WIP: list assignment

New type vars are introduced when programming new ndarray functions.

.clang-format

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

.gitignore

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

Cargo.lock

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

flake.lock

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

flake.nix

@@ -6,6 +6,7 @@
   outputs = { self, nixpkgs }:
     let
       pkgs = import nixpkgs { system = "x86_64-linux"; };
+      pkgs32 = import nixpkgs { system = "i686-linux"; };
     in rec {
       packages.x86_64-linux = rec {
         llvm-nac3 = pkgs.callPackage ./nix/llvm {};
@@ -13,9 +14,24 @@
           ''
           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 $out/bin/llvm-as-irrt
           '';
+        demo-linalg-stub = pkgs.rustPlatform.buildRustPackage {
+          name = "demo-linalg-stub";
+          src = ./nac3standalone/demo/linalg;
+          cargoLock = {
+            lockFile = ./nac3standalone/demo/linalg/Cargo.lock;
+          };
+          doCheck = false;
+        };
+        demo-linalg-stub32 = pkgs32.rustPlatform.buildRustPackage {
+          name = "demo-linalg-stub32";
+          src = ./nac3standalone/demo/linalg;
+          cargoLock = {
+            lockFile = ./nac3standalone/demo/linalg/Cargo.lock;
+          };
+          doCheck = false;
+        };
         nac3artiq = pkgs.python3Packages.toPythonModule (
           pkgs.rustPlatform.buildRustPackage rec {
             name = "nac3artiq";
@@ -24,9 +40,8 @@
             cargoLock = {
               lockFile = ./Cargo.lock;
             };
-            cargoTestFlags = [ "--features" "test" ];
             passthru.cargoLock = cargoLock;
-            nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_14.clang llvm-tools-irrt pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
+            nativeBuildInputs = [ pkgs.python3 (pkgs.wrapClangMulti pkgs.llvmPackages_14.clang) llvm-tools-irrt pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
             buildInputs = [ pkgs.python3 llvm-nac3 ];
             checkInputs = [ (pkgs.python3.withPackages(ps: [ ps.numpy ps.scipy ])) ];
             checkPhase =
@@ -34,7 +49,9 @@
               echo "Checking nac3standalone demos..."
               pushd nac3standalone/demo
               patchShebangs .
-              ./check_demos.sh
+              export DEMO_LINALG_STUB=${demo-linalg-stub}/lib/liblinalg.a
+              export DEMO_LINALG_STUB32=${demo-linalg-stub32}/lib/liblinalg.a
+              ./check_demos.sh -i686
               popd
               echo "Running Cargo tests..."
               cargoCheckHook
@@ -151,7 +168,7 @@
         buildInputs = with pkgs; [
           # build dependencies
           packages.x86_64-linux.llvm-nac3
-          llvmPackages_14.clang llvmPackages_14.llvm.out  # for running nac3standalone demos
+          (pkgs.wrapClangMulti llvmPackages_14.clang) llvmPackages_14.llvm.out  # for running nac3standalone demos
           packages.x86_64-linux.llvm-tools-irrt
           cargo
           rustc
@@ -163,10 +180,12 @@
           clippy
           pre-commit
           rustfmt
-          rust-analyzer
         ];
-        # https://nixos.wiki/wiki/Rust#Shell.nix_example
+        shellHook =
-        RUST_SRC_PATH = "${pkgs.rust.packages.stable.rustPlatform.rustLibSrc}";
+          ''
+          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";

nac3artiq/Cargo.toml

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

nac3artiq/demo/list_cmp.py

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

nac3artiq/demo/str_abi.py

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

nac3artiq/src/codegen.rs

@@ -1,8 +1,11 @@
 use nac3core::{
     codegen::{
+        classes::{ListValue, UntypedArrayLikeAccessor},
         expr::gen_call,
         llvm_intrinsics::{call_int_smax, call_stackrestore, call_stacksave},
-        stmt::{gen_block, gen_with},
+        model::*,
+        object::{any::AnyObject, ndarray::NDArrayObject, range::RangeObject, str::str_model},
+        stmt::{gen_block, gen_for_callback_incrementing, gen_if_callback, gen_with},
         CodeGenContext, CodeGenerator,
     },
     symbol_resolver::ValueEnum,
@@ -13,7 +16,11 @@ use nac3core::{
 use nac3parser::ast::{Expr, ExprKind, Located, Stmt, StmtKind, StrRef};
 
 use inkwell::{
-    context::Context, module::Linkage, types::IntType, values::BasicValueEnum, AddressSpace,
+    context::Context,
+    module::Linkage,
+    types::IntType,
+    values::{BasicValueEnum, PointerValue, StructValue},
+    AddressSpace, IntPredicate,
 };
 
 use pyo3::{
@@ -23,10 +30,12 @@ use pyo3::{
 
 use crate::{symbol_resolver::InnerResolver, timeline::TimeFns};
 
+use itertools::Itertools;
 use std::{
-    collections::hash_map::DefaultHasher,
+    collections::{hash_map::DefaultHasher, HashMap},
-    collections::HashMap,
     hash::{Hash, Hasher},
+    iter::once,
+    mem,
     sync::Arc,
 };
 
@@ -119,7 +128,7 @@ impl<'a> ArtiqCodeGenerator<'a> {
     /// (possibly indirect) `parallel` block.
     ///
     /// * `store_name` - The LLVM value name for the pointer to `end`. `.addr` will be appended to
-    /// the end of the provided value name.
+    ///   the end of the provided value name.
     fn timeline_update_end_max(
         &mut self,
         ctx: &mut CodeGenContext<'_, '_>,
@@ -386,7 +395,7 @@ fn gen_rpc_tag(
     } else {
         let ty_enum = ctx.unifier.get_ty(ty);
         match &*ty_enum {
-            TTuple { ty } => {
+            TTuple { ty, is_vararg_ctx: false } => {
                 buffer.push(b't');
                 buffer.push(ty.len() as u8);
                 for ty in ty {
@@ -414,7 +423,10 @@ fn gen_rpc_tag(
                 } else {
                     unreachable!()
                 };
-                assert!((0u64..=u64::from(u8::MAX)).contains(&ndarray_ndims));
+                assert!(
+                    (0u64..=u64::from(u8::MAX)).contains(&ndarray_ndims),
+                    "Only NDArrays of sizes between 0 and 255 can be RPCed"
+                );
 
                 buffer.push(b'a');
                 buffer.push((ndarray_ndims & 0xFF) as u8);
@@ -426,6 +438,77 @@ fn gen_rpc_tag(
     Ok(())
 }
 
+/// Formats an RPC argument to conform to the expected format required by `send_value`.
+///
+/// See `artiq/firmware/libproto_artiq/rpc_proto.rs` for the expected format.
+fn format_rpc_arg<'ctx>(
+    generator: &mut dyn CodeGenerator,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    (arg, arg_ty, arg_idx): (BasicValueEnum<'ctx>, Type, usize),
+) -> PointerValue<'ctx> {
+    let llvm_i8 = ctx.ctx.i8_type();
+    let llvm_pi8 = llvm_i8.ptr_type(AddressSpace::default());
+
+    let arg_slot = match &*ctx.unifier.get_ty_immutable(arg_ty) {
+        TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
+            // NAC3: NDArray = { usize, usize*, T* }
+            // libproto_artiq: NDArray = [data[..], dim_sz[..]]
+
+            let ndarray = AnyObject { ty: arg_ty, value: arg };
+            let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
+
+            let dtype = ctx.get_llvm_type(generator, ndarray.dtype);
+            let ndims = ndarray.ndims_llvm(generator, ctx.ctx);
+
+            // `ndarray.data` is possibly not contiguous, and we need it to be contiguous for
+            // the reader.
+            let carray = ndarray.make_contiguous_ndarray(generator, ctx, Any(dtype));
+
+            let sizeof_sizet = Int(SizeT).sizeof(generator, ctx.ctx);
+            let sizeof_sizet = Int(SizeT).truncate_or_bit_cast(generator, ctx, sizeof_sizet);
+
+            let sizeof_pdata = Ptr(Any(dtype)).sizeof(generator, ctx.ctx);
+            let sizeof_pdata = Int(SizeT).truncate_or_bit_cast(generator, ctx, sizeof_pdata);
+
+            let sizeof_buf_shape = sizeof_sizet.mul(ctx, ndims);
+            let sizeof_buf = sizeof_buf_shape.add(ctx, sizeof_pdata);
+
+            // buf = { data: void*, shape: [size_t; ndims]; }
+            let buf = Int(Byte).array_alloca(generator, ctx, sizeof_buf.value);
+            let buf_data = buf;
+            let buf_shape = buf_data.offset(ctx, sizeof_pdata.value);
+
+            // Write to `buf->data`
+            let carray_data = carray.get(generator, ctx, |f| f.data); // has type Ptr<Any>
+            let carray_data = carray_data.pointer_cast(generator, ctx, Int(Byte));
+            buf_data.copy_from(generator, ctx, carray_data, sizeof_pdata.value);
+
+            // Write to `buf->shape`
+            let carray_shape = ndarray.instance.get(generator, ctx, |f| f.shape);
+            let carray_shape_i8 = carray_shape.pointer_cast(generator, ctx, Int(Byte));
+            buf_shape.copy_from(generator, ctx, carray_shape_i8, sizeof_buf_shape.value);
+
+            buf.value
+        }
+
+        _ => {
+            let arg_slot = generator
+                .gen_var_alloc(ctx, arg.get_type(), Some(&format!("rpc.arg{arg_idx}")))
+                .unwrap();
+            ctx.builder.build_store(arg_slot, arg).unwrap();
+
+            ctx.builder
+                .build_bitcast(arg_slot, llvm_pi8, "rpc.arg")
+                .map(BasicValueEnum::into_pointer_value)
+                .unwrap()
+        }
+    };
+
+    debug_assert_eq!(arg_slot.get_type(), llvm_pi8);
+
+    arg_slot
+}
+
 fn rpc_codegen_callback_fn<'ctx>(
     ctx: &mut CodeGenContext<'ctx, '_>,
     obj: Option<(Type, ValueEnum<'ctx>)>,
@@ -433,10 +516,10 @@ fn rpc_codegen_callback_fn<'ctx>(
     args: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
     generator: &mut dyn CodeGenerator,
 ) -> Result<Option<BasicValueEnum<'ctx>>, String> {
-    let ptr_type = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
-    let size_type = generator.get_size_type(ctx.ctx);
     let int8 = ctx.ctx.i8_type();
     let int32 = ctx.ctx.i32_type();
+    let size_type = generator.get_size_type(ctx.ctx);
+    let ptr_type = int8.ptr_type(AddressSpace::default());
     let tag_ptr_type = ctx.ctx.struct_type(&[ptr_type.into(), size_type.into()], false);
 
     let service_id = int32.const_int(fun.1 .0 as u64, false);
@@ -509,22 +592,25 @@ fn rpc_codegen_callback_fn<'ctx>(
         .0
         .args
         .iter()
-        .map(|arg| mapping.remove(&arg.name).unwrap().to_basic_value_enum(ctx, generator, arg.ty))
+        .map(|arg| {
-        .collect::<Result<Vec<_>, _>>()?;
+            mapping
+                .remove(&arg.name)
+                .unwrap()
+                .to_basic_value_enum(ctx, generator, arg.ty)
+                .map(|llvm_val| (llvm_val, arg.ty))
+        })
+        .collect::<Result<Vec<(_, _)>, _>>()?;
     if let Some(obj) = obj {
-        if let ValueEnum::Static(obj) = obj.1 {
+        if let ValueEnum::Static(obj_val) = obj.1 {
-            real_params.insert(0, obj.get_const_obj(ctx, generator));
+            real_params.insert(0, (obj_val.get_const_obj(ctx, generator), obj.0));
         } else {
             // should be an error here...
             panic!("only host object is allowed");
         }
     }
 
