExprKind, Located, Stmt, StmtKind, StrRef},
parser::parse_program,
},
codegen::irrt::load_irrt,
codegen::{concrete_type::ConcreteTypeStore, CodeGenTask, WithCall, WorkerRegistry},
symbol_resolver::SymbolResolver,
toplevel::{
builtins::get_exn_constructor,
composer::{BuiltinFuncCreator, BuiltinFuncSpec, ComposerConfig, TopLevelComposer},
composer::{ComposerConfig, TopLevelComposer},
DefinitionId, GenCall, TopLevelDef,
},
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{into_var_map, FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
},
typecheck::typedef::{FunSignature, FuncArg},
typecheck::{type_inferencer::PrimitiveStore, typedef::Type},
};
use nac3ld::Linker;
use codegen::{
attributes_writeback, gen_core_log, gen_rtio_log, rpc_codegen_callback, ArtiqCodeGenerator,
use tempfile::{self, TempDir};
use crate::codegen::attributes_writeback;
use crate::{
codegen::{rpc_codegen_callback, ArtiqCodeGenerator},
symbol_resolver::{DeferredEvaluationStore, InnerResolver, PythonHelper, Resolver},
};
use symbol_resolver::{DeferredEvaluationStore, InnerResolver, PythonHelper, Resolver};
use timeline::TimeFns;
mod codegen;
mod symbol_resolver;
mod timeline;
use timeline::TimeFns;
#[derive(PartialEq, Clone, Copy)]
enum Isa {
Host,
@ -128,7 +127,7 @@ struct Nac3 {
isa: Isa,
time_fns: &'static (dyn TimeFns + Sync),
primitive: PrimitiveStore,
builtins: Vec<BuiltinFuncSpec>,
builtins: Vec<(StrRef, FunSignature, Arc<GenCall>)>,
pyid_to_def: Arc<RwLock<HashMap<u64, DefinitionId>>>,
primitive_ids: PrimitivePythonId,
working_directory: TempDir,
@ -195,8 +194,10 @@ impl Nac3 {
body.retain(|stmt| {
if let StmtKind::FunctionDef { ref decorator_list, .. } = stmt.node {
decorator_list.iter().any(|decorator| {
if let Some(id) = decorator_id_string(decorator) {
id == "kernel" || id == "portable" || id == "rpc"
if let ExprKind::Name { id, .. } = decorator.node {
id.to_string() == "kernel"
|| id.to_string() == "portable"
|| id.to_string() == "rpc"
} else {
false
}
@ -209,8 +210,9 @@ impl Nac3 {
}
StmtKind::FunctionDef { ref decorator_list, .. } => {
decorator_list.iter().any(|decorator| {
if let Some(id) = decorator_id_string(decorator) {
id == "extern" || id == "kernel" || id == "portable" || id == "rpc"
if let ExprKind::Name { id, .. } = decorator.node {
let id = id.to_string();
id == "extern" || id == "portable" || id == "kernel" || id == "rpc"
} else {
false
}
@ -263,7 +265,7 @@ 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() => {
@ -299,64 +301,6 @@ 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,
@ -369,7 +313,6 @@ 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,
);
@ -446,6 +389,7 @@ impl Nac3 {
pyid_to_type: pyid_to_type.clone(),
primitive_ids: self.primitive_ids.clone(),
global_value_ids: global_value_ids.clone(),
class_names: Mutex::default(),
name_to_pyid: name_to_pyid.clone(),
module: module.clone(),
id_to_pyval: RwLock::default(),
@ -476,25 +420,9 @@ impl Nac3 {
match &stmt.node {
StmtKind::FunctionDef { decorator_list, .. } => {
if decorator_list
.iter()
.any(|decorator| decorator_id_string(decorator) == Some("rpc".to_string()))
{
store_fun
.call1(
py,
(
def_id.0.into_py(py),
module.getattr(py, name.to_string().as_str()).unwrap(),
),
)
.unwrap();
let is_async = decorator_list.iter().any(|decorator| {
decorator_get_flags(decorator)
.iter()
.any(|constant| *constant == Constant::Str("async".into()))
});
rpc_ids.push((None, def_id, is_async));
if decorator_list.iter().any(|decorator| matches!(decorator.node, ExprKind::Name { id, .. } if id == "rpc".into())) {
store_fun.call1(py, (def_id.0.into_py(py), module.getattr(py, name.to_string().as_str()).unwrap())).unwrap();
rpc_ids.push((None, def_id));
}
}
StmtKind::ClassDef { name, body, .. } => {
@ -502,26 +430,19 @@ impl Nac3 {
let class_obj = module.getattr(py, class_name.as_str()).unwrap();
for stmt in body {
if let StmtKind::FunctionDef { name, decorator_list, .. } = &stmt.node {
if decorator_list.iter().any(|decorator| {
decorator_id_string(decorator) == Some("rpc".to_string())
}) {
let is_async = decorator_list.iter().any(|decorator| {
decorator_get_flags(decorator)
.iter()
.any(|constant| *constant == Constant::Str("async".into()))
});
if decorator_list.iter().any(|decorator| matches!(decorator.node, ExprKind::Name { id, .. } if id == "rpc".into())) {
if name == &"__init__".into() {
return Err(CompileError::new_err(format!(
"compilation failed\n----------\nThe constructor of class {} should not be decorated with rpc decorator (at {})",
class_name, stmt.location
)));
}
rpc_ids.push((Some((class_obj.clone(), *name)), def_id, is_async));
rpc_ids.push((Some((class_obj.clone(), *name)), def_id));
}
}
}
}
_ => (),
_ => ()
}
let id = *name_to_pyid.get(&name).unwrap();
@ -560,6 +481,7 @@ impl Nac3 {
pyid_to_type: pyid_to_type.clone(),
primitive_ids: self.primitive_ids.clone(),
global_value_ids: global_value_ids.clone(),
class_names: Mutex::default(),
id_to_pyval: RwLock::default(),
id_to_primitive: RwLock::default(),
field_to_val: RwLock::default(),
@ -576,10 +498,6 @@ impl Nac3 {
.register_top_level(synthesized.pop().unwrap(), Some(resolver.clone()), "", false)
.unwrap();
// Process IRRT
let context = Context::create();
let irrt = load_irrt(&context, resolver.as_ref());
let fun_signature =
FunSignature { args: vec![], ret: self.primitive.none, vars: VarMap::new() };
let mut store = ConcreteTypeStore::new();
@ -617,12 +535,13 @@ impl Nac3 {
let top_level = Arc::new(composer.make_top_level_context());
{
let rpc_codegen = rpc_codegen_callback();
let defs = top_level.definitions.read();
for (class_data, id, is_async) in &rpc_ids {
for (class_data, id) in &rpc_ids {
let mut def = defs[id.0].write();
match &mut *def {
TopLevelDef::Function { codegen_callback, .. } => {
*codegen_callback = Some(rpc_codegen_callback(*is_async));
*codegen_callback = Some(rpc_codegen.clone());
}
TopLevelDef::Class { methods, .. } => {
let (class_def, method_name) = class_data.as_ref().unwrap();
@ -633,7 +552,7 @@ impl Nac3 {
if let TopLevelDef::Function { codegen_callback, .. } =
&mut *defs[id.0].write()
{
*codegen_callback = Some(rpc_codegen_callback(*is_async));
*codegen_callback = Some(rpc_codegen.clone());
store_fun
.call1(
py,
@ -648,11 +567,6 @@ impl Nac3 {
}
}
}
TopLevelDef::Variable { .. } => {
return Err(CompileError::new_err(String::from(
"Unsupported @rpc annotation on global variable",
)))
}
}
}
}
@ -712,7 +626,7 @@ impl Nac3 {
let buffer = buffer.as_slice().into();
membuffer.lock().push(buffer);
})));
let size_t = context
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 };
@ -731,7 +645,7 @@ impl Nac3 {
let mut generator =
ArtiqCodeGenerator::new("attributes_writeback".to_string(), size_t, self.time_fns);
let context = Context::create();
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());
@ -754,7 +668,7 @@ impl Nac3 {
membuffer.lock().push(buffer);
});
// Link all modules into `main`.
let context = inkwell::context::Context::create();
let buffers = membuffers.lock();
let main = context
.create_module_from_ir(MemoryBuffer::create_from_memory_range(&buffers[0], "main"))
@ -783,7 +697,8 @@ impl Nac3 {
)
.unwrap();
main.link_in_module(irrt).map_err(|err| CompileError::new_err(err.to_string()))?;
main.link_in_module(load_irrt(&context))
.map_err(|err| CompileError::new_err(err.to_string()))?;
let mut function_iter = main.get_first_function();
while let Some(func) = function_iter {
@ -869,41 +784,6 @@ impl Nac3 {
}
}
/// Retrieves the Name.id from a decorator, supports decorators with arguments.
fn decorator_id_string(decorator: &Located<ExprKind>) -> Option<String> {
if let ExprKind::Name { id, .. } = decorator.node {
// Bare decorator
return Some(id.to_string());
} else if let ExprKind::Call { func, .. } = &decorator.node {
// Decorators that are calls (e.g. "@rpc()") have Call for the node,
// need to extract the id from within.
if let ExprKind::Name { id, .. } = func.node {
return Some(id.to_string());
}
}
None
}
/// Retrieves flags from a decorator, if any.
fn decorator_get_flags(decorator: &Located<ExprKind>) -> Vec<Constant> {
let mut flags = vec![];
if let ExprKind::Call { keywords, .. } = &decorator.node {
for keyword in keywords {
if keyword.node.arg != Some("flags".into()) {
continue;
}
if let ExprKind::Set { elts } = &keyword.node.value.node {
for elt in elts {
if let ExprKind::Constant { value, .. } = &elt.node {
flags.push(value.clone());
}
}
}
}
}
flags
}
fn link_with_lld(elf_filename: String, obj_filename: String) -> PyResult<()> {
let linker_args = vec![
"-shared".to_string(),
@ -973,7 +853,7 @@ impl Nac3 {
Isa::RiscV32IMA => &timeline::NOW_PINNING_TIME_FNS,
Isa::CortexA9 | Isa::Host => &timeline::EXTERN_TIME_FNS,
};
let (primitive, _) = TopLevelComposer::make_primitives(isa.get_size_type());
let primitive: PrimitiveStore = TopLevelComposer::make_primitives(isa.get_size_type()).0;
let builtins = vec![
(
"now_mu".into(),
@ -989,7 +869,6 @@ impl Nac3 {
name: "t".into(),
ty: primitive.int64,
default_value: None,
is_vararg: false,
}],
ret: primitive.none,
vars: VarMap::new(),
@ -1009,7 +888,6 @@ impl Nac3 {
name: "dt".into(),
ty: primitive.int64,
default_value: None,
is_vararg: false,
}],
ret: primitive.none,
vars: VarMap::new(),

62
nac3artiq/src/symbol_resolver.rs
View File

@ -1,30 +1,14 @@
use std::{
collections::{HashMap, HashSet},
sync::{
atomic::{AtomicBool, Ordering::Relaxed},
Arc,
},
use inkwell::{
types::{BasicType, BasicTypeEnum},
values::BasicValueEnum,
AddressSpace,
};
use itertools::Itertools;
use parking_lot::RwLock;
use pyo3::{
types::{PyDict, PyTuple},
PyAny, PyObject, PyResult, Python,
};
use nac3core::{
codegen::{
classes::{NDArrayType, ProxyType},
CodeGenContext, CodeGenerator,
},
inkwell::{
module::Linkage,
types::{BasicType, BasicTypeEnum},
values::BasicValueEnum,
AddressSpace,
},
nac3parser::ast::{self, StrRef},
symbol_resolver::{StaticValue, SymbolResolver, SymbolValue, ValueEnum},
toplevel::{
helper::PrimDef,
@ -36,8 +20,21 @@ use nac3core::{
typedef::{into_var_map, iter_type_vars, Type, TypeEnum, TypeVar, Unifier, VarMap},
},
};
use nac3parser::ast::{self, StrRef};
use parking_lot::{Mutex, RwLock};
use pyo3::{
types::{PyDict, PyTuple},
PyAny, PyObject, PyResult, Python,
};
use std::{
collections::{HashMap, HashSet},
sync::{
atomic::{AtomicBool, Ordering::Relaxed},
Arc,
},
};
use super::PrimitivePythonId;
use crate::PrimitivePythonId;
pub enum PrimitiveValue {
I32(i32),
@ -82,6 +79,7 @@ pub struct InnerResolver {
pub id_to_primitive: RwLock<HashMap<u64, PrimitiveValue>>,
pub field_to_val: RwLock<HashMap<ResolverField, Option<PyFieldHandle>>>,
pub global_value_ids: Arc<RwLock<HashMap<u64, PyObject>>>,
pub class_names: Mutex<HashMap<StrRef, Type>>,
pub pyid_to_def: Arc<RwLock<HashMap<u64, DefinitionId>>>,
pub pyid_to_type: Arc<RwLock<HashMap<u64, Type>>>,
pub primitive_ids: PrimitivePythonId,
@ -135,8 +133,6 @@ 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,
@ -167,7 +163,7 @@ impl StaticValue for PythonValue {
PrimitiveValue::Bool(val) => {
ctx.ctx.i8_type().const_int(u64::from(*val), false).into()
}
PrimitiveValue::Str(val) => ctx.gen_string(generator, val).into(),
PrimitiveValue::Str(val) => ctx.ctx.const_string(val.as_bytes(), true).into(),
});
}
if let Some(global) = ctx.module.get_global(&self.id.to_string()) {
@ -355,7 +351,7 @@ 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![], is_vararg_ctx: false }), false)))
Ok(Ok((unifier.add_ty(TypeEnum::TTuple { ty: vec![] }), false)))
} else if ty_id == self.primitive_ids.option {
Ok(Ok((primitives.option, false)))
} else if ty_id == self.primitive_ids.none {
@ -559,10 +555,7 @@ 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, is_vararg_ctx: false }),
true,
)))
Ok(Ok((unifier.add_ty(TypeEnum::TTuple { ty: args }), true)))
}
TypeEnum::TObj { params, obj_id, .. } => {
let subst = {
@ -804,9 +797,7 @@ 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, is_vararg_ctx: false })
}))
Ok(types.map(|types| unifier.add_ty(TypeEnum::TTuple { ty: types })))
}
// 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
@ -981,7 +972,7 @@ 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.gen_string(generator, val).into()))
Ok(Some(ctx.ctx.const_string(val.as_bytes(), true).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));
@ -1212,9 +1203,7 @@ impl InnerResolver {
Ok(Some(ndarray.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, is_vararg_ctx: false } = expected_ty_enum.as_ref() else {
unreachable!()
};
let TypeEnum::TTuple { ty } = expected_ty_enum.as_ref() else { unreachable!() };
let tup_tys = ty.iter();
let elements: &PyTuple = obj.downcast()?;
@ -1470,7 +1459,6 @@ impl SymbolResolver for Resolver {
&self,
id: StrRef,
_: &mut CodeGenContext<'ctx, '_>,
_: &mut dyn CodeGenerator,
) -> Option<ValueEnum<'ctx>> {
let sym_value = {
let id_to_val = self.0.id_to_pyval.read();

23
nac3artiq/src/timeline.rs
View File

@ -1,12 +1,9 @@
use itertools::Either;
use nac3core::{
codegen::CodeGenContext,
inkwell::{
values::{BasicValueEnum, CallSiteValue},
AddressSpace, AtomicOrdering,
},
use inkwell::{
values::{BasicValueEnum, CallSiteValue},
AddressSpace, AtomicOrdering,
};
use itertools::Either;
use nac3core::codegen::CodeGenContext;
/// Functions for manipulating the timeline.
pub trait TimeFns {
@ -34,7 +31,7 @@ impl TimeFns for NowPinningTimeFns64 {
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr = ctx
.builder
.build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
@ -83,7 +80,7 @@ impl TimeFns for NowPinningTimeFns64 {
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr = ctx
.builder
.build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
@ -112,7 +109,7 @@ impl TimeFns for NowPinningTimeFns64 {
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr = ctx
.builder
.build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
@ -210,7 +207,7 @@ impl TimeFns for NowPinningTimeFns {
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr = ctx
.builder
.build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
@ -261,7 +258,7 @@ impl TimeFns for NowPinningTimeFns {
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
let now_hiptr = ctx
.builder
.build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();

1
nac3ast/Cargo.toml
View File

@ -10,6 +10,7 @@ constant-optimization = ["fold"]
fold = []
[dependencies]
lazy_static = "1.5"
parking_lot = "0.12"
string-interner = "0.17"
fxhash = "0.2"

8
nac3ast/src/ast_gen.rs
View File

@ -5,12 +5,14 @@ pub use crate::location::Location;
use fxhash::FxBuildHasher;
use parking_lot::{Mutex, MutexGuard};
use std::{cell::RefCell, collections::HashMap, fmt, sync::LazyLock};
use std::{cell::RefCell, collections::HashMap, fmt};
use string_interner::{symbol::SymbolU32, DefaultBackend, StringInterner};
pub type Interner = StringInterner<DefaultBackend, FxBuildHasher>;
static INTERNER: LazyLock<Mutex<Interner>> =
LazyLock::new(|| Mutex::new(StringInterner::with_hasher(FxBuildHasher::default())));
lazy_static! {
static ref INTERNER: Mutex<Interner> =
Mutex::new(StringInterner::with_hasher(FxBuildHasher::default()));
}
thread_local! {
static LOCAL_INTERNER: RefCell<HashMap<String, StrRef>> = RefCell::default();

3
nac3ast/src/lib.rs
View File

@ -14,6 +14,9 @@
clippy::wildcard_imports
)]
#[macro_use]
extern crate lazy_static;
mod ast_gen;
mod constant;
#[cfg(feature = "fold")]

18
nac3core/Cargo.toml
View File

@ -1,26 +1,26 @@
[features]
test = []
[package]
name = "nac3core"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2021"
[features]
no-escape-analysis = []
[dependencies]
itertools = "0.13"
crossbeam = "0.8"
indexmap = "2.6"
indexmap = "2.2"
parking_lot = "0.12"
rayon = "1.10"
rayon = "1.8"
nac3parser = { path = "../nac3parser" }
strum = "0.26"
strum_macros = "0.26"
strum = "0.26.2"
strum_macros = "0.26.4"
[dependencies.inkwell]
version = "0.5"
version = "0.4"
default-features = false
features = ["llvm14-0-prefer-dynamic", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
[dev-dependencies]
test-case = "1.2.0"

94
nac3core/build.rs
View File

@ -1,32 +1,46 @@
use regex::Regex;
use std::{
env,
fs::File,
io::Write,
path::Path,
path::{Path, PathBuf},
process::{Command, Stdio},
};
use regex::Regex;
const CMD_IRRT_CLANG: &str = "clang-irrt";
const CMD_IRRT_CLANG_TEST: &str = "clang-irrt-test";
const CMD_IRRT_LLVM_AS: &str = "llvm-as-irrt";
fn main() {
let out_dir = env::var("OUT_DIR").unwrap();
let out_dir = Path::new(&out_dir);
let irrt_dir = Path::new("irrt");
fn get_out_dir() -> PathBuf {
PathBuf::from(env::var("OUT_DIR").unwrap())
}
let irrt_cpp_path = irrt_dir.join("irrt.cpp");
fn get_irrt_dir() -> &'static Path {
Path::new("irrt")
}
/// Compile `irrt.cpp` for use in `src/codegen`
fn compile_irrt_cpp() {
let out_dir = get_out_dir();
let irrt_dir = get_irrt_dir();
/*
* 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 mut flags: Vec<&str> = vec![
let irrt_cpp_path = irrt_dir.join("irrt.cpp");
let flags: &[&str] = &[
"--target=wasm32",
"-x",
"c++",
"-std=c++20",
"-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",
@ -38,22 +52,11 @@ fn main() {
irrt_cpp_path.to_str().unwrap(),
];
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")
let output = Command::new(CMD_IRRT_CLANG)
.args(flags)
.output()
.map(|o| {
@ -99,7 +102,9 @@ fn main() {
file.write_all(filtered_output.as_bytes()).unwrap();
}
let mut llvm_as = Command::new("llvm-as-irrt")
// Assemble the emitted and filtered IR to .bc
// That .bc will be integrated into nac3core's codegen
let mut llvm_as = Command::new(CMD_IRRT_LLVM_AS)
.stdin(Stdio::piped())
.arg("-o")
.arg(out_dir.join("irrt.bc"))
@ -108,3 +113,48 @@ fn main() {
llvm_as.stdin.as_mut().unwrap().write_all(filtered_output.as_bytes()).unwrap();
assert!(llvm_as.wait().unwrap().success());
}
/// Compile `irrt_test.cpp` for testing
fn compile_irrt_test_cpp() {
let out_dir = get_out_dir();
let irrt_dir = get_irrt_dir();
let exe_path = out_dir.join("irrt_test.out"); // Output path of the compiled test executable
let irrt_test_cpp_path = irrt_dir.join("irrt_test.cpp");
let flags: &[&str] = &[
irrt_test_cpp_path.to_str().unwrap(),
"-x",
"c++",
"-I",
irrt_dir.to_str().unwrap(),
"-g",
"-fno-discard-value-names",
"-O0",
"-Wall",
"-Wextra",
"-Werror=return-type",
"-lm", // for `tgamma()`, `lgamma()`
"-o",
exe_path.to_str().unwrap(),
];
Command::new(CMD_IRRT_CLANG_TEST)
.args(flags)
.output()
.map(|o| {
assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
o
})
.unwrap();
println!("cargo:rerun-if-changed={}", irrt_dir.to_str().unwrap());
}
fn main() {
compile_irrt_cpp();
// https://github.com/rust-lang/cargo/issues/2549
// `cargo test -F test` to also build `irrt_test.cpp
if cfg!(feature = "test") {
compile_irrt_test_cpp();
}
}

16
nac3core/irrt/irrt.cpp
View File

@ -1,6 +1,10 @@
#include "irrt/exception.hpp"
#include "irrt/int_types.hpp"
#include "irrt/list.hpp"
#include "irrt/math.hpp"
#include "irrt/ndarray.hpp"
#include "irrt/slice.hpp"
#define IRRT_DEFINE_TYPEDEF_INTS
#include <irrt_everything.hpp>
/*
* All IRRT implementations.
*
* We don't have pre-compiled objects, so we are writing all implementations in
* headers and concatenate them with `#include` into one massive source file that
* contains all the IRRT stuff.
*/

39
nac3core/irrt/irrt/artiq_defs.hpp Normal file
View File

@ -0,0 +1,39 @@
#pragma once
#include <irrt/int_defs.hpp>
/*
This file defines all ARTIQ-specific structures
*/
/**
* @brief ARTIQ's `cslice` object
*
* See https://docs.rs/cslice/0.3.0/src/cslice/lib.rs.html#33-37
*/
template <typename SizeT>
struct CSlice {
const char *base;
SizeT len;
};
/**
* @brief Int type of ARTIQ's `Exception` IDs.
*/
typedef uint32_t ExceptionId;
/**
* @brief ARTIQ's `Exception` object
*
* See https://github.com/m-labs/artiq/blob/b0d2705c385f64b6e6711c1726cd9178f40b598e/artiq/firmware/libeh/eh_artiq.rs#L1C1-L17C1
*/
template <typename SizeT>
struct Exception {
ExceptionId id;
CSlice<SizeT> file;
uint32_t line;
uint32_t column;
CSlice<SizeT> function;
CSlice<SizeT> message;
uint32_t param;
};

347
nac3core/irrt/irrt/core.hpp Normal file
View File

@ -0,0 +1,347 @@
#pragma once
#include <irrt/int_defs.hpp>
#include <irrt/slice.hpp>
#include <irrt/utils.hpp>
// NDArray indices are always `uint32_t`.
using NDIndexInt = uint32_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;
}
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, NDIndexInt* 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 NDIndexInt* 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 NDIndexInt* in_idx,
NDIndexInt* 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 + size + size + 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, NDIndexInt* 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, NDIndexInt* 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 NDIndexInt* 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 NDIndexInt* 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 NDIndexInt* in_idx,
NDIndexInt* 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 NDIndexInt* in_idx,
NDIndexInt* out_idx) {
__nac3_ndarray_calc_broadcast_idx_impl(src_dims, src_ndims, in_idx,
out_idx);
}
} // extern "C"

9
nac3core/irrt/irrt/cslice.hpp
View File

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

25
nac3core/irrt/irrt/debug.hpp
View File

@ -1,25 +0,0 @@
#pragma once
// Set in nac3core/build.rs
#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 constexpr (IRRT_DEBUG_ASSERT_BOOL) { \
if ((lhs) != (rhs)) { \
raise_debug_assert(SizeT, "LHS = {0}. RHS = {1}", lhs, rhs, NO_PARAM); \
} \
}
#define debug_assert(SizeT, expr) \
if constexpr (IRRT_DEBUG_ASSERT_BOOL) { \
if (!(expr)) { \
raise_debug_assert(SizeT, "Got false.", NO_PARAM, NO_PARAM, NO_PARAM); \
} \
}

92
nac3core/irrt/irrt/error_context.hpp Normal file
View File

@ -0,0 +1,92 @@
#pragma once
#include <irrt/artiq_defs.hpp>
#include <irrt/int_defs.hpp>
#include <irrt/utils.hpp>
namespace {
/**
* @brief A (limited) set of known Exception IDs usable in IRRT
*/
struct ErrorContextExceptions {
ExceptionId index_error;
ExceptionId value_error;
ExceptionId assertion_error;
ExceptionId runtime_error;
ExceptionId type_error;
};
/**
* @brief The IRRT error context object
*
* This object contains all the details needed to propagate Python-like Exceptions in
* IRRT - within IRRT itself or propagate out of extern calls from nac3core.
*/
struct ErrorContext {
const ErrorContextExceptions *exceptions;
// Exception thrown by IRRT
ExceptionId exception_id;
// Points to empty c-string if there is no thrown Exception
const char *msg;
uint64_t param1;
uint64_t param2;
uint64_t param3;
void initialize(const ErrorContextExceptions *exceptions) {
this->exceptions = exceptions;
clear_error();
}
void clear_error() {
// NOTE: Point the msg to an empty str.
// Don't set it to nullptr - to implement `has_exception`
this->msg = "";
}
void set_exception(ExceptionId exception_id, const char *msg,
uint64_t param1 = 0, uint64_t param2 = 0,
uint64_t param3 = 0) {
this->exception_id = exception_id;
this->msg = msg;
this->param1 = param1;
this->param2 = param2;
this->param3 = param3;
}
bool has_exception() { return !cstr_utils::is_empty(msg); }
template <typename SizeT>
void get_exception_str(CSlice<SizeT> *dst_str) {
dst_str->base = msg;
dst_str->len = (SizeT)cstr_utils::length(msg);
}
};
} // namespace
extern "C" {
void __nac3_error_context_initialize(ErrorContext *errctx,
ErrorContextExceptions *exceptions) {
errctx->initialize(exceptions);
}
bool __nac3_error_context_has_exception(ErrorContext *errctx) {
return errctx->has_exception();
}
void __nac3_error_context_get_exception_str(ErrorContext *errctx,
CSlice<int32_t> *dst_str) {
errctx->get_exception_str<int32_t>(dst_str);
}
void __nac3_error_context_get_exception_str64(ErrorContext *errctx,
CSlice<int64_t> *dst_str) {
errctx->get_exception_str<int64_t>(dst_str);
}
// Used for testing
void __nac3_error_dummy_raise(ErrorContext *errctx) {
errctx->set_exception(errctx->exceptions->runtime_error,
"Error thrown from __nac3_error_dummy_raise");
}
}

82
nac3core/irrt/irrt/exception.hpp
View File

@ -1,82 +0,0 @@
#pragma once
#include "irrt/cslice.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];
};
constexpr 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 = reinterpret_cast<const uint8_t*>(filename), .len = __builtin_strlen(filename)},
.line = line,
.column = 0,
.function = {.base = reinterpret_cast<const uint8_t*>(function), .len = __builtin_strlen(function)},
.msg = {.base = reinterpret_cast<const uint8_t*>(msg), .len = __builtin_strlen(msg)},
};
e.params[0] = param0;
e.params[1] = param1;
e.params[2] = param2;
__nac3_raise(reinterpret_cast<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` to `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

12
nac3core/irrt/irrt/int_defs.hpp Normal file
View File

@ -0,0 +1,12 @@
#pragma once
// This is made toggleable since `irrt_test.cpp` itself would include
// headers that define these typedefs
#ifdef IRRT_DEFINE_TYPEDEF_INTS
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);
#endif

22
nac3core/irrt/irrt/int_types.hpp
View File

@ -1,22 +0,0 @@
#pragma once
#if __STDC_VERSION__ >= 202000
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);
#else
using int8_t = _ExtInt(8);
using uint8_t = unsigned _ExtInt(8);
using int32_t = _ExtInt(32);
using uint32_t = unsigned _ExtInt(32);
using int64_t = _ExtInt(64);
using uint64_t = unsigned _ExtInt(64);
#endif
// 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;

75
nac3core/irrt/irrt/list.hpp
View File

@ -1,75 +0,0 @@
#pragma once
#include "irrt/int_types.hpp"
#include "irrt/math_util.hpp"
extern "C" {
// 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;
}
} // extern "C"

93
nac3core/irrt/irrt/math.hpp
View File

@ -1,93 +0,0 @@
#pragma once
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
#define DEF_nac3_int_exp_(T) \
T __nac3_int_exp_##T(T base, T exp) { \
return __nac3_int_exp_impl(base, exp); \
}
extern "C" {
// Putting semicolons here to make clang-format not reformat this into
// a stair shape.
DEF_nac3_int_exp_(int32_t);
DEF_nac3_int_exp_(int64_t);
DEF_nac3_int_exp_(uint32_t);
DEF_nac3_int_exp_(uint64_t);
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);
}
}

13
nac3core/irrt/irrt/math_util.hpp
View File

@ -1,13 +0,0 @@
#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

144
nac3core/irrt/irrt/ndarray.hpp
View File

@ -1,144 +0,0 @@
#pragma once
#include "irrt/int_types.hpp"
namespace {
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" {
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);
}
}