-    for (i, arg) in real_params.iter().enumerate() {
+    for (i, (arg, arg_ty)) in real_params.iter().enumerate() {
-        let arg_slot =
+        let arg_slot = format_rpc_arg(generator, ctx, (*arg, *arg_ty, i));
-            generator.gen_var_alloc(ctx, arg.get_type(), Some(&format!("rpc.arg{i}"))).unwrap();
-        ctx.builder.build_store(arg_slot, *arg).unwrap();
-        let arg_slot = ctx.builder.build_bitcast(arg_slot, ptr_type, "rpc.arg").unwrap();
         let arg_ptr = unsafe {
             ctx.builder.build_gep(
                 args_ptr,
@@ -700,6 +786,7 @@ pub fn attributes_writeback(
                     name: i.to_string().into(),
                     ty: *ty,
                     default_value: None,
+                    is_vararg: false,
                 })
                 .collect(),
             ret: ctx.primitives.none,
@@ -723,3 +810,470 @@ pub fn rpc_codegen_callback() -> Arc<GenCall> {
         rpc_codegen_callback_fn(ctx, obj, fun, args, generator)
     })))
 }
+
+/// Returns the `fprintf` format constant for the given [`llvm_int_t`][`IntType`] on a platform with
+/// [`llvm_usize`] as its native word size.
+///
+/// Note that, similar to format constants in `<inttypes.h>`, these constants need to be prepended
+/// with `%`.
+#[must_use]
+fn get_fprintf_format_constant<'ctx>(
+    llvm_usize: IntType<'ctx>,
+    llvm_int_t: IntType<'ctx>,
+    is_unsigned: bool,
+) -> String {
+    debug_assert!(matches!(llvm_usize.get_bit_width(), 8 | 16 | 32 | 64));
+
+    let conv_spec = if is_unsigned { 'u' } else { 'd' };
+
+    // https://en.cppreference.com/w/c/language/arithmetic_types
+    // Note that NAC3 does **not** support LP32 and LLP64 configurations
+    match llvm_int_t.get_bit_width() {
+        8 => format!("hh{conv_spec}"),
+        16 => format!("h{conv_spec}"),
+        32 => conv_spec.to_string(),
+        64 => format!("{}{conv_spec}", if llvm_usize.get_bit_width() == 64 { "l" } else { "ll" }),
+        _ => todo!(
+            "Not yet implemented for i{} on {}-bit platform",
+            llvm_int_t.get_bit_width(),
+            llvm_usize.get_bit_width()
+        ),
+    }
+}
+
+/// Prints one or more `values` to `core_log` or `rtio_log`.
+///
+/// * `separator` - The separator between multiple values.
+/// * `suffix` - String to terminate the printed string, if any.
+/// * `as_repr` - Whether the `repr()` output of values instead of `str()`.
+/// * `as_rtio` - Whether to print to `rtio_log` instead of `core_log`.
+fn polymorphic_print<'ctx>(
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    generator: &mut dyn CodeGenerator,
+    values: &[(Type, ValueEnum<'ctx>)],
+    separator: &str,
+    suffix: Option<&str>,
+    as_repr: bool,
+    as_rtio: bool,
+) -> Result<(), String> {
+    let printf = |ctx: &mut CodeGenContext<'ctx, '_>,
+                  generator: &mut dyn CodeGenerator,
+                  fmt: String,
+                  args: Vec<BasicValueEnum<'ctx>>| {
+        debug_assert!(!fmt.is_empty());
+        debug_assert_eq!(fmt.as_bytes().last().unwrap(), &0u8);
+
+        let fn_name = if as_rtio { "rtio_log" } else { "core_log" };
+        let print_fn = ctx.module.get_function(fn_name).unwrap_or_else(|| {
+            let llvm_pi8 = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
+            let fn_t = if as_rtio {
+                let llvm_void = ctx.ctx.void_type();
+                llvm_void.fn_type(&[llvm_pi8.into()], true)
+            } else {
+                let llvm_i32 = ctx.ctx.i32_type();
+                llvm_i32.fn_type(&[llvm_pi8.into()], true)
+            };
+            ctx.module.add_function(fn_name, fn_t, None)
+        });
+
+        let fmt = ctx.gen_string(generator, fmt);
+        let fmt = fmt.get_field(generator, ctx.ctx, |f| f.base);
+
+        ctx.builder
+            .build_call(
+                print_fn,
+                &once(fmt.value.into()).chain(args).map(BasicValueEnum::into).collect_vec(),
+                "",
+            )
+            .unwrap();
+    };
+
+    let llvm_i32 = ctx.ctx.i32_type();
+    let llvm_i64 = ctx.ctx.i64_type();
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+
+    let suffix = suffix.unwrap_or_default();
+
+    let mut fmt = String::new();
+    let mut args = Vec::new();
+
+    let flush = |ctx: &mut CodeGenContext<'ctx, '_>,
+                 generator: &mut dyn CodeGenerator,
+                 fmt: &mut String,
+                 args: &mut Vec<BasicValueEnum<'ctx>>| {
+        if !fmt.is_empty() {
+            fmt.push('\0');
+            printf(ctx, generator, mem::take(fmt), mem::take(args));
+        }
+    };
+
+    for (ty, value) in values {
+        let ty = *ty;
+        let value = value.clone().to_basic_value_enum(ctx, generator, ty).unwrap();
+
+        if !fmt.is_empty() {
+            fmt.push_str(separator);
+        }
+
+        match &*ctx.unifier.get_ty_immutable(ty) {
+            TypeEnum::TTuple { ty: tys, is_vararg_ctx: false } => {
+                let pvalue = {
+                    let pvalue = generator.gen_var_alloc(ctx, value.get_type(), None).unwrap();
+                    ctx.builder.build_store(pvalue, value).unwrap();
+                    pvalue
+                };
+
+                fmt.push('(');
+                flush(ctx, generator, &mut fmt, &mut args);
+
+                let tuple_vals = tys
+                    .iter()
+                    .enumerate()
+                    .map(|(i, ty)| {
+                        (*ty, {
+                            let pfield =
+                                ctx.builder.build_struct_gep(pvalue, i as u32, "").unwrap();
+
+                            ValueEnum::from(ctx.builder.build_load(pfield, "").unwrap())
+                        })
+                    })
+                    .collect_vec();
+
+                polymorphic_print(ctx, generator, &tuple_vals, ", ", None, true, as_rtio)?;
+
+                if tuple_vals.len() == 1 {
+                    fmt.push_str(",)");
+                } else {
+                    fmt.push(')');
+                }
+            }
+
+            TypeEnum::TFunc { .. } => todo!(),
+            TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::None.id() => {
+                fmt.push_str("None");
+            }
+
+            TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Bool.id() => {
+                fmt.push_str("%.*s");
+
+                let true_str = ctx.gen_string(generator, "True");
+                let false_str = ctx.gen_string(generator, "False");
+
+                let true_data = true_str.get_field(generator, ctx.ctx, |f| f.base);
+                let true_len = true_str.get_field(generator, ctx.ctx, |f| f.len);
+
+                let false_data = false_str.get_field(generator, ctx.ctx, |f| f.base);
+                let false_len = false_str.get_field(generator, ctx.ctx, |f| f.len);
+
+                let bool_val = generator.bool_to_i1(ctx, value.into_int_value());
+
+                args.extend([
+                    ctx.builder
+                        .build_select(bool_val, true_len.value, false_len.value, "")
+                        .unwrap(),
+                    ctx.builder
+                        .build_select(bool_val, true_data.value, false_data.value, "")
+                        .unwrap(),
+                ]);
+            }
+
+            TypeEnum::TObj { obj_id, .. }
+                if *obj_id == PrimDef::Int32.id()
+                    || *obj_id == PrimDef::Int64.id()
+                    || *obj_id == PrimDef::UInt32.id()
+                    || *obj_id == PrimDef::UInt64.id() =>
+            {
+                let is_unsigned =
+                    *obj_id == PrimDef::UInt32.id() || *obj_id == PrimDef::UInt64.id();
+
+                let llvm_int_t = value.get_type().into_int_type();
+                debug_assert!(matches!(llvm_usize.get_bit_width(), 32 | 64));
+                debug_assert!(matches!(llvm_int_t.get_bit_width(), 32 | 64));
+
+                let fmt_spec = format!(
+                    "%{}",
+                    get_fprintf_format_constant(llvm_usize, llvm_int_t, is_unsigned)
+                );
+
+                fmt.push_str(fmt_spec.as_str());
+                args.push(value);
+            }
+
+            TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Float.id() => {
+                fmt.push_str("%g");
+                args.push(value);
+            }
+
+            TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Str.id() => {
+                if as_repr {
+                    fmt.push_str("\"%.*s\"");
+                } else {
+                    fmt.push_str("%.*s");
+                }
+
+                let str = str_model().check_value(generator, ctx.ctx, value).unwrap();
+
+                let str_data = str.get_field(generator, ctx.ctx, |f| f.base);
+                let str_len = str.get_field(generator, ctx.ctx, |f| f.len);
+
+                args.extend(&[str_len.value.into(), str_data.value.into()]);
+            }
+
+            TypeEnum::TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
+                let elem_ty = *params.iter().next().unwrap().1;
+
+                fmt.push('[');
+                flush(ctx, generator, &mut fmt, &mut args);
+
+                let val = ListValue::from_ptr_val(value.into_pointer_value(), llvm_usize, None);
+                let len = val.load_size(ctx, None);
+                let last =
+                    ctx.builder.build_int_sub(len, llvm_usize.const_int(1, false), "").unwrap();
+
+                gen_for_callback_incrementing(
+                    generator,
+                    ctx,
+                    None,
+                    llvm_usize.const_zero(),
+                    (len, false),
+                    |generator, ctx, _, i| {
+                        let elem = unsafe { val.data().get_unchecked(ctx, generator, &i, None) };
+
+                        polymorphic_print(
+                            ctx,
+                            generator,
+                            &[(elem_ty, elem.into())],
+                            "",
+                            None,
+                            true,
+                            as_rtio,
+                        )?;
+
+                        gen_if_callback(
+                            generator,
+                            ctx,
+                            |_, ctx| {
+                                Ok(ctx
+                                    .builder
+                                    .build_int_compare(IntPredicate::ULT, i, last, "")
+                                    .unwrap())
+                            },
+                            |generator, ctx| {
+                                printf(ctx, generator, ", \0".into(), Vec::default());
+
+                                Ok(())
+                            },
+                            |_, _| Ok(()),
+                        )?;
+
+                        Ok(())
+                    },
+                    llvm_usize.const_int(1, false),
+                )?;
+
+                fmt.push(']');
+                flush(ctx, generator, &mut fmt, &mut args);
+            }
+
+            TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
+                fmt.push_str("array([");
+                flush(ctx, generator, &mut fmt, &mut args);
+
+                let ndarray = AnyObject { ty, value };
+                let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
+
+                let num_0 = Int(SizeT).const_0(generator, ctx.ctx);
+
+                // Print `ndarray` as a flat list delimited by interspersed with ", \0"
+                ndarray.foreach(generator, ctx, |generator, ctx, _, hdl| {
+                    let i = hdl.get_index(generator, ctx);
+                    let scalar = hdl.get_scalar(generator, ctx);
+
+                    // if (i != 0) { puts(", "); }
+                    gen_if_callback(
+                        generator,
+                        ctx,
+                        |_, ctx| {
+                            let not_first = i.compare(ctx, IntPredicate::NE, num_0);
+                            Ok(not_first.value)
+                        },
+                        |generator, ctx| {
+                            printf(ctx, generator, ", \0".into(), Vec::default());
+                            Ok(())
+                        },
+                        |_, _| Ok(()),
+                    )?;
+
+                    // Print element
+                    polymorphic_print(
+                        ctx,
+                        generator,
+                        &[(scalar.ty, scalar.value.into())],
+                        "",
+                        None,
+                        true,
+                        as_rtio,
+                    )?;
+                    Ok(())
+                })?;
+
+                fmt.push_str(")]");
+                flush(ctx, generator, &mut fmt, &mut args);
+            }
+
+            TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Range.id() => {
+                fmt.push_str("range(");
+                flush(ctx, generator, &mut fmt, &mut args);
+
+                let range = AnyObject { ty, value };
+                let range = RangeObject::from_object(generator, ctx, range);
+
+                let (start, stop, step) = range.instance.destructure(generator, ctx);
+                let start = start.value;
+                let stop = stop.value;
+                let step = step.value;
+
+                polymorphic_print(
+                    ctx,
+                    generator,
+                    &[
+                        (ctx.primitives.int32, start.into()),
+                        (ctx.primitives.int32, stop.into()),
+                        (ctx.primitives.int32, step.into()),
+                    ],
+                    ", ",
+                    None,
+                    false,
+                    as_rtio,
+                )?;
+
+                fmt.push(')');
+            }
+
+            TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::Exception.id() => {
+                let fmt_str = format!(
+                    "%{}(%{}, %{1:}, %{1:})",
+                    get_fprintf_format_constant(llvm_usize, llvm_i32, false),
+                    get_fprintf_format_constant(llvm_usize, llvm_i64, false),
+                );
+
+                let exn = value.into_pointer_value();
+                let name = ctx
+                    .build_in_bounds_gep_and_load(
+                        exn,
+                        &[llvm_i32.const_zero(), llvm_i32.const_zero()],
+                        None,
+                    )
+                    .into_int_value();
+                let param0 = ctx
+                    .build_in_bounds_gep_and_load(
+                        exn,
+                        &[llvm_i32.const_zero(), llvm_i32.const_int(6, false)],
+                        None,
+                    )
+                    .into_int_value();
+                let param1 = ctx
+                    .build_in_bounds_gep_and_load(
+                        exn,
+                        &[llvm_i32.const_zero(), llvm_i32.const_int(7, false)],
+                        None,
+                    )
+                    .into_int_value();
+                let param2 = ctx
+                    .build_in_bounds_gep_and_load(
+                        exn,
+                        &[llvm_i32.const_zero(), llvm_i32.const_int(8, false)],
+                        None,
+                    )
+                    .into_int_value();
+
+                fmt.push_str(fmt_str.as_str());
+                args.extend_from_slice(&[name.into(), param0.into(), param1.into(), param2.into()]);
+            }
+
+            _ => unreachable!(
+                "Unsupported object type for polymorphic_print: {}",
+                ctx.unifier.stringify(ty)
+            ),
+        }
+    }
+
+    fmt.push_str(suffix);
+    flush(ctx, generator, &mut fmt, &mut args);
+
+    Ok(())
+}
+
+/// Invokes the `core_log` intrinsic function.
+pub fn call_core_log_impl<'ctx>(
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    generator: &mut dyn CodeGenerator,
+    arg: (Type, BasicValueEnum<'ctx>),
+) -> Result<(), String> {
+    let (arg_ty, arg_val) = arg;
+
+    polymorphic_print(ctx, generator, &[(arg_ty, arg_val.into())], " ", Some("\n"), false, false)?;
+
+    Ok(())
+}
+
+/// Invokes the `rtio_log` intrinsic function.
+pub fn call_rtio_log_impl<'ctx>(
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    generator: &mut dyn CodeGenerator,
+    channel: StructValue<'ctx>,
+    arg: (Type, BasicValueEnum<'ctx>),
+) -> Result<(), String> {
+    let (arg_ty, arg_val) = arg;
+
+    polymorphic_print(
+        ctx,
+        generator,
+        &[(ctx.primitives.str, channel.into())],
+        " ",
+        Some("\x1E"),
+        false,
+        true,
+    )?;
+    polymorphic_print(ctx, generator, &[(arg_ty, arg_val.into())], " ", Some("\x1D"), false, true)?;
+
+    Ok(())
+}
+
+/// Generates a call to `core_log`.
+pub fn gen_core_log<'ctx>(
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    obj: &Option<(Type, ValueEnum<'ctx>)>,
+    fun: (&FunSignature, DefinitionId),
+    args: &[(Option<StrRef>, ValueEnum<'ctx>)],
+    generator: &mut dyn CodeGenerator,
+) -> Result<(), String> {
+    assert!(obj.is_none());
+    assert_eq!(args.len(), 1);
+
+    let value_ty = fun.0.args[0].ty;
+    let value_arg = args[0].1.clone().to_basic_value_enum(ctx, generator, value_ty)?;
+
+    call_core_log_impl(ctx, generator, (value_ty, value_arg))
+}
+
+/// Generates a call to `rtio_log`.
+pub fn gen_rtio_log<'ctx>(
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    obj: &Option<(Type, ValueEnum<'ctx>)>,
+    fun: (&FunSignature, DefinitionId),
+    args: &[(Option<StrRef>, ValueEnum<'ctx>)],
+    generator: &mut dyn CodeGenerator,
+) -> Result<(), String> {
+    assert!(obj.is_none());
+    assert_eq!(args.len(), 2);
+
+    let channel_ty = fun.0.args[0].ty;
+    assert!(ctx.unifier.unioned(channel_ty, ctx.primitives.str));
+    let channel_arg =
+        args[0].1.clone().to_basic_value_enum(ctx, generator, channel_ty)?.into_struct_value();
+    let value_ty = fun.0.args[1].ty;
+    let value_arg = args[1].1.clone().to_basic_value_enum(ctx, generator, value_ty)?;
+
+    call_rtio_log_impl(ctx, generator, channel_arg, (value_ty, value_arg))
+}

nac3artiq/src/lib.rs

@@ -24,6 +24,7 @@ use std::rc::Rc;
 use std::sync::Arc;
 
 use inkwell::{
+    context::Context,
     memory_buffer::MemoryBuffer,
     module::{Linkage, Module},
     passes::PassBuilderOptions,
@@ -32,9 +33,10 @@ use inkwell::{
     OptimizationLevel,
 };
 use itertools::Itertools;
+use nac3core::codegen::irrt::setup_irrt_exceptions;
 use nac3core::codegen::{gen_func_impl, CodeGenLLVMOptions, CodeGenTargetMachineOptions};
 use nac3core::toplevel::builtins::get_exn_constructor;
-use nac3core::typecheck::typedef::{TypeEnum, Unifier, VarMap};
+use nac3core::typecheck::typedef::{into_var_map, TypeEnum, Unifier, VarMap};
 use nac3parser::{
     ast::{ExprKind, Stmt, StmtKind, StrRef},
     parser::parse_program,
@@ -50,7 +52,7 @@ use nac3core::{
     codegen::{concrete_type::ConcreteTypeStore, CodeGenTask, WithCall, WorkerRegistry},
     symbol_resolver::SymbolResolver,
     toplevel::{
-        composer::{ComposerConfig, TopLevelComposer},
+        composer::{BuiltinFuncCreator, BuiltinFuncSpec, ComposerConfig, TopLevelComposer},
         DefinitionId, GenCall, TopLevelDef,
     },
     typecheck::typedef::{FunSignature, FuncArg},
@@ -59,13 +61,13 @@ use nac3core::{
 
 use nac3ld::Linker;
 
-use tempfile::{self, TempDir};
-
-use crate::codegen::attributes_writeback;
 use crate::{
-    codegen::{rpc_codegen_callback, ArtiqCodeGenerator},
+    codegen::{
+        attributes_writeback, gen_core_log, gen_rtio_log, rpc_codegen_callback, ArtiqCodeGenerator,
+    },
     symbol_resolver::{DeferredEvaluationStore, InnerResolver, PythonHelper, Resolver},
 };
+use tempfile::{self, TempDir};
 
 mod codegen;
 mod symbol_resolver;
@@ -126,7 +128,7 @@ struct Nac3 {
     isa: Isa,
     time_fns: &'static (dyn TimeFns + Sync),
     primitive: PrimitiveStore,
-    builtins: Vec<(StrRef, FunSignature, Arc<GenCall>)>,
+    builtins: Vec<BuiltinFuncSpec>,
     pyid_to_def: Arc<RwLock<HashMap<u64, DefinitionId>>>,
     primitive_ids: PrimitivePythonId,
     working_directory: TempDir,
@@ -264,7 +266,7 @@ impl Nac3 {
                     arg_names.len(),
                 ));
             }
-            for (i, FuncArg { ty, default_value, name }) in args.iter().enumerate() {
+            for (i, FuncArg { ty, default_value, name, .. }) in args.iter().enumerate() {
                 let in_name = match arg_names.get(i) {
                     Some(n) => n,
                     None if default_value.is_none() => {
@@ -300,6 +302,64 @@ impl Nac3 {
         None
     }
 
+    /// Returns a [`Vec`] of builtins that needs to be initialized during method compilation time.
+    fn get_lateinit_builtins() -> Vec<Box<BuiltinFuncCreator>> {
+        vec![
+            Box::new(|primitives, unifier| {
+                let arg_ty = unifier.get_fresh_var(Some("T".into()), None);
+
+                (
+                    "core_log".into(),
+                    FunSignature {
+                        args: vec![FuncArg {
+                            name: "arg".into(),
+                            ty: arg_ty.ty,
+                            default_value: None,
+                            is_vararg: false,
+                        }],
+                        ret: primitives.none,
+                        vars: into_var_map([arg_ty]),
+                    },
+                    Arc::new(GenCall::new(Box::new(move |ctx, obj, fun, args, generator| {
+                        gen_core_log(ctx, &obj, fun, &args, generator)?;
+
+                        Ok(None)
+                    }))),
+                )
+            }),
+            Box::new(|primitives, unifier| {
+                let arg_ty = unifier.get_fresh_var(Some("T".into()), None);
+
+                (
+                    "rtio_log".into(),
+                    FunSignature {
+                        args: vec![
+                            FuncArg {
+                                name: "channel".into(),
+                                ty: primitives.str,
+                                default_value: None,
+                                is_vararg: false,
+                            },
+                            FuncArg {
+                                name: "arg".into(),
+                                ty: arg_ty.ty,
+                                default_value: None,
+                                is_vararg: false,
+                            },
+                        ],
+                        ret: primitives.none,
+                        vars: into_var_map([arg_ty]),
+                    },
+                    Arc::new(GenCall::new(Box::new(move |ctx, obj, fun, args, generator| {
+                        gen_rtio_log(ctx, &obj, fun, &args, generator)?;
+
+                        Ok(None)
+                    }))),
+                )
+            }),
+        ]
+    }
+
     fn compile_method<T>(
         &self,
         obj: &PyAny,
@@ -312,6 +372,7 @@ impl Nac3 {
         let size_t = self.isa.get_size_type();
         let (mut composer, mut builtins_def, mut builtins_ty) = TopLevelComposer::new(
             self.builtins.clone(),
+            Self::get_lateinit_builtins(),
             ComposerConfig { kernel_ann: Some("Kernel"), kernel_invariant_ann: "KernelInvariant" },
             size_t,
         );
@@ -497,6 +558,11 @@ impl Nac3 {
             .register_top_level(synthesized.pop().unwrap(), Some(resolver.clone()), "", false)
             .unwrap();
 
+        // Process IRRT
+        let context = inkwell::context::Context::create();
+        let irrt = load_irrt(&context);
+        setup_irrt_exceptions(&context, &irrt, resolver.as_ref());
+
         let fun_signature =
             FunSignature { args: vec![], ret: self.primitive.none, vars: VarMap::new() };
         let mut store = ConcreteTypeStore::new();
@@ -625,7 +691,9 @@ impl Nac3 {
             let buffer = buffer.as_slice().into();
             membuffer.lock().push(buffer);
         })));
-        let size_t = if self.isa == Isa::Host { 64 } else { 32 };
+        let size_t = Context::create()
+            .ptr_sized_int_type(&self.get_llvm_target_machine().get_target_data(), None)
+            .get_bit_width();
         let num_threads = if is_multithreaded() { 4 } else { 1 };
         let thread_names: Vec<String> = (0..num_threads).map(|_| "main".to_string()).collect();
         let threads: Vec<_> = thread_names
@@ -644,6 +712,9 @@ impl Nac3 {
                 ArtiqCodeGenerator::new("attributes_writeback".to_string(), size_t, self.time_fns);
             let context = inkwell::context::Context::create();
             let module = context.create_module("attributes_writeback");
+            let target_machine = self.llvm_options.create_target_machine().unwrap();
+            module.set_data_layout(&target_machine.get_target_data().get_data_layout());
+            module.set_triple(&target_machine.get_triple());
             let builder = context.create_builder();
             let (_, module, _) = gen_func_impl(
                 &context,
@@ -662,7 +733,7 @@ impl Nac3 {
             membuffer.lock().push(buffer);
         });
 
-        let context = inkwell::context::Context::create();
+        // Link all modules into `main`.
         let buffers = membuffers.lock();
         let main = context
             .create_module_from_ir(MemoryBuffer::create_from_memory_range(&buffers[0], "main"))
@@ -691,8 +762,7 @@ impl Nac3 {
             )
             .unwrap();
 
-        main.link_in_module(load_irrt(&context))
+        main.link_in_module(irrt).map_err(|err| CompileError::new_err(err.to_string()))?;
-            .map_err(|err| CompileError::new_err(err.to_string()))?;
 
         let mut function_iter = main.get_first_function();
         while let Some(func) = function_iter {
@@ -847,7 +917,7 @@ impl Nac3 {
             Isa::RiscV32IMA => &timeline::NOW_PINNING_TIME_FNS,
             Isa::CortexA9 | Isa::Host => &timeline::EXTERN_TIME_FNS,
         };
-        let primitive: PrimitiveStore = TopLevelComposer::make_primitives(isa.get_size_type()).0;
+        let (primitive, _) = TopLevelComposer::make_primitives(isa.get_size_type());
         let builtins = vec![
             (
                 "now_mu".into(),
@@ -863,6 +933,7 @@ impl Nac3 {
                         name: "t".into(),
                         ty: primitive.int64,
                         default_value: None,
+                        is_vararg: false,
                     }],
                     ret: primitive.none,
                     vars: VarMap::new(),
@@ -882,6 +953,7 @@ impl Nac3 {
                         name: "dt".into(),
                         ty: primitive.int64,
                         default_value: None,
+                        is_vararg: false,
                     }],
                     ret: primitive.none,
                     vars: VarMap::new(),

nac3artiq/src/symbol_resolver.rs

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

nac3core/Cargo.toml

@@ -1,12 +1,12 @@
-[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"
@@ -14,8 +14,8 @@ indexmap = "2.2"
 parking_lot = "0.12"
 rayon = "1.8"
 nac3parser = { path = "../nac3parser" }
-strum = "0.26.2"
+strum = "0.26"
-strum_macros = "0.26.4"
+strum_macros = "0.26"
 
 [dependencies.inkwell]
 version = "0.4"

nac3core/build.rs

@@ -7,39 +7,51 @@ use std::{
     process::{Command, Stdio},
 };
 
-fn compile_irrt(irrt_dir: &Path, out_dir: &Path) {
+fn main() {
+    // Define relevant directories
+    let out_dir = env::var("OUT_DIR").unwrap();
+    let out_dir = Path::new(&out_dir);
+    let irrt_dir = Path::new("irrt");
+
     let irrt_cpp_path = irrt_dir.join("irrt.cpp");
 
     /*
      * HACK: Sadly, clang doesn't let us emit generic LLVM bitcode.
      * Compiling for WASM32 and filtering the output with regex is the closest we can get.
      */
-    let flags: &[&str] = &[
+    let mut flags: Vec<&str> = vec![
         "--target=wasm32",
-        irrt_cpp_path.to_str().unwrap(),
         "-x",
         "c++",
         "-fno-discard-value-names",
         "-fno-exceptions",
         "-fno-rtti",
-        match env::var("PROFILE").as_deref() {
-            Ok("debug") => "-O0",
-            Ok("release") => "-O3",
-            flavor => panic!("Unknown or missing build flavor {flavor:?}"),
-        },
         "-emit-llvm",
         "-S",
         "-Wall",
         "-Wextra",
-        "-Werror=return-type",
-        "-I",
-        irrt_dir.to_str().unwrap(),
         "-o",
         "-",
+        "-I",
+        irrt_dir.to_str().unwrap(),
+        irrt_cpp_path.to_str().unwrap(),
     ];
 
-    println!("cargo:rerun-if-changed={}", out_dir.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")
         .args(flags)
         .output()
@@ -53,11 +65,17 @@ fn compile_irrt(irrt_dir: &Path, out_dir: &Path) {
     let output = std::str::from_utf8(&output.stdout).unwrap().replace("\r\n", "\n");
     let mut filtered_output = String::with_capacity(output.len());
 
-    // (?ms:^define.*?\}$) to capture `define` blocks
+    // Filter out irrelevant IR
-    // (?m:^declare.*?$) to capture `declare` blocks
+    //
-    // (?m:^%.+?=\s*type\s*\{.+?\}$) to capture `type` declarations
+    // Regex:
-    let regex_filter =
+    // - `(?ms:^define.*?\}$)` captures LLVM `define` blocks
-        Regex::new(r"(?ms:^define.*?\}$)|(?m:^declare.*?$)|(?m:^%.+?=\s*type\s*\{.+?\}$)").unwrap();
+    // - `(?m:^declare.*?$)` captures LLVM `declare` lines
+    // - `(?m:^%.+?=\s*type\s*\{.+?\}$)` captures LLVM `type` declarations
+    // - `(?m:^@.+?=.+$)` captures global constants
+    let regex_filter = Regex::new(
+        r"(?ms:^define.*?\}$)|(?m:^declare.*?$)|(?m:^%.+?=\s*type\s*\{.+?\}$)|(?m:^@.+?=.+$)",
+    )
+    .unwrap();
     for f in regex_filter.captures_iter(&output) {
         assert_eq!(f.len(), 1);
         filtered_output.push_str(&f[0]);
@@ -68,14 +86,20 @@ fn compile_irrt(irrt_dir: &Path, out_dir: &Path) {
         .unwrap()
         .replace_all(&filtered_output, "");
 
-    println!("cargo:rerun-if-env-changed=DEBUG_DUMP_IRRT");
+    // For debugging
-    if env::var("DEBUG_DUMP_IRRT").is_ok() {
+    // Doing `DEBUG_DUMP_IRRT=1 cargo build -p nac3core` dumps the LLVM IR generated
+    const DEBUG_DUMP_IRRT: &str = "DEBUG_DUMP_IRRT";
+    println!("cargo:rerun-if-env-changed={DEBUG_DUMP_IRRT}");
+    if env::var(DEBUG_DUMP_IRRT).is_ok() {
         let mut file = File::create(out_dir.join("irrt.ll")).unwrap();
         file.write_all(output.as_bytes()).unwrap();
+
         let mut file = File::create(out_dir.join("irrt-filtered.ll")).unwrap();
         file.write_all(filtered_output.as_bytes()).unwrap();
     }
 