315
nac3core/irrt/irrt/ndarray/basic.hpp Normal file
View File

@ -0,0 +1,315 @@
#pragma once
#include <irrt/error_context.hpp>
#include <irrt/int_defs.hpp>
#include <irrt/ndarray/def.hpp>
namespace {
namespace ndarray {
namespace basic {
namespace util {
/**
* @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(ErrorContext* errctx, SizeT ndims,
const SizeT* shape) {
for (SizeT axis = 0; axis < ndims; axis++) {
if (shape[axis] < 0) {
errctx->set_exception(
errctx->exceptions->value_error,
"negative dimensions are not allowed; axis {0} "
"has dimension {1}",
axis, shape[axis]);
return;
}
}
}
/**
* @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 (int32_t i = 0; i < ndims; i++) {
int32_t axis = ndims - i - 1;
int32_t dim = shape[axis];
indices[axis] = nth % dim;
nth /= dim;
}
}
} // namespace util
/**
* @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 util::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 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++) {
int axis = ndarray->ndims - i - 1;
ndarray->strides[axis] = stride_product * ndarray->itemsize;
stride_product *= ndarray->shape[axis];
}
}
/**
* @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 Return the pointer to the nth (0-based) element in a flattened view of `ndarray`.
*/
template <typename SizeT>
uint8_t* get_nth_pelement(const NDArray<SizeT>* ndarray, SizeT nth) {
SizeT* indices = (SizeT*)__builtin_alloca(sizeof(SizeT) * ndarray->ndims);
util::set_indices_by_nth(ndarray->ndims, ndarray->shape, indices, nth);
return get_pelement_by_indices(ndarray, indices);
}
/**
* @brief Like `get_nth_pelement` but asserts that `nth` is in bounds.
*/
template <typename SizeT>
uint8_t* checked_get_nth_pelement(ErrorContext* errctx,
const NDArray<SizeT>* ndarray, SizeT nth) {
SizeT arr_size = ndarray->size();
if (!(0 <= nth && nth < arr_size)) {
errctx->set_exception(
errctx->exceptions->index_error,
"index {0} is out of bounds, valid range is {1} <= index < {2}",
nth, 0, arr_size);
return 0;
}
return get_nth_pelement(ndarray, nth);
}
/**
* @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 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>
void len(ErrorContext* errctx, const NDArray<SizeT>* ndarray,
SliceIndex* dst_length) {
// numpy prohibits `__len__` on unsized objects
if (ndarray->ndims == 0) {
errctx->set_exception(errctx->exceptions->type_error,
"len() of unsized object");
return;
}
*dst_length = (SliceIndex)ndarray->shape[0];
}
/**
* @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) {
__builtin_assume(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);
}
}
/**
* @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 = 0; i < ndarray->ndims - 1; i++) {
if (ndarray->strides[i] !=
ndarray->shape[i + 1] + ndarray->strides[i + 1]) {
return false;
}
}
return true;
}
} // namespace basic
} // namespace ndarray
} // namespace
extern "C" {
using namespace ndarray::basic;
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);
}
void __nac3_ndarray_len(ErrorContext* errctx, NDArray<int32_t>* ndarray,
SliceIndex* dst_len) {
return len(errctx, ndarray, dst_len);
}
void __nac3_ndarray_len64(ErrorContext* errctx, NDArray<int64_t>* ndarray,
SliceIndex* dst_len) {
return len(errctx, ndarray, dst_len);
}
void __nac3_ndarray_util_assert_shape_no_negative(ErrorContext* errctx,
int32_t ndims,
int32_t* shape) {
util::assert_shape_no_negative(errctx, ndims, shape);
}
void __nac3_ndarray_util_assert_shape_no_negative64(ErrorContext* errctx,
int64_t ndims,
int64_t* shape) {
util::assert_shape_no_negative(errctx, ndims, shape);
}
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);
}
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);
}
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);
}
uint8_t* __nac3_ndarray_get_nth_pelement(NDArray<int32_t>* ndarray,
int32_t index) {
return get_nth_pelement(ndarray, index);
}
uint8_t* __nac3_ndarray_get_nth_pelement64(NDArray<int64_t>* ndarray,
int64_t index) {
return get_nth_pelement(ndarray, index);
}
}

221
nac3core/irrt/irrt/ndarray/broadcast.hpp Normal file
View File

@ -0,0 +1,221 @@
#pragma once
#include <irrt/error_context.hpp>
#include <irrt/int_defs.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 {
namespace util {
/**
* @brief Return true if `src_shape` can broadcast to `dst_shape`.
*/
template <typename SizeT>
bool can_broadcast_shape_to(SizeT target_ndims, const SizeT* target_shape,
SizeT src_ndims, const SizeT* src_shape) {
/*
* // See https://numpy.org/doc/stable/user/basics.broadcasting.html
* This function handles this example:
* ```
* Image (3d array): 256 x 256 x 3
* Scale (1d array): 3
* Result (3d array): 256 x 256 x 3
* ```
* Other interesting examples to consider:
* - `can_broadcast_shape_to([3], [1, 1, 1, 1, 3]) == true`
* - `can_broadcast_shape_to([3], [3, 1]) == false`
* - `can_broadcast_shape_to([256, 256, 3], [256, 1, 3]) == true`
* In cases when the shapes contain zero(es):
* - `can_broadcast_shape_to([0], [1]) == true`
* - `can_broadcast_shape_to([0], [2]) == false`
* - `can_broadcast_shape_to([0, 4, 0, 0], [1]) == true`
* - `can_broadcast_shape_to([0, 4, 0, 0], [1, 1, 1, 1]) == true`
* - `can_broadcast_shape_to([0, 4, 0, 0], [1, 4, 1, 1]) == true`
* - `can_broadcast_shape_to([4, 3], [0, 3]) == false`
* - `can_broadcast_shape_to([4, 3], [0, 0]) == false`
*/
// This is essentially doing the following in Python:
// `for target_dim, src_dim in itertools.zip_longest(target_shape[::-1], src_shape[::-1], fillvalue=1)`
for (SizeT i = 0; i < max(target_ndims, src_ndims); i++) {
SizeT target_dim_i = target_ndims - i - 1;
SizeT src_dim_i = src_ndims - i - 1;
bool target_dim_exists = target_dim_i >= 0;
bool src_dim_exists = src_dim_i >= 0;
SizeT target_dim = target_dim_exists ? target_shape[target_dim_i] : 1;
SizeT src_dim = src_dim_exists ? src_shape[src_dim_i] : 1;
bool ok = src_dim == 1 || target_dim == src_dim;
if (!ok) return false;
}
return true;
}
/**
* @brief Performs `np.broadcast_shapes`
*/
template <typename SizeT>
void broadcast_shapes(ErrorContext* errctx, SizeT num_shapes,
const ShapeEntry<SizeT>* shapes, SizeT dst_ndims,
SizeT* 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 it should already know in order to allocate `dst_shape` in the first place.
// `dst_shape` must be pre-allocated.
// `dst_shape` does not have to be initialized
// TODO: Implementation is not obvious
// This is essentially a `mconcat` where the neutral element is `[1, 1, 1, 1, ...]`, and the operation is commutative.
// Set `dst_shape` to all `1`s.
for (SizeT dst_axis = 0; dst_axis < dst_ndims; dst_axis++) {
dst_shape[dst_axis] = 0;
}
for (SizeT i = 0; i < num_shapes; i++) {
ShapeEntry<SizeT> entry = shapes[i];
SizeT entry_axis = entry.ndims - i;
SizeT dst_axis = dst_ndims - i;
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) {
// Do nothing
} else if (entry_dim == dst_dim) {
// Do nothing
} else {
errctx->set_exception(errctx->exceptions->value_error,
"shape mismatch: objects cannot be broadcast "
"to a single shape.");
return;
}
}
}
} // namespace util
/**
* @brief Perform `np.broadcast_to(<ndarray>, <target_shape>)` and appropriate assertions.
*
* Cautious note on https://github.com/numpy/numpy/issues/21744..
*
* 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(ErrorContext* errctx, const NDArray<SizeT>* src_ndarray,
NDArray<SizeT>* dst_ndarray) {
/*
* Cautions:
* ```
* xs = np.zeros((4,))
* ys = np.zero((4, 1))
* ys[:] = xs # ok
*
* xs = np.zeros((1, 4))
* ys = np.zero((4,))
* ys[:] = xs # allowed
* # However `np.broadcast_to(xs, (4,))` would fails, as per numpy's broadcasting rule.
* # and apparently numpy will "deprecate" this? SEE https://github.com/numpy/numpy/issues/21744
* # This implementation will NOT support this assignment.
* ```
*/
if (!ndarray::broadcast::util::can_broadcast_shape_to(
dst_ndarray->ndims, dst_ndarray->shape, src_ndarray->ndims,
src_ndarray->shape)) {
errctx->set_exception(errctx->exceptions->value_error,
"operands could not be broadcast together");
return;
}
dst_ndarray->data = src_ndarray->data;
dst_ndarray->itemsize = src_ndarray->itemsize;
// TODO: Implementation is not obvious
SizeT stride_product = 1;
for (SizeT i = 0; i < max(src_ndarray->ndims, dst_ndarray->ndims); i++) {
SizeT src_ndarray_dim_i = src_ndarray->ndims - i - 1;
SizeT dst_dim_i = dst_ndarray->ndims - i - 1;
bool src_ndarray_dim_exists = src_ndarray_dim_i >= 0;
bool dst_dim_exists = dst_dim_i >= 0;
bool c1 = src_ndarray_dim_exists &&
src_ndarray->shape[src_ndarray_dim_i] == 1;
bool c2 = dst_dim_exists && dst_ndarray->shape[dst_dim_i] != 1;
if (!src_ndarray_dim_exists || (c1 && c2)) {
dst_ndarray->strides[dst_dim_i] = 0; // Freeze it in-place
} else {
dst_ndarray->strides[dst_dim_i] =
stride_product * src_ndarray->itemsize;
stride_product *= src_ndarray->shape[src_ndarray_dim_i];
}
}
}
} // namespace broadcast
} // namespace ndarray
} // namespace
extern "C" {
using namespace ndarray::broadcast;
void __nac3_ndarray_broadcast_to(ErrorContext* errctx,
NDArray<int32_t>* src_ndarray,
NDArray<int32_t>* dst_ndarray) {
broadcast_to(errctx, src_ndarray, dst_ndarray);
}
void __nac3_ndarray_broadcast_to64(ErrorContext* errctx,
NDArray<int64_t>* src_ndarray,
NDArray<int64_t>* dst_ndarray) {
broadcast_to(errctx, src_ndarray, dst_ndarray);
}
void __nac3_ndarray_broadcast_shapes(ErrorContext* errctx, int32_t num_shapes,
const ShapeEntry<int32_t>* shapes,
int32_t dst_ndims, int32_t* dst_shape) {
ndarray::broadcast::util::broadcast_shapes(errctx, num_shapes, shapes,
dst_ndims, dst_shape);
}
void __nac3_ndarray_broadcast_shapes64(ErrorContext* errctx, int64_t num_shapes,
const ShapeEntry<int64_t>* shapes,
int64_t dst_ndims, int64_t* dst_shape) {
ndarray::broadcast::util::broadcast_shapes(errctx, num_shapes, shapes,
dst_ndims, dst_shape);
}
}

44
nac3core/irrt/irrt/ndarray/def.hpp Normal file
View File

@ -0,0 +1,44 @@
#pragma once
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

200
nac3core/irrt/irrt/ndarray/indexing.hpp Normal file
View File

@ -0,0 +1,200 @@
#pragma once
#include <irrt/error_context.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
*
* See https://numpy.org/doc/stable/user/basics.indexing.html#single-element-indexing
*
* `data` points to a `SliceIndex`.
*/
const NDIndexType ND_INDEX_TYPE_SINGLE_ELEMENT = 0;
/**
* @brief A slice index
*
* See https://numpy.org/doc/stable/user/basics.indexing.html#slicing-and-striding
*
* `data` points to a `UserRange`.
*/
const NDIndexType ND_INDEX_TYPE_SLICE = 1;
/**
* @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 {
namespace util {
/**
* @brief Return the expected rank of the resulting ndarray
* created by indexing an ndarray of rank `ndims` using `indexes`.
*/
template <typename SizeT>
void deduce_ndims_after_indexing(ErrorContext* errctx, SizeT* final_ndims,
SizeT ndims, SizeT num_indexes,
const NDIndex* indexes) {
if (num_indexes > ndims) {
errctx->set_exception(errctx->exceptions->index_error,
"too many indices for array: array is "
"{0}-dimensional, but {1} were indexed",
ndims, num_indexes);
return;
}
*final_ndims = ndims;
for (SizeT i = 0; i < num_indexes; i++) {
if (indexes[i].type == ND_INDEX_TYPE_SINGLE_ELEMENT) {
// An index demotes the rank by 1
(*final_ndims)--;
}
}
}
} // namespace util
/**
* @brief Perform ndarray "basic indexing" (https://numpy.org/doc/stable/user/basics.indexing.html#basic-indexing)
*
* This is function very similar to performing `dst_ndarray = src_ndarray[indexes]` in Python (where the variables
* can all be found in the parameter of this function).
*
* In other words, this function takes in an ndarray (`src_ndarray`), index it with `indexes`, and return the
* indexed array (by writing the result to `dst_ndarray`).
*
* This function also does proper assertions on `indexes`.
*
* # 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 `indexes`.
* - `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 indexes Indexes 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(ErrorContext* errctx, SizeT num_indexes, const NDIndex* indexes,
const NDArray<SizeT>* src_ndarray, NDArray<SizeT>* dst_ndarray) {
// Reference code: https://github.com/wadetb/tinynumpy/blob/0d23d22e07062ffab2afa287374c7b366eebdda1/tinynumpy/tinynumpy.py#L652
dst_ndarray->data = src_ndarray->data;
dst_ndarray->itemsize = src_ndarray->itemsize;
SizeT src_axis = 0;
SizeT dst_axis = 0;
for (SliceIndex i = 0; i < num_indexes; i++) {
const NDIndex* index = &indexes[i];
if (index->type == ND_INDEX_TYPE_SINGLE_ELEMENT) {
SliceIndex input = *((SliceIndex*)index->data);
SliceIndex k = slice::resolve_index_in_length(
src_ndarray->shape[src_axis], input);
if (k == slice::OUT_OF_BOUNDS) {
errctx->set_exception(errctx->exceptions->index_error,
"index {0} is out of bounds for axis {1} "
"with size {2}",
input, src_axis,
src_ndarray->shape[src_axis]);
return;
}
dst_ndarray->data += k * src_ndarray->strides[src_axis];
src_axis++;
} else if (index->type == ND_INDEX_TYPE_SLICE) {
UserSlice* input = (UserSlice*)index->data;
Slice slice;
input->indices_checked(errctx, src_ndarray->shape[src_axis],
&slice);
if (errctx->has_exception()) {
return;
}
dst_ndarray->data +=
(SizeT)slice.start * src_ndarray->strides[src_axis];
dst_ndarray->strides[dst_axis] =
((SizeT)slice.step) * src_ndarray->strides[src_axis];
dst_ndarray->shape[dst_axis] = (SizeT)slice.len();
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];
}
}
} // namespace indexing
} // namespace ndarray
} // namespace
extern "C" {
using namespace ndarray::indexing;
void __nac3_ndarray_indexing_deduce_ndims_after_indexing(
ErrorContext* errctx, int32_t* result, int32_t ndims, int32_t num_indexes,
const NDIndex* indexes) {
ndarray::indexing::util::deduce_ndims_after_indexing(errctx, result, ndims,
num_indexes, indexes);
}
void __nac3_ndarray_indexing_deduce_ndims_after_indexing64(
ErrorContext* errctx, int64_t* result, int64_t ndims, int64_t num_indexes,
const NDIndex* indexes) {
ndarray::indexing::util::deduce_ndims_after_indexing(errctx, result, ndims,
num_indexes, indexes);
}
void __nac3_ndarray_index(ErrorContext* errctx, int32_t num_indexes,
NDIndex* indexes, NDArray<int32_t>* src_ndarray,
NDArray<int32_t>* dst_ndarray) {
index(errctx, num_indexes, indexes, src_ndarray, dst_ndarray);
}
void __nac3_ndarray_index64(ErrorContext* errctx, int64_t num_indexes,
NDIndex* indexes, NDArray<int64_t>* src_ndarray,
NDArray<int64_t>* dst_ndarray) {
index(errctx, num_indexes, indexes, src_ndarray, dst_ndarray);
}
}

117
nac3core/irrt/irrt/ndarray/reshape.hpp Normal file
View File

@ -0,0 +1,117 @@
#pragma once
#include <irrt/error_context.hpp>
#include <irrt/int_defs.hpp>
#include <irrt/ndarray/def.hpp>
namespace {
namespace ndarray {
namespace reshape {
namespace util {
/**
* @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(ErrorContext* errctx, 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.
errctx->set_exception(
errctx->exceptions->value_error,
"can only specify one unknown dimension");
return;
} 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...
errctx->set_exception(
errctx->exceptions->value_error,
"Found negative dimension {0} on axis {1}", dim, axis_i);
return;
}
} 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) {
errctx->set_exception(
errctx->exceptions->value_error,
"cannot reshape array of size {0} into given shape", size);
return;
}
}
} // namespace util
} // namespace reshape
} // namespace ndarray
} // namespace
extern "C" {
void __nac3_ndarray_resolve_and_check_new_shape(ErrorContext* errctx,
int32_t size, int32_t new_ndims,
int32_t* new_shape) {
ndarray::reshape::util::resolve_and_check_new_shape(errctx, size, new_ndims,
new_shape);
}
void __nac3_ndarray_resolve_and_check_new_shape64(ErrorContext* errctx,
int64_t size,
int64_t new_ndims,
int64_t* new_shape) {
ndarray::reshape::util::resolve_and_check_new_shape(errctx, size, new_ndims,
new_shape);
}
}

178
nac3core/irrt/irrt/slice.hpp
View File

@ -1,28 +1,166 @@
#pragma once
#include "irrt/int_types.hpp"
#include <irrt/error_context.hpp>
#include <irrt/int_defs.hpp>
#include <irrt/slice.hpp>
#include <irrt/utils.hpp>
extern "C" {
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;
}
// The type of an index or a value describing the length of a
// range/slice is always `int32_t`.
using SliceIndex = int32_t;
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;
namespace {
/**
* @brief A Python-like slice with resolved indices.
*
* "Resolved indices" means that `start` and `stop` must be positive and are
* bound to a known length.
*/
struct Slice {
SliceIndex start;
SliceIndex stop;
SliceIndex step;
/**
* @brief Calculate and return the length / the number of the slice.
*
* If this were a Python range, this function would be `len(range(start, stop, step))`.
*/
SliceIndex len() {
SliceIndex diff = stop - start;
if (diff > 0 && step > 0) {
return ((diff - 1) / step) + 1;
} else if (diff < 0 && step < 0) {
return ((diff + 1) / step) + 1;
} else {
return 0;
}
}
};
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).
*
*/
SliceIndex resolve_index_in_length_clamped(SliceIndex length,
SliceIndex index) {
if (index < 0) {
return max<SliceIndex>(length + index, 0);
} else {
return 0;
return min<SliceIndex>(length, index);
}
}
const SliceIndex OUT_OF_BOUNDS = -1;
/**
* @brief Like `resolve_index_in_length_clamped`, but returns `OUT_OF_BOUNDS`
* if `index` is out of bounds.
*/
SliceIndex resolve_index_in_length(SliceIndex length, SliceIndex index) {
SliceIndex resolved = index < 0 ? length + index : index;
if (0 <= resolved && resolved < length) {
return resolved;
} else {
return OUT_OF_BOUNDS;
}
}
} // namespace slice
/**
* @brief A Python-like slice with **unresolved** indices.
*/
struct UserSlice {
bool start_defined;
SliceIndex start;
bool stop_defined;
SliceIndex stop;
bool step_defined;
SliceIndex step;
UserSlice() { this->reset(); }
void reset() {
this->start_defined = false;
this->stop_defined = false;
this->step_defined = false;
}
void set_start(SliceIndex start) {
this->start_defined = true;
this->start = start;
}
void set_stop(SliceIndex stop) {
this->stop_defined = true;
this->stop = stop;
}
void set_step(SliceIndex step) {
this->step_defined = true;
this->step = step;
}
/**
* @brief Resolve this slice.
*
* In Python, this would be `slice(start, stop, step).indices(length)`.
*
* @return A `Slice` with the resolved indices.
*/
Slice indices(SliceIndex length) {
Slice result;
result.step = step_defined ? step : 1;
bool step_is_negative = result.step < 0;
if (start_defined) {
result.start =
slice::resolve_index_in_length_clamped(length, start);
} else {
result.start = step_is_negative ? length - 1 : 0;
}
if (stop_defined) {
result.stop = slice::resolve_index_in_length_clamped(length, stop);
} else {
result.stop = step_is_negative ? -1 : length;
}
return result;
}
/**
* @brief Like `.indices()` but with assertions.
*/
void indices_checked(ErrorContext* errctx, SliceIndex length,
Slice* result) {
if (length < 0) {
errctx->set_exception(errctx->exceptions->value_error,
"length should not be negative, got {0}",
length);
return;
}
if (this->step_defined && this->step == 0) {
errctx->set_exception(errctx->exceptions->value_error,
"slice step cannot be zero");
return;
}
*result = this->indices(length);
}
};
} // namespace

104
nac3core/irrt/irrt/utils.hpp Normal file
View File

@ -0,0 +1,104 @@
#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;
}
/**
* @brief Compare contents of two arrays with the same length.
*/
template <typename T>
bool arrays_match(int len, T* as, T* bs) {
for (int i = 0; i < len; i++) {
if (as[i] != bs[i]) return false;
}
return true;
}
namespace cstr_utils {
/**
* @brief Return true if `str` is empty.
*/
bool is_empty(const char* str) { return str[0] == '\0'; }
/**
* @brief Implementation of `strcmp()`
*/
int8_t compare(const char* a, const char* b) {
uint32_t i = 0;
while (true) {
if (a[i] < b[i]) {
return -1;
} else if (a[i] > b[i]) {
return 1;
} else {
if (a[i] == '\0') {
return 0;
} else {
i++;
}
}
}
}
/**
* @brief Return true two strings have the same content.
*/
int8_t equal(const char* a, const char* b) { return compare(a, b) == 0; }
/**
* @brief Implementation of `strlen()`.
*/
uint32_t length(const char* str) {
uint32_t length = 0;
while (*str != '\0') {
length++;
str++;
}
return length;
}
/**
* @brief Copy a null-terminated string to a buffer with limited size and guaranteed null-termination.
*
* `dst_max_size` must be greater than 0, otherwise this function has undefined behavior.
*
* This function attempts to copy everything from `src` from `dst`, and *always* null-terminates `dst`.
*
* If the size of `dst` is too small, the final byte (`dst[dst_max_size - 1]`) of `dst` will be set to
* the null terminator.
*
* @param src String to copy from.
* @param dst Buffer to copy string to.
* @param dst_max_size
* Number of bytes of this buffer, including the space needed for the null terminator.
* Must be greater than 0.
* @return If `dst` is too small to contain everything in `src`.
*/
bool copy(const char* src, char* dst, uint32_t dst_max_size) {
for (uint32_t i = 0; i < dst_max_size; i++) {
bool is_last = i + 1 == dst_max_size;
if (is_last && src[i] != '\0') {
dst[i] = '\0';
return false;
}
if (src[i] == '\0') {
dst[i] = '\0';
return true;
}
dst[i] = src[i];
}
__builtin_unreachable();
}
} // namespace cstr_utils
} // namespace

13
nac3core/irrt/irrt_everything.hpp Normal file
View File

@ -0,0 +1,13 @@
#pragma once
#include <irrt/artiq_defs.hpp>
#include <irrt/core.hpp>
#include <irrt/error_context.hpp>
#include <irrt/int_defs.hpp>
#include <irrt/ndarray/basic.hpp>
#include <irrt/ndarray/broadcast.hpp>
#include <irrt/ndarray/def.hpp>
#include <irrt/ndarray/indexing.hpp>
#include <irrt/ndarray/reshape.hpp>
#include <irrt/slice.hpp>
#include <irrt/utils.hpp>

20
nac3core/irrt/irrt_test.cpp Normal file
View File

@ -0,0 +1,20 @@
// This file will be compiled like a real C++ program,
// and we do have the luxury to use the standard libraries.
// That is if the nix flakes do not have issues... especially on msys2...
#include <cstdint>
#include <cstdio>
#include <cstdlib>
#include <test/test_core.hpp>
#include <test/test_ndarray_basic.hpp>
#include <test/test_ndarray_broadcast.hpp>
#include <test/test_ndarray_indexing.hpp>
#include <test/test_slice.hpp>
int main() {
test::core::run();
test::slice::run();
test::ndarray_basic::run();
test::ndarray_indexing::run();
test::ndarray_broadcast::run();
return 0;
}

11
nac3core/irrt/test/includes.hpp Normal file
View File

@ -0,0 +1,11 @@
#pragma once
#include <cstdint>
#include <cstdio>
#include <cstdlib>
#include <irrt_everything.hpp>
#include <test/util.hpp>
/*
Include this header for every test_*.cpp
*/

16
nac3core/irrt/test/test_core.hpp Normal file
View File

@ -0,0 +1,16 @@
#pragma once
#include <test/includes.hpp>
namespace test {
namespace core {
void test_int_exp() {
BEGIN_TEST();
assert_values_match(125, __nac3_int_exp_impl<int32_t>(5, 3));
assert_values_match(3125, __nac3_int_exp_impl<int32_t>(5, 5));
}
void run() { test_int_exp(); }
} // namespace core
} // namespace test

30
nac3core/irrt/test/test_ndarray_basic.hpp Normal file
View File

@ -0,0 +1,30 @@
#pragma once
#include <test/includes.hpp>
namespace test {
namespace ndarray_basic {
void test_calc_size_from_shape_normal() {
// Test shapes with normal values
BEGIN_TEST();
int32_t shape[4] = {2, 3, 5, 7};
assert_values_match(
210, ndarray::basic::util::calc_size_from_shape<int32_t>(4, shape));
}
void test_calc_size_from_shape_has_zero() {
// Test shapes with 0 in them
BEGIN_TEST();
int32_t shape[4] = {2, 0, 5, 7};
assert_values_match(
0, ndarray::basic::util::calc_size_from_shape<int32_t>(4, shape));
}
void run() {
test_calc_size_from_shape_normal();
test_calc_size_from_shape_has_zero();
}
} // namespace ndarray_basic
} // namespace test

129
nac3core/irrt/test/test_ndarray_broadcast.hpp Normal file
View File

@ -0,0 +1,129 @@
#pragma once
#include <test/includes.hpp>
namespace test {
namespace ndarray_broadcast {
void test_can_broadcast_shape() {
BEGIN_TEST();
assert_values_match(true,
ndarray::broadcast::util::can_broadcast_shape_to(
1, (int32_t[]){3}, 5, (int32_t[]){1, 1, 1, 1, 3}));
assert_values_match(false, ndarray::broadcast::util::can_broadcast_shape_to(
1, (int32_t[]){3}, 2, (int32_t[]){3, 1}));
assert_values_match(true, ndarray::broadcast::util::can_broadcast_shape_to(
1, (int32_t[]){3}, 1, (int32_t[]){3}));
assert_values_match(false, ndarray::broadcast::util::can_broadcast_shape_to(
1, (int32_t[]){1}, 1, (int32_t[]){3}));
assert_values_match(true, ndarray::broadcast::util::can_broadcast_shape_to(
1, (int32_t[]){1}, 1, (int32_t[]){1}));
assert_values_match(
true, ndarray::broadcast::util::can_broadcast_shape_to(
3, (int32_t[]){256, 256, 3}, 3, (int32_t[]){256, 1, 3}));
assert_values_match(true,
ndarray::broadcast::util::can_broadcast_shape_to(
3, (int32_t[]){256, 256, 3}, 1, (int32_t[]){3}));
assert_values_match(false,
ndarray::broadcast::util::can_broadcast_shape_to(
3, (int32_t[]){256, 256, 3}, 1, (int32_t[]){2}));
assert_values_match(true,
ndarray::broadcast::util::can_broadcast_shape_to(
3, (int32_t[]){256, 256, 3}, 1, (int32_t[]){1}));
// In cases when the shapes contain zero(es)
assert_values_match(true, ndarray::broadcast::util::can_broadcast_shape_to(
1, (int32_t[]){0}, 1, (int32_t[]){1}));
assert_values_match(false, ndarray::broadcast::util::can_broadcast_shape_to(
1, (int32_t[]){0}, 1, (int32_t[]){2}));
assert_values_match(true,
ndarray::broadcast::util::can_broadcast_shape_to(
4, (int32_t[]){0, 4, 0, 0}, 1, (int32_t[]){1}));
assert_values_match(
true, ndarray::broadcast::util::can_broadcast_shape_to(
4, (int32_t[]){0, 4, 0, 0}, 4, (int32_t[]){1, 1, 1, 1}));
assert_values_match(
true, ndarray::broadcast::util::can_broadcast_shape_to(
4, (int32_t[]){0, 4, 0, 0}, 4, (int32_t[]){1, 4, 1, 1}));
assert_values_match(false, ndarray::broadcast::util::can_broadcast_shape_to(
2, (int32_t[]){4, 3}, 2, (int32_t[]){0, 3}));
assert_values_match(false, ndarray::broadcast::util::can_broadcast_shape_to(
2, (int32_t[]){4, 3}, 2, (int32_t[]){0, 0}));
}
void test_ndarray_broadcast() {
/*
# array = np.array([[19.9, 29.9, 39.9, 49.9]], dtype=np.float64)
# >>> [[19.9 29.9 39.9 49.9]]
#
# array = np.broadcast_to(array, (2, 3, 4))
# >>> [[[19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]]
# >>> [[19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]]]
#
# assery array.strides == (0, 0, 8)
*/
BEGIN_TEST();
double in_data[4] = {19.9, 29.9, 39.9, 49.9};
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = {1, 4};
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {.data = (uint8_t*)in_data,
.itemsize = sizeof(double),
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides};
ndarray::basic::set_strides_by_shape(&ndarray);
const int32_t dst_ndims = 3;
int32_t dst_shape[dst_ndims] = {2, 3, 4};
int32_t dst_strides[dst_ndims] = {};
NDArray<int32_t> dst_ndarray = {
.ndims = dst_ndims, .shape = dst_shape, .strides = dst_strides};
ErrorContext errctx = create_testing_errctx();
ndarray::broadcast::broadcast_to(&errctx, &ndarray, &dst_ndarray);
assert_errctx_no_exception(&errctx);
assert_arrays_match(dst_ndims, ((int32_t[]){0, 0, 8}), dst_ndarray.strides);
assert_values_match(19.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){0, 0, 0}))));
assert_values_match(29.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){0, 0, 1}))));
assert_values_match(39.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){0, 0, 2}))));
assert_values_match(49.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){0, 0, 3}))));
assert_values_match(19.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){0, 1, 0}))));
assert_values_match(29.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){0, 1, 1}))));
assert_values_match(39.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){0, 1, 2}))));
assert_values_match(49.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){0, 1, 3}))));
assert_values_match(49.9,
*((double*)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, ((int32_t[]){1, 2, 3}))));
}
void run() {
test_can_broadcast_shape();
test_ndarray_broadcast();
}
} // namespace ndarray_broadcast
} // namespace test