+    // Assemble the emitted and filtered IR to .bc
+    // That .bc will be integrated into nac3core's codegen
     let mut llvm_as = Command::new("llvm-as-irrt")
         .stdin(Stdio::piped())
         .arg("-o")
@@ -85,50 +109,3 @@ fn compile_irrt(irrt_dir: &Path, out_dir: &Path) {
     llvm_as.stdin.as_mut().unwrap().write_all(filtered_output.as_bytes()).unwrap();
     assert!(llvm_as.wait().unwrap().success());
 }
-
-fn compile_irrt_test(irrt_dir: &Path, out_dir: &Path) {
-    let irrt_test_cpp_path = irrt_dir.join("irrt_test.cpp");
-    let exe_path = out_dir.join("irrt_test.out");
-
-    let flags: &[&str] = &[
-        irrt_test_cpp_path.to_str().unwrap(),
-        "-x",
-        "c++",
-        "-I",
-        irrt_dir.to_str().unwrap(),
-        "-g",
-        "-fno-discard-value-names",
-        "-O0",
-        "-Wall",
-        "-Wextra",
-        "-Werror=return-type",
-        "-lm", // for `tgamma()`, `lgamma()`
-        "-o",
-        exe_path.to_str().unwrap(),
-    ];
-
-    Command::new("clang-irrt-test")
-        .args(flags)
-        .output()
-        .map(|o| {
-            assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
-            o
-        })
-        .unwrap();
-    println!("cargo:rerun-if-changed={}", out_dir.to_str().unwrap());
-}
-
-fn main() {
-    let out_dir = env::var("OUT_DIR").unwrap();
-    let out_dir = Path::new(&out_dir);
-
-    let irrt_dir = Path::new("./irrt");
-
-    compile_irrt(irrt_dir, out_dir);
-
-    // https://github.com/rust-lang/cargo/issues/2549
-    // `cargo test -F test` to also build `irrt_test.cpp
-    if cfg!(feature = "test") {
-        compile_irrt_test(irrt_dir, out_dir);
-    }
-}

nac3core/irrt/irrt.cpp

@@ -1,5 +1,16 @@
-#include "irrt_everything.hpp"
+#include <irrt/exception.hpp>
-
+#include <irrt/int_types.hpp>
-/*
+#include <irrt/list.hpp>
-    This file will be read by `clang-irrt` to conveniently produce LLVM IR for `nac3core/codegen`.
+#include <irrt/math_util.hpp>
-*/
+#include <irrt/ndarray/array.hpp>
+#include <irrt/ndarray/basic.hpp>
+#include <irrt/ndarray/broadcast.hpp>
+#include <irrt/ndarray/def.hpp>
+#include <irrt/ndarray/indexing.hpp>
+#include <irrt/ndarray/iter.hpp>
+#include <irrt/ndarray/matmul.hpp>
+#include <irrt/ndarray/reshape.hpp>
+#include <irrt/ndarray/transpose.hpp>
+#include <irrt/original.hpp>
+#include <irrt/range.hpp>
+#include <irrt/slice.hpp>

nac3core/irrt/irrt.hpp

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

nac3core/irrt/irrt/cslice.hpp

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

nac3core/irrt/irrt/cstr_util.hpp

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

nac3core/irrt/irrt/debug.hpp

@@ -0,0 +1,22 @@
+#pragma once
+
+#ifdef IRRT_DEBUG_ASSERT
+#define IRRT_DEBUG_ASSERT_BOOL true
+#else
+#define IRRT_DEBUG_ASSERT_BOOL false
+#endif
+
+#define raise_debug_assert(SizeT, msg, param1, param2, param3)                                                         \
+    raise_exception(SizeT, EXN_ASSERTION_ERROR, "IRRT debug assert failed: " msg, param1, param2, param3);
+
+#define debug_assert_eq(SizeT, lhs, rhs)                                                                               \
+    if (IRRT_DEBUG_ASSERT_BOOL && (lhs) != (rhs))                                                                      \
+    {                                                                                                                  \
+        raise_debug_assert(SizeT, "LHS = {0}. RHS = {1}", lhs, rhs, NO_PARAM);                                         \
+    }
+
+#define debug_assert(SizeT, expr)                                                                                      \
+    if (IRRT_DEBUG_ASSERT_BOOL && !(expr))                                                                             \
+    {                                                                                                                  \
+        raise_debug_assert(SizeT, "Got false.", NO_PARAM, NO_PARAM, NO_PARAM);                                         \
+    }

nac3core/irrt/irrt/exception.hpp

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

nac3core/irrt/irrt/int_types.hpp

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

nac3core/irrt/irrt/list.hpp

@@ -0,0 +1,59 @@
+#pragma once
+
+#include <irrt/debug.hpp>
+#include <irrt/exception.hpp>
+#include <irrt/int_types.hpp>
+#include <irrt/slice.hpp>
+
+namespace
+{
+/**
+ * @brief A list in NAC3.
+ *
+ * The `items` field is opaque. You must rely on external contexts to
+ * know how to interpret it.
+ */
+template <typename SizeT> struct List
+{
+    uint8_t *items;
+    SizeT len;
+};
+
+namespace list
+{
+template <typename SizeT> void range_assign(List<SizeT> *dst, SizeT itemsize, Range<SizeT> *range, List<SizeT> *src)
+{
+    debug_assert(range->step != 0);
+    SizeT assign_len = range->len();
+
+    if (assign_len < src->len)
+    {
+        // Encountered things like
+        // ```
+        // xs = [1, 2, 3, 4, 5]
+        // xs[1:3] = [999, 1000, 1001, 1002] # Note that step has to be 1.
+        // xs = [1, 999, 1000, 1001, 1002, 4, 5] # xs is longer
+        // ```
+        //
+        // We do not support extending lists since that requires allocation.
+        raise_exception(SizeT, EXN_VALUE_ERROR,
+                        "List assignment does not support list extension. Attempting to assign {0} item(s) into a "
+                        "space of {1} item(s).",
+                        src->len, assign_len, NO_PARAM);
+    }
+
+    if (range->step == 1)
+    {
+        // Assigning into a contiguous region. Optimized with memmove.
+
+        uint8_t* p1 = dst->items + range->start * itemsize;
+        uint8_t* p2 = dst->items + range->start * itemsize + assign_len * itemsize;
+
+        __builtin_memmove(cursor, src->items, assign_len * itemsize);
+        cursor += range_len * itemsize;
+
+        __builtin_memmove(cursor, );
+    }
+}
+} // namespace list
+} // namespace

nac3core/irrt/irrt/math_util.hpp

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

nac3core/irrt/irrt/ndarray/array.hpp

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

nac3core/irrt/irrt/ndarray/basic.hpp

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

nac3core/irrt/irrt/ndarray/broadcast.hpp

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

nac3core/irrt/irrt/ndarray/def.hpp

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

nac3core/irrt/irrt/ndarray/indexing.hpp

@@ -0,0 +1,243 @@
+#pragma once
+
+#include <irrt/exception.hpp>
+#include <irrt/int_types.hpp>
+#include <irrt/ndarray/basic.hpp>
+#include <irrt/ndarray/def.hpp>
+#include <irrt/range.hpp>
+#include <irrt/slice.hpp>
+
+namespace
+{
+typedef uint8_t NDIndexType;
+
+/**
+ * @brief A single element index
+ *
+ * `data` points to a `int32_t`.
+ */
+
+const NDIndexType ND_INDEX_TYPE_SINGLE_ELEMENT = 0;
+/**
+ * @brief A slice index
+ *
+ * `data` points to a `Slice<int32_t>`.
+ */
+const NDIndexType ND_INDEX_TYPE_SLICE = 1;
+
+/**
+ * @brief `np.newaxis` / `None`
+ *
+ * `data` is unused.
+ */
+const NDIndexType ND_INDEX_TYPE_NEWAXIS = 2;
+
+/**
+ * @brief `Ellipsis` / `...`
+ *
+ * `data` is unused.
+ */
+const NDIndexType ND_INDEX_TYPE_ELLIPSIS = 3;
+
+/**
+ * @brief An index used in ndarray indexing
+ */
+struct NDIndex
+{
+    /**
+     * @brief Enum tag to specify the type of index.
+     *
+     * Please see comments of each enum constant.
+     */
+    NDIndexType type;
+
+    /**
+     * @brief The accompanying data associated with `type`.
+     *
+     * Please see comments of each enum constant.
+     */
+    uint8_t *data;
+};
+} // namespace
+
+namespace
+{
+namespace ndarray
+{
+namespace indexing
+{
+/**
+ * @brief Perform ndarray "basic indexing" (https://numpy.org/doc/stable/user/basics.indexing.html#basic-indexing)
+ *
+ * This function is very similar to performing `dst_ndarray = src_ndarray[indices]` in Python.
+ *
+ * This function also does proper assertions on `indices` to check for out of bounds access.
+ *
+ * # Notes on `dst_ndarray`
+ * The caller is responsible for allocating space for the resulting ndarray.
+ * Here is what this function expects from `dst_ndarray` when called:
+ *   - `dst_ndarray->data` does not have to be initialized.
+ *   - `dst_ndarray->itemsize` does not have to be initialized.
+ *   - `dst_ndarray->ndims` must be initialized, and it must be equal to the expected `ndims` of the `dst_ndarray` after
+ *       indexing `src_ndarray` with `indices`.
+ *   - `dst_ndarray->shape` must be allocated, through it can contain uninitialized values.
+ *   - `dst_ndarray->strides` must be allocated, through it can contain uninitialized values.
+ * When this function call ends:
+ *   - `dst_ndarray->data` is set to `src_ndarray->data` (`dst_ndarray` is just a view to `src_ndarray`)
+ *   - `dst_ndarray->itemsize` is set to `src_ndarray->itemsize`
+ *   - `dst_ndarray->ndims` is unchanged.
+ *   - `dst_ndarray->shape` is updated according to how `src_ndarray` is indexed.
+ *   - `dst_ndarray->strides` is updated accordingly by how ndarray indexing works.
+ *
+ * @param indices indices to index `src_ndarray`, ordered in the same way you would write them in Python.
+ * @param src_ndarray The NDArray to be indexed.
+ * @param dst_ndarray The resulting NDArray after indexing. Further details in the comments above,
+ */
+template <typename SizeT>
+void index(SizeT num_indices, const NDIndex *indices, const NDArray<SizeT> *src_ndarray, NDArray<SizeT> *dst_ndarray)
+{
+    // Validate `indices`.
+
+    // Expected value of `dst_ndarray->ndims`.
+    SizeT expected_dst_ndims = src_ndarray->ndims;
+    // To check for "too many indices for array: array is ?-dimensional, but ? were indexed"
+    SizeT num_indexed = 0;
+    // There may be ellipsis `...` in `indices`. There can only be 0 or 1 ellipsis.
+    SizeT num_ellipsis = 0;
+
+    for (SizeT i = 0; i < num_indices; i++)
+    {
+        if (indices[i].type == ND_INDEX_TYPE_SINGLE_ELEMENT)
+        {
+            expected_dst_ndims--;
+            num_indexed++;
+        }
+        else if (indices[i].type == ND_INDEX_TYPE_SLICE)
+        {
+            num_indexed++;
+        }
+        else if (indices[i].type == ND_INDEX_TYPE_NEWAXIS)
+        {
+            expected_dst_ndims++;
+        }
+        else if (indices[i].type == ND_INDEX_TYPE_ELLIPSIS)
+        {
+            num_ellipsis++;
+            if (num_ellipsis > 1)
+            {
+                raise_exception(SizeT, EXN_INDEX_ERROR, "an index can only have a single ellipsis ('...')", NO_PARAM,
+                                NO_PARAM, NO_PARAM);
+            }
+        }
+        else
+        {
+            __builtin_unreachable();
+        }
+    }
+
+    debug_assert_eq(SizeT, expected_dst_ndims, dst_ndarray->ndims);
+
+    if (src_ndarray->ndims - num_indexed < 0)
+    {
+        raise_exception(SizeT, EXN_INDEX_ERROR,
+                        "too many indices for array: array is {0}-dimensional, "
+                        "but {1} were indexed",
+                        src_ndarray->ndims, num_indices, NO_PARAM);
+    }
+
+    dst_ndarray->data = src_ndarray->data;
+    dst_ndarray->itemsize = src_ndarray->itemsize;
+
+    // Reference code: https://github.com/wadetb/tinynumpy/blob/0d23d22e07062ffab2afa287374c7b366eebdda1/tinynumpy/tinynumpy.py#L652
+    SizeT src_axis = 0;
+    SizeT dst_axis = 0;
+
+    for (int32_t i = 0; i < num_indices; i++)
+    {
+        const NDIndex *index = &indices[i];
+        if (index->type == ND_INDEX_TYPE_SINGLE_ELEMENT)
+        {
+            SizeT input = (SizeT) * ((int32_t *)index->data);
+            SizeT k = slice::resolve_index_in_length(src_ndarray->shape[src_axis], input);
+
+            if (k == -1)
+            {
+                raise_exception(SizeT, EXN_INDEX_ERROR,
+                                "index {0} is out of bounds for axis {1} "
+                                "with size {2}",
+                                input, src_axis, src_ndarray->shape[src_axis]);
+            }
+
+            dst_ndarray->data += k * src_ndarray->strides[src_axis];
+
+            src_axis++;
+        }
+        else if (index->type == ND_INDEX_TYPE_SLICE)
+        {
+            Slice<int32_t> *slice = (Slice<int32_t> *)index->data;
+
+            Range<int32_t> range = slice->indices_checked<SizeT>(src_ndarray->shape[src_axis]);
+
+            dst_ndarray->data += (SizeT)range.start * src_ndarray->strides[src_axis];
+            dst_ndarray->strides[dst_axis] = ((SizeT)range.step) * src_ndarray->strides[src_axis];
+            dst_ndarray->shape[dst_axis] = (SizeT)range.len<SizeT>();
+
+            dst_axis++;
+            src_axis++;
+        }
+        else if (index->type == ND_INDEX_TYPE_NEWAXIS)
+        {
+            dst_ndarray->strides[dst_axis] = 0;
+            dst_ndarray->shape[dst_axis] = 1;
+
+            dst_axis++;
+        }
+        else if (index->type == ND_INDEX_TYPE_ELLIPSIS)
+        {
+            // The number of ':' entries this '...' implies.
+            SizeT ellipsis_size = src_ndarray->ndims - num_indexed;
+
+            for (SizeT j = 0; j < ellipsis_size; j++)
+            {
+                dst_ndarray->strides[dst_axis] = src_ndarray->strides[src_axis];
+                dst_ndarray->shape[dst_axis] = src_ndarray->shape[src_axis];
+
+                dst_axis++;
+                src_axis++;
+            }
+        }
+        else
+        {
+            __builtin_unreachable();
+        }
+    }
+
+    for (; dst_axis < dst_ndarray->ndims; dst_axis++, src_axis++)
+    {
+        dst_ndarray->shape[dst_axis] = src_ndarray->shape[src_axis];
+        dst_ndarray->strides[dst_axis] = src_ndarray->strides[src_axis];
+    }
+
+    debug_assert_eq(SizeT, src_ndarray->ndims, src_axis);
+    debug_assert_eq(SizeT, dst_ndarray->ndims, dst_axis);
+}
+} // namespace indexing
+} // namespace ndarray
+} // namespace
+
+extern "C"
+{
+    using namespace ndarray::indexing;
+
+    void __nac3_ndarray_index(int32_t num_indices, NDIndex *indices, NDArray<int32_t> *src_ndarray,
+                              NDArray<int32_t> *dst_ndarray)
+    {
+        index(num_indices, indices, src_ndarray, dst_ndarray);
+    }
+
+    void __nac3_ndarray_index64(int64_t num_indices, NDIndex *indices, NDArray<int64_t> *src_ndarray,
+                                NDArray<int64_t> *dst_ndarray)
+    {
+        index(num_indices, indices, src_ndarray, dst_ndarray);
+    }
+}

nac3core/irrt/irrt/ndarray/iter.hpp

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

nac3core/irrt/irrt/ndarray/matmul.hpp

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

nac3core/irrt/irrt/ndarray/reshape.hpp

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

nac3core/irrt/irrt/ndarray/transpose.hpp

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

nac3core/irrt/irrt/original.hpp

@@ -0,0 +1,215 @@
+#pragma once
+
+#include <irrt/int_types.hpp>
+#include <irrt/math_util.hpp>
+
+// The type of an index or a value describing the length of a range/slice is always `int32_t`.
+using SliceIndex = int32_t;
+
+namespace
+{
+// adapted from GNU Scientific Library: https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
+// need to make sure `exp >= 0` before calling this function
+template <typename T> T __nac3_int_exp_impl(T base, T exp)
+{
+    T res = 1;
+    /* repeated squaring method */
+    do
+    {
+        if (exp & 1)
+        {
+            res *= base; /* for n odd */
+        }
+        exp >>= 1;
+        base *= base;
+    } while (exp);
+    return res;
+}
+} // namespace
+
+extern "C"
+{
+#define DEF_nac3_int_exp_(T)                                                                                           \
+    T __nac3_int_exp_##T(T base, T exp)                                                                                \
+    {                                                                                                                  \
+        return __nac3_int_exp_impl(base, exp);                                                                         \
+    }
+
+    DEF_nac3_int_exp_(int32_t) DEF_nac3_int_exp_(int64_t) DEF_nac3_int_exp_(uint32_t) DEF_nac3_int_exp_(uint64_t)
+
+        SliceIndex __nac3_slice_index_bound(SliceIndex i, const SliceIndex len)
+    {
+        if (i < 0)
+        {
+            i = len + i;
+        }
+        if (i < 0)
+        {
+            return 0;
+        }
+        else if (i > len)
+        {
+            return len;
+        }
+        return i;
+    }
+
+    SliceIndex __nac3_range_slice_len(const SliceIndex start, const SliceIndex end, const SliceIndex step)
+    {
+        SliceIndex diff = end - start;
+        if (diff > 0 && step > 0)
+        {
+            return ((diff - 1) / step) + 1;
+        }
+        else if (diff < 0 && step < 0)
+        {
+            return ((diff + 1) / step) + 1;
+        }
+        else
+        {
+            return 0;
+        }
+    }
+
+    // Handle list assignment and dropping part of the list when
+    // both dest_step and src_step are +1.
+    // - All the index must *not* be out-of-bound or negative,
+    // - The end index is *inclusive*,
+    // - The length of src and dest slice size should already
+    // be checked: if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest)
+    SliceIndex __nac3_list_slice_assign_var_size(SliceIndex dest_start, SliceIndex dest_end, SliceIndex dest_step,
+                                                 uint8_t *dest_arr, SliceIndex dest_arr_len, SliceIndex src_start,
+                                                 SliceIndex src_end, SliceIndex src_step, uint8_t *src_arr,
+                                                 SliceIndex src_arr_len, const SliceIndex size)
+    {
+        /* if dest_arr_len == 0, do nothing since we do not support extending list */
+        if (dest_arr_len == 0)
+            return dest_arr_len;
+        /* if both step is 1, memmove directly, handle the dropping of the list, and shrink size */
+        if (src_step == dest_step && dest_step == 1)
+        {
+            const SliceIndex src_len = (src_end >= src_start) ? (src_end - src_start + 1) : 0;
+            const SliceIndex dest_len = (dest_end >= dest_start) ? (dest_end - dest_start + 1) : 0;
+            if (src_len > 0)
+            {
+                __builtin_memmove(dest_arr + dest_start * size, src_arr + src_start * size, src_len * size);
+            }
+            if (dest_len > 0)
+            {
+                /* dropping */
+                __builtin_memmove(dest_arr + (dest_start + src_len) * size, dest_arr + (dest_end + 1) * size,
+                                  (dest_arr_len - dest_end - 1) * size);
+            }
+            /* shrink size */
+            return dest_arr_len - (dest_len - src_len);
+        }
+        /* if two range overlaps, need alloca */
+        uint8_t need_alloca = (dest_arr == src_arr) && !(max(dest_start, dest_end) < min(src_start, src_end) ||
+                                                         max(src_start, src_end) < min(dest_start, dest_end));
+        if (need_alloca)
+        {
+            uint8_t *tmp = reinterpret_cast<uint8_t *>(__builtin_alloca(src_arr_len * size));
+            __builtin_memcpy(tmp, src_arr, src_arr_len * size);
+            src_arr = tmp;
+        }
+        SliceIndex src_ind = src_start;
+        SliceIndex dest_ind = dest_start;
+        for (; (src_step > 0) ? (src_ind <= src_end) : (src_ind >= src_end); src_ind += src_step, dest_ind += dest_step)
+        {
+            /* for constant optimization */
+            if (size == 1)
+            {
+                __builtin_memcpy(dest_arr + dest_ind, src_arr + src_ind, 1);
+            }
+            else if (size == 4)
+            {
+                __builtin_memcpy(dest_arr + dest_ind * 4, src_arr + src_ind * 4, 4);
+            }
+            else if (size == 8)
+            {
+                __builtin_memcpy(dest_arr + dest_ind * 8, src_arr + src_ind * 8, 8);
+            }
+            else
+            {
+                /* memcpy for var size, cannot overlap after previous alloca */
+                __builtin_memcpy(dest_arr + dest_ind * size, src_arr + src_ind * size, size);
+            }
+        }
+        /* only dest_step == 1 can we shrink the dest list. */
+        /* size should be ensured prior to calling this function */
+        if (dest_step == 1 && dest_end >= dest_start)
+        {
+            __builtin_memmove(dest_arr + dest_ind * size, dest_arr + (dest_end + 1) * size,
+                              (dest_arr_len - dest_end - 1) * size);
+            return dest_arr_len - (dest_end - dest_ind) - 1;
+        }
+        return dest_arr_len;
+    }
+
+    int32_t __nac3_isinf(double x)
+    {
+        return __builtin_isinf(x);
+    }
+
+    int32_t __nac3_isnan(double x)
+    {
+        return __builtin_isnan(x);
+    }
+
+    double tgamma(double arg);
+
+    double __nac3_gamma(double z)
+    {
+        // Handling for denormals
+        //     | x                 | Python gamma(x) | C tgamma(x) |
+        // --- | ----------------- | --------------- | ----------- |
+        // (1) | nan               | nan             | nan         |
+        // (2) | -inf              | -inf            | inf         |
+        // (3) | inf               | inf             | inf         |
+        // (4) | 0.0               | inf             | inf         |
+        // (5) | {-1.0, -2.0, ...} | inf             | nan         |
+
+        // (1)-(3)
+        if (__builtin_isinf(z) || __builtin_isnan(z))
+        {
+            return z;
+        }
+
+        double v = tgamma(z);
+
+        // (4)-(5)
+        return __builtin_isinf(v) || __builtin_isnan(v) ? __builtin_inf() : v;
+    }
+
+    double lgamma(double arg);
+
+    double __nac3_gammaln(double x)
+    {
+        // libm's handling of value overflows differs from scipy:
+        // - scipy: gammaln(-inf) -> -inf
+        // - libm : lgamma(-inf) -> inf
+
+        if (__builtin_isinf(x))
+        {
+            return x;
+        }
+
+        return lgamma(x);
+    }
+
+    double j0(double x);
+
+    double __nac3_j0(double x)
+    {
+        // libm's handling of value overflows differs from scipy:
+        // - scipy: j0(inf) -> nan
+        // - libm : j0(inf) -> 0.0
+
+        if (__builtin_isinf(x))
+        {
+            return __builtin_nan("");
+        }
+
+        return j0(x);
+    }
+} // extern "C"

nac3core/irrt/irrt/range.hpp

@@ -0,0 +1,54 @@
+#pragma once
+
+#include <irrt/debug.hpp>
+#include <irrt/int_types.hpp>
+
+namespace
+{
+namespace range
+{
+template <typename T> T len(T start, T stop, T step)
+{
+    // Reference:
+    // https://github.com/python/cpython/blob/9dbd12375561a393eaec4b21ee4ac568a407cdb0/Objects/rangeobject.c#L933
+    if (step > 0 && start < stop)
+        return 1 + (stop - 1 - start) / step;
+    else if (step < 0 && start > stop)
+        return 1 + (start - 1 - stop) / (-step);
+    else
+        return 0;
+}
+} // namespace range
+
+/**
+ * @brief A Python range.
+ */
+template <typename T> struct Range
+{
+    T start;
+    T stop;
+    T step;
+
+    /**
+   * @brief Calculate the `len()` of this range.
+   */
+    template <typename SizeT> T len()
+    {
+        debug_assert(SizeT, step != 0);
+        return range::len(start, stop, step);
+    }
+};
+} // namespace
+
+extern "C"
+{
+    int32_t __nac3_range_len_i32(int32_t start, int32_t stop, int32_t step)
+    {
+        return range::len(start, stop, step);
+    }
+
+    int64_t __nac3_range_len_i64(int64_t start, int64_t stop, int64_t step)
+    {
+        return range::len(start, stop, step);
+    }
+}

nac3core/irrt/irrt/slice.hpp

@@ -0,0 +1,202 @@
+#pragma once
+
+#include <irrt/debug.hpp>
+#include <irrt/exception.hpp>
+#include <irrt/int_types.hpp>
+#include <irrt/math_util.hpp>
+#include <irrt/range.hpp>
+
+namespace
+{
+namespace slice
+{
+/**
+ * @brief Resolve a slice index under a given length like Python indexing.
+ *
+ * In Python, if you have a `list` of length 100, `list[-1]` resolves to
+ * `list[99]`, so `resolve_index_in_length_clamped(100, -1)` returns `99`.
+ *
+ * If `length` is 0, 0 is returned for any value of `index`.
+ *
+ * If `index` is out of bounds, clamps the returned value between `0` and
+ * `length - 1` (inclusive).
+ *
+ */
+template <typename T> T resolve_index_in_length_clamped(T length, T index)
+{
+    if (index < 0)
+    {
+        return max<T>(length + index, 0);
+    }
+    else
+    {
+        return min<T>(length, index);
+    }
+}
+
+/**
+ * @brief Like `resolve_index_in_length_clamped`, but returns `-1` if `index` is
+ * out of bounds.
+ */
+template <typename T> T resolve_index_in_length(T length, T index)
+{
+    T resolved = index < 0 ? length + index : index;
+    if (0 <= resolved && resolved < length)
+    {
+        return resolved;
+    }
+    else
+    {
+        return -1;
+    }
+}
+
+/**
+ * @brief Resolve a slice as a range.
+ *
+ * In Python, this would be `range(*slice(start, stop, step).indices(length))`.
+ */
+template <typename T>
+void indices(bool start_defined, T start, bool stop_defined, T stop, bool step_defined, T step, T length,
+             T *range_start, T *range_stop, T *range_step)
+{
+    // Reference:
+    // https://github.com/python/cpython/blob/main/Objects/sliceobject.c#L388
+    *range_step = step_defined ? step : 1;
+    bool step_is_negative = *range_step < 0;
+
+    T lower, upper;
+    if (step_is_negative)
+    {
+        lower = -1;
+        upper = length - 1;
+    }
+    else
+    {
+        lower = 0;
+        upper = length;
+    }
+
+    if (start_defined)
+    {
+        *range_start = start < 0 ? max(lower, start + length) : min(upper, start);
+    }
+    else
+    {
+        *range_start = step_is_negative ? upper : lower;
+    }
+
+    if (stop_defined)
+    {
+        *range_stop = stop < 0 ? max(lower, stop + length) : min(upper, stop);
+    }
+    else
+    {
+        *range_stop = step_is_negative ? lower : upper;
+    }
+}
+} // namespace slice
+
+/**
+ * @brief A Python-like slice with **unresolved** indices.
+ */
+template <typename T> struct Slice
+{
+    bool start_defined;
+    T start;
+
+    bool stop_defined;
+    T stop;
+
+    bool step_defined;
+    T step;
+
+    Slice()
+    {
+        this->reset();
+    }
+
+    void reset()
+    {
+        this->start_defined = false;
+        this->stop_defined = false;
+        this->step_defined = false;
+    }
+
+    void set_start(T start)
+    {
+        this->start_defined = true;
+        this->start = start;
+    }
+
+    void set_stop(T stop)
+    {
+        this->stop_defined = true;
+        this->stop = stop;
+    }
+
+    void set_step(T step)
+    {
+        this->step_defined = true;
+        this->step = step;
+    }
+
+    /**
+   * @brief Resolve this slice as a range.
+   *
+   * In Python, this would be `range(*slice(start, stop,
+   * step).indices(length))`.
+   */
+    template <typename SizeT> Range<T> indices(T length)
+    {
+        // Reference:
+        // https://github.com/python/cpython/blob/main/Objects/sliceobject.c#L388
+        debug_assert(SizeT, length >= 0);
+
+        Range<T> result;
+        slice::indices(start_defined, start, stop_defined, stop, step_defined, step, length, &result.start,
+                       &result.stop, &result.step);
+        return result;
+    }
+
+    /**
+   * @brief Like `.indices()` but with assertions.
+   */
+    template <typename SizeT> Range<T> indices_checked(T length)
+    {
+        // TODO: Switch to `SizeT length`
+
+        if (length < 0)
+        {
+            raise_exception(SizeT, EXN_VALUE_ERROR, "length should not be negative, got {0}", length, NO_PARAM,
+                            NO_PARAM);
+        }
+
+        if (this->step_defined && this->step == 0)
+        {
+            raise_exception(SizeT, EXN_VALUE_ERROR, "slice step cannot be zero", NO_PARAM, NO_PARAM, NO_PARAM);
+        }
+
+        return this->indices<SizeT>(length);
+    }
+};
+} // namespace
+
+extern "C"
+{
+    void __nac3_slice_indices_i32(bool start_defined, int32_t start, bool stop_defined, int32_t stop, bool step_defined,
+                                  int32_t step, int32_t length, int32_t *range_start, int32_t *range_stop,
+                                  int32_t *range_step)
+    {
+        slice::indices(start_defined, start, stop_defined, stop, step_defined, step, length, range_start, range_stop,
+                       range_step);
+    }
+
+    void __nac3_slice_indices_i64(bool start_defined, int64_t start, bool stop_defined, int64_t stop, bool step_defined,
+                                  int64_t step, int64_t length, int64_t *range_start, int64_t *range_stop,
+                                  int64_t *range_step)
+    {
+        slice::indices(start_defined, start, stop_defined, stop, step_defined, step, length, range_start, range_stop,
+                       range_step);
+    }
+}