220
nac3core/irrt/test/test_ndarray_indexing.hpp Normal file
View File

@ -0,0 +1,220 @@
#pragma once
#include <test/includes.hpp>
namespace test {
namespace ndarray_indexing {
void test_normal_1() {
/*
Reference Python code:
```python
ndarray = np.arange(12, dtype=np.float64).reshape((3, 4));
# array([[ 0., 1., 2., 3.],
# [ 4., 5., 6., 7.],
# [ 8., 9., 10., 11.]])
dst_ndarray = ndarray[-2:, 1::2]
# array([[ 5., 7.],
# [ 9., 11.]])
assert dst_ndarray.shape == (2, 2)
assert dst_ndarray.strides == (32, 16)
assert dst_ndarray[0, 0] == 5.0
assert dst_ndarray[0, 1] == 7.0
assert dst_ndarray[1, 0] == 9.0
assert dst_ndarray[1, 1] == 11.0
```
*/
BEGIN_TEST();
// Prepare src_ndarray
double src_data[12] = {0.0, 1.0, 2.0, 3.0, 4.0, 5.0,
6.0, 7.0, 8.0, 9.0, 10.0, 11.0};
int32_t src_itemsize = sizeof(double);
const int32_t src_ndims = 2;
int32_t src_shape[src_ndims] = {3, 4};
int32_t src_strides[src_ndims] = {};
NDArray<int32_t> src_ndarray = {.data = (uint8_t *)src_data,
.itemsize = src_itemsize,
.ndims = src_ndims,
.shape = src_shape,
.strides = src_strides};
ndarray::basic::set_strides_by_shape(&src_ndarray);
// Prepare dst_ndarray
const int32_t dst_ndims = 2;
int32_t dst_shape[dst_ndims] = {999, 999}; // Empty values
int32_t dst_strides[dst_ndims] = {999, 999}; // Empty values
NDArray<int32_t> dst_ndarray = {.data = nullptr,
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides};
// Create the subscripts in `ndarray[-2::, 1::2]`
UserSlice subscript_1;
subscript_1.set_start(-2);
UserSlice subscript_2;
subscript_2.set_start(1);
subscript_2.set_step(2);
const int32_t num_indexes = 2;
NDIndex indexes[num_indexes] = {
{.type = ND_INDEX_TYPE_SLICE, .data = (uint8_t *)&subscript_1},
{.type = ND_INDEX_TYPE_SLICE, .data = (uint8_t *)&subscript_2}};
ErrorContext errctx = create_testing_errctx();
ndarray::indexing::index(&errctx, num_indexes, indexes, &src_ndarray,
&dst_ndarray);
assert_errctx_no_exception(&errctx);
int32_t expected_shape[dst_ndims] = {2, 2};
int32_t expected_strides[dst_ndims] = {32, 16};
assert_arrays_match(dst_ndims, expected_shape, dst_ndarray.shape);
assert_arrays_match(dst_ndims, expected_strides, dst_ndarray.strides);
// dst_ndarray[0, 0]
assert_values_match(5.0,
*((double *)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, (int32_t[dst_ndims]){0, 0})));
// dst_ndarray[0, 1]
assert_values_match(7.0,
*((double *)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, (int32_t[dst_ndims]){0, 1})));
// dst_ndarray[1, 0]
assert_values_match(9.0,
*((double *)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, (int32_t[dst_ndims]){1, 0})));
// dst_ndarray[1, 1]
assert_values_match(11.0,
*((double *)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, (int32_t[dst_ndims]){1, 1})));
}
void test_normal_2() {
/*
```python
ndarray = np.arange(12, dtype=np.float64).reshape((3, 4))
# array([[ 0., 1., 2., 3.],
# [ 4., 5., 6., 7.],
# [ 8., 9., 10., 11.]])
dst_ndarray = ndarray[2, ::-2]
# array([11., 9.])
assert dst_ndarray.shape == (2,)
assert dst_ndarray.strides == (-16,)
assert dst_ndarray[0] == 11.0
assert dst_ndarray[1] == 9.0
```
*/
BEGIN_TEST();
// Prepare src_ndarray
double src_data[12] = {0.0, 1.0, 2.0, 3.0, 4.0, 5.0,
6.0, 7.0, 8.0, 9.0, 10.0, 11.0};
int32_t src_itemsize = sizeof(double);
const int32_t src_ndims = 2;
int32_t src_shape[src_ndims] = {3, 4};
int32_t src_strides[src_ndims] = {};
NDArray<int32_t> src_ndarray = {.data = (uint8_t *)src_data,
.itemsize = src_itemsize,
.ndims = src_ndims,
.shape = src_shape,
.strides = src_strides};
ndarray::basic::set_strides_by_shape(&src_ndarray);
// Prepare dst_ndarray
const int32_t dst_ndims = 1;
int32_t dst_shape[dst_ndims] = {999}; // Empty values
int32_t dst_strides[dst_ndims] = {999}; // Empty values
NDArray<int32_t> dst_ndarray = {.data = nullptr,
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides};
// Create the subscripts in `ndarray[2, ::-2]`
int32_t subscript_1 = 2;
UserSlice subscript_2;
subscript_2.set_step(-2);
const int32_t num_indexes = 2;
NDIndex indexes[num_indexes] = {
{.type = ND_INDEX_TYPE_SINGLE_ELEMENT, .data = (uint8_t *)&subscript_1},
{.type = ND_INDEX_TYPE_SLICE, .data = (uint8_t *)&subscript_2}};
ErrorContext errctx = create_testing_errctx();
ndarray::indexing::index(&errctx, num_indexes, indexes, &src_ndarray,
&dst_ndarray);
assert_errctx_no_exception(&errctx);
int32_t expected_shape[dst_ndims] = {2};
int32_t expected_strides[dst_ndims] = {-16};
assert_arrays_match(dst_ndims, expected_shape, dst_ndarray.shape);
assert_arrays_match(dst_ndims, expected_strides, dst_ndarray.strides);
assert_values_match(11.0,
*((double *)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, (int32_t[dst_ndims]){0})));
assert_values_match(9.0,
*((double *)ndarray::basic::get_pelement_by_indices(
&dst_ndarray, (int32_t[dst_ndims]){1})));
}
void test_index_subscript_out_of_bounds() {
/*
# Consider `my_array`
print(my_array.shape)
# (4, 5, 6)
my_array[2, 100] # error, index subscript at axis 1 is out of bounds
*/
BEGIN_TEST();
// Prepare src_ndarray
const int32_t src_ndims = 2;
int32_t src_shape[src_ndims] = {3, 4};
int32_t src_strides[src_ndims] = {};
NDArray<int32_t> src_ndarray = {
.data = (uint8_t *)nullptr, // placeholder, we wouldn't access it
.itemsize = sizeof(double), // placeholder
.ndims = src_ndims,
.shape = src_shape,
.strides = src_strides};
ndarray::basic::set_strides_by_shape(&src_ndarray);
// Create the subscripts in `my_array[2, 100]`
int32_t subscript_1 = 2;
int32_t subscript_2 = 100;
const int32_t num_indexes = 2;
NDIndex indexes[num_indexes] = {
{.type = ND_INDEX_TYPE_SINGLE_ELEMENT, .data = (uint8_t *)&subscript_1},
{.type = ND_INDEX_TYPE_SINGLE_ELEMENT,
.data = (uint8_t *)&subscript_2}};
// Prepare dst_ndarray
const int32_t dst_ndims = 0;
int32_t dst_shape[dst_ndims] = {};
int32_t dst_strides[dst_ndims] = {};
NDArray<int32_t> dst_ndarray = {.data = nullptr, // placehloder
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides};
ErrorContext errctx = create_testing_errctx();
ndarray::indexing::index(&errctx, num_indexes, indexes, &src_ndarray,
&dst_ndarray);
assert_errctx_has_exception(&errctx, errctx.exceptions->index_error);
}
void run() {
test_normal_1();
test_normal_2();
test_index_subscript_out_of_bounds();
}
} // namespace ndarray_indexing
} // namespace test

92
nac3core/irrt/test/test_slice.hpp Normal file
View File

@ -0,0 +1,92 @@
#pragma once
#include <irrt_everything.hpp>
#include <test/includes.hpp>
namespace test {
namespace slice {
void test_slice_normal() {
// Normal situation
BEGIN_TEST();
UserSlice user_slice;
user_slice.set_stop(5);
Slice slice = user_slice.indices(100);
printf("%d, %d, %d\n", slice.start, slice.stop, slice.step);
assert_values_match(0, slice.start);
assert_values_match(5, slice.stop);
assert_values_match(1, slice.step);
}
void test_slice_start_too_large() {
// Start is too large and should be clamped to length
BEGIN_TEST();
UserSlice user_slice;
user_slice.set_start(400);
Slice slice = user_slice.indices(100);
assert_values_match(100, slice.start);
assert_values_match(100, slice.stop);
assert_values_match(1, slice.step);
}
void test_slice_negative_start_stop() {
// Negative start/stop should be resolved
BEGIN_TEST();
UserSlice user_slice;
user_slice.set_start(-10);
user_slice.set_stop(-5);
Slice slice = user_slice.indices(100);
assert_values_match(90, slice.start);
assert_values_match(95, slice.stop);
assert_values_match(1, slice.step);
}
void test_slice_only_negative_step() {
// Things like `[::-5]` should be handled correctly
BEGIN_TEST();
UserSlice user_slice;
user_slice.set_step(-5);
Slice slice = user_slice.indices(100);
assert_values_match(99, slice.start);
assert_values_match(-1, slice.stop);
assert_values_match(-5, slice.step);
}
void test_slice_step_zero() {
// Step = 0 is a value error
BEGIN_TEST();
ErrorContext errctx = create_testing_errctx();
UserSlice user_slice;
user_slice.set_start(2);
user_slice.set_stop(12);
user_slice.set_step(0);
Slice slice;
user_slice.indices_checked(&errctx, 100, &slice);
assert_errctx_has_exception(&errctx, errctx.exceptions->value_error);
}
void run() {
test_slice_normal();
test_slice_start_too_large();
test_slice_negative_start_stop();
test_slice_only_negative_step();
test_slice_step_zero();
}
} // namespace slice
} // namespace test

188
nac3core/irrt/test/util.hpp Normal file
View File

@ -0,0 +1,188 @@
#pragma once
#include <cstdio>
#include <cstdlib>
template <class T>
void print_value(const T& value);
template <>
void print_value(const bool& value) {
printf("%s", value ? "true" : "false");
}
template <>
void print_value(const int8_t& value) {
printf("%d", value);
}
template <>
void print_value(const int32_t& value) {
printf("%d", value);
}
template <>
void print_value(const uint8_t& value) {
printf("%u", value);
}
template <>
void print_value(const uint32_t& value) {
printf("%u", value);
}
template <>
void print_value(const float& value) {
printf("%f", value);
}
template <>
void print_value(const double& value) {
printf("%f", value);
}
void __begin_test(const char* function_name, const char* file, int line) {
printf("######### Running %s @ %s:%d\n", function_name, file, line);
}
#define BEGIN_TEST() __begin_test(__FUNCTION__, __FILE__, __LINE__)
void test_fail() {
printf("[!] Test failed. Exiting with status code 1.\n");
exit(1);
}
template <typename T>
void debug_print_array(int len, const T* as) {
printf("[");
for (int i = 0; i < len; i++) {
if (i != 0) printf(", ");
print_value(as[i]);
}
printf("]");
}
void print_assertion_passed(const char* file, int line) {
printf("[*] Assertion passed on %s:%d\n", file, line);
}
void print_assertion_failed(const char* file, int line) {
printf("[!] Assertion failed on %s:%d\n", file, line);
}
void __assert_true(const char* file, int line, bool cond) {
if (cond) {
print_assertion_passed(file, line);
} else {
print_assertion_failed(file, line);
test_fail();
}
}
#define assert_true(cond) __assert_true(__FILE__, __LINE__, cond)
template <typename T>
void __assert_arrays_match(const char* file, int line, int len,
const T* expected, const T* got) {
if (arrays_match(len, expected, got)) {
print_assertion_passed(file, line);
} else {
print_assertion_failed(file, line);
printf("Expect = ");
debug_print_array(len, expected);
printf("\n");
printf(" Got = ");
debug_print_array(len, got);
printf("\n");
test_fail();
}
}
#define assert_arrays_match(len, expected, got) \
__assert_arrays_match(__FILE__, __LINE__, len, expected, got)
template <typename T>
void __assert_values_match(const char* file, int line, T expected, T got) {
if (expected == got) {
print_assertion_passed(file, line);
} else {
print_assertion_failed(file, line);
printf("Expect = ");
print_value(expected);
printf("\n");
printf(" Got = ");
print_value(got);
printf("\n");
test_fail();
}
}
#define assert_values_match(expected, got) \
__assert_values_match(__FILE__, __LINE__, expected, got)
// A fake set of ExceptionIds for testing only
const ErrorContextExceptions TEST_ERROR_CONTEXT_EXCEPTIONS = {
.index_error = 0,
.value_error = 1,
.assertion_error = 2,
.runtime_error = 3,
.type_error = 4,
};
ErrorContext create_testing_errctx() {
// Everything is global so it is fine to directly return a struct
// ErrorContext
ErrorContext errctx;
errctx.initialize(&TEST_ERROR_CONTEXT_EXCEPTIONS);
return errctx;
}
void print_errctx_content(ErrorContext* errctx) {
if (errctx->has_exception()) {
printf(
"(Exception ID %d): %s ... where param1 = %ld, param2 = %ld, "
"param3 = "
"%ld\n",
errctx->exception_id, errctx->msg, errctx->param1, errctx->param2,
errctx->param3);
} else {
printf("<no exception>\n");
}
}
void __assert_errctx_no_exception(const char* file, int line,
ErrorContext* errctx) {
if (errctx->has_exception()) {
print_assertion_failed(file, line);
printf("Expecting no exception but caught the following:\n\n");
print_errctx_content(errctx);
test_fail();
}
}
#define assert_errctx_no_exception(errctx) \
__assert_errctx_no_exception(__FILE__, __LINE__, errctx)
void __assert_errctx_has_exception(const char* file, int line,
ErrorContext* errctx,
ExceptionId expected_exception_id) {
if (errctx->has_exception()) {
if (errctx->exception_id != expected_exception_id) {
print_assertion_failed(file, line);
printf(
"Expecting exception id %d but got exception id %d. Error "
"caught:\n\n",
expected_exception_id, errctx->exception_id);
print_errctx_content(errctx);
test_fail();
}
} else {
print_assertion_failed(file, line);
printf("Expecting an exception, but there is none.");
test_fail();
}
}
#define assert_errctx_has_exception(errctx, expected_exception_id) \
__assert_errctx_has_exception(__FILE__, __LINE__, errctx, \
expected_exception_id)

631
nac3core/src/codegen/builtin_fns.rs
View File

@ -1,102 +1,26 @@
use inkwell::{
types::BasicTypeEnum,
values::{BasicValue, BasicValueEnum, IntValue, PointerValue},
FloatPredicate, IntPredicate, OptimizationLevel,
};
use inkwell::types::BasicTypeEnum;
use inkwell::values::BasicValueEnum;
use inkwell::{FloatPredicate, IntPredicate, OptimizationLevel};
use itertools::Itertools;
use super::{
classes::{
ArrayLikeValue, NDArrayValue, ProxyValue, RangeValue, TypedArrayLikeAccessor,
UntypedArrayLikeAccessor, UntypedArrayLikeMutator,
},
expr::destructure_range,
extern_fns, irrt,
irrt::calculate_len_for_slice_range,
llvm_intrinsics,
macros::codegen_unreachable,
numpy,
numpy::ndarray_elementwise_unaryop_impl,
stmt::gen_for_callback_incrementing,
CodeGenContext, CodeGenerator,
};
use crate::{
toplevel::{helper::PrimDef, numpy::unpack_ndarray_var_tys},
typecheck::typedef::{Type, TypeEnum},
};
use crate::codegen::classes::{NDArrayValue, ProxyValue, UntypedArrayLikeAccessor};
use crate::codegen::numpy::ndarray_elementwise_unaryop_impl;
use crate::codegen::stmt::gen_for_callback_incrementing;
use crate::codegen::{extern_fns, irrt, llvm_intrinsics, numpy, CodeGenContext, CodeGenerator};
use crate::toplevel::helper::PrimDef;
use crate::toplevel::numpy::unpack_ndarray_var_tys;
use crate::typecheck::typedef::Type;
/// Shorthand for [`unreachable!()`] when a type of argument is not supported.
///
/// The generated message will contain the function name and the name of the unsupported type.
fn unsupported_type(ctx: &CodeGenContext<'_, '_>, fn_name: &str, tys: &[Type]) -> ! {
codegen_unreachable!(
ctx,
unreachable!(
"{fn_name}() not supported for '{}'",
tys.iter().map(|ty| format!("'{}'", ctx.unifier.stringify(*ty))).join(", "),
)
}
/// Invokes the `len` builtin function.
pub fn call_len<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
) -> Result<IntValue<'ctx>, String> {
let llvm_i32 = ctx.ctx.i32_type();
let range_ty = ctx.primitives.range;
let (arg_ty, arg) = n;
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);
calculate_len_for_slice_range(generator, ctx, start, end, step)
} else {
match &*ctx.unifier.get_ty_immutable(arg_ty) {
TypeEnum::TTuple { ty, .. } => llvm_i32.const_int(ty.len() as u64, false),
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::List.id() => {
let zero = llvm_i32.const_zero();
let len = ctx
.build_gep_and_load(
arg.into_pointer_value(),
&[zero, llvm_i32.const_int(1, false)],
None,
)
.into_int_value();
ctx.builder.build_int_truncate_or_bit_cast(len, llvm_i32, "len").unwrap()
}
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
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",
"len() of unsized object",
[None, None, None],
ctx.current_loc,
);
let len = unsafe {
arg.dim_sizes().get_typed_unchecked(
ctx,
generator,
&llvm_usize.const_zero(),
None,
)
};
ctx.builder.build_int_truncate_or_bit_cast(len, llvm_i32, "len").unwrap()
}
_ => codegen_unreachable!(ctx),
}
})
}
/// Invokes the `int32` builtin function.
pub fn call_int32<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
@ -107,6 +31,7 @@ pub fn call_int32<'ctx, G: CodeGenerator + ?Sized>(
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
Ok(match n {
BasicValueEnum::IntValue(n) if matches!(n.get_type().get_bit_width(), 1 | 8) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.bool));
@ -677,7 +602,7 @@ pub fn call_ceil<'ctx, G: CodeGenerator + ?Sized>(
ret_elem_ty,
None,
NDArrayValue::from_ptr_val(n, llvm_usize, None),
|generator, ctx, val| call_ceil(generator, ctx, (elem_ty, val), ret_elem_ty),
|generator, ctx, val| call_floor(generator, ctx, (elem_ty, val), ret_elem_ty),
)?;
ndarray.as_base_value().into()
@ -794,7 +719,7 @@ pub fn call_numpy_minimum<'ctx, G: CodeGenerator + ?Sized>(
} else if is_ndarray2 {
unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty).0
} else {
codegen_unreachable!(ctx)
unreachable!()
};
let x1_scalar_ty = if is_ndarray1 { dtype } else { x1_ty };
@ -898,7 +823,7 @@ pub fn call_numpy_max_min<'ctx, G: CodeGenerator + ?Sized>(
match fn_name {
"np_argmin" | "np_argmax" => llvm_int64.const_zero().into(),
"np_max" | "np_min" => a,
_ => codegen_unreachable!(ctx),
_ => unreachable!(),
}
}
BasicValueEnum::PointerValue(n)
@ -953,7 +878,7 @@ pub fn call_numpy_max_min<'ctx, G: CodeGenerator + ?Sized>(
"np_argmax" | "np_max" => {
call_max(ctx, (elem_ty, accumulator), (elem_ty, elem))
}
_ => codegen_unreachable!(ctx),
_ => unreachable!(),
};
let updated_idx = match (accumulator, result) {
@ -990,7 +915,7 @@ pub fn call_numpy_max_min<'ctx, G: CodeGenerator + ?Sized>(
match fn_name {
"np_argmin" | "np_argmax" => ctx.builder.build_load(res_idx, "").unwrap(),
"np_max" | "np_min" => ctx.builder.build_load(accumulator_addr, "").unwrap(),
_ => codegen_unreachable!(ctx),
_ => unreachable!(),
}
}
@ -1056,7 +981,7 @@ pub fn call_numpy_maximum<'ctx, G: CodeGenerator + ?Sized>(
} else if is_ndarray2 {
unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty).0
} else {
codegen_unreachable!(ctx)
unreachable!()
};
let x1_scalar_ty = if is_ndarray1 { dtype } else { x1_ty };
@ -1086,9 +1011,9 @@ pub fn call_numpy_maximum<'ctx, G: CodeGenerator + ?Sized>(
/// * `(arg_ty, arg_val)`: The [`Type`] and llvm value of the input argument.
/// * `fn_name`: The name of the function, only used when throwing an error with [`unsupported_type`]
/// * `get_ret_elem_type`: A function that takes in the input scalar [`Type`], and returns the function's return scalar [`Type`].
/// Return a constant [`Type`] here if the return type does not depend on the input type.
/// Return a constant [`Type`] here if the return type does not depend on the input type.
/// * `on_scalar`: The function that acts on the scalars of the input. Returns [`Option::None`]
/// if the scalar type & value are faulty and should panic with [`unsupported_type`].
/// if the scalar type & value are faulty and should panic with [`unsupported_type`].
fn helper_call_numpy_unary_elementwise<'ctx, OnScalarFn, RetElemFn, G>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
@ -1199,9 +1124,9 @@ pub fn call_abs<'ctx, G: CodeGenerator + ?Sized>(
/// * `$name:ident`: The identifier of the rust function to be generated.
/// * `$fn_name:literal`: To be passed to the `fn_name` parameter of [`helper_call_numpy_unary_elementwise`]
/// * `$get_ret_elem_type:expr`: To be passed to the `get_ret_elem_type` parameter of [`helper_call_numpy_unary_elementwise`].
/// But there is no need to make it a reference.
/// But there is no need to make it a reference.
/// * `$on_scalar:expr`: To be passed to the `on_scalar` parameter of [`helper_call_numpy_unary_elementwise`].
/// But there is no need to make it a reference.
/// But there is no need to make it a reference.
macro_rules! create_helper_call_numpy_unary_elementwise {
($name:ident, $fn_name:literal, $get_ret_elem_type:expr, $on_scalar:expr) => {
#[allow(clippy::redundant_closure_call)]
@ -1228,7 +1153,7 @@ macro_rules! create_helper_call_numpy_unary_elementwise {
/// * `$name:ident`: The identifier of the rust function to be generated.
/// * `$fn_name:literal`: To be passed to the `fn_name` parameter of [`helper_call_numpy_unary_elementwise`].
/// * `$on_scalar:expr`: The closure (see below for its type) that acts on float scalar values and returns
/// the boolean results of LLVM type `i1`. The returned `i1` value will be converted into an `i8`.
/// the boolean results of LLVM type `i1`. The returned `i1` value will be converted into an `i8`.
///
/// ```ignore
/// // Type of `$on_scalar:expr`
@ -1496,7 +1421,7 @@ pub fn call_numpy_arctan2<'ctx, G: CodeGenerator + ?Sized>(
} else if is_ndarray2 {
unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty).0
} else {
codegen_unreachable!(ctx)
unreachable!()
};
let x1_scalar_ty = if is_ndarray1 { dtype } else { x1_ty };
@ -1563,7 +1488,7 @@ pub fn call_numpy_copysign<'ctx, G: CodeGenerator + ?Sized>(
} else if is_ndarray2 {
unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty).0
} else {
codegen_unreachable!(ctx)
unreachable!()
};
let x1_scalar_ty = if is_ndarray1 { dtype } else { x1_ty };
@ -1630,7 +1555,7 @@ pub fn call_numpy_fmax<'ctx, G: CodeGenerator + ?Sized>(
} else if is_ndarray2 {
unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty).0
} else {
codegen_unreachable!(ctx)
unreachable!()
};
let x1_scalar_ty = if is_ndarray1 { dtype } else { x1_ty };
@ -1697,7 +1622,7 @@ pub fn call_numpy_fmin<'ctx, G: CodeGenerator + ?Sized>(
} else if is_ndarray2 {
unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty).0
} else {
codegen_unreachable!(ctx)
unreachable!()
};
let x1_scalar_ty = if is_ndarray1 { dtype } else { x1_ty };
@ -1820,7 +1745,7 @@ pub fn call_numpy_hypot<'ctx, G: CodeGenerator + ?Sized>(
} else if is_ndarray2 {
unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty).0
} else {
codegen_unreachable!(ctx)
unreachable!()
};
let x1_scalar_ty = if is_ndarray1 { dtype } else { x1_ty };
@ -1887,7 +1812,7 @@ pub fn call_numpy_nextafter<'ctx, G: CodeGenerator + ?Sized>(
} else if is_ndarray2 {
unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty).0
} else {
codegen_unreachable!(ctx)
unreachable!()
};
let x1_scalar_ty = if is_ndarray1 { dtype } else { x1_ty };
@ -1911,501 +1836,3 @@ pub fn call_numpy_nextafter<'ctx, G: CodeGenerator + ?Sized>(
_ => unsupported_type(ctx, FN_NAME, &[x1_ty, x2_ty]),
})
}
/// Allocates a struct with the fields specified by `out_matrices` and returns a pointer to it
fn build_output_struct<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
out_matrices: Vec<BasicValueEnum<'ctx>>,
) -> PointerValue<'ctx> {
let field_ty =
out_matrices.iter().map(BasicValueEnum::get_type).collect::<Vec<BasicTypeEnum>>();
let out_ty = ctx.ctx.struct_type(&field_ty, false);
let out_ptr = ctx.builder.build_alloca(out_ty, "").unwrap();
for (i, v) in out_matrices.into_iter().enumerate() {
unsafe {
let ptr = ctx
.builder
.build_in_bounds_gep(
out_ptr,
&[
ctx.ctx.i32_type().const_zero(),
ctx.ctx.i32_type().const_int(i as u64, false),
],
"",
)
.unwrap();
ctx.builder.build_store(ptr, v).unwrap();
}
}
out_ptr
}
/// Invokes the `np_linalg_cholesky` linalg function
pub fn call_np_linalg_cholesky<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_cholesky";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty]);
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let out = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_cholesky(ctx, x1, out, None);
Ok(out)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_qr` linalg function
pub fn call_np_linalg_qr<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_qr";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let k = llvm_intrinsics::call_int_smin(ctx, dim0, dim1, None);
let out_q = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, k])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_r = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[k, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_qr(ctx, x1, out_q, out_r, None);
let out_ptr = build_output_struct(ctx, vec![out_q, out_r]);
Ok(ctx.builder.build_load(out_ptr, "QR_Factorization_result").map(Into::into).unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_svd` linalg function
pub fn call_np_linalg_svd<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_svd";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty]);
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let k = llvm_intrinsics::call_int_smin(ctx, dim0, dim1, None);
let out_u = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_s = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[k])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_vh = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim1, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_svd(ctx, x1, out_u, out_s, out_vh, None);
let out_ptr = build_output_struct(ctx, vec![out_u, out_s, out_vh]);
Ok(ctx.builder.build_load(out_ptr, "SVD_Factorization_result").map(Into::into).unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_inv` linalg function
pub fn call_np_linalg_inv<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_inv";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty]);
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let out = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_inv(ctx, x1, out, None);
Ok(out)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_pinv` linalg function
pub fn call_np_linalg_pinv<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_pinv";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty]);
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let out = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim1, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_pinv(ctx, x1, out, None);
Ok(out)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `sp_linalg_lu` linalg function
pub fn call_sp_linalg_lu<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "sp_linalg_lu";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty]);
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let k = llvm_intrinsics::call_int_smin(ctx, dim0, dim1, None);
let out_l = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, k])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_u = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[k, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_sp_linalg_lu(ctx, x1, out_l, out_u, None);
let out_ptr = build_output_struct(ctx, vec![out_l, out_u]);
Ok(ctx.builder.build_load(out_ptr, "LU_Factorization_result").map(Into::into).unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_matrix_power` linalg function
pub fn call_np_linalg_matrix_power<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
x2: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_matrix_power";
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
let x2 = call_float(generator, ctx, (x2_ty, x2)).unwrap();
let llvm_usize = generator.get_size_type(ctx.ctx);
if let (BasicValueEnum::PointerValue(n1), BasicValueEnum::FloatValue(n2)) = (x1, x2) {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty, x2_ty]);
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
// Changing second parameter to a `NDArray` for uniformity in function call
let n2_array = numpy::create_ndarray_const_shape(
generator,
ctx,
elem_ty,
&[llvm_usize.const_int(1, false)],
)
.unwrap();
unsafe {
n2_array.data().set_unchecked(
ctx,
generator,
&llvm_usize.const_zero(),
n2.as_basic_value_enum(),
);
};
let n2_array = n2_array.as_base_value().as_basic_value_enum();
let outdim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let outdim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let out = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[outdim0, outdim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_matrix_power(ctx, x1, n2_array, out, None);
Ok(out)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty, x2_ty])
}
}
/// Invokes the `np_linalg_det` linalg function
pub fn call_np_linalg_det<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_matrix_power";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(_) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty]);
};
// Changing second parameter to a `NDArray` for uniformity in function call
let out = numpy::create_ndarray_const_shape(
generator,
ctx,
elem_ty,
&[llvm_usize.const_int(1, false)],
)
.unwrap();
extern_fns::call_np_linalg_det(ctx, x1, out.as_base_value().as_basic_value_enum(), None);
let res =
unsafe { out.data().get_unchecked(ctx, generator, &llvm_usize.const_zero(), None) };
Ok(res)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `sp_linalg_schur` linalg function
pub fn call_sp_linalg_schur<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "sp_linalg_schur";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty]);
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let out_t = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_z = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_sp_linalg_schur(ctx, x1, out_t, out_z, None);
let out_ptr = build_output_struct(ctx, vec![out_t, out_z]);
Ok(ctx.builder.build_load(out_ptr, "Schur_Factorization_result").map(Into::into).unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `sp_linalg_hessenberg` linalg function
pub fn call_sp_linalg_hessenberg<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "sp_linalg_hessenberg";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unsupported_type(ctx, FN_NAME, &[x1_ty]);
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let out_h = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_q = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_sp_linalg_hessenberg(ctx, x1, out_h, out_q, None);
let out_ptr = build_output_struct(ctx, vec![out_h, out_q]);
Ok(ctx
.builder
.build_load(out_ptr, "Hessenberg_decomposition_result")
.map(Into::into)
.unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}

23
nac3core/src/codegen/classes.rs
View File

@ -1,16 +1,17 @@
use inkwell::{
context::Context,
types::{AnyTypeEnum, ArrayType, BasicType, BasicTypeEnum, IntType, PointerType, StructType},
values::{ArrayValue, BasicValue, BasicValueEnum, IntValue, PointerValue, StructValue},
AddressSpace, IntPredicate,
};
use super::{
use crate::codegen::{
irrt::{call_ndarray_calc_size, call_ndarray_flatten_index},
llvm_intrinsics::call_int_umin,
stmt::gen_for_callback_incrementing,
CodeGenContext, CodeGenerator,
};
use inkwell::context::Context;
use inkwell::types::{ArrayType, BasicType, StructType};
use inkwell::values::{ArrayValue, BasicValue, StructValue};
use inkwell::{
types::{AnyTypeEnum, BasicTypeEnum, IntType, PointerType},
values::{BasicValueEnum, IntValue, PointerValue},
AddressSpace, IntPredicate,
};
/// A LLVM type that is used to represent a non-primitive type in NAC3.
pub trait ProxyType<'ctx>: Into<Self::Base> {
@ -1249,13 +1250,11 @@ impl<'ctx> NDArrayType<'ctx> {
/// Returns the element type of this `ndarray` type.
#[must_use]
pub fn element_type(&self) -> AnyTypeEnum<'ctx> {
pub fn element_type(&self) -> BasicTypeEnum<'ctx> {
self.as_base_type()
.get_element_type()
.into_struct_type()
.get_field_type_at_index(2)
.map(BasicTypeEnum::into_pointer_type)
.map(PointerType::get_element_type)
.unwrap()
}
}
@ -1405,7 +1404,7 @@ impl<'ctx> NDArrayValue<'ctx> {
/// Returns the double-indirection pointer to the `data` array, as if by calling `getelementptr`
/// on the field.
pub fn ptr_to_data(&self, ctx: &CodeGenContext<'ctx, '_>) -> PointerValue<'ctx> {
fn ptr_to_data(&self, ctx: &CodeGenContext<'ctx, '_>) -> PointerValue<'ctx> {
let llvm_i32 = ctx.ctx.i32_type();
let var_name = self.name.map(|v| format!("{v}.data.addr")).unwrap_or_default();

29
nac3core/src/codegen/concrete_type.rs
View File

@ -1,9 +1,3 @@
use std::collections::HashMap;
use indexmap::IndexMap;
use nac3parser::ast::StrRef;
use crate::{
symbol_resolver::SymbolValue,
toplevel::DefinitionId,
@ -15,6 +9,10 @@ use crate::{
},
};
use indexmap::IndexMap;
use nac3parser::ast::StrRef;
use std::collections::HashMap;
pub struct ConcreteTypeStore {
store: Vec<ConcreteTypeEnum>,
}
@ -27,7 +25,6 @@ pub struct ConcreteFuncArg {
pub name: StrRef,
pub ty: ConcreteType,
pub default_value: Option<SymbolValue>,
pub is_vararg: bool,
}
#[derive(Clone, Debug)]
@ -49,7 +46,6 @@ pub enum ConcreteTypeEnum {
TPrimitive(Primitive),
TTuple {
ty: Vec<ConcreteType>,
is_vararg_ctx: bool,
},
TObj {
obj_id: DefinitionId,
@ -106,16 +102,8 @@ impl ConcreteTypeStore {
.iter()
.map(|arg| ConcreteFuncArg {
name: arg.name,
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)
},
ty: 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),
@ -170,12 +158,11 @@ impl ConcreteTypeStore {
cache.insert(ty, None);
let ty_enum = unifier.get_ty(ty);
let result = match &*ty_enum {
TypeEnum::TTuple { ty, is_vararg_ctx } => ConcreteTypeEnum::TTuple {
TypeEnum::TTuple { ty } => 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,
@ -261,12 +248,11 @@ impl ConcreteTypeStore {
*cache.get_mut(&cty).unwrap() = Some(ty);
return ty;
}
ConcreteTypeEnum::TTuple { ty, is_vararg_ctx } => TypeEnum::TTuple {
ConcreteTypeEnum::TTuple { ty } => 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) }
@ -291,7 +277,6 @@ 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),