nac3core/irrt/irrt_basic.hpp

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

nac3core/irrt/irrt_everything.hpp

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

nac3core/irrt/irrt_numpy_ndarray.hpp

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

nac3core/irrt/irrt_slice.hpp

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

nac3core/irrt/irrt_test.cpp

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

nac3core/irrt/irrt_typedefs.hpp

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

nac3core/irrt/irrt_utils.hpp

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

nac3core/src/codegen/concrete_type.rs

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

nac3core/src/codegen/extern_fns.rs

@@ -13,11 +13,11 @@ use crate::codegen::CodeGenContext;
 /// * `$extern_fn:literal`: Name of underlying extern function
 ///
 /// Optional Arguments:
-/// * `$(,$attributes:literal)*)`: Attributes linked with the extern function
+/// * `$(,$attributes:literal)*)`: Attributes linked with the extern function.
-/// The default attributes are "mustprogress", "nofree", "nounwind", "willreturn", and "writeonly"
+///   The default attributes are "mustprogress", "nofree", "nounwind", "willreturn", and "writeonly".
-/// These will be used unless other attributes are specified
+///   These will be used unless other attributes are specified
 /// * `$(,$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, $extern_fn:literal) => {
@@ -130,3 +130,62 @@ pub fn call_ldexp<'ctx>(
         .map(Either::unwrap_left)
         .unwrap()
 }
+
+/// Macro to generate `np_linalg` and `sp_linalg` functions
+/// The function takes as input `NDArray` and returns ()
+///
+/// Arguments:
+/// * `$fn_name:ident`: The identifier of the rust function to be generated
+/// * `$extern_fn:literal`: Name of underlying extern function
+/// * (2/3/4): Number of `NDArray` that function takes as input
+///
+/// Note:
+/// The operands and resulting `NDArray` are both passed as input to the funcion
+/// It is the responsibility of caller to ensure that output `NDArray` is properly allocated on stack
+/// The function changes the content of the output `NDArray` in-place
+macro_rules! generate_linalg_extern_fn {
+    ($fn_name:ident, $extern_fn:literal, 2) => {
+        generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2);
+    };
+    ($fn_name:ident, $extern_fn:literal, 3) => {
+        generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2, mat3);
+    };
+    ($fn_name:ident, $extern_fn:literal, 4) => {
+        generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2, mat3, mat4);
+    };
+    ($fn_name:ident, $extern_fn:literal $(,$input_matrix:ident)*) => {
+        #[doc = concat!("Invokes the linalg `", stringify!($extern_fn), " function." )]
+        pub fn $fn_name<'ctx>(
+            ctx: &mut CodeGenContext<'ctx, '_>
+            $(,$input_matrix: BasicValueEnum<'ctx>)*,
+            name: Option<&str>,
+        ){
+            const FN_NAME: &str = $extern_fn;
+            let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+                let fn_type = ctx.ctx.void_type().fn_type(&[$($input_matrix.get_type().into()),*], false);
+
+                let func = ctx.module.add_function(FN_NAME, fn_type, None);
+                for attr in ["mustprogress", "nofree", "nounwind", "willreturn", "writeonly"] {
+                    func.add_attribute(
+                        AttributeLoc::Function,
+                        ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
+                    );
+                }
+                func
+            });
+
+            ctx.builder.build_call(extern_fn, &[$($input_matrix.into(),)*], name.unwrap_or_default()).unwrap();
+        }
+    };
+}
+
+generate_linalg_extern_fn!(call_np_linalg_cholesky, "np_linalg_cholesky", 2);
+generate_linalg_extern_fn!(call_np_linalg_qr, "np_linalg_qr", 3);
+generate_linalg_extern_fn!(call_np_linalg_svd, "np_linalg_svd", 4);
+generate_linalg_extern_fn!(call_np_linalg_inv, "np_linalg_inv", 2);
+generate_linalg_extern_fn!(call_np_linalg_pinv, "np_linalg_pinv", 2);
+generate_linalg_extern_fn!(call_np_linalg_matrix_power, "np_linalg_matrix_power", 3);
+generate_linalg_extern_fn!(call_np_linalg_det, "np_linalg_det", 2);
+generate_linalg_extern_fn!(call_sp_linalg_lu, "sp_linalg_lu", 3);
+generate_linalg_extern_fn!(call_sp_linalg_schur, "sp_linalg_schur", 3);
+generate_linalg_extern_fn!(call_sp_linalg_hessenberg, "sp_linalg_hessenberg", 3);

nac3core/src/codegen/generator.rs

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

nac3core/src/codegen/irrt/mod.rs

@@ -1,23 +1,22 @@
-use crate::{typecheck::typedef::Type, util::SizeVariant};
+use crate::{symbol_resolver::SymbolResolver, typecheck::typedef::Type};
-
-mod test;
 
 use super::{
-    classes::{
+    classes::{ArrayLikeValue, ListValue},
-        ArrayLikeIndexer, ArrayLikeValue, ArraySliceValue, ListValue, NDArrayValue, NpArrayType,
+    model::*,
-        NpArrayValue, TypedArrayLikeAdapter, UntypedArrayLikeAccessor,
+    object::{
+        list::List,
+        ndarray::{broadcast::ShapeEntry, indexing::NDIndex, nditer::NDIter, NDArray},
     },
-    llvm_intrinsics, CodeGenContext, CodeGenerator,
+    CodeGenContext, CodeGenerator,
 };
-use crate::codegen::classes::TypedArrayLikeAccessor;
+use function::CallFunction;
-use crate::codegen::stmt::gen_for_callback_incrementing;
 use inkwell::{
     attributes::{Attribute, AttributeLoc},
     context::Context,
     memory_buffer::MemoryBuffer,
     module::Module,
-    types::{BasicType, BasicTypeEnum, FunctionType, IntType, PointerType},
+    types::BasicTypeEnum,
-    values::{BasicValueEnum, CallSiteValue, FloatValue, FunctionValue, IntValue},
+    values::{BasicValue, BasicValueEnum, CallSiteValue, FloatValue, IntValue},
     AddressSpace, IntPredicate,
 };
 use itertools::Either;
@@ -565,427 +564,383 @@ pub fn call_j0<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> Flo
         .unwrap()
 }
 
-/// Generates a call to `__nac3_ndarray_calc_size`. Returns an [`IntValue`] representing the
+// When [`TypeContext::size_type`] is 32-bits, the function name is "{fn_name}".
-/// calculated total size.
+// When [`TypeContext::size_type`] is 64-bits, the function name is "{fn_name}64".
-///
+#[must_use]
-/// * `dims` - An [`ArrayLikeIndexer`] containing the size of each dimension.
+pub fn get_sizet_dependent_function_name<G: CodeGenerator + ?Sized>(
-/// * `range` - The dimension index to begin and end (exclusively) calculating the dimensions for,
+    generator: &mut G,
-/// or [`None`] if starting from the first dimension and ending at the last dimension respectively.
+    ctx: &CodeGenContext<'_, '_>,
-pub fn call_ndarray_calc_size<'ctx, G, Dims>(
+    name: &str,
-    generator: &G,
+) -> String {
-    ctx: &CodeGenContext<'ctx, '_>,
+    let mut name = name.to_owned();
-    dims: &Dims,
+    match generator.get_size_type(ctx.ctx).get_bit_width() {
-    (begin, end): (Option<IntValue<'ctx>>, Option<IntValue<'ctx>>),
+        32 => {}
-) -> IntValue<'ctx>
+        64 => name.push_str("64"),
-where
+        bit_width => {
-    G: CodeGenerator + ?Sized,
+            panic!("Unsupported int type bit width {bit_width}, must be either 32-bits or 64-bits")
-    Dims: ArrayLikeIndexer<'ctx>,
+        }
-{
+    }
-    let llvm_usize = generator.get_size_type(ctx.ctx);
+    name
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    let ndarray_calc_size_fn_name = match llvm_usize.get_bit_width() {
-        32 => "__nac3_ndarray_calc_size",
-        64 => "__nac3_ndarray_calc_size64",
-        bw => unreachable!("Unsupported size type bit width: {}", bw),
-    };
-    let ndarray_calc_size_fn_t = llvm_usize.fn_type(
-        &[llvm_pusize.into(), llvm_usize.into(), llvm_usize.into(), llvm_usize.into()],
-        false,
-    );
-    let ndarray_calc_size_fn =
-        ctx.module.get_function(ndarray_calc_size_fn_name).unwrap_or_else(|| {
-            ctx.module.add_function(ndarray_calc_size_fn_name, ndarray_calc_size_fn_t, None)
-        });
-
-    let begin = begin.unwrap_or_else(|| llvm_usize.const_zero());
-    let end = end.unwrap_or_else(|| dims.size(ctx, generator));
-    ctx.builder
-        .build_call(
-            ndarray_calc_size_fn,
-            &[
-                dims.base_ptr(ctx, generator).into(),
-                dims.size(ctx, generator).into(),
-                begin.into(),
-                end.into(),
-            ],
-            "",
-        )
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_int_value))
-        .map(Either::unwrap_left)
-        .unwrap()
 }
 
-/// Generates a call to `__nac3_ndarray_calc_nd_indices`. Returns a [`TypeArrayLikeAdpater`]
+/// Initialize all global `EXN_*` exception IDs in IRRT with the [`SymbolResolver`].
-/// containing `i32` indices of the flattened index.
+pub fn setup_irrt_exceptions<'ctx>(
-///
+    ctx: &'ctx Context,
-/// * `index` - The index to compute the multidimensional index for.
+    module: &Module<'ctx>,
-/// * `ndarray` - LLVM pointer to the `NDArray`. This value must be the LLVM representation of an
+    symbol_resolver: &dyn SymbolResolver,
-/// `NDArray`.
+) {
-pub fn call_ndarray_calc_nd_indices<'ctx, G: CodeGenerator + ?Sized>(
+    let exn_id_type = ctx.i32_type();
-    generator: &G,
-    ctx: &mut CodeGenContext<'ctx, '_>,
-    index: IntValue<'ctx>,
-    ndarray: NDArrayValue<'ctx>,
-) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
-    let llvm_void = ctx.ctx.void_type();
-    let llvm_i32 = ctx.ctx.i32_type();
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
 
-    let ndarray_calc_nd_indices_fn_name = match llvm_usize.get_bit_width() {
+    let errors = &[
-        32 => "__nac3_ndarray_calc_nd_indices",
+        ("EXN_INDEX_ERROR", "0:IndexError"),
-        64 => "__nac3_ndarray_calc_nd_indices64",
+        ("EXN_VALUE_ERROR", "0:ValueError"),
-        bw => unreachable!("Unsupported size type bit width: {}", bw),
+        ("EXN_ASSERTION_ERROR", "0:AssertionError"),
-    };
+        ("EXN_TYPE_ERROR", "0:TypeError"),
-    let ndarray_calc_nd_indices_fn =
+    ];
-        ctx.module.get_function(ndarray_calc_nd_indices_fn_name).unwrap_or_else(|| {
-            let fn_type = llvm_void.fn_type(
-                &[llvm_usize.into(), llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into()],
-                false,
-            );
 
-            ctx.module.add_function(ndarray_calc_nd_indices_fn_name, fn_type, None)
+    for (irrt_name, symbol_name) in errors {
+        let exn_id = symbol_resolver.get_string_id(symbol_name);
+        let exn_id = exn_id_type.const_int(exn_id as u64, false).as_basic_value_enum();
+
+        let global = module.get_global(irrt_name).unwrap_or_else(|| {
+            panic!("Exception symbol name '{irrt_name}' should exist in the IRRT LLVM module")
         });
-
+        global.set_initializer(&exn_id);
-    let ndarray_num_dims = ndarray.load_ndims(ctx);
+    }
-    let ndarray_dims = ndarray.dim_sizes();
-
-    let indices = ctx.builder.build_array_alloca(llvm_i32, ndarray_num_dims, "").unwrap();
-
-    ctx.builder
-        .build_call(
-            ndarray_calc_nd_indices_fn,
-            &[
-                index.into(),
-                ndarray_dims.base_ptr(ctx, generator).into(),
-                ndarray_num_dims.into(),
-                indices.into(),
-            ],
-            "",
-        )
-        .unwrap();
-
-    TypedArrayLikeAdapter::from(
-        ArraySliceValue::from_ptr_val(indices, ndarray_num_dims, None),
-        Box::new(|_, v| v.into_int_value()),
-        Box::new(|_, v| v.into()),
-    )
 }
 
-fn call_ndarray_flatten_index_impl<'ctx, G, Indices>(
+pub fn call_nac3_range_len<'ctx, G: CodeGenerator + ?Sized, N: IntKind<'ctx>>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-    indices: &Indices,
-) -> IntValue<'ctx>
-where
-    G: CodeGenerator + ?Sized,
-    Indices: ArrayLikeIndexer<'ctx>,
-{
-    let llvm_i32 = ctx.ctx.i32_type();
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-
-    let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    debug_assert_eq!(
-        IntType::try_from(indices.element_type(ctx, generator))
-            .map(IntType::get_bit_width)
-            .unwrap_or_default(),
-        llvm_i32.get_bit_width(),
-        "Expected i32 value for argument `indices` to `call_ndarray_flatten_index_impl`"
-    );
-    debug_assert_eq!(
-        indices.size(ctx, generator).get_type().get_bit_width(),
-        llvm_usize.get_bit_width(),
-        "Expected usize integer value for argument `indices_size` to `call_ndarray_flatten_index_impl`"
-    );
-
-    let ndarray_flatten_index_fn_name = match llvm_usize.get_bit_width() {
-        32 => "__nac3_ndarray_flatten_index",
-        64 => "__nac3_ndarray_flatten_index64",
-        bw => unreachable!("Unsupported size type bit width: {}", bw),
-    };
-    let ndarray_flatten_index_fn =
-        ctx.module.get_function(ndarray_flatten_index_fn_name).unwrap_or_else(|| {
-            let fn_type = llvm_usize.fn_type(
-                &[llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into(), llvm_usize.into()],
-                false,
-            );
-
-            ctx.module.add_function(ndarray_flatten_index_fn_name, fn_type, None)
-        });
-
-    let ndarray_num_dims = ndarray.load_ndims(ctx);
-    let ndarray_dims = ndarray.dim_sizes();
-
-    let index = ctx
-        .builder
-        .build_call(
-            ndarray_flatten_index_fn,
-            &[
-                ndarray_dims.base_ptr(ctx, generator).into(),
-                ndarray_num_dims.into(),
-                indices.base_ptr(ctx, generator).into(),
-                indices.size(ctx, generator).into(),
-            ],
-            "",
-        )
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_int_value))
-        .map(Either::unwrap_left)
-        .unwrap();
-
-    index
-}
-
-/// Generates a call to `__nac3_ndarray_flatten_index`. Returns the flattened index for the
-/// multidimensional index.
-///
-/// * `ndarray` - LLVM pointer to the `NDArray`. This value must be the LLVM representation of an
-/// `NDArray`.
-/// * `indices` - The multidimensional index to compute the flattened index for.
-pub fn call_ndarray_flatten_index<'ctx, G, Index>(
     generator: &mut G,
     ctx: &mut CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
+    int_kind: N,
-    indices: &Index,
+    start: Instance<'ctx, Int<N>>,
-) -> IntValue<'ctx>
+    stop: Instance<'ctx, Int<N>>,
-where
+    step: Instance<'ctx, Int<N>>,
-    G: CodeGenerator + ?Sized,
+) -> Instance<'ctx, Int<N>> {
-    Index: ArrayLikeIndexer<'ctx>,
+    let bit_width = int_kind.get_int_type(generator, ctx.ctx).get_bit_width();
-{
+    let func_name = match bit_width {
-    call_ndarray_flatten_index_impl(generator, ctx, ndarray, indices)
+        32 => "__nac3_range_len_i32",
+        64 => "__nac3_range_len_i64",
+        _ => panic!("{bit_width}-bits ints not supported"), // We could add more variants when necessary.
+    };
+
+    CallFunction::begin(generator, ctx, func_name)
+        .arg(start)
+        .arg(stop)
+        .arg(step)
+        .returning("range_len", Int(int_kind))
 }
 
-/// Generates a call to `__nac3_ndarray_calc_broadcast`. Returns a tuple containing the number of
+#[allow(clippy::too_many_arguments)]
-/// dimension and size of each dimension of the resultant `ndarray`.
+pub fn call_nac3_slice_indices<'ctx, G: CodeGenerator + ?Sized, N: IntKind<'ctx>>(
-pub fn call_ndarray_calc_broadcast<'ctx, G: CodeGenerator + ?Sized>(
     generator: &mut G,
     ctx: &mut CodeGenContext<'ctx, '_>,
-    lhs: NDArrayValue<'ctx>,
+    int_kind: N,
-    rhs: NDArrayValue<'ctx>,
+    start_defined: Instance<'ctx, Int<Bool>>,
-) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
+    start: Instance<'ctx, Int<N>>,
-    let llvm_usize = generator.get_size_type(ctx.ctx);
+    stop_defined: Instance<'ctx, Int<Bool>>,
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
+    stop: Instance<'ctx, Int<N>>,
-
+    step_defined: Instance<'ctx, Int<Bool>>,
-    let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
+    step: Instance<'ctx, Int<N>>,
-        32 => "__nac3_ndarray_calc_broadcast",
+    length: Instance<'ctx, Int<N>>,
-        64 => "__nac3_ndarray_calc_broadcast64",
+    range_start: Instance<'ctx, Ptr<Int<N>>>,
-        bw => unreachable!("Unsupported size type bit width: {}", bw),
+    range_stop: Instance<'ctx, Ptr<Int<N>>>,
+    range_step: Instance<'ctx, Ptr<Int<N>>>,
+) -> Instance<'ctx, Int<N>> {
+    let bit_width = int_kind.get_int_type(generator, ctx.ctx).get_bit_width();
+    let func_name = match bit_width {
+        32 => "__nac3_slice_indices_i32",
+        64 => "__nac3_slice_indices_i64",
+        _ => panic!("{bit_width}-bits ints not supported"), // We could add more variants when necessary.
     };
-    let ndarray_calc_broadcast_fn =
-        ctx.module.get_function(ndarray_calc_broadcast_fn_name).unwrap_or_else(|| {
-            let fn_type = llvm_usize.fn_type(
-                &[
-                    llvm_pusize.into(),
-                    llvm_usize.into(),
-                    llvm_pusize.into(),
-                    llvm_usize.into(),
-                    llvm_pusize.into(),
-                ],
-                false,
-            );
 
-            ctx.module.add_function(ndarray_calc_broadcast_fn_name, fn_type, None)
+    CallFunction::begin(generator, ctx, func_name)
-        });
+        .arg(start_defined)
+        .arg(start)
+        .arg(stop_defined)
+        .arg(stop)
+        .arg(step_defined)
+        .arg(step)
+        .arg(length)
+        .arg(range_start)
+        .arg(range_stop)
+        .arg(range_step)
+        .returning("range_len", Int(int_kind))
+}
 
-    let lhs_ndims = lhs.load_ndims(ctx);
+pub fn call_nac3_ndarray_util_assert_shape_no_negative<'ctx, G: CodeGenerator + ?Sized>(
-    let rhs_ndims = rhs.load_ndims(ctx);
+    generator: &mut G,
-    let min_ndims = llvm_intrinsics::call_int_umin(ctx, lhs_ndims, rhs_ndims, None);
+    ctx: &mut CodeGenContext<'ctx, '_>,
-
+    ndims: Instance<'ctx, Int<SizeT>>,
-    gen_for_callback_incrementing(
+    shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(
         generator,
         ctx,
-        llvm_usize.const_zero(),
+        "__nac3_ndarray_util_assert_shape_no_negative",
-        (min_ndims, false),
+    );
-        |generator, ctx, _, idx| {
+    CallFunction::begin(generator, ctx, &name).arg(ndims).arg(shape).returning_void();
-            let idx = ctx.builder.build_int_sub(min_ndims, idx, "").unwrap();
-            let (lhs_dim_sz, rhs_dim_sz) = unsafe {
-                (
-                    lhs.dim_sizes().get_typed_unchecked(ctx, generator, &idx, None),
-                    rhs.dim_sizes().get_typed_unchecked(ctx, generator, &idx, None),
-                )
-            };
-
-            let llvm_usize_const_one = llvm_usize.const_int(1, false);
-            let lhs_eqz = ctx
-                .builder
-                .build_int_compare(IntPredicate::EQ, lhs_dim_sz, llvm_usize_const_one, "")
-                .unwrap();
-            let rhs_eqz = ctx
-                .builder
-                .build_int_compare(IntPredicate::EQ, rhs_dim_sz, llvm_usize_const_one, "")
-                .unwrap();
-            let lhs_or_rhs_eqz = ctx.builder.build_or(lhs_eqz, rhs_eqz, "").unwrap();
-
-            let lhs_eq_rhs = ctx
-                .builder
-                .build_int_compare(IntPredicate::EQ, lhs_dim_sz, rhs_dim_sz, "")
-                .unwrap();
-
-            let is_compatible = ctx.builder.build_or(lhs_or_rhs_eqz, lhs_eq_rhs, "").unwrap();
-
-            ctx.make_assert(
-                generator,
-                is_compatible,
-                "0:ValueError",
-                "operands could not be broadcast together",
-                [None, None, None],
-                ctx.current_loc,
-            );
-
-            Ok(())
-        },
-        llvm_usize.const_int(1, false),
-    )
-    .unwrap();
-
-    let max_ndims = llvm_intrinsics::call_int_umax(ctx, lhs_ndims, rhs_ndims, None);
-    let lhs_dims = lhs.dim_sizes().base_ptr(ctx, generator);
-    let lhs_ndims = lhs.load_ndims(ctx);
-    let rhs_dims = rhs.dim_sizes().base_ptr(ctx, generator);
-    let rhs_ndims = rhs.load_ndims(ctx);
-    let out_dims = ctx.builder.build_array_alloca(llvm_usize, max_ndims, "").unwrap();
-    let out_dims = ArraySliceValue::from_ptr_val(out_dims, max_ndims, None);
-
-    ctx.builder
-        .build_call(
-            ndarray_calc_broadcast_fn,
-            &[
-                lhs_dims.into(),
-                lhs_ndims.into(),
-                rhs_dims.into(),
-                rhs_ndims.into(),
-                out_dims.base_ptr(ctx, generator).into(),
-            ],
-            "",
-        )
-        .unwrap();
-
-    TypedArrayLikeAdapter::from(
-        out_dims,
-        Box::new(|_, v| v.into_int_value()),
-        Box::new(|_, v| v.into()),
-    )
 }
 
-/// Generates a call to `__nac3_ndarray_calc_broadcast_idx`. Returns an [`ArrayAllocaValue`]
+pub fn call_nac3_ndarray_util_assert_output_shape_same<'ctx, G: CodeGenerator + ?Sized>(
-/// containing the indices used for accessing `array` corresponding to the index of the broadcasted
-/// array `broadcast_idx`.
-pub fn call_ndarray_calc_broadcast_index<
-    'ctx,
-    G: CodeGenerator + ?Sized,
-    BroadcastIdx: UntypedArrayLikeAccessor<'ctx>,
->(
     generator: &mut G,
     ctx: &mut CodeGenContext<'ctx, '_>,
-    array: NDArrayValue<'ctx>,
+    ndarray_ndims: Instance<'ctx, Int<SizeT>>,
-    broadcast_idx: &BroadcastIdx,
+    ndarray_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
-) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
+    output_ndims: Instance<'ctx, Int<SizeT>>,
-    let llvm_i32 = ctx.ctx.i32_type();
+    output_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
-    let llvm_usize = generator.get_size_type(ctx.ctx);
+) {
-    let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
+    let name = get_sizet_dependent_function_name(
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
+        generator,
-
+        ctx,
-    let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
+        "__nac3_ndarray_util_assert_output_shape_same",
-        32 => "__nac3_ndarray_calc_broadcast_idx",
+    );
-        64 => "__nac3_ndarray_calc_broadcast_idx64",
+    CallFunction::begin(generator, ctx, &name)
-        bw => unreachable!("Unsupported size type bit width: {}", bw),
+        .arg(ndarray_ndims)
-    };
+        .arg(ndarray_shape)
-    let ndarray_calc_broadcast_fn =
+        .arg(output_ndims)
-        ctx.module.get_function(ndarray_calc_broadcast_fn_name).unwrap_or_else(|| {
+        .arg(output_shape)
-            let fn_type = llvm_usize.fn_type(
+        .returning_void();
-                &[llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into(), llvm_pi32.into()],
-                false,
-            );
-
-            ctx.module.add_function(ndarray_calc_broadcast_fn_name, fn_type, None)
-        });
-
-    let broadcast_size = broadcast_idx.size(ctx, generator);
-    let out_idx = ctx.builder.build_array_alloca(llvm_i32, broadcast_size, "").unwrap();
-
-    let array_dims = array.dim_sizes().base_ptr(ctx, generator);
-    let array_ndims = array.load_ndims(ctx);
-    let broadcast_idx_ptr = unsafe {
-        broadcast_idx.ptr_offset_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
-    };
-
-    ctx.builder
-        .build_call(
-            ndarray_calc_broadcast_fn,
-            &[array_dims.into(), array_ndims.into(), broadcast_idx_ptr.into(), out_idx.into()],
-            "",
-        )
-        .unwrap();
-
-    TypedArrayLikeAdapter::from(
-        ArraySliceValue::from_ptr_val(out_idx, broadcast_size, None),
-        Box::new(|_, v| v.into_int_value()),
-        Box::new(|_, v| v.into()),
-    )
 }
 
-fn get_size_variant<'ctx>(ty: IntType<'ctx>) -> SizeVariant {
+pub fn call_nac3_ndarray_size<'ctx, G: CodeGenerator + ?Sized>(
-    match ty.get_bit_width() {
+    generator: &mut G,
-        32 => SizeVariant::Bits32,
+    ctx: &mut CodeGenContext<'ctx, '_>,
-        64 => SizeVariant::Bits64,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
-        _ => unreachable!("Unsupported int type bit width {}", ty.get_bit_width()),
+) -> Instance<'ctx, Int<SizeT>> {
-    }
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_size");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("size")
 }
 
-fn get_size_type_dependent_function<'ctx, BuildFuncTypeFn>(
+pub fn call_nac3_ndarray_nbytes<'ctx, G: CodeGenerator + ?Sized>(
-    ctx: &CodeGenContext<'ctx, '_>,
+    generator: &mut G,
-    size_type: IntType<'ctx>,
+    ctx: &mut CodeGenContext<'ctx, '_>,
-    base_name: &str,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
-    build_func_type: BuildFuncTypeFn,
+) -> Instance<'ctx, Int<SizeT>> {
-) -> FunctionValue<'ctx>
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_nbytes");
-where
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("nbytes")
-    BuildFuncTypeFn: Fn() -> FunctionType<'ctx>,
-{
-    let mut fn_name = base_name.to_owned();
-    match get_size_variant(size_type) {
-        SizeVariant::Bits32 => {
-            // The original fn_name is the correct function name
-        }
-        SizeVariant::Bits64 => {
-            // Append "64" at the end, this is the naming convention for 64-bit
-            fn_name.push_str("64");
-        }
-    }
-
-    // Get (or declare then get if does not exist) the corresponding function
-    ctx.module.get_function(&fn_name).unwrap_or_else(|| {
-        let fn_type = build_func_type();
-        ctx.module.add_function(&fn_name, fn_type, None)
-    })
 }
 