1430
nac3core/src/codegen/expr.rs
View File

File diff suppressed because it is too large Load Diff

75
nac3core/src/codegen/extern_fns.rs
View File

@ -1,10 +1,8 @@
use inkwell::{
attributes::{Attribute, AttributeLoc},
values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue},
};
use inkwell::attributes::{Attribute, AttributeLoc};
use inkwell::values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue};
use itertools::Either;
use super::CodeGenContext;
use crate::codegen::CodeGenContext;
/// Macro to generate extern function
/// Both function return type and function parameter type are `FloatValue`
@ -15,11 +13,11 @@ use super::CodeGenContext;
/// * `$extern_fn:literal`: Name of underlying extern function
///
/// Optional Arguments:
/// * `$(,$attributes:literal)*)`: Attributes linked with the extern function.
/// The default attributes are "mustprogress", "nofree", "nounwind", "willreturn", and "writeonly".
/// These will be used unless other attributes are specified
/// * `$(,$attributes:literal)*)`: Attributes linked with the extern function
/// The default attributes are "mustprogress", "nofree", "nounwind", "willreturn", and "writeonly"
/// These will be used unless other attributes are specified
/// * `$(,$args:ident)*`: Operands of the extern function
/// The data type of these operands will be set to `FloatValue`
/// The data type of these operands will be set to `FloatValue`
///
macro_rules! generate_extern_fn {
("unary", $fn_name:ident, $extern_fn:literal) => {
@ -132,62 +130,3 @@ pub fn call_ldexp<'ctx>(
.map(Either::unwrap_left)
.unwrap()
}
/// Macro to generate `np_linalg` and `sp_linalg` functions
/// The function takes as input `NDArray` and returns ()
///
/// Arguments:
/// * `$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);

48
nac3core/src/codegen/generator.rs
View File

@ -1,18 +1,16 @@
use crate::{
codegen::{bool_to_i1, bool_to_i8, classes::ArraySliceValue, expr::*, stmt::*, CodeGenContext},
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, TopLevelDef},
typecheck::typedef::{FunSignature, Type},
};
use inkwell::{
context::Context,
types::{BasicTypeEnum, IntType},
values::{BasicValueEnum, IntValue, PointerValue},
};
use nac3parser::ast::{Expr, Stmt, StrRef};
use super::{bool_to_i1, bool_to_i8, classes::ArraySliceValue, expr::*, stmt::*, CodeGenContext};
use crate::{
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, TopLevelDef},
typecheck::typedef::{FunSignature, Type},
};
pub trait CodeGenerator {
/// Return the module name for the code generator.
fn get_name(&self) -> &str;
@ -59,7 +57,6 @@ 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>(
@ -134,39 +131,6 @@ pub trait CodeGenerator {
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.
/// Return true if the while loop must early return
fn gen_while(

198
nac3core/src/codegen/irrt/error_context.rs Normal file
View File

@ -0,0 +1,198 @@
use super::util::{function::CallFunction, get_sizet_dependent_function_name};
use crate::codegen::{
model::*,
structure::{cslice::CSlice, exception::ExceptionId},
CodeGenContext, CodeGenerator,
};
#[allow(clippy::struct_field_names)]
pub struct ErrorContextExceptionsFields<F: FieldVisitor> {
pub index_error: F::Field<IntModel<ExceptionId>>,
pub value_error: F::Field<IntModel<ExceptionId>>,
pub assertion_error: F::Field<IntModel<ExceptionId>>,
pub runtime_error: F::Field<IntModel<ExceptionId>>,
pub type_error: F::Field<IntModel<ExceptionId>>,
}
/// Corresponds to IRRT's `struct ErrorContextExceptions`
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
pub struct ErrorContextExceptions;
impl StructKind for ErrorContextExceptions {
type Fields<F: FieldVisitor> = ErrorContextExceptionsFields<F>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F> {
Self::Fields {
index_error: visitor.add("index_error"),
value_error: visitor.add("value_error"),
assertion_error: visitor.add("assertion_error"),
runtime_error: visitor.add("runtime_error"),
type_error: visitor.add("type_error"),
}
}
}
pub struct ErrorContextFields<F: FieldVisitor> {
pub exceptions: F::Field<PtrModel<StructModel<ErrorContextExceptions>>>,
pub exception_id: F::Field<IntModel<ExceptionId>>,
pub msg: F::Field<PtrModel<IntModel<Byte>>>,
pub param1: F::Field<IntModel<Int64>>,
pub param2: F::Field<IntModel<Int64>>,
pub param3: F::Field<IntModel<Int64>>,
}
/// Corresponds to IRRT's `struct ErrorContext`
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
pub struct ErrorContext;
impl StructKind for ErrorContext {
type Fields<F: FieldVisitor> = ErrorContextFields<F>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F> {
Self::Fields {
exceptions: visitor.add("exceptions"),
exception_id: visitor.add("exception_id"),
msg: visitor.add("msg"),
param1: visitor.add("param1"),
param2: visitor.add("param2"),
param3: visitor.add("param3"),
}
}
}
/// Build an [`ErrorContextExceptions`] loaded with resolved [`ExceptionID`]s according to the [`SymbolResolver`].
fn build_error_context_exceptions<'ctx>(
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
) -> Ptr<'ctx, StructModel<ErrorContextExceptions>> {
let exceptions =
StructModel(ErrorContextExceptions).alloca(tyctx, ctx, "error_context_exceptions");
let i32_model = IntModel(Int32);
let get_string_id = |string_id| {
i32_model.constant(tyctx, ctx.ctx, ctx.resolver.get_string_id(string_id) as u64)
};
exceptions.gep(ctx, |f| f.index_error).store(ctx, get_string_id("0:IndexError"));
exceptions.gep(ctx, |f| f.value_error).store(ctx, get_string_id("0:ValueError"));
exceptions.gep(ctx, |f| f.assertion_error).store(ctx, get_string_id("0:AssertionError"));
exceptions.gep(ctx, |f| f.runtime_error).store(ctx, get_string_id("0:RuntimeError"));
exceptions.gep(ctx, |f| f.type_error).store(ctx, get_string_id("0:TypeError"));
exceptions
}
pub fn call_nac3_error_context_initialize<'ctx>(
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
perrctx: Ptr<'ctx, StructModel<ErrorContext>>,
pexceptions: Ptr<'ctx, StructModel<ErrorContextExceptions>>,
) {
CallFunction::begin(tyctx, ctx, "__nac3_error_context_initialize")
.arg("errctx", perrctx)
.arg("exceptions", pexceptions)
.returning_void();
}
pub fn call_nac3_error_context_has_exception<'ctx>(
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
perrctx: Ptr<'ctx, StructModel<ErrorContext>>,
) -> Int<'ctx, Bool> {
CallFunction::begin(tyctx, ctx, "__nac3_error_context_has_exception")
.arg("errctx", perrctx)
.returning("has_exception")
}
pub fn call_nac3_error_context_get_exception_str<'ctx>(
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
perrctx: Ptr<'ctx, StructModel<ErrorContext>>,
dst_str: Ptr<'ctx, StructModel<CSlice>>,
) {
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_error_context_get_exception_str"),
)
.arg("errctx", perrctx)
.arg("dst_str", dst_str)
.returning_void();
}
/// Setup a [`ErrorContext`] that could be passed to IRRT functions taking in a `ErrorContext* errctx`
/// for error reporting purposes.
///
/// Also see: [`check_error_context`]
pub fn setup_error_context<'ctx>(
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
) -> Ptr<'ctx, StructModel<ErrorContext>> {
let errctx_model = StructModel(ErrorContext);
let exceptions = build_error_context_exceptions(tyctx, ctx);
let errctx_ptr = errctx_model.alloca(tyctx, ctx, "errctx");
call_nac3_error_context_initialize(tyctx, ctx, errctx_ptr, exceptions);
errctx_ptr
}
/// Check a [`ErrorContext`] to see if it contains error. **If there is an error,
/// a Pythonic exception will be raised in the firmware**.
pub fn check_error_context<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
perrctx: Ptr<'ctx, StructModel<ErrorContext>>,
) {
let tyctx = generator.type_context(ctx.ctx);
let cslice_model = StructModel(CSlice);
let current_bb = ctx.builder.get_insert_block().unwrap();
let irrt_has_exception_bb = ctx.ctx.insert_basic_block_after(current_bb, "irrt_has_exception");
let end_bb = ctx.ctx.insert_basic_block_after(irrt_has_exception_bb, "end");
// Inserting into `current_bb`
let has_exception = call_nac3_error_context_has_exception(tyctx, ctx, perrctx);
ctx.builder
.build_conditional_branch(has_exception.value, irrt_has_exception_bb, end_bb)
.unwrap();
// Inserting into `irrt_has_exception_bb`
ctx.builder.position_at_end(irrt_has_exception_bb);
// Load all the values for `ctx.make_assert_impl_by_id`
let pexception_str = cslice_model.alloca(tyctx, ctx, "exception_str");
call_nac3_error_context_get_exception_str(tyctx, ctx, perrctx, pexception_str);
let exception_id = perrctx.gep(ctx, |f| f.exception_id).load(tyctx, ctx, "exception_id");
let msg = pexception_str.load(tyctx, ctx, "msg");
let param1 = perrctx.gep(ctx, |f| f.param1).load(tyctx, ctx, "param1");
let param2 = perrctx.gep(ctx, |f| f.param2).load(tyctx, ctx, "param2");
let param3 = perrctx.gep(ctx, |f| f.param3).load(tyctx, ctx, "param3");
ctx.raise_exn_impl(
generator,
exception_id,
msg,
[Some(param1), Some(param2), Some(param3)],
ctx.current_loc,
);
// Position to `end_bb` for continuation
ctx.builder.position_at_end(end_bb);
}
pub fn call_nac3_dummy_raise<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext,
) {
let tyctx = generator.type_context(ctx.ctx);
let errctx = setup_error_context(tyctx, ctx);
CallFunction::begin(tyctx, ctx, "__nac3_error_dummy_raise")
.arg("errctx", errctx)
.returning_void();
check_error_context(generator, ctx, errctx);
}

106
nac3core/src/codegen/irrt/mod.rs
View File

@ -1,30 +1,35 @@
use crate::typecheck::typedef::Type;
pub mod error_context;
pub mod ndarray;
pub mod slice;
mod test;
mod util;
use super::model::*;
use super::{
classes::{
ArrayLikeIndexer, ArrayLikeValue, ArraySliceValue, ListValue, NDArrayValue,
TypedArrayLikeAdapter, UntypedArrayLikeAccessor,
},
llvm_intrinsics, CodeGenContext, CodeGenerator,
};
use crate::codegen::classes::TypedArrayLikeAccessor;
use crate::codegen::stmt::gen_for_callback_incrementing;
use inkwell::{
attributes::{Attribute, AttributeLoc},
context::Context,
memory_buffer::MemoryBuffer,
module::Module,
types::{BasicTypeEnum, IntType},
values::{BasicValue, BasicValueEnum, CallSiteValue, FloatValue, IntValue},
values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue},
AddressSpace, IntPredicate,
};
use itertools::Either;
use nac3parser::ast::Expr;
use super::{
classes::{
ArrayLikeIndexer, ArrayLikeValue, ArraySliceValue, ListValue, NDArrayValue,
TypedArrayLikeAccessor, TypedArrayLikeAdapter, UntypedArrayLikeAccessor,
},
llvm_intrinsics,
macros::codegen_unreachable,
stmt::gen_for_callback_incrementing,
CodeGenContext, CodeGenerator,
};
use crate::{symbol_resolver::SymbolResolver, typecheck::typedef::Type};
#[must_use]
pub fn load_irrt<'ctx>(ctx: &'ctx Context, symbol_resolver: &dyn SymbolResolver) -> Module<'ctx> {
pub fn load_irrt(ctx: &Context) -> Module {
let bitcode_buf = MemoryBuffer::create_from_memory_range(
include_bytes!(concat!(env!("OUT_DIR"), "/irrt.bc")),
"irrt_bitcode_buffer",
@ -40,25 +45,6 @@ pub fn load_irrt<'ctx>(ctx: &'ctx Context, symbol_resolver: &dyn SymbolResolver)
let function = irrt_mod.get_function(symbol).unwrap();
function.add_attribute(AttributeLoc::Function, ctx.create_enum_attribute(inline_attr, 0));
}
// Initialize all global `EXN_*` exception IDs in IRRT with the [`SymbolResolver`].
let exn_id_type = ctx.i32_type();
let errors = &[
("EXN_INDEX_ERROR", "0:IndexError"),
("EXN_VALUE_ERROR", "0:ValueError"),
("EXN_ASSERTION_ERROR", "0:AssertionError"),
("EXN_TYPE_ERROR", "0:TypeError"),
];
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 = irrt_mod.get_global(irrt_name).unwrap_or_else(|| {
panic!("Exception symbol name '{irrt_name}' should exist in the IRRT LLVM module")
});
global.set_initializer(&exn_id);
}
irrt_mod
}
@ -76,7 +62,7 @@ pub fn integer_power<'ctx, G: CodeGenerator + ?Sized>(
(64, 64, true) => "__nac3_int_exp_int64_t",
(32, 32, false) => "__nac3_int_exp_uint32_t",
(64, 64, false) => "__nac3_int_exp_uint64_t",
_ => codegen_unreachable!(ctx),
_ => unreachable!(),
};
let base_type = base.get_type();
let pow_fun = ctx.module.get_function(symbol).unwrap_or_else(|| {
@ -435,14 +421,29 @@ pub fn list_slice_assignment<'ctx, G: CodeGenerator + ?Sized>(
.unwrap();
let cond_1 = ctx.builder.build_and(dest_step_eq_one, src_slt_dest, "slice_cond_1").unwrap();
let cond = ctx.builder.build_or(src_eq_dest, cond_1, "slice_cond").unwrap();
ctx.make_assert(
generator,
cond,
"0:ValueError",
"attempt to assign sequence of size {0} to slice of size {1} with step size {2}",
[Some(src_slice_len), Some(dest_slice_len), Some(dest_idx.2)],
ctx.current_loc,
);
// TODO: Temporary fix. Rewrite `list_slice_assignment` later
// Exception params should have been i64
{
let type_context = generator.type_context(ctx.ctx);
let param_model = IntModel(Int64);
let src_slice_len =
param_model.s_extend_or_bit_cast(type_context, ctx, src_slice_len, "src_slice_len");
let dest_slice_len =
param_model.s_extend_or_bit_cast(type_context, ctx, dest_slice_len, "dest_slice_len");
let dest_idx_2 =
param_model.s_extend_or_bit_cast(type_context, ctx, dest_idx.2, "dest_idx_2");
ctx.make_assert(
generator,
cond,
"0:ValueError",
"attempt to assign sequence of size {0} to slice of size {1} with step size {2}",
[Some(src_slice_len.value), Some(dest_slice_len.value), Some(dest_idx_2.value)],
ctx.current_loc,
);
}
let new_len = {
let args = vec![
@ -462,7 +463,7 @@ pub fn list_slice_assignment<'ctx, G: CodeGenerator + ?Sized>(
BasicTypeEnum::IntType(t) => t.size_of(),
BasicTypeEnum::PointerType(t) => t.size_of(),
BasicTypeEnum::StructType(t) => t.size_of().unwrap(),
_ => codegen_unreachable!(ctx),
_ => unreachable!(),
};
ctx.builder.build_int_truncate_or_bit_cast(s, int32, "size").unwrap()
}
@ -589,8 +590,7 @@ pub fn call_j0<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> Flo
///
/// * `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.
/// 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, '_>,
@ -607,7 +607,7 @@ where
let ndarray_calc_size_fn_name = match llvm_usize.get_bit_width() {
32 => "__nac3_ndarray_calc_size",
64 => "__nac3_ndarray_calc_size64",
bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
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()],
@ -642,7 +642,7 @@ where
///
/// * `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`.
/// `NDArray`.
pub fn call_ndarray_calc_nd_indices<'ctx, G: CodeGenerator + ?Sized>(
generator: &G,
ctx: &mut CodeGenContext<'ctx, '_>,
@ -658,7 +658,7 @@ pub fn call_ndarray_calc_nd_indices<'ctx, G: CodeGenerator + ?Sized>(
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 => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
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(|| {
@ -727,7 +727,7 @@ where
let ndarray_flatten_index_fn_name = match llvm_usize.get_bit_width() {
32 => "__nac3_ndarray_flatten_index",
64 => "__nac3_ndarray_flatten_index64",
bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
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(|| {
@ -766,7 +766,7 @@ where
/// multidimensional index.
///
/// * `ndarray` - LLVM pointer to the `NDArray`. This value must be the LLVM representation of an
/// `NDArray`.
/// `NDArray`.
/// * `indices` - The multidimensional index to compute the flattened index for.
pub fn call_ndarray_flatten_index<'ctx, G, Index>(
generator: &mut G,
@ -795,7 +795,7 @@ pub fn call_ndarray_calc_broadcast<'ctx, G: CodeGenerator + ?Sized>(
let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
32 => "__nac3_ndarray_calc_broadcast",
64 => "__nac3_ndarray_calc_broadcast64",
bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
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(|| {
@ -915,7 +915,7 @@ pub fn call_ndarray_calc_broadcast_index<
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 => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
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(|| {

153
nac3core/src/codegen/irrt/ndarray/basic.rs Normal file
View File

@ -0,0 +1,153 @@
use crate::codegen::irrt::error_context::{check_error_context, setup_error_context};
use crate::codegen::irrt::slice::SliceIndex;
use crate::codegen::irrt::util::function::CallFunction;
use crate::codegen::irrt::util::get_sizet_dependent_function_name;
use crate::codegen::model::*;
use crate::codegen::structure::ndarray::NpArray;
use crate::codegen::{CodeGenContext, CodeGenerator};
pub fn call_nac3_ndarray_size<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray_ptr: Ptr<'ctx, StructModel<NpArray>>,
) -> Int<'ctx, SizeT> {
let tyctx = generator.type_context(ctx.ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_size"),
)
.arg("ndarray", ndarray_ptr)
.returning("size")
}
pub fn call_nac3_ndarray_nbytes<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray_ptr: Ptr<'ctx, StructModel<NpArray>>,
) -> Int<'ctx, SizeT> {
let tyctx = generator.type_context(ctx.ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_nbytes"),
)
.arg("ndarray", ndarray_ptr)
.returning("nbytes")
}
pub fn call_nac3_ndarray_len<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray_ptr: Ptr<'ctx, StructModel<NpArray>>,
) -> Int<'ctx, SliceIndex> {
let tyctx = generator.type_context(ctx.ctx);
let slice_index_model = IntModel(SliceIndex::default());
let dst_len = slice_index_model.alloca(tyctx, ctx, "dst_len");
let errctx = setup_error_context(tyctx, ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_len"),
)
.arg("errctx", errctx)
.arg("ndarray", ndarray_ptr)
.arg("dst_len", dst_len)
.returning_void();
check_error_context(generator, ctx, errctx);
dst_len.load(tyctx, ctx, "len")
}
pub fn call_nac3_ndarray_util_assert_shape_no_negative<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndims: Int<'ctx, SizeT>,
shape: Ptr<'ctx, IntModel<SizeT>>,
) {
let tyctx = generator.type_context(ctx.ctx);
let errctx = setup_error_context(tyctx, ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_util_assert_shape_no_negative"),
)
.arg("errctx", errctx)
.arg("ndims", ndims)
.arg("shape", shape)
.returning_void();
check_error_context(generator, ctx, errctx);
}
pub fn call_nac3_ndarray_set_strides_by_shape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray_ptr: Ptr<'ctx, StructModel<NpArray>>,
) {
let tyctx = generator.type_context(ctx.ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_set_strides_by_shape"),
)
.arg("ndarray", ndarray_ptr)
.returning_void();
}
pub fn call_nac3_ndarray_is_c_contiguous<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarray_ptr: Ptr<'ctx, StructModel<NpArray>>,
) -> Int<'ctx, Bool> {
let tyctx = generator.type_context(ctx.ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_is_c_contiguous"),
)
.arg("ndarray", ndarray_ptr)
.returning("is_c_contiguous")
}
pub fn call_nac3_ndarray_copy_data<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_ndarray: Ptr<'ctx, StructModel<NpArray>>,
dst_ndarray: Ptr<'ctx, StructModel<NpArray>>,
) {
let tyctx = generator.type_context(ctx.ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_copy_data"),
)
.arg("src_ndarray", src_ndarray)
.arg("dst_ndarray", dst_ndarray)
.returning_void();
}
pub fn call_nac3_ndarray_get_nth_pelement<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
pndarray: Ptr<'ctx, StructModel<NpArray>>,
index: Int<'ctx, SizeT>,
) -> Ptr<'ctx, IntModel<Byte>> {
let tyctx = generator.type_context(ctx.ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_get_nth_pelement"),
)
.arg("ndarray", pndarray)
.arg("index", index)
.returning("pelement")
}

74
nac3core/src/codegen/irrt/ndarray/broadcast.rs Normal file
View File

@ -0,0 +1,74 @@
use crate::codegen::{
irrt::{
error_context::{check_error_context, setup_error_context},
util::{function::CallFunction, get_sizet_dependent_function_name},
},
model::*,
structure::ndarray::NpArray,
CodeGenContext, CodeGenerator,
};
pub fn call_nac3_ndarray_broadcast_to<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_ndarray: Ptr<'ctx, StructModel<NpArray>>,
dst_ndarray: Ptr<'ctx, StructModel<NpArray>>,
) {
let tyctx = generator.type_context(ctx.ctx);
let perrctx = setup_error_context(tyctx, ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_broadcast_to"),
)
.arg("errctx", perrctx)
.arg("src_ndarray", src_ndarray)
.arg("dst_ndarray", dst_ndarray)
.returning_void();
check_error_context(generator, ctx, perrctx);
}
/// Fields of [`ShapeEntry`]
pub struct ShapeEntryFields<F: FieldVisitor> {
pub ndims: F::Field<IntModel<SizeT>>,
pub shape: F::Field<PtrModel<IntModel<SizeT>>>,
}
#[derive(Debug, Clone, Copy, Default)]
pub struct ShapeEntry;
impl StructKind for ShapeEntry {
type Fields<F: FieldVisitor> = ShapeEntryFields<F>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F> {
Self::Fields { ndims: visitor.add("ndims"), shape: visitor.add("shape") }
}
}
pub fn call_nac3_ndarray_broadcast_shapes<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
num_shape_entries: Int<'ctx, SizeT>,
shape_entries: Ptr<'ctx, StructModel<ShapeEntry>>,
dst_ndims: Int<'ctx, SizeT>,
dst_shape: Ptr<'ctx, IntModel<SizeT>>,
) {
let tyctx = generator.type_context(ctx.ctx);
let perrctx = setup_error_context(tyctx, ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_broadcast_shapes"),
)
.arg("errctx", perrctx)
.arg("num_shapes", num_shape_entries)
.arg("shapes", shape_entries)
.arg("dst_ndims", dst_ndims)
.arg("dst_shape", dst_shape)
.returning_void();
check_error_context(generator, ctx, perrctx);
}

170
nac3core/src/codegen/irrt/ndarray/indexing.rs Normal file
View File

@ -0,0 +1,170 @@
use crate::codegen::{
irrt::{
error_context::{check_error_context, setup_error_context},
slice::{RustUserSlice, SliceIndex, UserSlice},
util::{function::CallFunction, get_sizet_dependent_function_name},
},
model::*,
structure::ndarray::NpArray,
CodeGenContext, CodeGenerator,
};
pub type NDIndexType = Byte;
#[derive(Debug, Clone, Copy)]
pub struct NDIndexFields<F: FieldVisitor> {
pub type_: F::Field<IntModel<NDIndexType>>, // Defined to be uint8_t in IRRT
pub data: F::Field<PtrModel<IntModel<Byte>>>,
}
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
pub struct NDIndex;
impl StructKind for NDIndex {
type Fields<F: FieldVisitor> = NDIndexFields<F>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F> {
Self::Fields { type_: visitor.add("type"), data: visitor.add("data") }
}
}
// An enum variant to store the content
// and type of an NDIndex in high level.
#[derive(Debug, Clone)]
pub enum RustNDIndex<'ctx> {
SingleElement(Int<'ctx, SliceIndex>),
Slice(RustUserSlice<'ctx>),
}
impl<'ctx> RustNDIndex<'ctx> {
fn get_type_id(&self) -> u64 {
// Defined in IRRT, must be in sync
match self {
RustNDIndex::SingleElement(_) => 0,
RustNDIndex::Slice(_) => 1,
}
}
fn write_to_ndindex(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
dst_ndindex_ptr: Ptr<'ctx, StructModel<NDIndex>>,
) {
let ndindex_type_model = IntModel(NDIndexType::default());
let slice_index_model = IntModel(SliceIndex::default());
let user_slice_model = StructModel(UserSlice);
// Set `dst_ndindex_ptr->type`
dst_ndindex_ptr
.gep(ctx, |f| f.type_)
.store(ctx, ndindex_type_model.constant(tyctx, ctx.ctx, self.get_type_id()));
// Set `dst_ndindex_ptr->data`
let data = match self {
RustNDIndex::SingleElement(in_index) => {
let index_ptr = slice_index_model.alloca(tyctx, ctx, "index");
index_ptr.store(ctx, *in_index);
index_ptr.transmute(tyctx, ctx, IntModel(Byte), "")
}
RustNDIndex::Slice(in_rust_slice) => {
let user_slice_ptr = user_slice_model.alloca(tyctx, ctx, "user_slice");
in_rust_slice.write_to_user_slice(tyctx, ctx, user_slice_ptr);
user_slice_ptr.transmute(tyctx, ctx, IntModel(Byte), "")
}
};
dst_ndindex_ptr.gep(ctx, |f| f.data).store(ctx, data);
}
/// Allocate an array of `NDIndex`es on the stack and return its stack pointer.
pub fn alloca_ndindexes(
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
in_ndindexes: &[RustNDIndex<'ctx>],
) -> (Int<'ctx, SizeT>, Ptr<'ctx, StructModel<NDIndex>>) {
let sizet_model = IntModel(SizeT);
let ndindex_model = StructModel(NDIndex);
let num_ndindexes = sizet_model.constant(tyctx, ctx.ctx, in_ndindexes.len() as u64);
let ndindexes = ndindex_model.array_alloca(tyctx, ctx, num_ndindexes.value, "ndindexes");
for (i, in_ndindex) in in_ndindexes.iter().enumerate() {
let i = sizet_model.constant(tyctx, ctx.ctx, i as u64);
let pndindex = ndindexes.offset(tyctx, ctx, i.value, "");
in_ndindex.write_to_ndindex(tyctx, ctx, pndindex);
}
(num_ndindexes, ndindexes)
}
#[must_use]
pub fn deduce_ndims_after_indexing(indices: &[RustNDIndex], original_ndims: u64) -> u64 {
let mut final_ndims = original_ndims;
for index in indices {
match index {
RustNDIndex::SingleElement(_) => {
final_ndims -= 1;
}
RustNDIndex::Slice(_) => {}
}
}
final_ndims
}
}
pub fn call_nac3_ndarray_indexing_deduce_ndims_after_indexing<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndims: Int<'ctx, SizeT>,
num_ndindexes: Int<'ctx, SizeT>,
ndindexs: Ptr<'ctx, StructModel<NDIndex>>,
) -> Int<'ctx, SizeT> {
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
let pfinal_ndims = sizet_model.alloca(tyctx, ctx, "pfinal_ndims");
let errctx_ptr = setup_error_context(tyctx, ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(
tyctx,
"__nac3_ndarray_indexing_deduce_ndims_after_indexing",
),
)
.arg("errctx", errctx_ptr)
.arg("result", pfinal_ndims)
.arg("ndims", ndims)
.arg("num_ndindexs", num_ndindexes)
.arg("ndindexs", ndindexs)
.returning_void();
check_error_context(generator, ctx, errctx_ptr);
pfinal_ndims.load(tyctx, ctx, "final_ndims")
}
pub fn call_nac3_ndarray_index<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
num_indexes: Int<'ctx, SizeT>,
indexes: Ptr<'ctx, StructModel<NDIndex>>,
src_ndarray: Ptr<'ctx, StructModel<NpArray>>,
dst_ndarray: Ptr<'ctx, StructModel<NpArray>>,
) {
let tyctx = generator.type_context(ctx.ctx);
let perrctx = setup_error_context(tyctx, ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_index"),
)
.arg("errctx", perrctx)
.arg("num_indexes", num_indexes)
.arg("indexes", indexes)
.arg("src_ndarray", src_ndarray)
.arg("dst_ndarray", dst_ndarray)
.returning_void();
check_error_context(generator, ctx, perrctx);
}

4
nac3core/src/codegen/irrt/ndarray/mod.rs Normal file
View File

@ -0,0 +1,4 @@
pub mod basic;
pub mod broadcast;
pub mod indexing;
pub mod reshape;

31
nac3core/src/codegen/irrt/ndarray/reshape.rs Normal file
View File

@ -0,0 +1,31 @@
use crate::codegen::{
irrt::{
error_context::{check_error_context, setup_error_context},
util::{function::CallFunction, get_sizet_dependent_function_name},
},
model::*,
CodeGenContext, CodeGenerator,
};
pub fn call_nac3_ndarray_resolve_and_check_new_shape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
size: Int<'ctx, SizeT>,
new_ndims: Int<'ctx, SizeT>,
new_shape: Ptr<'ctx, IntModel<SizeT>>,
) {
let tyctx = generator.type_context(ctx.ctx);
let perrctx = setup_error_context(tyctx, ctx);
CallFunction::begin(
tyctx,
ctx,
&get_sizet_dependent_function_name(tyctx, "__nac3_ndarray_resolve_and_check_new_shape"),
)
.arg("errctx", perrctx)
.arg("size", size)
.arg("new_ndims", new_ndims)
.arg("new_shape", new_shape)
.returning_void();
check_error_context(generator, ctx, perrctx);
}

81
nac3core/src/codegen/irrt/slice.rs Normal file
View File

@ -0,0 +1,81 @@
use crate::codegen::{model::*, CodeGenContext};
// nac3core's slicing index/length values are always int32_t
pub type SliceIndex = Int32;
#[derive(Debug, Clone)]
pub struct UserSliceFields<F: FieldVisitor> {
pub start_defined: F::Field<IntModel<Bool>>,
pub start: F::Field<IntModel<SliceIndex>>,
pub stop_defined: F::Field<IntModel<Bool>>,
pub stop: F::Field<IntModel<SliceIndex>>,
pub step_defined: F::Field<IntModel<Bool>>,
pub step: F::Field<IntModel<SliceIndex>>,
}
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
pub struct UserSlice;
impl StructKind for UserSlice {
type Fields<F: FieldVisitor> = UserSliceFields<F>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F> {
Self::Fields {
start_defined: visitor.add("start_defined"),
start: visitor.add("start"),
stop_defined: visitor.add("stop_defined"),
stop: visitor.add("stop"),
step_defined: visitor.add("step_defined"),
step: visitor.add("step"),
}
}
}
#[derive(Debug, Clone)]
pub struct RustUserSlice<'ctx> {
pub start: Option<Int<'ctx, SliceIndex>>,
pub stop: Option<Int<'ctx, SliceIndex>>,
pub step: Option<Int<'ctx, SliceIndex>>,
}
impl<'ctx> RustUserSlice<'ctx> {
// Set the values of an LLVM UserSlice
// in the format of Python's `slice()`
pub fn write_to_user_slice(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
dst_slice_ptr: Ptr<'ctx, StructModel<UserSlice>>,
) {
let bool_model = IntModel(Bool);
let false_ = bool_model.constant(tyctx, ctx.ctx, 0);
let true_ = bool_model.constant(tyctx, ctx.ctx, 1);
// TODO: Code duplication. Probably okay...?
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_),
}
}
}

26
nac3core/src/codegen/irrt/test.rs Normal file
View File

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

103
nac3core/src/codegen/irrt/util.rs Normal file
View File

@ -0,0 +1,103 @@
use crate::codegen::model::*;
// 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(tyctx: TypeContext<'_>, name: &str) -> String {
let mut name = name.to_owned();
match tyctx.size_type.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
}
pub mod function {
use crate::codegen::{model::*, CodeGenContext};
use inkwell::{
types::{BasicMetadataTypeEnum, BasicType, FunctionType},
values::{AnyValue, BasicMetadataValueEnum, BasicValue, BasicValueEnum, CallSiteValue},
};
use itertools::Itertools;
#[derive(Debug, Clone, Copy)]
struct Arg<'ctx> {
ty: BasicMetadataTypeEnum<'ctx>,
val: BasicMetadataValueEnum<'ctx>,
}
/// Helper structure to reduce IRRT Inkwell function call boilerplate
/// TODO: Optimize
pub struct CallFunction<'ctx, 'a, 'b, 'c> {
tyctx: TypeContext<'ctx>,
ctx: &'b CodeGenContext<'ctx, 'a>,
/// Function name
name: &'c str,
/// Call arguments
args: Vec<Arg<'ctx>>,
}
impl<'ctx, 'a, 'b, 'c> CallFunction<'ctx, 'a, 'b, 'c> {
pub fn begin(
tyctx: TypeContext<'ctx>,
ctx: &'b CodeGenContext<'ctx, 'a>,
name: &'c str,
) -> Self {
CallFunction { tyctx, ctx, name, args: Vec::new() }
}
/// Push a call argument to the function call.
///
/// The `_name` parameter is there for self-documentation purposes.
#[allow(clippy::needless_pass_by_value)]
pub fn arg<M: Model>(mut self, _name: &str, arg: Instance<'ctx, M>) -> Self {
let arg = Arg {
ty: arg.model.get_type(self.tyctx, self.ctx.ctx).as_basic_type_enum().into(),
val: arg.value.as_basic_value_enum().into(),
};
self.args.push(arg);
self
}
/// Like [`CallFunction::returning_`] but `return_model` is automatically inferred.
pub fn returning<M: Model>(self, name: &str) -> Instance<'ctx, M> {
self.returning_(name, M::default())
}
/// Call the function and expect the function to return a value of type of `return_model`.
pub fn returning_<M: Model>(self, name: &str, return_model: M) -> Instance<'ctx, M> {
let ret_ty = return_model.get_type(self.tyctx, 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.tyctx, self.ctx.ctx, ret).unwrap(); // Must work
ret
}
/// 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, declare the function if it doesn't exist - it will be defined by other
// components of NAC3.
let func = self.ctx.module.get_function(self.name).unwrap_or_else(|| {
let tys = self.args.iter().map(|arg| arg.ty).collect_vec();
let fn_type = make_fn_type(&tys);
self.ctx.module.add_function(self.name, fn_type, None)
});
let vals = self.args.iter().map(|arg| arg.val).collect_vec();
self.ctx.builder.build_call(func, &vals, return_value_name).unwrap()
}
}
}

63
nac3core/src/codegen/llvm_intrinsics.rs
View File

@ -1,14 +1,12 @@
use inkwell::{
context::Context,
intrinsics::Intrinsic,
types::{AnyTypeEnum::IntType, FloatType},
values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue, PointerValue},
AddressSpace,
};
use crate::codegen::CodeGenContext;
use inkwell::context::Context;
use inkwell::intrinsics::Intrinsic;
use inkwell::types::AnyTypeEnum::IntType;
use inkwell::types::FloatType;
use inkwell::values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue, PointerValue};
use inkwell::AddressSpace;
use itertools::Either;
use super::CodeGenContext;
/// Returns the string representation for the floating-point type `ft` when used in intrinsic
/// functions.
fn get_float_intrinsic_repr(ctx: &Context, ft: FloatType) -> &'static str {
@ -37,40 +35,6 @@ 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>(
@ -185,7 +149,7 @@ pub fn call_memcpy_generic<'ctx>(
dest
} else {
ctx.builder
.build_bit_cast(dest, llvm_p0i8, "")
.build_bitcast(dest, llvm_p0i8, "")
.map(BasicValueEnum::into_pointer_value)
.unwrap()
};
@ -193,7 +157,7 @@ pub fn call_memcpy_generic<'ctx>(
src
} else {
ctx.builder
.build_bit_cast(src, llvm_p0i8, "")
.build_bitcast(src, llvm_p0i8, "")
.map(BasicValueEnum::into_pointer_value)
.unwrap()
};
@ -207,9 +171,8 @@ 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
/// * `,($val:ident)*`: Comma separated list of operands
macro_rules! generate_llvm_intrinsic_fn_body {
@ -225,8 +188,8 @@ macro_rules! generate_llvm_intrinsic_fn_body {
/// Arguments:
/// * `float/int`: Indicates the return and argument type of the function
/// * `$fn_name:ident`: The identifier of the rust function to be generated
/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function.
/// Omit "llvm." prefix from the function name i.e. use "ceil" instead of "llvm.ceil"
/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function
/// Omit "llvm." prefix from the function name i.e. use "ceil" instead of "llvm.ceil"
/// * `$val:ident`: The operand for unary operations
/// * `$val1:ident`, `$val2:ident`: The operands for binary operations
macro_rules! generate_llvm_intrinsic_fn {

174
nac3core/src/codegen/mod.rs
View File

@ -1,12 +1,12 @@
use std::{
collections::{HashMap, HashSet},
sync::{
atomic::{AtomicBool, Ordering},
Arc,
use crate::{
codegen::classes::{ListType, ProxyType, RangeType},
symbol_resolver::{StaticValue, SymbolResolver},
toplevel::{helper::PrimDef, TopLevelContext, TopLevelDef},
typecheck::{
type_inferencer::{CodeLocation, PrimitiveStore},
typedef::{CallId, FuncArg, Type, TypeEnum, Unifier},
},
thread,
};
use crossbeam::channel::{unbounded, Receiver, Sender};
use inkwell::{
attributes::{Attribute, AttributeLoc},
@ -24,21 +24,16 @@ use inkwell::{
AddressSpace, IntPredicate, OptimizationLevel,
};
use itertools::Itertools;
use parking_lot::{Condvar, Mutex};
use model::*;
use nac3parser::ast::{Location, Stmt, StrRef};
use crate::{
symbol_resolver::{StaticValue, SymbolResolver},
toplevel::{helper::PrimDef, numpy::unpack_ndarray_var_tys, TopLevelContext, TopLevelDef},
typecheck::{
type_inferencer::{CodeLocation, PrimitiveStore},
typedef::{CallId, FuncArg, Type, TypeEnum, Unifier},
},
use parking_lot::{Condvar, Mutex};
use std::collections::{HashMap, HashSet};
use std::sync::{
atomic::{AtomicBool, Ordering},
Arc,
};
use classes::{ListType, NDArrayType, ProxyType, RangeType};
use concrete_type::{ConcreteType, ConcreteTypeEnum, ConcreteTypeStore};
pub use generator::{CodeGenerator, DefaultCodeGenerator};
use std::thread;
use structure::{cslice::CSlice, exception::Exception, ndarray::NpArray};
pub mod builtin_fns;
pub mod classes;
@ -48,27 +43,18 @@ pub mod extern_fns;
mod generator;
pub mod irrt;
pub mod llvm_intrinsics;
pub mod model;
pub mod numpy;
pub mod numpy_new;
pub mod stmt;
pub mod structure;
pub mod util;
#[cfg(test)]
mod test;
mod macros {
/// Codegen-variant of [`std::unreachable`] which accepts an instance of [`CodeGenContext`] as
/// its first argument to provide Python source information to indicate the codegen location
/// causing the assertion.
macro_rules! codegen_unreachable {
($ctx:expr $(,)?) => {
std::unreachable!("unreachable code while processing {}", &$ctx.current_loc)
};
($ctx:expr, $($arg:tt)*) => {
std::unreachable!("unreachable code while processing {}: {}", &$ctx.current_loc, std::format!("{}", std::format_args!($($arg)+)))
};
}
pub(crate) use codegen_unreachable;
}
use concrete_type::{ConcreteType, ConcreteTypeEnum, ConcreteTypeStore};
pub use generator::{CodeGenerator, DefaultCodeGenerator};
#[derive(Default)]
pub struct StaticValueStore {
@ -188,11 +174,11 @@ pub struct CodeGenContext<'ctx, 'a> {
pub registry: &'a WorkerRegistry,
/// Cache for constant strings.
pub const_strings: HashMap<String, BasicValueEnum<'ctx>>,
pub const_strings: HashMap<String, Struct<'ctx, CSlice>>,
/// [`BasicBlock`] containing all `alloca` statements for the current function.
pub init_bb: BasicBlock<'ctx>,
pub exception_val: Option<PointerValue<'ctx>>,
pub exception_val: Option<Ptr<'ctx, StructModel<Exception>>>,
/// The header and exit basic blocks of a loop in this context. See
/// <https://llvm.org/docs/LoopTerminology.html> for explanation of these terminology.
@ -464,7 +450,7 @@ pub struct CodeGenTask {
fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
ctx: &'ctx Context,
module: &Module<'ctx>,
generator: &G,
generator: &mut G,
unifier: &mut Unifier,
top_level: &TopLevelContext,
type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
@ -509,12 +495,9 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
}
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()
let tyctx = generator.type_context(ctx);
let pndarray_model = PtrModel(StructModel(NpArray));
pndarray_model.get_type(tyctx, ctx).into()
}
_ => unreachable!(
@ -558,10 +541,8 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
};
return ty;
}
TTuple { ty, is_vararg_ctx } => {
TTuple { ty } => {
// 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| {
@ -591,7 +572,7 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
fn get_llvm_abi_type<'ctx, G: CodeGenerator + ?Sized>(
ctx: &'ctx Context,
module: &Module<'ctx>,
generator: &G,
generator: &mut G,
unifier: &mut Unifier,
top_level: &TopLevelContext,
type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
@ -600,11 +581,11 @@ 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.
if unifier.unioned(ty, primitives.bool) {
return 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`.
@ -629,40 +610,6 @@ 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,
@ -720,43 +667,20 @@ pub fn gen_func_impl<
..primitives
};
let mut type_cache: HashMap<_, _> = [
let type_context = generator.type_context(context);
let cslice_model = StructModel(CSlice);
let pexn_model = PtrModel(StructModel(Exception));
let mut type_cache: HashMap<_, BasicTypeEnum<'ctx>> = [
(primitives.int32, context.i32_type().into()),
(primitives.int64, context.i64_type().into()),
(primitives.uint32, context.i32_type().into()),
(primitives.uint64, context.i64_type().into()),
(primitives.float, context.f64_type().into()),
(primitives.bool, context.i8_type().into()),
(primitives.str, {
let name = "str";
match module.get_struct_type(name) {
None => {
let str_type = context.opaque_struct_type("str");
let fields = [
context.i8_type().ptr_type(AddressSpace::default()).into(),
generator.get_size_type(context).into(),
];
str_type.set_body(&fields, false);
str_type.into()
}
Some(t) => t.as_basic_type_enum(),
}
}),
(primitives.str, cslice_model.get_type(type_context, context).into()),
(primitives.range, RangeType::new(context).as_base_type().into()),
(primitives.exception, {
let name = "Exception";
if let Some(t) = module.get_struct_type(name) {
t.ptr_type(AddressSpace::default()).as_basic_type_enum()
} else {
let exception = context.opaque_struct_type("Exception");
let int32 = context.i32_type().into();
let int64 = context.i64_type().into();
let str_ty = module.get_struct_type("str").unwrap().as_basic_type_enum();
let fields = [int32, str_ty, int32, int32, str_ty, str_ty, int64, int64, int64];
exception.set_body(&fields, false);
exception.ptr_type(AddressSpace::default()).as_basic_type_enum()
}
}),
(primitives.exception, pexn_model.get_type(type_context, context).into()),
]
.iter()
.copied()
@ -774,7 +698,6 @@ 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),
@ -797,10 +720,7 @@ 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,
@ -819,12 +739,9 @@ 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, vararg_arg.is_some()),
_ => context.void_type().fn_type(&params, vararg_arg.is_some()),
Some(ret_type) if !has_sret => ret_type.fn_type(&params, false),
_ => context.void_type().fn_type(&params, false),
};
let symbol = &task.symbol_name;
@ -852,10 +769,9 @@ pub fn gen_func_impl<
builder.position_at_end(init_bb);
let body_bb = context.append_basic_block(fn_val, "body");
// Store non-vararg argument values into local variables
let mut var_assignment = HashMap::new();
let offset = u32::from(has_sret);
for (n, arg) in args.iter().enumerate().filter(|(_, arg)| !arg.is_vararg) {
for (n, arg) in args.iter().enumerate() {
let param = fn_val.get_nth_param((n as u32) + offset).unwrap();
let local_type = get_llvm_type(
context,
@ -888,8 +804,6 @@ 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 {
@ -1112,9 +1026,3 @@ 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()
}

161
nac3core/src/codegen/model/core.rs Normal file
View File

@ -0,0 +1,161 @@
use std::fmt;
use inkwell::{context::Context, types::*, values::*};
use super::*;
use crate::codegen::{CodeGenContext, CodeGenerator};
#[derive(Clone, Copy)]
pub struct TypeContext<'ctx> {
pub size_type: IntType<'ctx>,
}
pub trait HasTypeContext {
fn type_context<'ctx>(&self, ctx: &'ctx Context) -> TypeContext<'ctx>;
}
impl<T: CodeGenerator + ?Sized> HasTypeContext for T {
fn type_context<'ctx>(&self, ctx: &'ctx Context) -> TypeContext<'ctx> {
TypeContext { size_type: self.get_size_type(ctx) }
}
}
#[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
}
}
/// A [`Model`] is a singleton object that uniquely identifies a [`BasicType`]
/// solely from a [`CodeGenerator`] and a [`Context`].
pub trait Model: CheckType + fmt::Debug + Clone + Copy + Default {
type Value<'ctx>: BasicValue<'ctx> + TryFrom<BasicValueEnum<'ctx>>;
type Type<'ctx>: BasicType<'ctx>;
/// Return the [`BasicType`] of this model.
fn get_type<'ctx>(&self, tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> Self::Type<'ctx>;
/// Check if a [`BasicType`] is the same type of this model.
fn check_type<'ctx, T: BasicType<'ctx>>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
ty: T,
) -> Result<(), ModelError> {
let ty = ty.as_basic_type_enum();
self.check_type_impl(tyctx, ctx, ty.as_basic_type_enum())
}
/// 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.
fn believe_value<'ctx>(&self, value: Self::Value<'ctx>) -> 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<'ctx, V: BasicValue<'ctx>>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
value: V,
) -> Result<Instance<'ctx, Self>, ModelError> {
let value = value.as_basic_value_enum();
self.check_type(tyctx, ctx, value.get_type())
.map_err(|err| err.under_context("the value {value:?}"))?;
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<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
name: &str,
) -> Ptr<'ctx, Self> {
let pmodel = PtrModel(*self);
let p = ctx.builder.build_alloca(self.get_type(tyctx, ctx.ctx), name).unwrap();
pmodel.believe_value(p)
}
// Allocate an array on the stack and return its pointer.
fn array_alloca<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
len: IntValue<'ctx>,
name: &str,
) -> Ptr<'ctx, Self> {
let pmodel = PtrModel(*self);
let p = ctx.builder.build_array_alloca(self.get_type(tyctx, ctx.ctx), len, name).unwrap();
pmodel.believe_value(p)
}
fn var_alloca<'ctx, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
name: Option<&str>,
) -> Result<Ptr<'ctx, Self>, String> {
let tyctx = generator.type_context(ctx.ctx);
let pmodel = PtrModel(*self);
let p = generator.gen_var_alloc(
ctx,
self.get_type(tyctx, ctx.ctx).as_basic_type_enum(),
name,
)?;
Ok(pmodel.believe_value(p))
}
fn array_var_alloca<'ctx, G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
len: IntValue<'ctx>,
name: Option<&'ctx str>,
) -> Result<Ptr<'ctx, Self>, String> {
let tyctx = generator.type_context(ctx.ctx);
// TODO: Remove ArraySliceValue
let pmodel = PtrModel(*self);
let p = generator.gen_array_var_alloc(
ctx,
self.get_type(tyctx, ctx.ctx).as_basic_type_enum(),
len,
name,
)?;
Ok(pmodel.believe_value(PointerValue::from(p)))
}
}
#[derive(Debug, Clone, Copy)]
pub struct Instance<'ctx, M: Model> {
/// 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<'ctx>,
}
// NOTE: Must be Rust object-safe - This must be typeable for a Rust trait object.
pub trait CheckType {
fn check_type_impl<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
ty: BasicTypeEnum<'ctx>,
) -> Result<(), ModelError>;
}

228
nac3core/src/codegen/model/int.rs Normal file
View File

@ -0,0 +1,228 @@
use std::fmt;
use inkwell::{
context::Context,
types::{BasicTypeEnum, IntType},
values::IntValue,
IntPredicate,
};
use crate::codegen::{CodeGenContext, CodeGenerator};
use super::*;
pub trait IntKind: fmt::Debug + Clone + Copy + Default {
fn get_int_type<'ctx>(&self, tyctx: TypeContext<'ctx>, 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 IntKind for Bool {
fn get_int_type<'ctx>(&self, _tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> IntType<'ctx> {
ctx.bool_type()
}
}
impl IntKind for Byte {
fn get_int_type<'ctx>(&self, _tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> IntType<'ctx> {
ctx.i8_type()
}
}
impl IntKind for Int32 {
fn get_int_type<'ctx>(&self, _tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> IntType<'ctx> {
ctx.i32_type()
}
}
impl IntKind for Int64 {
fn get_int_type<'ctx>(&self, _tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> IntType<'ctx> {
ctx.i64_type()
}
}
impl IntKind for SizeT {
fn get_int_type<'ctx>(&self, tyctx: TypeContext<'ctx>, _ctx: &'ctx Context) -> IntType<'ctx> {
tyctx.size_type
}
}
#[derive(Debug, Clone, Copy, Default)]
pub struct IntModel<N: IntKind>(pub N);
pub type Int<'ctx, N> = Instance<'ctx, IntModel<N>>;
impl<N: IntKind> CheckType for IntModel<N> {
fn check_type_impl<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
ty: BasicTypeEnum<'ctx>,
) -> Result<(), ModelError> {
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(tyctx, 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<N: IntKind> Model for IntModel<N> {
type Value<'ctx> = IntValue<'ctx>;
type Type<'ctx> = IntType<'ctx>;
#[must_use]
fn get_type<'ctx>(&self, tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> Self::Type<'ctx> {
self.0.get_int_type(tyctx, ctx)
}
}
impl<N: IntKind> IntModel<N> {
pub fn constant<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
value: u64,
) -> Int<'ctx, N> {
let value = self.get_type(tyctx, ctx).const_int(value, false);
self.believe_value(value)
}
pub fn const_0<'ctx>(&self, tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> Int<'ctx, N> {
self.constant(tyctx, ctx, 0)
}
pub fn const_1<'ctx>(&self, tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> Int<'ctx, N> {
self.constant(tyctx, ctx, 1)
}
pub fn s_extend_or_bit_cast<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
name: &str,
) -> Int<'ctx, N> {
let value = ctx
.builder
.build_int_s_extend_or_bit_cast(value, self.get_type(tyctx, ctx.ctx), name)
.unwrap();
self.believe_value(value)
}
pub fn truncate<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
name: &str,
) -> Int<'ctx, N> {
let value =
ctx.builder.build_int_truncate(value, self.get_type(tyctx, ctx.ctx), name).unwrap();
self.believe_value(value)
}
}
impl IntModel<Bool> {
#[must_use]
pub fn const_false<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
) -> Int<'ctx, Bool> {
self.constant(tyctx, ctx, 0)
}
#[must_use]
pub fn const_true<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
) -> Int<'ctx, Bool> {
self.constant(tyctx, ctx, 1)
}
}
impl<'ctx, N: IntKind> Int<'ctx, N> {
pub fn s_extend_or_bit_cast<NewN: IntKind, G: CodeGenerator + ?Sized>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
to_int_kind: NewN,
name: &str,
) -> Int<'ctx, NewN> {
IntModel(to_int_kind).s_extend_or_bit_cast(tyctx, ctx, self.value, name)
}
pub fn truncate<NewN: IntKind, G: CodeGenerator + ?Sized>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
to_int_kind: NewN,
name: &str,
) -> Int<'ctx, NewN> {
IntModel(to_int_kind).truncate(tyctx, ctx, self.value, name)
}
#[must_use]
pub fn add<G: CodeGenerator + ?Sized>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
other: Int<'ctx, N>,
name: &str,
) -> Int<'ctx, N> {
let value = ctx.builder.build_int_add(self.value, other.value, name).unwrap();
self.model.believe_value(value)
}
#[must_use]
pub fn sub<G: CodeGenerator + ?Sized>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
other: Int<'ctx, N>,
name: &str,
) -> Int<'ctx, N> {
let value = ctx.builder.build_int_sub(self.value, other.value, name).unwrap();
self.model.believe_value(value)
}
#[must_use]
pub fn mul<G: CodeGenerator + ?Sized>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
other: Int<'ctx, N>,
name: &str,
) -> Int<'ctx, N> {
let value = ctx.builder.build_int_mul(self.value, other.value, name).unwrap();
self.model.believe_value(value)
}
pub fn compare<G: CodeGenerator + ?Sized>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
op: IntPredicate,
other: Int<'ctx, N>,
name: &str,
) -> Int<'ctx, Bool> {
let bool_model = IntModel(Bool);
let value = ctx.builder.build_int_compare(op, self.value, other.value, name).unwrap();
bool_model.believe_value(value)
}
}

12
nac3core/src/codegen/model/mod.rs Normal file
View File

@ -0,0 +1,12 @@
mod core;
mod int;
mod ptr;
mod slice;
mod structure;
pub mod util;
pub use core::*;
pub use int::*;
pub use ptr::*;
pub use slice::*;
pub use structure::*;

142
nac3core/src/codegen/model/ptr.rs Normal file
View File

@ -0,0 +1,142 @@
use inkwell::{
context::Context,
types::{BasicType, BasicTypeEnum, PointerType},
values::{IntValue, PointerValue},
AddressSpace,
};
use crate::codegen::CodeGenContext;
use super::*;
#[derive(Debug, Clone, Copy, Default)]
pub struct PtrModel<Element>(pub Element);
pub type Ptr<'ctx, Element> = Instance<'ctx, PtrModel<Element>>;
impl<Element: CheckType> CheckType for PtrModel<Element> {
fn check_type_impl<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
ty: BasicTypeEnum<'ctx>,
) -> Result<(), super::ModelError> {
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_impl(tyctx, ctx, elem_ty)
.map_err(|err| err.under_context("a PointerType"))?;
Ok(())
}
}
impl<Element: Model> Model for PtrModel<Element> {
type Value<'ctx> = PointerValue<'ctx>;
type Type<'ctx> = PointerType<'ctx>;
fn get_type<'ctx>(&self, tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> Self::Type<'ctx> {
self.0.get_type(tyctx, ctx).ptr_type(AddressSpace::default())
}
}
impl<Element: Model> PtrModel<Element> {
/// Return a ***constant*** nullptr.
pub fn nullptr<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
) -> Ptr<'ctx, Element> {
let ptr = self.get_type(tyctx, ctx).const_null();
self.believe_value(ptr)
}
pub fn transmute<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
ptr: PointerValue<'ctx>,
name: &str,
) -> Ptr<'ctx, Element> {
let ptr = ctx.builder.build_pointer_cast(ptr, self.get_type(tyctx, ctx.ctx), name).unwrap();
self.believe_value(ptr)
}
}
impl<'ctx, Element: Model> Ptr<'ctx, Element> {
/// Offset the pointer by [`inkwell::builder::Builder::build_in_bounds_gep`].
#[must_use]
pub fn offset(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
offset: IntValue<'ctx>,
name: &str,
) -> Ptr<'ctx, Element> {
let new_ptr =
unsafe { ctx.builder.build_in_bounds_gep(self.value, &[offset], name).unwrap() };
self.model.check_value(tyctx, ctx.ctx, new_ptr).unwrap()
}
// Load the `i`-th element (0-based) on the array with [`inkwell::builder::Builder::build_in_bounds_gep`].
pub fn ix(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
i: IntValue<'ctx>,
name: &str,
) -> Instance<'ctx, Element> {
self.offset(tyctx, ctx, i, name).load(tyctx, ctx, name)
}
/// Load the value with [`inkwell::builder::Builder::build_load`].
pub fn load(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
name: &str,
) -> Instance<'ctx, Element> {
let value = ctx.builder.build_load(self.value, name).unwrap();
self.model.0.check_value(tyctx, 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, Element>) {
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 transmute<NewElement: Model>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
new_model: NewElement,
name: &str,
) -> Ptr<'ctx, NewElement> {
PtrModel(new_model).transmute(tyctx, ctx, self.value, name)
}
/// Check if the pointer is null with [`inkwell::builder::Builder::build_is_null`].
pub fn is_null(&self, ctx: &CodeGenContext<'ctx, '_>, name: &str) -> Int<'ctx, Bool> {
let bool_model = IntModel(Bool);
let value = ctx.builder.build_is_null(self.value, name).unwrap();
bool_model.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, '_>, name: &str) -> Int<'ctx, Bool> {
let bool_model = IntModel(Bool);
let value = ctx.builder.build_is_not_null(self.value, name).unwrap();
bool_model.believe_value(value)
}
}

72
nac3core/src/codegen/model/slice.rs Normal file
View File

@ -0,0 +1,72 @@
use crate::codegen::{CodeGenContext, CodeGenerator};
use super::*;
/// A slice - literally just a pointer and a length value.
///
/// NOTE: This is NOT a [`Model`].
pub struct ArraySlice<'ctx, Len: IntKind, Item: Model> {
pub base: Ptr<'ctx, Item>,
pub len: Int<'ctx, Len>,
}
impl<'ctx, Len: IntKind, Item: Model> ArraySlice<'ctx, Len, Item> {
/// Get the `idx`-nth element of this [`ArraySlice`], but doesn't do an assertion to see if `idx` is out of bounds or not.
///
/// Also see [`ArraySlice::ix`].
pub fn ix_unchecked(
&self,
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
idx: Int<'ctx, Len>,
name: &str,
) -> Ptr<'ctx, Item> {
let element_ptr = unsafe {
ctx.builder.build_in_bounds_gep(self.base.value, &[idx.value], name).unwrap()
};
self.base.model.check_value(tyctx, ctx.ctx, element_ptr).unwrap()
}
/// Call [`ArraySlice::ix_unchecked`], but checks if `idx` is in bounds, otherwise a runtime `IndexError` will be thrown.
pub fn ix<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
idx: Int<'ctx, Len>,
name: &str,
) -> Ptr<'ctx, Item> {
let tyctx = generator.type_context(ctx.ctx);
let len_model = IntModel(Len::default());
// Assert `0 <= idx < length` and throw an Exception if `idx` is out of bounds
let lower_bounded = ctx
.builder
.build_int_compare(
inkwell::IntPredicate::SLE,
len_model.constant(tyctx, ctx.ctx, 0).value,
idx.value,
"lower_bounded",
)
.unwrap();
let upper_bounded = ctx
.builder
.build_int_compare(
inkwell::IntPredicate::SLT,
idx.value,
self.len.value,
"upper_bounded",
)
.unwrap();
let bounded = ctx.builder.build_and(lower_bounded, upper_bounded, "bounded").unwrap();
ctx.make_assert(
generator,
bounded,
"0:IndexError",
"nac3core LLVM codegen attempting to access out of bounds array index {0}. Must satisfy 0 <= index < {2}",
[ Some(idx.value), Some(self.len.value), None],
ctx.current_loc
);
self.ix_unchecked(tyctx, ctx, idx, name)
}
}

174
nac3core/src/codegen/model/structure.rs Normal file
View File

@ -0,0 +1,174 @@
use std::fmt;
use inkwell::{
context::Context,
types::{BasicType, BasicTypeEnum, StructType},
values::StructValue,
};
use itertools::izip;
use crate::codegen::CodeGenContext;
use super::*;
#[derive(Debug, Clone, Copy)]
pub struct GepField<M: Model> {
pub gep_index: u64,
pub name: &'static str,
pub model: M,
}
pub trait FieldVisitor {
type Field<M: Model + 'static>;
fn add<M: Model + 'static>(&mut self, name: &'static str) -> Self::Field<M>;
}
pub struct GepFieldVisitor {
gep_index_counter: u64,
}
impl FieldVisitor for GepFieldVisitor {
type Field<M: Model + 'static> = GepField<M>;
fn add<M: Model + 'static>(&mut self, name: &'static str) -> Self::Field<M> {
let gep_index = self.gep_index_counter;
self.gep_index_counter += 1;
Self::Field { gep_index, name, model: M::default() }
}
}
struct TypeFieldVisitor<'ctx> {
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
field_types: Vec<BasicTypeEnum<'ctx>>,
}
impl<'ctx> FieldVisitor for TypeFieldVisitor<'ctx> {
type Field<M: Model + 'static> = ();
fn add<M: Model + 'static>(&mut self, _name: &'static str) -> Self::Field<M> {
self.field_types.push(M::default().get_type(self.tyctx, self.ctx).as_basic_type_enum());
}
}
struct CheckTypeEntry {
check_type: Box<dyn CheckType + 'static>,
name: &'static str,
}
struct CheckTypeFieldVisitor<'ctx> {
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
check_types: Vec<CheckTypeEntry>,
}
impl<'ctx> FieldVisitor for CheckTypeFieldVisitor<'ctx> {
type Field<M: Model + 'static> = ();
fn add<M: Model + 'static>(&mut self, name: &'static str) -> Self::Field<M> {
self.check_types.push(CheckTypeEntry { check_type: Box::<M>::default(), name });
}
}
pub trait StructKind: fmt::Debug + Clone + Copy + Default {
type Fields<F: FieldVisitor>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F>;
fn fields(&self) -> Self::Fields<GepFieldVisitor> {
self.visit_fields(&mut GepFieldVisitor { gep_index_counter: 0 })
}
fn get_struct_type<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
) -> StructType<'ctx> {
let mut visitor = TypeFieldVisitor { tyctx, ctx, field_types: Vec::new() };
self.visit_fields(&mut visitor);
ctx.struct_type(&visitor.field_types, false)
}
}
#[derive(Debug, Clone, Copy, Default)]
pub struct StructModel<S: StructKind>(pub S);
pub type Struct<'ctx, S> = Instance<'ctx, StructModel<S>>;
impl<S: StructKind> CheckType for StructModel<S> {
fn check_type_impl<'ctx>(
&self,
tyctx: TypeContext<'ctx>,
ctx: &'ctx Context,
ty: BasicTypeEnum<'ctx>,
) -> 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 field_types = ty.get_field_types();
let check_types = {
let mut builder = CheckTypeFieldVisitor { tyctx, ctx, check_types: Vec::new() };
self.0.visit_fields(&mut builder);
builder.check_types
};
if check_types.len() != field_types.len() {
return Err(ModelError(format!(
"Expecting StructType to have {} field(s), but got {} field(s)",
check_types.len(),
field_types.len()
)));
}
for (field_i, (entry, field_type)) in izip!(check_types, field_types).enumerate() {
let field_at = field_i + 1;
entry.check_type.check_type_impl(tyctx, ctx, field_type).map_err(|err| {
err.under_context(format!("struct field #{field_at} '{}'", entry.name).as_str())
})?;
}
Ok(())
}
}
impl<S: StructKind> Model for StructModel<S> {
type Value<'ctx> = StructValue<'ctx>;
type Type<'ctx> = StructType<'ctx>;
fn get_type<'ctx>(&self, tyctx: TypeContext<'ctx>, ctx: &'ctx Context) -> Self::Type<'ctx> {
self.0.get_struct_type(tyctx, ctx)
}
}
impl<'ctx, S: StructKind> Ptr<'ctx, StructModel<S>> {
pub fn gep<M, GetField>(
&self,
ctx: &CodeGenContext<'ctx, '_>,
get_field: GetField,
) -> Ptr<'ctx, M>
where
M: Model,
GetField: FnOnce(S::Fields<GepFieldVisitor>) -> GepField<M>,
{
let field = get_field(self.model.0 .0.fields());
let llvm_i32 = ctx.ctx.i32_type(); // must be i32, if its i64 then rust segfaults
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()
};
let ptr_model = PtrModel(field.model);
ptr_model.believe_value(ptr)
}
}

23
nac3core/src/codegen/model/util.rs Normal file
View File

@ -0,0 +1,23 @@
use inkwell::{types::BasicType, values::IntValue};
use crate::codegen::{llvm_intrinsics::call_memcpy_generic, CodeGenContext};
use super::*;
pub fn gen_model_memcpy<'ctx, M: Model>(
tyctx: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
dst: Ptr<'ctx, M>,
src: Ptr<'ctx, M>,
num_elements: IntValue<'ctx>,
volatile: bool,
) {
let bool_model = IntModel(Bool);
let itemsize = M::default().get_type(tyctx, ctx.ctx).size_of().unwrap();
let totalsize =
ctx.builder.build_int_mul(itemsize, num_elements, "model_memcpy_totalsize").unwrap();
let is_volatile = bool_model.constant(tyctx, ctx.ctx, u64::from(volatile));
call_memcpy_generic(ctx, dst.value, src.value, totalsize, is_volatile.value);
}

579
nac3core/src/codegen/numpy.rs
View File

@ -1,28 +1,20 @@
use inkwell::{
types::{AnyTypeEnum, BasicType, BasicTypeEnum, PointerType},
values::{BasicValue, BasicValueEnum, IntValue, PointerValue},
AddressSpace, IntPredicate, OptimizationLevel,
};
use nac3parser::ast::{Operator, StrRef};
use super::{
classes::{
ArrayLikeIndexer, ArrayLikeValue, ListType, ListValue, NDArrayType, NDArrayValue,
ProxyType, ProxyValue, TypedArrayLikeAccessor, TypedArrayLikeAdapter,
TypedArrayLikeMutator, UntypedArrayLikeAccessor, UntypedArrayLikeMutator,
},
expr::gen_binop_expr_with_values,
irrt::{
calculate_len_for_slice_range, call_ndarray_calc_broadcast,
call_ndarray_calc_broadcast_index, call_ndarray_calc_nd_indices, call_ndarray_calc_size,
},
llvm_intrinsics::{self, call_memcpy_generic},
macros::codegen_unreachable,
stmt::{gen_for_callback_incrementing, gen_for_range_callback, gen_if_else_expr_callback},
CodeGenContext, CodeGenerator,
};
use crate::{
codegen::{
classes::{
ArrayLikeIndexer, ArrayLikeValue, ListType, ListValue, NDArrayType, NDArrayValue,
ProxyType, ProxyValue, TypedArrayLikeAccessor, TypedArrayLikeAdapter,
TypedArrayLikeMutator, UntypedArrayLikeAccessor, UntypedArrayLikeMutator,
},
expr::gen_binop_expr_with_values,
irrt::{
calculate_len_for_slice_range, call_ndarray_calc_broadcast,
call_ndarray_calc_broadcast_index, call_ndarray_calc_nd_indices,
call_ndarray_calc_size,
},
llvm_intrinsics::{self, call_memcpy_generic},
stmt::{gen_for_callback_incrementing, gen_for_range_callback, gen_if_else_expr_callback},
CodeGenContext, CodeGenerator,
},
symbol_resolver::ValueEnum,
toplevel::{
helper::PrimDef,
@ -34,6 +26,13 @@ use crate::{
typedef::{FunSignature, Type, TypeEnum},
},
};
use inkwell::types::{AnyTypeEnum, BasicTypeEnum, PointerType};
use inkwell::{
types::BasicType,
values::{BasicValueEnum, IntValue, PointerValue},
AddressSpace, IntPredicate, OptimizationLevel,
};
use nac3parser::ast::{Operator, StrRef};
/// Creates an uninitialized `NDArray` instance.
fn create_ndarray_uninitialized<'ctx, G: CodeGenerator + ?Sized>(
@ -160,7 +159,7 @@ where
///
/// * `elem_ty` - The element type of the `NDArray`.
/// * `shape` - The shape of the `NDArray`, represented am array of [`IntValue`]s.
pub fn create_ndarray_const_shape<'ctx, G: CodeGenerator + ?Sized>(
fn create_ndarray_const_shape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
elem_ty: Type,
@ -255,9 +254,9 @@ fn ndarray_zero_value<'ctx, G: CodeGenerator + ?Sized>(
} else if ctx.unifier.unioned(elem_ty, ctx.primitives.bool) {
ctx.ctx.bool_type().const_zero().into()
} else if ctx.unifier.unioned(elem_ty, ctx.primitives.str) {
ctx.gen_string(generator, "").into()
ctx.gen_string(generator, "").value.into()
} else {
codegen_unreachable!(ctx)
unreachable!()
}
}
@ -283,9 +282,9 @@ fn ndarray_one_value<'ctx, G: CodeGenerator + ?Sized>(
} else if ctx.unifier.unioned(elem_ty, ctx.primitives.bool) {
ctx.ctx.bool_type().const_int(1, false).into()
} else if ctx.unifier.unioned(elem_ty, ctx.primitives.str) {
ctx.gen_string(generator, "1").into()
ctx.gen_string(generator, "1").value.into()
} else {
codegen_unreachable!(ctx)
unreachable!()
}
}
@ -353,7 +352,7 @@ fn call_ndarray_empty_impl<'ctx, G: CodeGenerator + ?Sized>(
create_ndarray_const_shape(generator, ctx, elem_ty, &[shape_int])
}
_ => codegen_unreachable!(ctx),
_ => unreachable!(),
}
}
@ -624,7 +623,7 @@ fn call_ndarray_full_impl<'ctx, G: CodeGenerator + ?Sized>(
} else if fill_value.is_int_value() || fill_value.is_float_value() {
fill_value
} else {
codegen_unreachable!(ctx)
unreachable!()
};
Ok(value)
@ -939,7 +938,7 @@ fn call_ndarray_array_impl<'ctx, G: CodeGenerator + ?Sized>(
.build_store(
lst,
ctx.builder
.build_bit_cast(object.as_base_value(), llvm_plist_i8, "")
.build_bitcast(object.as_base_value(), llvm_plist_i8, "")
.unwrap(),
)
.unwrap();
@ -961,7 +960,7 @@ fn call_ndarray_array_impl<'ctx, G: CodeGenerator + ?Sized>(
.builder
.build_load(lst, "")
.map(BasicValueEnum::into_pointer_value)
.map(|v| ctx.builder.build_bit_cast(v, plist_plist_i8, "").unwrap())
.map(|v| ctx.builder.build_bitcast(v, plist_plist_i8, "").unwrap())
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let this_dim = ListValue::from_ptr_val(this_dim, llvm_usize, None);
@ -980,9 +979,7 @@ fn call_ndarray_array_impl<'ctx, G: CodeGenerator + ?Sized>(
ctx.builder
.build_store(
lst,
ctx.builder
.build_bit_cast(next_dim, llvm_plist_i8, "")
.unwrap(),
ctx.builder.build_bitcast(next_dim, llvm_plist_i8, "").unwrap(),
)
.unwrap();
@ -1071,15 +1068,15 @@ fn call_ndarray_eye_impl<'ctx, G: CodeGenerator + ?Sized>(
/// Copies a slice of an [`NDArrayValue`] to another.
///
/// - `dst_arr`: The [`NDArrayValue`] instance of the destination array. The `ndims` and `dim_sz`
/// fields should be populated before calling this function.
/// fields should be populated before calling this function.
/// - `dst_slice_ptr`: The [`PointerValue`] to the first element of the currently processing
/// dimensional slice in the destination array.
/// dimensional slice in the destination array.
/// - `src_arr`: The [`NDArrayValue`] instance of the source array.
/// - `src_slice_ptr`: The [`PointerValue`] to the first element of the currently processing
/// dimensional slice in the source array.
/// dimensional slice in the source array.
/// - `dim`: The index of the currently processing dimension.
/// - `slices`: List of all slices, with the first element corresponding to the slice applicable to
/// this dimension. The `start`/`stop` values of each slice must be non-negative indices.
/// this dimension. The `start`/`stop` values of each slice must be non-negative indices.
fn ndarray_sliced_copyto_impl<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
@ -1184,7 +1181,7 @@ fn ndarray_sliced_copyto_impl<'ctx, G: CodeGenerator + ?Sized>(
///
/// * `elem_ty` - The element type of the `NDArray`.
/// - `slices`: List of all slices, with the first element corresponding to the slice applicable to
/// this dimension. The `start`/`stop` values of each slice must be positive indices.
/// this dimension. The `start`/`stop` values of each slice must be positive indices.
pub fn ndarray_sliced_copy<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
@ -1349,7 +1346,7 @@ where
///
/// * `elem_ty` - The element type of the `NDArray`.
/// * `res` - The `ndarray` instance to write results into, or [`None`] if the result should be
/// written to a new `ndarray`.
/// written to a new `ndarray`.
/// * `value_fn` - Function mapping the two input elements into the result.
///
/// # Panic
@ -1436,7 +1433,7 @@ where
///
/// * `elem_ty` - The element type of the `NDArray`.
/// * `res` - The `ndarray` instance to write results into, or [`None`] if the result should be
/// written to a new `ndarray`.
/// written to a new `ndarray`.
pub fn ndarray_matmul_2d<'ctx, G: CodeGenerator>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
@ -2020,7 +2017,7 @@ pub fn gen_ndarray_fill<'ctx>(
} else if value_arg.is_int_value() || value_arg.is_float_value() {
value_arg
} else {
codegen_unreachable!(ctx)
unreachable!()
};
Ok(value)
@ -2029,497 +2026,3 @@ pub fn gen_ndarray_fill<'ctx>(
Ok(())
}
/// Generates LLVM IR for `ndarray.transpose`.
pub fn ndarray_transpose<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "ndarray_transpose";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let n_sz = call_ndarray_calc_size(generator, ctx, &n1.dim_sizes(), (None, None));
// Dimensions are reversed in the transposed array
let out = create_ndarray_dyn_shape(
generator,
ctx,
elem_ty,
&n1,
|_, ctx, n| Ok(n.load_ndims(ctx)),
|generator, ctx, n, idx| {
let new_idx = ctx.builder.build_int_sub(n.load_ndims(ctx), idx, "").unwrap();
let new_idx = ctx
.builder
.build_int_sub(new_idx, new_idx.get_type().const_int(1, false), "")
.unwrap();
unsafe { Ok(n.dim_sizes().get_typed_unchecked(ctx, generator, &new_idx, None)) }
},
)
.unwrap();
gen_for_callback_incrementing(
generator,
ctx,
None,
llvm_usize.const_zero(),
(n_sz, false),
|generator, ctx, _, idx| {
let elem = unsafe { n1.data().get_unchecked(ctx, generator, &idx, None) };
let new_idx = generator.gen_var_alloc(ctx, llvm_usize.into(), None)?;
let rem_idx = generator.gen_var_alloc(ctx, llvm_usize.into(), None)?;
ctx.builder.build_store(new_idx, llvm_usize.const_zero()).unwrap();
ctx.builder.build_store(rem_idx, idx).unwrap();
// Incrementally calculate the new index in the transposed array
// For each index, we first decompose it into the n-dims and use those to reconstruct the new index
// The formula used for indexing is:
// idx = dim_n * ( ... (dim2 * (dim0 * dim1) + dim1) + dim2 ... ) + dim_n
gen_for_callback_incrementing(
generator,
ctx,
None,
llvm_usize.const_zero(),
(n1.load_ndims(ctx), false),
|generator, ctx, _, ndim| {
let ndim_rev =
ctx.builder.build_int_sub(n1.load_ndims(ctx), ndim, "").unwrap();
let ndim_rev = ctx
.builder
.build_int_sub(ndim_rev, llvm_usize.const_int(1, false), "")
.unwrap();
let dim = unsafe {
n1.dim_sizes().get_typed_unchecked(ctx, generator, &ndim_rev, None)
};
let rem_idx_val =
ctx.builder.build_load(rem_idx, "").unwrap().into_int_value();
let new_idx_val =
ctx.builder.build_load(new_idx, "").unwrap().into_int_value();
let add_component =
ctx.builder.build_int_unsigned_rem(rem_idx_val, dim, "").unwrap();
let rem_idx_val =
ctx.builder.build_int_unsigned_div(rem_idx_val, dim, "").unwrap();
let new_idx_val = ctx.builder.build_int_mul(new_idx_val, dim, "").unwrap();
let new_idx_val =
ctx.builder.build_int_add(new_idx_val, add_component, "").unwrap();
ctx.builder.build_store(rem_idx, rem_idx_val).unwrap();
ctx.builder.build_store(new_idx, new_idx_val).unwrap();
Ok(())
},
llvm_usize.const_int(1, false),
)?;
let new_idx_val = ctx.builder.build_load(new_idx, "").unwrap().into_int_value();
unsafe { out.data().set_unchecked(ctx, generator, &new_idx_val, elem) };
Ok(())
},
llvm_usize.const_int(1, false),
)?;
Ok(out.as_base_value().into())
} else {
codegen_unreachable!(
ctx,
"{FN_NAME}() not supported for '{}'",
format!("'{}'", ctx.unifier.stringify(x1_ty))
)
}
}
/// LLVM-typed implementation for generating the implementation for `ndarray.reshape`.
///
/// * `x1` - `NDArray` to reshape.
/// * `shape` - The `shape` parameter used to construct the new `NDArray`.
/// Just like numpy, the `shape` argument can be:
/// 1. A list of `int32`; e.g., `np.reshape(arr, [600, -1, 3])`
/// 2. A tuple of `int32`; e.g., `np.reshape(arr, (-1, 800, 3))`
/// 3. A scalar `int32`; e.g., `np.reshape(arr, 3)`
///
/// Note that unlike other generating functions, one of the dimensions in the shape can be negative.
pub fn ndarray_reshape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
shape: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "ndarray_reshape";
let (x1_ty, x1) = x1;
let (_, shape) = shape;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let n_sz = call_ndarray_calc_size(generator, ctx, &n1.dim_sizes(), (None, None));
let acc = generator.gen_var_alloc(ctx, llvm_usize.into(), None)?;
let num_neg = generator.gen_var_alloc(ctx, llvm_usize.into(), None)?;
ctx.builder.build_store(acc, llvm_usize.const_int(1, false)).unwrap();
ctx.builder.build_store(num_neg, llvm_usize.const_zero()).unwrap();
let out = match shape {
BasicValueEnum::PointerValue(shape_list_ptr)
if ListValue::is_instance(shape_list_ptr, llvm_usize).is_ok() =>
{
// 1. A list of ints; e.g., `np.reshape(arr, [int64(600), int64(800, -1])`
let shape_list = ListValue::from_ptr_val(shape_list_ptr, llvm_usize, None);
// Check for -1 in dimensions
gen_for_callback_incrementing(
generator,
ctx,
None,
llvm_usize.const_zero(),
(shape_list.load_size(ctx, None), false),
|generator, ctx, _, idx| {
let ele =
shape_list.data().get(ctx, generator, &idx, None).into_int_value();
let ele = ctx.builder.build_int_s_extend(ele, llvm_usize, "").unwrap();
gen_if_else_expr_callback(
generator,
ctx,
|_, ctx| {
Ok(ctx
.builder
.build_int_compare(
IntPredicate::SLT,
ele,
llvm_usize.const_zero(),
"",
)
.unwrap())
},
|_, ctx| -> Result<Option<IntValue>, String> {
let num_neg_value =
ctx.builder.build_load(num_neg, "").unwrap().into_int_value();
let num_neg_value = ctx
.builder
.build_int_add(
num_neg_value,
llvm_usize.const_int(1, false),
"",
)
.unwrap();
ctx.builder.build_store(num_neg, num_neg_value).unwrap();
Ok(None)
},
|_, ctx| {
let acc_value =
ctx.builder.build_load(acc, "").unwrap().into_int_value();
let acc_value =
ctx.builder.build_int_mul(acc_value, ele, "").unwrap();
ctx.builder.build_store(acc, acc_value).unwrap();
Ok(None)
},
)?;
Ok(())
},
llvm_usize.const_int(1, false),
)?;
let acc_val = ctx.builder.build_load(acc, "").unwrap().into_int_value();
let rem = ctx.builder.build_int_unsigned_div(n_sz, acc_val, "").unwrap();
// Generate the output shape by filling -1 with `rem`
create_ndarray_dyn_shape(
generator,
ctx,
elem_ty,
&shape_list,
|_, ctx, _| Ok(shape_list.load_size(ctx, None)),
|generator, ctx, shape_list, idx| {
let dim =
shape_list.data().get(ctx, generator, &idx, None).into_int_value();
let dim = ctx.builder.build_int_s_extend(dim, llvm_usize, "").unwrap();
Ok(gen_if_else_expr_callback(
generator,
ctx,
|_, ctx| {
Ok(ctx
.builder
.build_int_compare(
IntPredicate::SLT,
dim,
llvm_usize.const_zero(),
"",
)
.unwrap())
},
|_, _| Ok(Some(rem)),
|_, _| Ok(Some(dim)),
)?
.unwrap()
.into_int_value())
},
)
}
BasicValueEnum::StructValue(shape_tuple) => {
// 2. A tuple of `int32`; e.g., `np.reshape(arr, (-1, 800, 3))`
let ndims = shape_tuple.get_type().count_fields();
// Check for -1 in dims
for dim_i in 0..ndims {
let dim = ctx
.builder
.build_extract_value(shape_tuple, dim_i, "")
.unwrap()
.into_int_value();
let dim = ctx.builder.build_int_s_extend(dim, llvm_usize, "").unwrap();
gen_if_else_expr_callback(
generator,
ctx,
|_, ctx| {
Ok(ctx
.builder
.build_int_compare(
IntPredicate::SLT,
dim,
llvm_usize.const_zero(),
"",
)
.unwrap())
},
|_, ctx| -> Result<Option<IntValue>, String> {
let num_negs =
ctx.builder.build_load(num_neg, "").unwrap().into_int_value();
let num_negs = ctx
.builder
.build_int_add(num_negs, llvm_usize.const_int(1, false), "")
.unwrap();
ctx.builder.build_store(num_neg, num_negs).unwrap();
Ok(None)
},
|_, ctx| {
let acc_val = ctx.builder.build_load(acc, "").unwrap().into_int_value();
let acc_val = ctx.builder.build_int_mul(acc_val, dim, "").unwrap();
ctx.builder.build_store(acc, acc_val).unwrap();
Ok(None)
},
)?;
}
let acc_val = ctx.builder.build_load(acc, "").unwrap().into_int_value();
let rem = ctx.builder.build_int_unsigned_div(n_sz, acc_val, "").unwrap();
let mut shape = Vec::with_capacity(ndims as usize);
// Reconstruct shape filling negatives with rem
for dim_i in 0..ndims {
let dim = ctx
.builder
.build_extract_value(shape_tuple, dim_i, "")
.unwrap()
.into_int_value();
let dim = ctx.builder.build_int_s_extend(dim, llvm_usize, "").unwrap();
let dim = gen_if_else_expr_callback(
generator,
ctx,
|_, ctx| {
Ok(ctx
.builder
.build_int_compare(
IntPredicate::SLT,
dim,
llvm_usize.const_zero(),
"",
)
.unwrap())
},
|_, _| Ok(Some(rem)),
|_, _| Ok(Some(dim)),
)?
.unwrap()
.into_int_value();
shape.push(dim);
}
create_ndarray_const_shape(generator, ctx, elem_ty, shape.as_slice())
}
BasicValueEnum::IntValue(shape_int) => {
// 3. A scalar `int32`; e.g., `np.reshape(arr, 3)`
let shape_int = gen_if_else_expr_callback(
generator,
ctx,
|_, ctx| {
Ok(ctx
.builder
.build_int_compare(
IntPredicate::SLT,
shape_int,
llvm_usize.const_zero(),
"",
)
.unwrap())
},
|_, _| Ok(Some(n_sz)),
|_, ctx| {
Ok(Some(ctx.builder.build_int_s_extend(shape_int, llvm_usize, "").unwrap()))
},
)?
.unwrap()
.into_int_value();
create_ndarray_const_shape(generator, ctx, elem_ty, &[shape_int])
}
_ => codegen_unreachable!(ctx),
}
.unwrap();
// Only allow one dimension to be negative
let num_negs = ctx.builder.build_load(num_neg, "").unwrap().into_int_value();
ctx.make_assert(
generator,
ctx.builder
.build_int_compare(IntPredicate::ULT, num_negs, llvm_usize.const_int(2, false), "")
.unwrap(),
"0:ValueError",
"can only specify one unknown dimension",
[None, None, None],
ctx.current_loc,
);
// The new shape must be compatible with the old shape
let out_sz = call_ndarray_calc_size(generator, ctx, &out.dim_sizes(), (None, None));
ctx.make_assert(
generator,
ctx.builder.build_int_compare(IntPredicate::EQ, out_sz, n_sz, "").unwrap(),
"0:ValueError",
"cannot reshape array of size {0} into provided shape of size {1}",
[Some(n_sz), Some(out_sz), None],
ctx.current_loc,
);
gen_for_callback_incrementing(
generator,
ctx,
None,
llvm_usize.const_zero(),
(n_sz, false),
|generator, ctx, _, idx| {
let elem = unsafe { n1.data().get_unchecked(ctx, generator, &idx, None) };
unsafe { out.data().set_unchecked(ctx, generator, &idx, elem) };
Ok(())
},
llvm_usize.const_int(1, false),
)?;
Ok(out.as_base_value().into())
} else {
codegen_unreachable!(
ctx,
"{FN_NAME}() not supported for '{}'",
format!("'{}'", ctx.unifier.stringify(x1_ty))
)
}
}
/// Generates LLVM IR for `ndarray.dot`.
/// Calculate inner product of two vectors or literals
/// For matrix multiplication use `np_matmul`
///
/// The input `NDArray` are flattened and treated as 1D
/// The operation is equivalent to `np.dot(arr1.ravel(), arr2.ravel())`
pub fn ndarray_dot<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
x2: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "ndarray_dot";
let (x1_ty, x1) = x1;
let (_, x2) = x2;
let llvm_usize = generator.get_size_type(ctx.ctx);
match (x1, x2) {
(BasicValueEnum::PointerValue(n1), BasicValueEnum::PointerValue(n2)) => {
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let n2 = NDArrayValue::from_ptr_val(n2, llvm_usize, None);
let n1_sz = call_ndarray_calc_size(generator, ctx, &n1.dim_sizes(), (None, None));
let n2_sz = call_ndarray_calc_size(generator, ctx, &n1.dim_sizes(), (None, None));
ctx.make_assert(
generator,
ctx.builder.build_int_compare(IntPredicate::EQ, n1_sz, n2_sz, "").unwrap(),
"0:ValueError",
"shapes ({0}), ({1}) not aligned",
[Some(n1_sz), Some(n2_sz), None],
ctx.current_loc,
);
let identity =
unsafe { n1.data().get_unchecked(ctx, generator, &llvm_usize.const_zero(), None) };
let acc = ctx.builder.build_alloca(identity.get_type(), "").unwrap();
ctx.builder.build_store(acc, identity.get_type().const_zero()).unwrap();
gen_for_callback_incrementing(
generator,
ctx,
None,
llvm_usize.const_zero(),
(n1_sz, false),
|generator, ctx, _, idx| {
let elem1 = unsafe { n1.data().get_unchecked(ctx, generator, &idx, None) };
let elem2 = unsafe { n2.data().get_unchecked(ctx, generator, &idx, None) };
let product = match elem1 {
BasicValueEnum::IntValue(e1) => ctx
.builder
.build_int_mul(e1, elem2.into_int_value(), "")
.unwrap()
.as_basic_value_enum(),
BasicValueEnum::FloatValue(e1) => ctx
.builder
.build_float_mul(e1, elem2.into_float_value(), "")
.unwrap()
.as_basic_value_enum(),
_ => codegen_unreachable!(ctx),
};
let acc_val = ctx.builder.build_load(acc, "").unwrap();
let acc_val = match acc_val {
BasicValueEnum::IntValue(e1) => ctx
.builder
.build_int_add(e1, product.into_int_value(), "")
.unwrap()
.as_basic_value_enum(),
BasicValueEnum::FloatValue(e1) => ctx
.builder
.build_float_add(e1, product.into_float_value(), "")
.unwrap()
.as_basic_value_enum(),
_ => codegen_unreachable!(ctx),
};
ctx.builder.build_store(acc, acc_val).unwrap();
Ok(())
},
llvm_usize.const_int(1, false),
)?;
let acc_val = ctx.builder.build_load(acc, "").unwrap();
Ok(acc_val)
}
(BasicValueEnum::IntValue(e1), BasicValueEnum::IntValue(e2)) => {
Ok(ctx.builder.build_int_mul(e1, e2, "").unwrap().as_basic_value_enum())
}
(BasicValueEnum::FloatValue(e1), BasicValueEnum::FloatValue(e2)) => {
Ok(ctx.builder.build_float_mul(e1, e2, "").unwrap().as_basic_value_enum())
}
_ => codegen_unreachable!(
ctx,
"{FN_NAME}() not supported for '{}'",
format!("'{}'", ctx.unifier.stringify(x1_ty))
),
}
}

113
nac3core/src/codegen/numpy_new/broadcast.rs Normal file
View File

@ -0,0 +1,113 @@
use itertools::Itertools;
use crate::{
codegen::{
irrt::ndarray::broadcast::{
call_nac3_ndarray_broadcast_shapes, call_nac3_ndarray_broadcast_to, ShapeEntry,
},
model::*,
numpy_new::util::{create_ndims, extract_ndims},
CodeGenContext, CodeGenerator,
},
typecheck::typedef::Type,
};
use super::object::NDArrayObject;
#[derive(Debug, Clone)]
pub struct BroadcastAllResult<'ctx> {
/// The statically known `ndims` of the broadcast result.
pub ndims: u64,
/// The broadcasting shape.
pub shape: Ptr<'ctx, IntModel<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>>,
}
// TODO: DOCUMENT: Behaves like `np.broadcast()`, except returns results differently.
pub fn broadcast_all_ndarrays<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndarrays: Vec<NDArrayObject<'ctx>>,
) -> BroadcastAllResult<'ctx> {
assert!(!ndarrays.is_empty());
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
let shape_model = StructModel(ShapeEntry);
// We can deduce the final ndims statically and immediately.
// It should be `max([ ndarray.ndims for ndarray in ndarrays ])`.
let broadcast_ndims =
ndarrays.iter().map(|ndarray| extract_ndims(&ctx.unifier, ndarray.ndims)).max().unwrap();
let broadcast_ndims_ty = create_ndims(&mut ctx.unifier, broadcast_ndims);
// NOTE: Now prepare before calling `call_nac3_ndarray_broadcast_shapes`
// Prepare input shape entries
let num_shape_entries =
sizet_model.constant(tyctx, ctx.ctx, u64::try_from(ndarrays.len()).unwrap());
let shape_entries =
shape_model.array_alloca(tyctx, ctx, num_shape_entries.value, "shape_entries");
for (i, ndarray) in ndarrays.iter().enumerate() {
let i = sizet_model.constant(tyctx, ctx.ctx, i as u64).value;
let this_shape = ndarray.instance.gep(ctx, |f| f.shape).load(tyctx, ctx, "this_shape");
let this_ndims = ndarray.instance.gep(ctx, |f| f.ndims).load(tyctx, ctx, "this_ndims");
let shape_entry = shape_entries.offset(tyctx, ctx, i, "shape_entry");
shape_entry.gep(ctx, |f| f.shape).store(ctx, this_shape);
shape_entry.gep(ctx, |f| f.ndims).store(ctx, this_ndims);
}
// Prepare destination
let dst_ndims = sizet_model.constant(tyctx, ctx.ctx, broadcast_ndims);
let dst_shape = sizet_model.array_alloca(tyctx, ctx, dst_ndims.value, "dst_shape");
call_nac3_ndarray_broadcast_shapes(
generator,
ctx,
num_shape_entries,
shape_entries,
dst_ndims,
dst_shape,
);
// Now that we know about the broadcasting shape, broadcast all the inputs.
// Broadcast all the inputs to shape `dst_shape`
let broadcasted_ndarrays = ndarrays
.into_iter()
.map(|ndarray| ndarray.broadcast_to(generator, ctx, broadcast_ndims_ty, dst_shape))
.collect_vec();
BroadcastAllResult { ndims: broadcast_ndims, shape: dst_shape, ndarrays: broadcasted_ndarrays }
}
impl<'ctx> NDArrayObject<'ctx> {
/// Broadcast an ndarray to a target shape.
#[must_use]
pub fn broadcast_to<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
target_ndims_ty: Type,
target_shape: Ptr<'ctx, IntModel<SizeT>>,
) -> Self {
// Please see comment in IRRT on how the caller should prepare `dst_ndarray`
let dst_ndarray = NDArrayObject::alloca(
generator,
ctx,
target_ndims_ty,
self.dtype,
"broadcast_ndarray_to_dst",
);
dst_ndarray.copy_shape(generator, ctx, target_shape);
call_nac3_ndarray_broadcast_to(generator, ctx, self.instance, dst_ndarray.instance);
dst_ndarray
}
}

217
nac3core/src/codegen/numpy_new/factory.rs Normal file
View File

@ -0,0 +1,217 @@
use inkwell::{
types::BasicType,
values::{BasicValue, BasicValueEnum, PointerValue},
AddressSpace,
};
use nac3parser::ast::StrRef;
use crate::{
codegen::{
model::*,
numpy_new::util::{alloca_ndarray, init_ndarray_data_by_alloca, init_ndarray_shape},
structure::ndarray::NpArray,
util::shape::make_shape_writer,
CodeGenContext, CodeGenerator,
},
symbol_resolver::ValueEnum,
toplevel::DefinitionId,
typecheck::typedef::{FunSignature, Type},
};
use super::util::gen_foreach_ndarray_elements;
/// Helper function to create an ndarray with uninitialized values
///
/// * `elem_ty` - The [`Type`] of the ndarray elements
/// * `shape` - The user input shape argument
/// * `shape_ty` - The [`Type`] of the shape argument
/// * `name` - LLVM IR name of the returned ndarray
fn create_empty_ndarray<'ctx, G>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
elem_ty: Type,
shape: BasicValueEnum<'ctx>,
shape_ty: Type,
name: &str,
) -> Result<Ptr<'ctx, StructModel<NpArray>>, String>
where
G: CodeGenerator + ?Sized,
{
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
let shape_writer = make_shape_writer(generator, ctx, shape, shape_ty);
let ndims = shape_writer.len;
let ndarray = alloca_ndarray(generator, ctx, ndims, name);
init_ndarray_shape(generator, ctx, ndarray, &shape_writer)?;
let itemsize = ctx.get_llvm_type(generator, elem_ty).size_of().unwrap();
let itemsize = sizet_model.check_value(tyctx, ctx.ctx, itemsize).unwrap();
ndarray.gep(ctx, |f| f.itemsize).store(ctx, itemsize);
// Needs `itemsize` and `shape` initialized
init_ndarray_data_by_alloca(generator, ctx, ndarray);
Ok(ndarray)
}
/// Helper function to create an ndarray full of a value.
///
/// * `elem_ty` - The [`Type`] of the ndarray elements and the fill value
/// * `shape` - The user input shape argument
/// * `shape_ty` - The [`Type`] of the shape argument
/// * `fill_value` - The user specified fill value
/// * `name` - LLVM IR name of the returned ndarray
fn create_full_ndarray<'ctx, G>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
elem_ty: Type,
shape: BasicValueEnum<'ctx>,
shape_ty: Type,
fill_value: BasicValueEnum<'ctx>,
name: &str,
) -> Result<Ptr<'ctx, StructModel<NpArray>>, String>
where
G: CodeGenerator + ?Sized,
{
let pndarray = create_empty_ndarray(generator, ctx, elem_ty, shape, shape_ty, name)?;
gen_foreach_ndarray_elements(
generator,
ctx,
pndarray,
|_generator, ctx, _hooks, _i, pelement| {
// Cannot use Model here, fill_value's type is not statically known.
let pfill_value_ty = fill_value.get_type().ptr_type(AddressSpace::default());
let pelement =
ctx.builder.build_pointer_cast(pelement.value, pfill_value_ty, "pelement").unwrap();
ctx.builder.build_store(pelement, fill_value).unwrap();
Ok(())
},
)?;
Ok(pndarray)
}
/// Generates LLVM IR for `np.empty`.
pub fn gen_ndarray_empty<'ctx>(
context: &mut CodeGenContext<'ctx, '_>,
obj: &Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: &[(Option<StrRef>, ValueEnum<'ctx>)],
generator: &mut dyn CodeGenerator,
) -> Result<PointerValue<'ctx>, String> {
assert!(obj.is_none());
assert_eq!(args.len(), 1);
// Parse arguments
let shape_ty = fun.0.args[0].ty;
let shape = args[0].1.clone().to_basic_value_enum(context, generator, shape_ty)?;
// Implementation
let ndarray_ptr = create_empty_ndarray(
generator,
context,
context.primitives.float,
shape,
shape_ty,
"ndarray",
)?;
Ok(ndarray_ptr.value)
}
/// Generates LLVM IR for `np.zeros`.
pub fn gen_ndarray_zeros<'ctx>(
context: &mut CodeGenContext<'ctx, '_>,
obj: &Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: &[(Option<StrRef>, ValueEnum<'ctx>)],
generator: &mut dyn CodeGenerator,
) -> Result<PointerValue<'ctx>, String> {
assert!(obj.is_none());
assert_eq!(args.len(), 1);
// Parse arguments
let shape_ty = fun.0.args[0].ty;
let shape = args[0].1.clone().to_basic_value_enum(context, generator, shape_ty)?;
// Implementation
// NOTE: Currently nac3's `np.zeros` is always `float64`.
let float64_ty = context.primitives.float;
let float64_llvm_type = context.get_llvm_type(generator, float64_ty).into_float_type();
let ndarray_ptr = create_full_ndarray(
generator,
context,
float64_ty, // `elem_ty` is always `float64`
shape,
shape_ty,
float64_llvm_type.const_zero().as_basic_value_enum(),
"ndarray",
)?;
Ok(ndarray_ptr.value)
}
/// Generates LLVM IR for `np.ones`.
pub fn gen_ndarray_ones<'ctx>(
context: &mut CodeGenContext<'ctx, '_>,
obj: &Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: &[(Option<StrRef>, ValueEnum<'ctx>)],
generator: &mut dyn CodeGenerator,
) -> Result<PointerValue<'ctx>, String> {
assert!(obj.is_none());
assert_eq!(args.len(), 1);
// Parse arguments
let shape_ty = fun.0.args[0].ty;
let shape = args[0].1.clone().to_basic_value_enum(context, generator, shape_ty)?;
// Implementation
// NOTE: Currently nac3's `np.ones` is always `float64`.
let float64_ty = context.primitives.float;
let float64_llvm_type = context.get_llvm_type(generator, float64_ty).into_float_type();
let ndarray_ptr = create_full_ndarray(
generator,
context,
float64_ty, // `elem_ty` is always `float64`
shape,
shape_ty,
float64_llvm_type.const_float(1.0).as_basic_value_enum(),
"ndarray",
)?;
Ok(ndarray_ptr.value)
}
/// Generates LLVM IR for `np.full`.
pub fn gen_ndarray_full<'ctx>(
context: &mut CodeGenContext<'ctx, '_>,
obj: &Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: &[(Option<StrRef>, ValueEnum<'ctx>)],
generator: &mut dyn CodeGenerator,
) -> Result<PointerValue<'ctx>, String> {
assert!(obj.is_none());
assert_eq!(args.len(), 2);
// Parse argument #1 shape
let shape_ty = fun.0.args[0].ty;
let shape_arg = args[0].1.clone().to_basic_value_enum(context, generator, shape_ty)?;
// Parse argument #2 fill_value
let fill_value_ty = fun.0.args[1].ty;
let fill_value_arg =
args[1].1.clone().to_basic_value_enum(context, generator, fill_value_ty)?;
// Implementation
let ndarray_ptr = create_full_ndarray(
generator,
context,
fill_value_ty,
shape_arg,
shape_ty,
fill_value_arg,
"ndarray",
)?;
Ok(ndarray_ptr.value)
}

76
nac3core/src/codegen/numpy_new/indexing.rs Normal file
View File

@ -0,0 +1,76 @@
use crate::{
codegen::{
irrt::ndarray::indexing::{call_nac3_ndarray_index, RustNDIndex},
model::*,
CodeGenContext, CodeGenerator,
},
typecheck::typedef::{Type, Unifier},
};
use super::{
object::{NDArrayObject, ScalarObject, ScalarOrNDArray},
util::{create_ndims, extract_ndims},
};
impl<'ctx> NDArrayObject<'ctx> {
pub fn deduce_ndims_after_indexing_with(
&self,
unifier: &mut Unifier,
indexes: &[RustNDIndex<'ctx>],
) -> Type {
let ndims = extract_ndims(unifier, self.ndims);
let new_ndims = RustNDIndex::deduce_ndims_after_indexing(indexes, ndims);
create_ndims(unifier, new_ndims)
}
#[must_use]
pub fn index_always_ndarray<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
indexes: &[RustNDIndex<'ctx>],
name: &str,
) -> Self {
let tyctx = generator.type_context(ctx.ctx);
let dst_ndims = self.deduce_ndims_after_indexing_with(&mut ctx.unifier, indexes);
let dst_ndarray = NDArrayObject::alloca(generator, ctx, dst_ndims, self.dtype, name);
let (num_indexes, indexes) = RustNDIndex::alloca_ndindexes(tyctx, ctx, indexes);
call_nac3_ndarray_index(
generator,
ctx,
num_indexes,
indexes,
self.instance,
dst_ndarray.instance,
);
dst_ndarray
}
pub fn index<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
indexes: &[RustNDIndex<'ctx>],
name: &str,
) -> ScalarOrNDArray<'ctx> {
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
let subndarray = self.index_always_ndarray(generator, ctx, indexes, name);
if subndarray.is_unsized(&ctx.unifier) {
// TODO: This actually never fails, don't use the `checked_` version.
let value = subndarray.checked_get_nth_element(
generator,
ctx,
sizet_model.const_0(tyctx, ctx.ctx),
name,
);
ScalarOrNDArray::Scalar(ScalarObject { dtype: self.dtype, value })
} else {
ScalarOrNDArray::NDArray(subndarray)
}
}
}

6
nac3core/src/codegen/numpy_new/mod.rs Normal file
View File

@ -0,0 +1,6 @@
pub mod broadcast;
pub mod factory;
pub mod indexing;
pub mod object;
pub mod util;
pub mod view;

69
nac3core/src/codegen/numpy_new/object.rs Normal file
View File

@ -0,0 +1,69 @@
use inkwell::values::{BasicValue, BasicValueEnum};
use crate::{
codegen::{model::*, structure::ndarray::NpArray, CodeGenContext},
toplevel::numpy::unpack_ndarray_var_tys,
typecheck::typedef::{Type, TypeEnum},
};
/// An LLVM ndarray instance with its typechecker [`Type`]s.
#[derive(Debug, Clone, Copy)]
pub struct NDArrayObject<'ctx> {
pub dtype: Type,
pub ndims: Type,
pub instance: Ptr<'ctx, StructModel<NpArray>>,
}
/// An LLVM numpy scalar with its [`Type`].
#[derive(Debug, Clone, Copy)]
pub struct ScalarObject<'ctx> {
pub dtype: Type,
pub value: BasicValueEnum<'ctx>,
}
#[derive(Debug, Clone, Copy)]
pub enum ScalarOrNDArray<'ctx> {
Scalar(ScalarObject<'ctx>),
NDArray(NDArrayObject<'ctx>),
}
impl<'ctx> ScalarOrNDArray<'ctx> {
/// 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(),
}
}
}
impl<'ctx> From<ScalarOrNDArray<'ctx>> for BasicValueEnum<'ctx> {
fn from(input: ScalarOrNDArray<'ctx>) -> BasicValueEnum<'ctx> {
input.to_basic_value_enum()
}
}
/// Split an [`BasicValueEnum<'ctx>`] into a [`ScalarOrNDArray`] depending
/// on its [`Type`].
pub fn split_scalar_or_ndarray<'ctx>(
tyctx: TypeContext<'ctx>,
ctx: &mut CodeGenContext<'ctx, '_>,
input: BasicValueEnum<'ctx>,
input_ty: Type,
) -> ScalarOrNDArray<'ctx> {
let pndarray_model = PtrModel(StructModel(NpArray));
let input_ty_enum = ctx.unifier.get_ty(input_ty);
match &*input_ty_enum {
TypeEnum::TObj { obj_id, .. }
if *obj_id == ctx.primitives.ndarray.obj_id(&ctx.unifier).unwrap() =>
{
let value = pndarray_model.check_value(tyctx, ctx.ctx, input).unwrap();
let (dtype, ndims) = unpack_ndarray_var_tys(&mut ctx.unifier, input_ty);
ScalarOrNDArray::NDArray(NDArrayObject { dtype, ndims, instance: value })
}
_ => ScalarOrNDArray::Scalar(ScalarObject { dtype: input_ty, value: input }),
}
}

328
nac3core/src/codegen/numpy_new/util.rs Normal file
View File

@ -0,0 +1,328 @@
use inkwell::{
types::BasicType,
values::{BasicValueEnum, PointerValue},
AddressSpace,
};
use util::gen_model_memcpy;
use crate::{
codegen::{
irrt::ndarray::basic::{
call_nac3_ndarray_copy_data, call_nac3_ndarray_get_nth_pelement,
call_nac3_ndarray_is_c_contiguous, call_nac3_ndarray_nbytes,
call_nac3_ndarray_set_strides_by_shape, call_nac3_ndarray_size,
call_nac3_ndarray_util_assert_shape_no_negative,
},
model::*,
stmt::BreakContinueHooks,
structure::ndarray::NpArray,
util::{array_writer::ArrayWriter, control::gen_model_for},
CodeGenContext, CodeGenerator,
},
symbol_resolver::SymbolValue,
typecheck::typedef::{Type, TypeEnum, Unifier},
};
use super::object::{NDArrayObject, ScalarOrNDArray};
/// Extract an ndarray's `ndims` [type][`Type`] in `u64`. Panic if not possible.
#[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)
}
/// Allocate an ndarray on the stack given its `ndims`.
///
/// `shape` and `strides` will be automatically allocated on the stack.
///
/// The returned ndarray's content will be:
/// - `data`: `nullptr`
/// - `itemsize`: **uninitialized** value
/// - `ndims`: initialized value, set to the input `ndims`
/// - `shape`: initialized pointer to an allocated stack with **uninitialized** values
/// - `strides`: initialized pointer to an allocated stack with **uninitialized** values
pub fn alloca_ndarray<'ctx, G>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndims: Int<'ctx, SizeT>,
name: &str,
) -> Ptr<'ctx, StructModel<NpArray>>
where
G: CodeGenerator + ?Sized,
{
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
let ndarray_model = StructModel(NpArray);
let ndarray_data_model = PtrModel(IntModel(Byte));
// Setup ndarray
let ndarray_ptr = ndarray_model.alloca(tyctx, ctx, name);
let shape = sizet_model.array_alloca(tyctx, ctx, ndims.value, "shape");
let strides = sizet_model.array_alloca(tyctx, ctx, ndims.value, "strides");
ndarray_ptr.gep(ctx, |f| f.data).store(ctx, ndarray_data_model.nullptr(tyctx, ctx.ctx));
ndarray_ptr.gep(ctx, |f| f.ndims).store(ctx, ndims);
ndarray_ptr.gep(ctx, |f| f.shape).store(ctx, shape);
ndarray_ptr.gep(ctx, |f| f.strides).store(ctx, strides);
ndarray_ptr
}
/// Initialize an ndarray's `shape` and asserts on.
/// `shape`'s values and prohibit illegal inputs like negative dimensions.
pub fn init_ndarray_shape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
pndarray: Ptr<'ctx, StructModel<NpArray>>,
shape_writer: &ArrayWriter<'ctx, G, SizeT, IntModel<SizeT>>,
) -> Result<(), String> {
let tyctx = generator.type_context(ctx.ctx);
let shape = pndarray.gep(ctx, |f| f.shape).load(tyctx, ctx, "shape");
(shape_writer.write)(generator, ctx, shape)?;
call_nac3_ndarray_util_assert_shape_no_negative(generator, ctx, shape_writer.len, shape);
Ok(())
}
/// 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 init_ndarray_data_by_alloca<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
pndarray: Ptr<'ctx, StructModel<NpArray>>,
) {
let tyctx = generator.type_context(ctx.ctx);
let ndarray_data_model = IntModel(Byte);
let nbytes = call_nac3_ndarray_nbytes(generator, ctx, pndarray);
let data = ndarray_data_model.array_alloca(tyctx, ctx, nbytes.value, "data");
pndarray.gep(ctx, |f| f.data).store(ctx, data);
call_nac3_ndarray_set_strides_by_shape(generator, ctx, pndarray);
}
/// Iterate through all elements in an ndarray.
///
/// `body` is given the index of an element and an opaque pointer (as an `uint8_t*`, you might want to cast it) to the element.
///
/// Short-circuiting is possible with the given [`BreakContinueHooks`].
pub fn gen_foreach_ndarray_elements<'ctx, G, F>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
pndarray: Ptr<'ctx, StructModel<NpArray>>,
body: F,
) -> Result<(), String>
where
G: CodeGenerator + ?Sized,
F: Fn(
&mut G,
&mut CodeGenContext<'ctx, '_>,
BreakContinueHooks<'ctx>,
Int<'ctx, SizeT>,
Ptr<'ctx, IntModel<Byte>>,
) -> Result<(), String>,
{
// TODO: Make this more efficient - use a special NDArray iterator?
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
let size = call_nac3_ndarray_size(generator, ctx, pndarray);
gen_model_for(
generator,
ctx,
sizet_model.const_0(tyctx, ctx.ctx),
size,
sizet_model.const_1(tyctx, ctx.ctx),
|generator, ctx, hooks, index| {
let pelement = call_nac3_ndarray_get_nth_pelement(generator, ctx, pndarray, index);
body(generator, ctx, hooks, index, pelement)
},
)
}
impl<'ctx> ScalarOrNDArray<'ctx> {
/// Convert `input` to an ndarray - behaves like `np.asarray`.
pub fn as_ndarray<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> NDArrayObject<'ctx> {
match self {
ScalarOrNDArray::NDArray(ndarray) => *ndarray,
ScalarOrNDArray::Scalar(scalar) => {
let tyctx = generator.type_context(ctx.ctx);
let pbyte_model = PtrModel(IntModel(Byte));
// We have to put the value on the stack to get a data pointer.
let data =
ctx.builder.build_alloca(scalar.value.get_type(), "as_ndarray_scalar").unwrap();
ctx.builder.build_store(data, scalar.value).unwrap();
let data = pbyte_model.transmute(tyctx, ctx, data, "data");
let ndims_ty = create_ndims(&mut ctx.unifier, 0);
let ndarray = NDArrayObject::alloca(
generator,
ctx,
ndims_ty,
scalar.dtype,
"scalar_as_ndarray",
);
ndarray.instance.gep(ctx, |f| f.data).store(ctx, data);
// No need to initialize/setup strides or shapes - because `ndims` is 0.
// So we only have to set `data`, `itemsize`, and `ndims = 0`.
ndarray
}
}
}
}
impl<'ctx> NDArrayObject<'ctx> {
pub fn alloca<G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
ndims: Type,
dtype: Type,
name: &str,
) -> Self {
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
let ndims_int = sizet_model.constant(tyctx, ctx.ctx, extract_ndims(&ctx.unifier, ndims));
let instance = alloca_ndarray(generator, ctx, ndims_int, name);
// Set itemsize
let dtype_ty = ctx.get_llvm_type(generator, dtype);
let itemsize = dtype_ty.size_of().unwrap();
let itemsize = sizet_model.s_extend_or_bit_cast(tyctx, ctx, itemsize, "itemsize");
instance.gep(ctx, |f| f.itemsize).store(ctx, itemsize);
NDArrayObject { dtype, ndims, instance }
}
pub fn copy_shape<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_shape: Ptr<'ctx, IntModel<SizeT>>,
) {
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
let self_shape = self.instance.gep(ctx, |f| f.shape).load(tyctx, ctx, "self_shape");
let ndims_int =
sizet_model.constant(tyctx, ctx.ctx, extract_ndims(&ctx.unifier, self.ndims));
gen_model_memcpy(tyctx, ctx, self_shape, src_shape, ndims_int.value, false);
}
pub fn copy_shape_from<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
src_ndarray: NDArrayObject<'ctx>,
) {
let tyctx = generator.type_context(ctx.ctx);
let src_shape = src_ndarray.instance.gep(ctx, |f| f.shape).load(tyctx, ctx, "src_shape");
self.copy_shape(generator, ctx, src_shape);
}
pub fn update_strides_by_shape<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) {
call_nac3_ndarray_set_strides_by_shape(generator, ctx, self.instance);
}
pub fn checked_get_nth_pelement<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
i: Int<'ctx, SizeT>,
name: &str,
) -> PointerValue<'ctx> {
let elem_ty = ctx.get_llvm_type(generator, self.dtype);
let p = call_nac3_ndarray_get_nth_pelement(generator, ctx, self.instance, i);
ctx.builder
.build_pointer_cast(p.value, elem_ty.ptr_type(AddressSpace::default()), name)
.unwrap()
}
pub fn checked_get_nth_element<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
i: Int<'ctx, SizeT>,
name: &str,
) -> BasicValueEnum<'ctx> {
let pelement = self.checked_get_nth_pelement(generator, ctx, i, "pelement");
ctx.builder.build_load(pelement, name).unwrap()
}
#[must_use]
pub fn is_unsized(&self, unifier: &Unifier) -> bool {
extract_ndims(unifier, self.ndims) == 0
}
pub fn size<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Int<'ctx, SizeT> {
call_nac3_ndarray_size(generator, ctx, self.instance)
}
pub fn nbytes<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Int<'ctx, SizeT> {
call_nac3_ndarray_nbytes(generator, ctx, self.instance)
}
pub fn is_c_contiguous<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) -> Int<'ctx, Bool> {
call_nac3_ndarray_is_c_contiguous(generator, ctx, self.instance)
}
pub fn alloca_owned_data<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
) {
init_ndarray_data_by_alloca(generator, ctx, self.instance);
}
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);
}
}

114
nac3core/src/codegen/numpy_new/view.rs Normal file
View File

@ -0,0 +1,114 @@
use inkwell::values::PointerValue;
use nac3parser::ast::StrRef;
use crate::{
codegen::{
irrt::ndarray::reshape::call_nac3_ndarray_resolve_and_check_new_shape,
model::*,
numpy_new::{object::split_scalar_or_ndarray, util::extract_ndims},
util::shape::make_shape_writer,
CodeGenContext, CodeGenerator,
},
symbol_resolver::ValueEnum,
toplevel::{numpy::unpack_ndarray_var_tys, DefinitionId},
typecheck::typedef::{FunSignature, Type},
};
use super::object::NDArrayObject;
impl<'ctx> NDArrayObject<'ctx> {
#[must_use]
pub fn reshape_or_copy<G: CodeGenerator + ?Sized>(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
new_ndims: Type,
new_shape: Ptr<'ctx, IntModel<SizeT>>,
) -> Self {
let tyctx = generator.type_context(ctx.ctx);
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, new_ndims, self.dtype, "reshaped_ndarray");
dst_ndarray.copy_shape(generator, ctx, new_shape);
dst_ndarray.update_strides_by_shape(generator, ctx);
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
.instance
.gep(ctx, |f| f.data)
.store(ctx, dst_ndarray.instance.gep(ctx, |f| f.data).load(tyctx, ctx, "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.alloca_owned_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
}
}
/// Generates LLVM IR for `np.reshape`.
pub fn gen_ndarray_reshape<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
obj: &Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: &[(Option<StrRef>, ValueEnum<'ctx>)],
generator: &mut dyn CodeGenerator,
) -> Result<PointerValue<'ctx>, String> {
assert!(obj.is_none());
assert_eq!(args.len(), 2);
// Parse argument #1 input
let input_ty = fun.0.args[0].ty;
let input_arg = args[0].1.clone().to_basic_value_enum(ctx, generator, input_ty)?;
// Parse argument #2 shape
let shape_ty = fun.0.args[1].ty;
let shape_arg = args[1].1.clone().to_basic_value_enum(ctx, generator, shape_ty)?;
// Define models
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
// Extract reshaped_ndims
let (_, reshaped_ndims) = unpack_ndarray_var_tys(&mut ctx.unifier, fun.0.ret);
let reshaped_ndims_int = extract_ndims(&ctx.unifier, reshaped_ndims);
// Process `input`
let ndarray =
split_scalar_or_ndarray(tyctx, ctx, input_arg, input_ty).as_ndarray(generator, ctx);
// Process the shape input from user and resolve negative indices
let new_shape = make_shape_writer(generator, ctx, shape_arg, shape_ty).alloca_array_and_write(
generator,
ctx,
"new_shape",
)?;
let size = ndarray.size(generator, ctx);
call_nac3_ndarray_resolve_and_check_new_shape(
generator,
ctx,
size,
sizet_model.constant(tyctx, ctx.ctx, reshaped_ndims_int),
new_shape,
);
// Reshape
let reshaped_ndarray = ndarray.reshape_or_copy(generator, ctx, reshaped_ndims, new_shape);
Ok(reshaped_ndarray.instance.value)
}

860
nac3core/src/codegen/stmt.rs
View File

File diff suppressed because it is too large Load Diff

43
nac3core/src/codegen/structure/cslice.rs Normal file
View File

@ -0,0 +1,43 @@
use crate::codegen::{model::*, CodeGenContext};
/// Fields of [`CSlice<'ctx>`].
pub struct CSliceFields<F: FieldVisitor> {
/// Pointer to the data.
pub base: F::Field<PtrModel<IntModel<Byte>>>,
/// Number of bytes of the data.
pub len: F::Field<IntModel<SizeT>>,
}
/// See <https://crates.io/crates/cslice>.
///
/// Additionally, see <https://github.com/m-labs/artiq/blob/b0d2705c385f64b6e6711c1726cd9178f40b598e/artiq/firmware/libeh/eh_artiq.rs>)
/// for ARTIQ-specific notes.
#[derive(Debug, Clone, Copy, Default)]
pub struct CSlice;
impl StructKind for CSlice {
type Fields<F: FieldVisitor> = CSliceFields<F>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F> {
Self::Fields { base: visitor.add("base"), len: visitor.add("len") }
}
}
impl StructModel<CSlice> {
/// Create a [`CSlice`].
///
/// `base` and `len` must be LLVM global constants.
pub fn create_const<'ctx>(
&self,
type_context: TypeContext<'ctx>,
ctx: &CodeGenContext<'ctx, '_>,
base: Ptr<'ctx, IntModel<Byte>>,
len: Int<'ctx, SizeT>,
) -> Struct<'ctx, CSlice> {
let value = self
.0
.get_struct_type(type_context, ctx.ctx)
.const_named_struct(&[base.value.into(), len.value.into()]);
self.believe_value(value)
}
}

57
nac3core/src/codegen/structure/exception.rs Normal file
View File

@ -0,0 +1,57 @@
use crate::codegen::model::*;
use super::cslice::CSlice;
/// The LLVM int type of an Exception ID.
pub type ExceptionId = Int32;
/// Fields of [`Exception<'ctx>`]
///
/// The definition came from `pub struct Exception<'a>` in
/// <https://github.com/m-labs/artiq/blob/master/artiq/firmware/libeh/eh_artiq.rs>.
pub struct ExceptionFields<F: FieldVisitor> {
/// nac3core's ID of the exception
pub id: F::Field<IntModel<ExceptionId>>,
/// The name of the file this `Exception` was raised in.
pub filename: F::Field<StructModel<CSlice>>,
/// The line number in the file this `Exception` was raised in.
pub line: F::Field<IntModel<Int32>>,
/// The column number in the file this `Exception` was raised in.
pub column: F::Field<IntModel<Int32>>,
/// The name of the Python function this `Exception` was raised in.
pub function: F::Field<StructModel<CSlice>>,
/// The message of this Exception.
///
/// The message can optionally contain integer parameters `{0}`, `{1}`, and `{2}` in its string,
/// where they will be substituted by `params[0]`, `params[1]`, and `params[2]` respectively (as `int64_t`s).
/// Here is an example:
///
/// ```ignore
/// "Index {0} is out of bounds! List only has {1} element(s)."
/// ```
///
/// In this case, `params[0]` and `params[1]` must be specified, and `params[2]` is ***unused***.
/// Having only 3 parameters is a constraint in ARTIQ.
pub msg: F::Field<StructModel<CSlice>>,
pub params: [F::Field<IntModel<Int64>>; 3],
}
/// nac3core & ARTIQ's Exception
#[derive(Debug, Clone, Copy, Default)]
pub struct Exception;
impl StructKind for Exception {
type Fields<F: FieldVisitor> = ExceptionFields<F>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F> {
Self::Fields {
id: visitor.add("id"),
filename: visitor.add("filename"),
line: visitor.add("line"),
column: visitor.add("column"),
function: visitor.add("function"),
msg: visitor.add("msg"),
params: [visitor.add("params[0]"), visitor.add("params[1]"), visitor.add("params[2]")],
}
}
}

3
nac3core/src/codegen/structure/mod.rs Normal file
View File

@ -0,0 +1,3 @@
pub mod cslice;
pub mod exception;
pub mod ndarray;

27
nac3core/src/codegen/structure/ndarray.rs Normal file
View File

@ -0,0 +1,27 @@
use crate::codegen::*;
pub struct NpArrayFields<F: FieldVisitor> {
pub data: F::Field<PtrModel<IntModel<Byte>>>,
pub itemsize: F::Field<IntModel<SizeT>>,
pub ndims: F::Field<IntModel<SizeT>>,
pub shape: F::Field<PtrModel<IntModel<SizeT>>>,
pub strides: F::Field<PtrModel<IntModel<SizeT>>>,
}
// TODO: Rename to `NDArray` when the old NDArray is removed.
#[derive(Debug, Clone, Copy, Default)]
pub struct NpArray;
impl StructKind for NpArray {
type Fields<F: FieldVisitor> = NpArrayFields<F>;
fn visit_fields<F: FieldVisitor>(&self, visitor: &mut F) -> Self::Fields<F> {
Self::Fields {
data: visitor.add("data"),
itemsize: visitor.add("itemsize"),
ndims: visitor.add("ndims"),
shape: visitor.add("shape"),
strides: visitor.add("strides"),
}
}
}

77
nac3core/src/codegen/test.rs
View File

@ -1,37 +1,34 @@
use std::{
collections::{HashMap, HashSet},
sync::Arc,
};
use indexmap::IndexMap;
use indoc::indoc;
use inkwell::{
targets::{InitializationConfig, Target},
OptimizationLevel,
};
use nac3parser::{
ast::{fold::Fold, FileName, StrRef},
parser::parse_program,
};
use parking_lot::RwLock;
use super::{
classes::{ListType, NDArrayType, ProxyType, RangeType},
concrete_type::ConcreteTypeStore,
CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask, CodeGenerator,
DefaultCodeGenerator, WithCall, WorkerRegistry,
};
use crate::{
codegen::{
classes::{ListType, NDArrayType, ProxyType, RangeType},
concrete_type::ConcreteTypeStore,
CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask,
CodeGenerator, DefaultCodeGenerator, WithCall, WorkerRegistry,
},
symbol_resolver::{SymbolResolver, ValueEnum},
toplevel::{
composer::{ComposerConfig, TopLevelComposer},
DefinitionId, FunInstance, TopLevelContext, TopLevelDef,
},
typecheck::{
type_inferencer::{FunctionData, IdentifierInfo, Inferencer, PrimitiveStore},
type_inferencer::{FunctionData, Inferencer, PrimitiveStore},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
},
};
use indexmap::IndexMap;
use indoc::indoc;
use inkwell::{
targets::{InitializationConfig, Target},
OptimizationLevel,
};
use nac3parser::ast::FileName;
use nac3parser::{
ast::{fold::Fold, StrRef},
parser::parse_program,
};
use parking_lot::RwLock;
use std::collections::{HashMap, HashSet};
use std::sync::Arc;
struct Resolver {
id_to_type: HashMap<StrRef, Type>,
@ -67,7 +64,6 @@ impl SymbolResolver for Resolver {
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, '_>,
_: &mut dyn CodeGenerator,
) -> Option<ValueEnum<'ctx>> {
unimplemented!()
}
@ -98,7 +94,7 @@ fn test_primitives() {
"};
let statements = parse_program(source, FileName::default()).unwrap();
let composer = TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 32).0;
let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
let mut unifier = composer.unifier.clone();
let primitives = composer.primitives_ty;
let top_level = Arc::new(composer.make_top_level_context());
@ -113,18 +109,8 @@ 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,
is_vararg: false,
},
FuncArg {
name: "b".into(),
ty: primitives.int32,
default_value: None,
is_vararg: false,
},
FuncArg { name: "a".into(), ty: primitives.int32, default_value: None },
FuncArg { name: "b".into(), ty: primitives.int32, default_value: None },
],
ret: primitives.int32,
vars: VarMap::new(),
@ -142,8 +128,7 @@ fn test_primitives() {
};
let mut virtual_checks = Vec::new();
let mut calls = HashMap::new();
let mut identifiers: HashMap<_, _> =
["a".into(), "b".into()].map(|id| (id, IdentifierInfo::default())).into();
let mut identifiers: HashSet<_> = ["a".into(), "b".into()].into();
let mut inferencer = Inferencer {
top_level: &top_level,
function_data: &mut function_data,
@ -263,19 +248,14 @@ fn test_simple_call() {
"};
let statements_2 = parse_program(source_2, FileName::default()).unwrap();
let composer = TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 32).0;
let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
let mut unifier = composer.unifier.clone();
let primitives = composer.primitives_ty;
let top_level = Arc::new(composer.make_top_level_context());
unifier.top_level = Some(top_level.clone());
let signature = FunSignature {
args: vec![FuncArg {
name: "a".into(),
ty: primitives.int32,
default_value: None,
is_vararg: false,
}],
args: vec![FuncArg { name: "a".into(), ty: primitives.int32, default_value: None }],
ret: primitives.int32,
vars: VarMap::new(),
};
@ -322,8 +302,7 @@ fn test_simple_call() {
};
let mut virtual_checks = Vec::new();
let mut calls = HashMap::new();
let mut identifiers: HashMap<_, _> =
["a".into(), "foo".into()].map(|id| (id, IdentifierInfo::default())).into();
let mut identifiers: HashSet<_> = ["a".into(), "foo".into()].into();
let mut inferencer = Inferencer {
top_level: &top_level,
function_data: &mut function_data,

34
nac3core/src/codegen/util/array_writer.rs Normal file
View File

@ -0,0 +1,34 @@
use crate::codegen::{model::*, CodeGenContext, CodeGenerator};
/// A closure containing details on how to write to/initialize an array.
#[allow(clippy::type_complexity)]
pub struct ArrayWriter<'ctx, G: CodeGenerator + ?Sized, Len: IntKind, Item: Model> {
/// Number of items to write
pub len: Int<'ctx, Len>,
/// Implementation to write to an array given its base pointer.
pub write: Box<
dyn Fn(
&mut G,
&mut CodeGenContext<'ctx, '_>,
Ptr<'ctx, Item>, // Base pointer
) -> Result<(), String>
+ 'ctx,
>,
}
impl<'ctx, G: CodeGenerator + ?Sized, Len: IntKind, Item: Model> ArrayWriter<'ctx, G, Len, Item> {
pub fn alloca_array_and_write(
&self,
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
name: &str,
) -> Result<Ptr<'ctx, Item>, String> {
let tyctx = generator.type_context(ctx.ctx);
let item_model = Item::default();
let item_array = item_model.array_alloca(tyctx, ctx, self.len.value, name);
(self.write)(generator, ctx, item_array)?;
Ok(item_array)
}
}

42
nac3core/src/codegen/util/control.rs Normal file
View File

@ -0,0 +1,42 @@
use crate::codegen::{
model::*,
stmt::{gen_for_callback_incrementing, BreakContinueHooks},
CodeGenContext, CodeGenerator,
};
// TODO: Document
// TODO: Rename function
/// Only allows positive steps
pub fn gen_model_for<'ctx, 'a, G, F, I>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, 'a>,
start: Int<'ctx, I>,
stop: Int<'ctx, I>,
step: Int<'ctx, I>,
body: F,
) -> Result<(), String>
where
G: CodeGenerator + ?Sized,
F: FnOnce(
&mut G,
&mut CodeGenContext<'ctx, 'a>,
BreakContinueHooks<'ctx>,
Int<'ctx, I>,
) -> Result<(), String>,
I: IntKind,
{
let int_model = IntModel(I::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,
)
}

3
nac3core/src/codegen/util/mod.rs Normal file
View File

@ -0,0 +1,3 @@
pub mod array_writer;
pub mod control;
pub mod shape;

127
nac3core/src/codegen/util/shape.rs Normal file
View File

@ -0,0 +1,127 @@
use inkwell::values::BasicValueEnum;
use crate::{
codegen::{
classes::{ListValue, UntypedArrayLikeAccessor},
model::*,
CodeGenContext, CodeGenerator,
},
typecheck::typedef::{Type, TypeEnum},
};
use super::{array_writer::ArrayWriter, control::gen_model_for};
// TODO: Generalize to complex iterables under a common interface
/// Create an [`ArrayWriter`] from a NumPy-like `shape` argument input.
/// * `shape` - The `shape` parameter.
/// * `shape_ty` - The element type of the `NDArray`.
///
/// The `shape` 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])`
///
/// The `int32` values will be sign-extended to `SizeT`
pub fn make_shape_writer<'ctx, G>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
shape: BasicValueEnum<'ctx>,
shape_ty: Type,
) -> ArrayWriter<'ctx, G, SizeT, IntModel<SizeT>>
where
G: CodeGenerator + ?Sized,
{
let tyctx = generator.type_context(ctx.ctx);
let sizet_model = IntModel(SizeT);
match &*ctx.unifier.get_ty(shape_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])`
// TODO: Remove ListValue with Model
let shape = ListValue::from_ptr_val(shape.into_pointer_value(), tyctx.size_type, None);
let len =
sizet_model.check_value(tyctx, ctx.ctx, shape.load_size(ctx, Some("len"))).unwrap();
ArrayWriter {
len,
write: Box::new(move |generator, ctx, dst_array| {
gen_model_for(
generator,
ctx,
sizet_model.constant(tyctx, ctx.ctx, 0),
len,
sizet_model.constant(tyctx, ctx.ctx, 1),
|generator, ctx, _hooks, i| {
let dim =
shape.data().get(ctx, generator, &i.value, None).into_int_value();
let dim = sizet_model.s_extend_or_bit_cast(tyctx, ctx, dim, "");
dst_array.offset(tyctx, ctx, i.value, "pdim").store(ctx, dim);
Ok(())
},
)
}),
}
}
TypeEnum::TTuple { ty: tuple_types } => {
// 2. A tuple of ints; e.g., `np.empty((600, 800, 3))`
let ndims = tuple_types.len();
// A tuple has to be a StructValue
// Read [`codegen::expr::gen_expr`] to see how `nac3core` translates a Python tuple into LLVM.
let shape = shape.into_struct_value();
ArrayWriter {
len: sizet_model.constant(tyctx, ctx.ctx, ndims as u64),
write: Box::new(move |_generator, ctx, dst_array| {
for axis in 0..ndims {
let dim = ctx
.builder
.build_extract_value(shape, axis as u32, format!("dim{axis}").as_str())
.unwrap()
.into_int_value();
let dim = sizet_model.s_extend_or_bit_cast(tyctx, ctx, dim, "");
dst_array
.offset(
tyctx,
ctx,
sizet_model.constant(tyctx, ctx.ctx, axis as u64).value,
"pdim",
)
.store(ctx, dim);
}
Ok(())
}),
}
}
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])`
// The value has to be an integer
let shape_int = shape.into_int_value();
ArrayWriter {
len: sizet_model.constant(tyctx, ctx.ctx, 1),
write: Box::new(move |_generator, ctx, dst_array| {
let dim = sizet_model.s_extend_or_bit_cast(tyctx, ctx, shape_int, "");
// Set shape[0] = shape_int
dst_array
.offset(tyctx, ctx, sizet_model.constant(tyctx, ctx.ctx, 0).value, "pdim")
.store(ctx, dim);
Ok(())
}),
}
}
_ => panic!("encountered shape type"),
}
}

4
nac3core/src/lib.rs
View File

@ -19,10 +19,6 @@
clippy::wildcard_imports
)]
// users of nac3core need to use the same version of these dependencies, so expose them as nac3core::*
pub use inkwell;
pub use nac3parser;
pub mod codegen;
pub mod symbol_resolver;
pub mod toplevel;

29
nac3core/src/symbol_resolver.rs
View File

@ -1,15 +1,7 @@
use std::{
collections::{HashMap, HashSet},
fmt::{Debug, Display},
rc::Rc,
sync::Arc,
};
use inkwell::values::{BasicValueEnum, FloatValue, IntValue, PointerValue, StructValue};
use itertools::{chain, izip, Itertools};
use parking_lot::RwLock;
use nac3parser::ast::{Constant, Expr, Location, StrRef};
use std::fmt::Debug;
use std::rc::Rc;
use std::sync::Arc;
use std::{collections::HashMap, collections::HashSet, fmt::Display};
use crate::{
codegen::{CodeGenContext, CodeGenerator},
@ -19,6 +11,10 @@ use crate::{
typedef::{Type, TypeEnum, Unifier, VarMap},
},
};
use inkwell::values::{BasicValueEnum, FloatValue, IntValue, PointerValue, StructValue};
use itertools::{chain, izip, Itertools};
use nac3parser::ast::{Constant, Expr, Location, StrRef};
use parking_lot::RwLock;
#[derive(Clone, PartialEq, Debug)]
pub enum SymbolValue {
@ -82,14 +78,14 @@ impl SymbolValue {
}
Constant::Tuple(t) => {
let expected_ty = unifier.get_ty(expected_ty);
let TypeEnum::TTuple { ty, is_vararg_ctx } = expected_ty.as_ref() else {
let TypeEnum::TTuple { ty } = expected_ty.as_ref() else {
return Err(format!(
"Expected {:?}, but got Tuple",
expected_ty.get_type_name()
));
};
assert!(*is_vararg_ctx || ty.len() == t.len());
assert_eq!(ty.len(), t.len());
let elems = t
.iter()
@ -159,7 +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, is_vararg_ctx: false })
unifier.add_ty(TypeEnum::TTuple { ty: vs_tys })
}
SymbolValue::OptionSome(_) | SymbolValue::OptionNone => primitives.option,
}
@ -369,7 +365,6 @@ pub trait SymbolResolver {
&self,
str: StrRef,
ctx: &mut CodeGenContext<'ctx, '_>,
generator: &mut dyn CodeGenerator,
) -> Option<ValueEnum<'ctx>>;
fn get_default_param_value(&self, expr: &Expr) -> Option<SymbolValue>;
@ -487,7 +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, is_vararg_ctx: false }))
Ok(unifier.add_ty(TypeEnum::TTuple { ty }))
} else {
Err(HashSet::from(["Expected multiple elements for tuple".into()]))
}

451
nac3core/src/toplevel/builtins.rs
View File

@ -1,5 +1,6 @@
use std::iter::once;
use helper::{debug_assert_prim_is_allowed, make_exception_fields, PrimDefDetails};
use indexmap::IndexMap;
use inkwell::{
attributes::{Attribute, AttributeLoc},
@ -8,24 +9,28 @@ use inkwell::{
IntPredicate,
};
use itertools::Either;
use ndarray::basic::call_nac3_ndarray_len;
use strum::IntoEnumIterator;
use super::{
helper::{debug_assert_prim_is_allowed, make_exception_fields, PrimDef, PrimDefDetails},
numpy::make_ndarray_ty,
*,
};
use crate::{
codegen::{
builtin_fns,
classes::{ProxyValue, RangeValue},
expr::destructure_range,
irrt::*,
model::*,
numpy::*,
numpy_new,
stmt::exn_constructor,
structure::ndarray::NpArray,
},
symbol_resolver::SymbolValue,
toplevel::{helper::PrimDef, numpy::make_ndarray_ty},
typecheck::typedef::{into_var_map, iter_type_vars, TypeVar, VarMap},
};
use super::*;
type BuiltinInfo = Vec<(Arc<RwLock<TopLevelDef>>, Option<Stmt>)>;
pub fn get_exn_constructor(
@ -44,26 +49,10 @@ 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),
@ -113,7 +102,7 @@ pub fn get_exn_constructor(
/// * `name`: The name of the implemented NumPy function.
/// * `ret_ty`: The return type of this function.
/// * `param_ty`: The parameters accepted by this function, represented by a tuple of the
/// [parameter type][Type] and the parameter symbol name.
/// [parameter type][Type] and the parameter symbol name.
/// * `codegen_callback`: A lambda generating LLVM IR for the implementation of this function.
fn create_fn_by_codegen(
unifier: &mut Unifier,
@ -129,12 +118,7 @@ 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,
is_vararg: false,
})
.map(|p| FuncArg { name: p.1.into(), ty: p.0, default_value: None })
.collect(),
ret: ret_ty,
vars: var_map.clone(),
@ -153,7 +137,7 @@ fn create_fn_by_codegen(
/// * `name`: The name of the implemented NumPy function.
/// * `ret_ty`: The return type of this function.
/// * `param_ty`: The parameters accepted by this function, represented by a tuple of the
/// [parameter type][Type] and the parameter symbol name.
/// [parameter type][Type] and the parameter symbol name.
/// * `intrinsic_fn`: The fully-qualified name of the LLVM intrinsic function.
fn create_fn_by_intrinsic(
unifier: &mut Unifier,
@ -215,10 +199,10 @@ fn create_fn_by_intrinsic(
/// * `name`: The name of the implemented NumPy function.
/// * `ret_ty`: The return type of this function.
/// * `param_ty`: The parameters accepted by this function, represented by a tuple of the
/// [parameter type][Type] and the parameter symbol name.
/// [parameter type][Type] and the parameter symbol name.
/// * `extern_fn`: The fully-qualified name of the extern function used as the implementation.
/// * `attrs`: The list of attributes to apply to this function declaration. Note that `nounwind` is
/// already implied by the C ABI.
/// already implied by the C ABI.
fn create_fn_by_extern(
unifier: &mut Unifier,
var_map: &VarMap,
@ -512,6 +496,8 @@ impl<'a> BuiltinBuilder<'a> {
| PrimDef::FunNpEye
| PrimDef::FunNpIdentity => self.build_ndarray_other_factory_function(prim),
PrimDef::FunNpReshape => self.build_ndarray_view_functions(prim),
PrimDef::FunStr => self.build_str_function(),
PrimDef::FunFloor | PrimDef::FunFloor64 | PrimDef::FunCeil | PrimDef::FunCeil64 => {
@ -576,22 +562,6 @@ impl<'a> BuiltinBuilder<'a> {
| PrimDef::FunNpLdExp
| PrimDef::FunNpHypot
| PrimDef::FunNpNextAfter => self.build_np_2ary_function(prim),
PrimDef::FunNpTranspose | PrimDef::FunNpReshape => {
self.build_np_sp_ndarray_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) {
@ -649,24 +619,17 @@ 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,
is_vararg: false,
},
FuncArg { name: "start".into(), ty: int32, default_value: None },
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,
@ -922,7 +885,6 @@ 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]),
@ -1057,7 +1019,6 @@ 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(),
@ -1247,9 +1208,11 @@ impl<'a> BuiltinBuilder<'a> {
&[(self.ndarray_factory_fn_shape_arg_tvar.ty, "shape")],
Box::new(move |ctx, obj, fun, args, generator| {
let func = match prim {
PrimDef::FunNpNDArray | PrimDef::FunNpEmpty => gen_ndarray_empty,
PrimDef::FunNpZeros => gen_ndarray_zeros,
PrimDef::FunNpOnes => gen_ndarray_ones,
PrimDef::FunNpNDArray | PrimDef::FunNpEmpty => {
numpy_new::factory::gen_ndarray_empty
}
PrimDef::FunNpZeros => numpy_new::factory::gen_ndarray_zeros,
PrimDef::FunNpOnes => numpy_new::factory::gen_ndarray_ones,
_ => unreachable!(),
};
func(ctx, &obj, fun, &args, generator).map(|val| Some(val.as_basic_value_enum()))
@ -1277,23 +1240,16 @@ 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,
is_vararg: false,
},
FuncArg { name: "object".into(), ty: tv.ty, default_value: None },
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,
@ -1324,7 +1280,7 @@ impl<'a> BuiltinBuilder<'a> {
// type variable
&[(self.list_int32, "shape"), (tv.ty, "fill_value")],
Box::new(move |ctx, obj, fun, args, generator| {
gen_ndarray_full(ctx, &obj, fun, &args, generator)
numpy_new::factory::gen_ndarray_full(ctx, &obj, fun, &args, generator)
.map(|val| Some(val.as_basic_value_enum()))
}),
)
@ -1335,24 +1291,17 @@ 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,
is_vararg: false,
},
FuncArg { name: "N".into(), ty: int32, default_value: None },
// 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,
@ -1386,6 +1335,41 @@ impl<'a> BuiltinBuilder<'a> {
}
}
// Build functions related to NDArray views
fn build_ndarray_view_functions(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(prim, &[PrimDef::FunNpReshape]);
match prim {
PrimDef::FunNpReshape => {
// TODO: Support scalar inputs, e.g., `np.reshape(99, (1, 1, 1, 1))`
let new_ndim_ty = self.unifier.get_fresh_var(Some("NewNDim".into()), None);
let returned_ndarray_ty = make_ndarray_ty(
self.unifier,
self.primitives,
Some(self.ndarray_dtype_tvar.ty),
Some(new_ndim_ty.ty),
);
create_fn_by_codegen(
self.unifier,
&into_var_map([self.ndarray_dtype_tvar, self.ndarray_ndims_tvar, new_ndim_ty]),
prim.name(),
returned_ndarray_ty,
&[
(self.primitives.ndarray, "array"),
(self.ndarray_factory_fn_shape_arg_tvar.ty, "shape"),
],
Box::new(|ctx, obj, fun, args, generator| {
numpy_new::view::gen_ndarray_reshape(ctx, &obj, fun, &args, generator)
.map(|val| Some(val.as_basic_value_enum()))
}),
)
}
_ => unreachable!(),
}
}
/// Build the `str()` function.
fn build_str_function(&mut self) -> TopLevelDef {
let prim = PrimDef::FunStr;
@ -1396,12 +1380,7 @@ 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,
is_vararg: false,
}],
args: vec![FuncArg { name: "s".into(), ty: str, default_value: None }],
ret: str,
vars: VarMap::default(),
})),
@ -1465,21 +1444,31 @@ impl<'a> BuiltinBuilder<'a> {
fn build_len_function(&mut self) -> TopLevelDef {
let prim = PrimDef::FunLen;
// Type handled in [`Inferencer::try_fold_special_call`]
let arg_tvar = self.unifier.get_dummy_var();
let PrimitiveStore { uint64, int32, .. } = *self.primitives;
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: "obj".into(),
ty: arg_tvar.ty,
default_value: None,
is_vararg: false,
}],
ret: self.primitives.int32,
vars: into_var_map([arg_tvar]),
args: vec![FuncArg { name: "ls".into(), ty: arg_ty.ty, default_value: None }],
ret: int32,
vars: into_var_map([tvar, arg_ty]),
})),
var_id: Vec::default(),
instance_to_symbol: HashMap::default(),
@ -1487,10 +1476,48 @@ 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 tyctx = generator.type_context(ctx.ctx);
let pndarray_model = PtrModel(StructModel(NpArray));
builtin_fns::call_len(generator, ctx, (arg_ty, arg)).map(|ret| Some(ret.into()))
let ndarray =
pndarray_model.check_value(tyctx, ctx.ctx, arg).unwrap();
let len = call_nac3_ndarray_len(generator, ctx, ndarray);
Some(len.value.as_basic_value_enum())
}
_ => unreachable!(),
}
})
},
)))),
loc: None,
@ -1506,18 +1533,8 @@ 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,
is_vararg: false,
},
FuncArg {
name: "n".into(),
ty: self.num_ty.ty,
default_value: None,
is_vararg: false,
},
FuncArg { name: "m".into(), ty: self.num_ty.ty, default_value: None },
FuncArg { name: "n".into(), ty: self.num_ty.ty, default_value: None },
],
ret: self.num_ty.ty,
vars: self.num_var_map.clone(),
@ -1599,12 +1616,7 @@ 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,
is_vararg: false,
})
.map(|p| FuncArg { name: p.1.into(), ty: p.0, default_value: None })
.collect(),
ret: ret_ty.ty,
vars: into_var_map([x1_ty, x2_ty, ret_ty]),
@ -1645,7 +1657,6 @@ 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(),
@ -1834,12 +1845,7 @@ 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,
is_vararg: false,
})
.map(|p| FuncArg { name: p.1.into(), ty: p.0, default_value: None })
.collect(),
ret: ret_ty.ty,
vars: into_var_map([x1_ty, x2_ty, ret_ty]),
@ -1873,207 +1879,6 @@ impl<'a> BuiltinBuilder<'a> {
}
}
/// Build np/sp functions that take as input `NDArray` only
fn build_np_sp_ndarray_function(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(prim, &[PrimDef::FunNpTranspose, PrimDef::FunNpReshape]);
match prim {
PrimDef::FunNpTranspose => {
let ndarray_ty = self.unifier.get_fresh_var_with_range(
&[self.ndarray_num_ty],
Some("T".into()),
None,
);
create_fn_by_codegen(
self.unifier,
&into_var_map([ndarray_ty]),
prim.name(),
ndarray_ty.ty,
&[(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)?;
Ok(Some(ndarray_transpose(generator, ctx, (arg_ty, arg_val))?))
}),
)
}
// 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::FunNpReshape => create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
self.ndarray_num_ty,
&[(self.ndarray_num_ty, "x"), (self.ndarray_factory_fn_shape_arg_tvar.ty, "shape")],
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(ndarray_reshape(generator, ctx, (x1_ty, x1_val), (x2_ty, x2_val))?))
}),
),
_ => unreachable!(),
}
}
/// 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)?;
Ok(Some(ndarray_dot(generator, ctx, (x1_ty, x1_val), (x2_ty, x2_val))?))
}),
),
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())
}

803
nac3core/src/toplevel/composer.rs
View File

File diff suppressed because it is too large Load Diff

209
nac3core/src/toplevel/helper.rs
View File

@ -1,15 +1,13 @@
use std::convert::TryInto;
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::numpy::unpack_ndarray_var_tys;
use crate::typecheck::typedef::{into_var_map, iter_type_vars, Mapping, TypeVarId, VarMap};
use nac3parser::ast::{Constant, Location};
use strum::IntoEnumIterator;
use strum_macros::EnumIter;
use nac3parser::ast::{Constant, ExprKind, Location};
use super::{numpy::unpack_ndarray_var_tys, *};
use crate::{
symbol_resolver::SymbolValue,
typecheck::typedef::{into_var_map, iter_type_vars, Mapping, TypeVarId, VarMap},
};
use super::*;
/// All primitive types and functions in nac3core.
#[derive(Clone, Copy, Debug, EnumIter, PartialEq, Eq)]
@ -54,6 +52,9 @@ pub enum PrimDef {
FunNpEye,
FunNpIdentity,
// NumPy view functions
FunNpReshape,
// Miscellaneous NumPy & SciPy functions
FunNpRound,
FunNpFloor,
@ -101,21 +102,6 @@ pub enum PrimDef {
FunNpLdExp,
FunNpHypot,
FunNpNextAfter,
FunNpTranspose,
FunNpReshape,
// Linalg functions
FunNpDot,
FunNpLinalgCholesky,
FunNpLinalgQr,
FunNpLinalgSvd,
FunNpLinalgInv,
FunNpLinalgPinv,
FunNpLinalgMatrixPower,
FunNpLinalgDet,
FunSpLinalgLu,
FunSpLinalgSchur,
FunSpLinalgHessenberg,
// Miscellaneous Python & NAC3 functions
FunInt32,
@ -240,6 +226,9 @@ impl PrimDef {
PrimDef::FunNpEye => fun("np_eye", None),
PrimDef::FunNpIdentity => fun("np_identity", None),
// NumPy view functions
PrimDef::FunNpReshape => fun("np_reshape", None),
// Miscellaneous NumPy & SciPy functions
PrimDef::FunNpRound => fun("np_round", None),
PrimDef::FunNpFloor => fun("np_floor", None),
@ -287,21 +276,6 @@ impl PrimDef {
PrimDef::FunNpLdExp => fun("np_ldexp", None),
PrimDef::FunNpHypot => fun("np_hypot", None),
PrimDef::FunNpNextAfter => fun("np_nextafter", None),
PrimDef::FunNpTranspose => fun("np_transpose", None),
PrimDef::FunNpReshape => fun("np_reshape", 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),
@ -389,9 +363,6 @@ impl TopLevelDef {
r
}
),
TopLevelDef::Variable { name, ty, .. } => {
format!("Variable {{ name: {name:?}, ty: {:?} }}", unifier.stringify(*ty),)
}
}
}
}
@ -510,7 +481,6 @@ 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]),
@ -593,18 +563,6 @@ impl TopLevelComposer {
}
}
#[must_use]
pub fn make_top_level_variable_def(
name: String,
simple_name: StrRef,
ty: Type,
ty_decl: Option<Expr>,
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
loc: Option<Location>,
) -> TopLevelDef {
TopLevelDef::Variable { name, simple_name, ty, ty_decl, resolver, loc }
}
#[must_use]
pub fn make_class_method_name(mut class_name: String, method_name: &str) -> String {
class_name.push('.');
@ -750,16 +708,7 @@ impl TopLevelComposer {
)
}
/// This function returns the fields that have been initialized in the `__init__` function of a class
/// The function takes as input:
/// * `class_id`: The `object_id` of the class whose function is being evaluated (check `TopLevelDef::Class`)
/// * `definition_ast_list`: A list of ast definitions and statements defined in `TopLevelComposer`
/// * `stmts`: The body of function being parsed. Each statment is analyzed to check varaible initialization statements
pub fn get_all_assigned_field(
class_id: usize,
definition_ast_list: &Vec<DefAst>,
stmts: &[Stmt<()>],
) -> Result<HashSet<StrRef>, HashSet<String>> {
pub fn get_all_assigned_field(stmts: &[Stmt<()>]) -> Result<HashSet<StrRef>, HashSet<String>> {
let mut result = HashSet::new();
for s in stmts {
match &s.node {
@ -795,138 +744,30 @@ impl TopLevelComposer {
// TODO: do not check for For and While?
ast::StmtKind::For { body, orelse, .. }
| ast::StmtKind::While { body, orelse, .. } => {
result.extend(Self::get_all_assigned_field(
class_id,
definition_ast_list,
body.as_slice(),
)?);
result.extend(Self::get_all_assigned_field(
class_id,
definition_ast_list,
orelse.as_slice(),
)?);
result.extend(Self::get_all_assigned_field(body.as_slice())?);
result.extend(Self::get_all_assigned_field(orelse.as_slice())?);
}
ast::StmtKind::If { body, orelse, .. } => {
let inited_for_sure = Self::get_all_assigned_field(
class_id,
definition_ast_list,
body.as_slice(),
)?
.intersection(&Self::get_all_assigned_field(
class_id,
definition_ast_list,
orelse.as_slice(),
)?)
.copied()
.collect::<HashSet<_>>();
let inited_for_sure = Self::get_all_assigned_field(body.as_slice())?
.intersection(&Self::get_all_assigned_field(orelse.as_slice())?)
.copied()
.collect::<HashSet<_>>();
result.extend(inited_for_sure);
}
ast::StmtKind::Try { body, orelse, finalbody, .. } => {
let inited_for_sure = Self::get_all_assigned_field(
class_id,
definition_ast_list,
body.as_slice(),
)?
.intersection(&Self::get_all_assigned_field(
class_id,
definition_ast_list,
orelse.as_slice(),
)?)
.copied()
.collect::<HashSet<_>>();
let inited_for_sure = Self::get_all_assigned_field(body.as_slice())?
.intersection(&Self::get_all_assigned_field(orelse.as_slice())?)
.copied()
.collect::<HashSet<_>>();
result.extend(inited_for_sure);
result.extend(Self::get_all_assigned_field(
class_id,
definition_ast_list,
finalbody.as_slice(),
)?);
result.extend(Self::get_all_assigned_field(finalbody.as_slice())?);
}
ast::StmtKind::With { body, .. } => {
result.extend(Self::get_all_assigned_field(
class_id,
definition_ast_list,
body.as_slice(),
)?);
}
// Variables Initialized in function calls
ast::StmtKind::Expr { value, .. } => {
let ExprKind::Call { func, .. } = &value.node else {
continue;
};
let ExprKind::Attribute { value, attr, .. } = &func.node else {
continue;
};
let ExprKind::Name { id, .. } = &value.node else {
continue;
};
// Need to consider the two cases:
// Case 1) Call to class function i.e. id = `self`
// Case 2) Call to class ancestor function i.e. id = ancestor_name
// We leave checking whether function in case 2 belonged to class ancestor or not to type checker
//
// According to current handling of `self`, function definition are fixed and do not change regardless
// of which object is passed as `self` i.e. virtual polymorphism is not supported
// Therefore, we change class id for case 2 to reflect behavior of our compiler
let class_name = if *id == "self".into() {
let ast::StmtKind::ClassDef { name, .. } =
&definition_ast_list[class_id].1.as_ref().unwrap().node
else {
unreachable!()
};
name
} else {
id
};
let parent_method = definition_ast_list.iter().find_map(|def| {
let (
class_def,
Some(ast::Located {
node: ast::StmtKind::ClassDef { name, body, .. },
..
}),
) = &def
else {
return None;
};
let TopLevelDef::Class { object_id: class_id, .. } = &*class_def.read()
else {
unreachable!()
};
if name == class_name {
body.iter().find_map(|m| {
let ast::StmtKind::FunctionDef { name, body, .. } = &m.node else {
return None;
};
if *name == *attr {
return Some((body.clone(), class_id.0));
}
None
})
} else {
None
}
});
// If method body is none then method does not exist
if let Some((method_body, class_id)) = parent_method {
result.extend(Self::get_all_assigned_field(
class_id,
definition_ast_list,
method_body.as_slice(),
)?);
} else {
return Err(HashSet::from([format!(
"{}.{} not found in class {class_name} at {}",
*id, *attr, value.location
)]));
}
result.extend(Self::get_all_assigned_field(body.as_slice())?);
}
ast::StmtKind::Pass { .. }
| ast::StmtKind::Assert { .. }
| ast::StmtKind::AnnAssign { .. } => {}
| ast::StmtKind::Expr { .. } => {}
_ => {
unimplemented!()

61
nac3core/src/toplevel/mod.rs
View File

@ -6,36 +6,36 @@ use std::{
sync::Arc,
};
use inkwell::values::BasicValueEnum;
use itertools::Itertools;
use parking_lot::RwLock;
use nac3parser::ast::{self, Expr, Location, Stmt, StrRef};
use super::codegen::CodeGenContext;
use super::typecheck::type_inferencer::PrimitiveStore;
use super::typecheck::typedef::{
FunSignature, FuncArg, SharedUnifier, Type, TypeEnum, Unifier, VarMap,
};
use crate::{
codegen::{CodeGenContext, CodeGenerator},
codegen::CodeGenerator,
symbol_resolver::{SymbolResolver, ValueEnum},
typecheck::{
type_inferencer::{CodeLocation, PrimitiveStore},
typedef::{
CallId, FunSignature, FuncArg, SharedUnifier, Type, TypeEnum, TypeVarId, Unifier,
VarMap,
},
type_inferencer::CodeLocation,
typedef::{CallId, TypeVarId},
},
};
use composer::*;
use type_annotation::*;
use inkwell::values::BasicValueEnum;
use itertools::Itertools;
use nac3parser::ast::{self, Location, Stmt, StrRef};
use parking_lot::RwLock;
#[derive(PartialEq, Eq, PartialOrd, Ord, Clone, Copy, Hash, Debug)]
pub struct DefinitionId(pub usize);
pub mod builtins;
pub mod composer;
pub mod helper;
pub mod numpy;
pub mod type_annotation;
use composer::*;
use type_annotation::*;
#[cfg(test)]
mod test;
pub mod type_annotation;
#[derive(PartialEq, Eq, PartialOrd, Ord, Clone, Copy, Hash, Debug)]
pub struct DefinitionId(pub usize);
type GenCallCallback = dyn for<'ctx, 'a> Fn(
&mut CodeGenContext<'ctx, 'a>,
@ -130,14 +130,14 @@ pub enum TopLevelDef {
/// Function instance to symbol mapping
///
/// * Key: String representation of type variable values, sorted by variable ID in ascending
/// order, including type variables associated with the class.
/// order, including type variables associated with the class.
/// * Value: Function symbol name.
instance_to_symbol: HashMap<String, String>,
/// Function instances to annotated AST mapping
///
/// * Key: String representation of type variable values, sorted by variable ID in ascending
/// order, including type variables associated with the class. Excluding rigid type
/// variables.
/// order, including type variables associated with the class. Excluding rigid type
/// variables.
///
/// Rigid type variables that would be substituted when the function is instantiated.
instance_to_stmt: HashMap<String, FunInstance>,
@ -148,25 +148,6 @@ pub enum TopLevelDef {
/// Definition location.
loc: Option<Location>,
},
Variable {
/// Qualified name of the global variable, should be unique globally.
name: String,
/// Simple name, the same as in method/function definition.
simple_name: StrRef,
/// Type of the global variable.
ty: Type,
/// The declared type of the global variable, or [`None`] if no type annotation is provided.
ty_decl: Option<Expr>,
/// Symbol resolver of the module defined the class.
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
/// Definition location.
loc: Option<Location>,
},
}
pub struct TopLevelContext {

21
nac3core/src/toplevel/numpy.rs
View File

@ -1,17 +1,18 @@
use itertools::Itertools;
use super::helper::PrimDef;
use crate::typecheck::{
type_inferencer::PrimitiveStore,
typedef::{Type, TypeEnum, TypeVarId, Unifier, VarMap},
use crate::{
toplevel::helper::PrimDef,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{Type, TypeEnum, TypeVarId, Unifier, VarMap},
},
};
use itertools::Itertools;
/// Creates a `ndarray` [`Type`] with the given type arguments.
///
/// * `dtype` - The element type of the `ndarray`, or [`None`] if the type variable is not
/// specialized.
/// specialized.
/// * `ndims` - The number of dimensions of the `ndarray`, or [`None`] if the type variable is not
/// specialized.
/// specialized.
pub fn make_ndarray_ty(
unifier: &mut Unifier,
primitives: &PrimitiveStore,
@ -24,9 +25,9 @@ pub fn make_ndarray_ty(
/// Substitutes type variables in `ndarray`.
///
/// * `dtype` - The element type of the `ndarray`, or [`None`] if the type variable is not
/// specialized.
/// specialized.
/// * `ndims` - The number of dimensions of the `ndarray`, or [`None`] if the type variable is not
/// specialized.
/// specialized.
pub fn subst_ndarray_tvars(
unifier: &mut Unifier,
ndarray: Type,

2
nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__generic_class.snap
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(241)]\n}\n",
"Function {\nname: \"Generic_A.fun\",\nsig: \"fn[[a:int32], V]\",\nvar_id: [TypeVarId(248)]\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",

2
nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__inheritance_override.snap
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[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",
"Class {\nname: \"B\",\nancestors: [\"B[typevar237]\", \"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: [\"typevar237\"]\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",

4
nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__list_tuple_generic.snap
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(243)]\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(248)]\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[v:V], none]\",\nvar_id: [TypeVarId(250)]\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(255)]\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",

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