-fn get_ndarray_struct_ptr<'ctx>(ctx: &'ctx Context, size_type: IntType<'ctx>) -> PointerType<'ctx> {
+pub fn call_nac3_ndarray_len<'ctx, G: CodeGenerator + ?Sized>(
-    let i8_type = ctx.i8_type();
+    generator: &mut G,
-
+    ctx: &mut CodeGenContext<'ctx, '_>,
-    let ndarray_ty = NpArrayType { size_type, elem_type: i8_type.as_basic_type_enum() };
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
-    let struct_ty = ndarray_ty.fields().whole_struct.as_struct_type(ctx);
+) -> Instance<'ctx, Int<SizeT>> {
-    struct_ty.ptr_type(AddressSpace::default())
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_len");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("len")
 }
 
-pub fn call_nac3_ndarray_size<'ctx>(
+pub fn call_nac3_ndarray_is_c_contiguous<'ctx, G: CodeGenerator + ?Sized>(
-    ctx: &CodeGenContext<'ctx, '_>,
+    generator: &mut G,
-    ndarray: NpArrayValue<'ctx>,
+    ctx: &mut CodeGenContext<'ctx, '_>,
-) -> IntValue<'ctx> {
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
-    let size_type = ndarray.ty.size_type;
+) -> Instance<'ctx, Int<Bool>> {
-    let function = get_size_type_dependent_function(ctx, size_type, "__nac3_ndarray_size", || {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_is_c_contiguous");
-        size_type.fn_type(&[get_ndarray_struct_ptr(ctx.ctx, size_type).into()], false)
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("is_c_contiguous")
-    });
+}
-
+
-    ctx.builder
+pub fn call_nac3_ndarray_get_nth_pelement<'ctx, G: CodeGenerator + ?Sized>(
-        .build_call(function, &[ndarray.ptr.into()], "size")
+    generator: &mut G,
-        .unwrap()
+    ctx: &mut CodeGenContext<'ctx, '_>,
-        .try_as_basic_value()
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
-        .unwrap_left()
+    index: Instance<'ctx, Int<SizeT>>,
-        .into_int_value()
+) -> Instance<'ctx, Ptr<Int<Byte>>> {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_get_nth_pelement");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).arg(index).returning_auto("pelement")
+}
+
+pub fn call_nac3_ndarray_get_pelement_by_indices<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    indices: Instance<'ctx, Ptr<Int<SizeT>>>,
+) -> Instance<'ctx, Ptr<Int<Byte>>> {
+    let name =
+        get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_get_pelement_by_indices");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).arg(indices).returning_auto("pelement")
+}
+
+pub fn call_nac3_ndarray_set_strides_by_shape<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name =
+        get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_set_strides_by_shape");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_void();
+}
+
+pub fn call_nac3_ndarray_copy_data<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_copy_data");
+    CallFunction::begin(generator, ctx, &name).arg(src_ndarray).arg(dst_ndarray).returning_void();
+}
+
+pub fn call_nac3_nditer_initialize<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    indices: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_initialize");
+    CallFunction::begin(generator, ctx, &name).arg(iter).arg(ndarray).arg(indices).returning_void();
+}
+
+pub fn call_nac3_nditer_has_next<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
+) -> Instance<'ctx, Int<Bool>> {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_has_next");
+    CallFunction::begin(generator, ctx, &name).arg(iter).returning_auto("has_next")
+}
+
+pub fn call_nac3_nditer_next<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_next");
+    CallFunction::begin(generator, ctx, &name).arg(iter).returning_void();
+}
+
+pub fn call_nac3_ndarray_index<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    num_indices: Instance<'ctx, Int<SizeT>>,
+    indices: Instance<'ctx, Ptr<Struct<NDIndex>>>,
+    src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_index");
+    CallFunction::begin(generator, ctx, &name)
+        .arg(num_indices)
+        .arg(indices)
+        .arg(src_ndarray)
+        .arg(dst_ndarray)
+        .returning_void();
+}
+
+pub fn call_nac3_ndarray_array_set_and_validate_list_shape<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    list: Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>>,
+    ndims: Instance<'ctx, Int<SizeT>>,
+    shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(
+        generator,
+        ctx,
+        "__nac3_ndarray_array_set_and_validate_list_shape",
+    );
+    CallFunction::begin(generator, ctx, &name).arg(list).arg(ndims).arg(shape).returning_void();
+}
+
+pub fn call_nac3_ndarray_array_write_list_to_array<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    list: Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name = get_sizet_dependent_function_name(
+        generator,
+        ctx,
+        "__nac3_ndarray_array_write_list_to_array",
+    );
+    CallFunction::begin(generator, ctx, &name).arg(list).arg(ndarray).returning_void();
+}
+
+pub fn call_nac3_ndarray_reshape_resolve_and_check_new_shape<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    size: Instance<'ctx, Int<SizeT>>,
+    new_ndims: Instance<'ctx, Int<SizeT>>,
+    new_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(
+        generator,
+        ctx,
+        "__nac3_ndarray_reshape_resolve_and_check_new_shape",
+    );
+    CallFunction::begin(generator, ctx, &name)
+        .arg(size)
+        .arg(new_ndims)
+        .arg(new_shape)
+        .returning_void();
+}
+
+pub fn call_nac3_ndarray_broadcast_to<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_broadcast_to");
+    CallFunction::begin(generator, ctx, &name).arg(src_ndarray).arg(dst_ndarray).returning_void();
+}
+
+pub fn call_nac3_ndarray_broadcast_shapes<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    num_shape_entries: Instance<'ctx, Int<SizeT>>,
+    shape_entries: Instance<'ctx, Ptr<Struct<ShapeEntry>>>,
+    dst_ndims: Instance<'ctx, Int<SizeT>>,
+    dst_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_broadcast_shapes");
+    CallFunction::begin(generator, ctx, &name)
+        .arg(num_shape_entries)
+        .arg(shape_entries)
+        .arg(dst_ndims)
+        .arg(dst_shape)
+        .returning_void();
+}
+
+pub fn call_nac3_ndarray_transpose<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    num_axes: Instance<'ctx, Int<SizeT>>,
+    axes: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_transpose");
+    CallFunction::begin(generator, ctx, &name)
+        .arg(src_ndarray)
+        .arg(dst_ndarray)
+        .arg(num_axes)
+        .arg(axes)
+        .returning_void();
+}
+
+#[allow(clippy::too_many_arguments)]
+pub fn call_nac3_ndarray_matmul_calculate_shapes<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    a_ndims: Instance<'ctx, Int<SizeT>>,
+    a_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    b_ndims: Instance<'ctx, Int<SizeT>>,
+    b_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    final_ndims: Instance<'ctx, Int<SizeT>>,
+    new_a_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    new_b_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    dst_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name =
+        get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_matmul_calculate_shapes");
+    CallFunction::begin(generator, ctx, &name)
+        .arg(a_ndims)
+        .arg(a_shape)
+        .arg(b_ndims)
+        .arg(b_shape)
+        .arg(final_ndims)
+        .arg(new_a_shape)
+        .arg(new_b_shape)
+        .arg(dst_shape)
+        .returning_void();
 }

nac3core/src/codegen/irrt/test.rs

@@ -1,26 +0,0 @@
-#[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");
-        }
-    }
-}

nac3core/src/codegen/llvm_intrinsics.rs

@@ -35,6 +35,40 @@ fn get_float_intrinsic_repr(ctx: &Context, ft: FloatType) -> &'static str {
     unreachable!()
 }
 
+/// Invokes the [`llvm.va_start`](https://llvm.org/docs/LangRef.html#llvm-va-start-intrinsic)
+/// intrinsic.
+pub fn call_va_start<'ctx>(ctx: &CodeGenContext<'ctx, '_>, arglist: PointerValue<'ctx>) {
+    const FN_NAME: &str = "llvm.va_start";
+
+    let intrinsic_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+        let llvm_void = ctx.ctx.void_type();
+        let llvm_i8 = ctx.ctx.i8_type();
+        let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
+        let fn_type = llvm_void.fn_type(&[llvm_p0i8.into()], false);
+
+        ctx.module.add_function(FN_NAME, fn_type, None)
+    });
+
+    ctx.builder.build_call(intrinsic_fn, &[arglist.into()], "").unwrap();
+}
+
+/// Invokes the [`llvm.va_start`](https://llvm.org/docs/LangRef.html#llvm-va-start-intrinsic)
+/// intrinsic.
+pub fn call_va_end<'ctx>(ctx: &CodeGenContext<'ctx, '_>, arglist: PointerValue<'ctx>) {
+    const FN_NAME: &str = "llvm.va_end";
+
+    let intrinsic_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+        let llvm_void = ctx.ctx.void_type();
+        let llvm_i8 = ctx.ctx.i8_type();
+        let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
+        let fn_type = llvm_void.fn_type(&[llvm_p0i8.into()], false);
+
+        ctx.module.add_function(FN_NAME, fn_type, None)
+    });
+
+    ctx.builder.build_call(intrinsic_fn, &[arglist.into()], "").unwrap();
+}
+
 /// Invokes the [`llvm.stacksave`](https://llvm.org/docs/LangRef.html#llvm-stacksave-intrinsic)
 /// intrinsic.
 pub fn call_stacksave<'ctx>(
@@ -171,8 +205,9 @@ pub fn call_memcpy_generic<'ctx>(
 /// * `$ctx:ident`: Reference to the current Code Generation Context
 /// * `$name:ident`: Optional name to be assigned to the llvm build call (Option<&str>)
 /// * `$llvm_name:literal`: Name of underlying llvm intrinsic function
-/// * `$map_fn:ident`: Mapping function to be applied on `BasicValue` (`BasicValue` -> Function Return Type)
+/// * `$map_fn:ident`: Mapping function to be applied on `BasicValue` (`BasicValue` -> Function Return Type).
-/// Use `BasicValueEnum::into_int_value` for Integer return type and `BasicValueEnum::into_float_value` for Float return type
+///   Use `BasicValueEnum::into_int_value` for Integer return type and
+///   `BasicValueEnum::into_float_value` for Float return type
 /// * `$llvm_ty:ident`: Type of first operand
 /// * `,($val:ident)*`: Comma separated list of operands
 macro_rules! generate_llvm_intrinsic_fn_body {
@@ -188,8 +223,8 @@ macro_rules! generate_llvm_intrinsic_fn_body {
 /// Arguments:
 /// * `float/int`: Indicates the return and argument type of the function
 /// * `$fn_name:ident`: The identifier of the rust function to be generated
-/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function
+/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function.
-/// Omit "llvm." prefix from the function name i.e. use "ceil" instead of "llvm.ceil"
+///   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 {

nac3core/src/codegen/mod.rs

@@ -1,7 +1,7 @@
 use crate::{
-    codegen::classes::{ListType, NDArrayType, ProxyType, RangeType},
+    codegen::classes::{ListType, ProxyType},
     symbol_resolver::{StaticValue, SymbolResolver},
-    toplevel::{helper::PrimDef, numpy::unpack_ndarray_var_tys, TopLevelContext, TopLevelDef},
+    toplevel::{helper::PrimDef, TopLevelContext, TopLevelDef},
     typecheck::{
         type_inferencer::{CodeLocation, PrimitiveStore},
         typedef::{CallId, FuncArg, Type, TypeEnum, Unifier},
@@ -24,7 +24,14 @@ use inkwell::{
     AddressSpace, IntPredicate, OptimizationLevel,
 };
 use itertools::Itertools;
+use model::*;
 use nac3parser::ast::{Location, Stmt, StrRef};
+use object::{
+    exception::Exception,
+    ndarray::NDArray,
+    range::range_model,
+    str::{str_model, Str},
+};
 use parking_lot::{Condvar, Mutex};
 use std::collections::{HashMap, HashSet};
 use std::sync::{
@@ -41,7 +48,9 @@ pub mod extern_fns;
 mod generator;
 pub mod irrt;
 pub mod llvm_intrinsics;
+pub mod model;
 pub mod numpy;
+pub mod object;
 pub mod stmt;
 
 #[cfg(test)]
@@ -68,6 +77,16 @@ pub struct CodeGenLLVMOptions {
     pub target: CodeGenTargetMachineOptions,
 }
 
+impl CodeGenLLVMOptions {
+    /// Creates a [`TargetMachine`] using the target options specified by this struct.
+    ///
+    /// See [`Target::create_target_machine`].
+    #[must_use]
+    pub fn create_target_machine(&self) -> Option<TargetMachine> {
+        self.target.create_target_machine(self.opt_level)
+    }
+}
+
 /// Additional options for code generation for the target machine.
 #[derive(Clone, Debug, Eq, PartialEq)]
 pub struct CodeGenTargetMachineOptions {
@@ -158,11 +177,11 @@ 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, Instance<'ctx, Str>>,
 
     /// [`BasicBlock`] containing all `alloca` statements for the current function.
     pub init_bb: BasicBlock<'ctx>,
-    pub exception_val: Option<PointerValue<'ctx>>,
+    pub exception_val: Option<Instance<'ctx, Ptr<Struct<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.
@@ -338,6 +357,10 @@ impl WorkerRegistry {
         let mut builder = context.create_builder();
         let mut module = context.create_module(generator.get_name());
 
+        let target_machine = self.llvm_options.create_target_machine().unwrap();
+        module.set_data_layout(&target_machine.get_target_data().get_data_layout());
+        module.set_triple(&target_machine.get_triple());
+
         module.add_basic_value_flag(
             "Debug Info Version",
             inkwell::module::FlagBehavior::Warning,
@@ -361,6 +384,10 @@ impl WorkerRegistry {
                     errors.insert(e);
                     // create a new empty module just to continue codegen and collect errors
                     module = context.create_module(&format!("{}_recover", generator.get_name()));
+
+                    let target_machine = self.llvm_options.create_target_machine().unwrap();
+                    module.set_data_layout(&target_machine.get_target_data().get_data_layout());
+                    module.set_triple(&target_machine.get_triple());
                 }
             }
             *self.task_count.lock() -= 1;
@@ -426,7 +453,7 @@ pub struct CodeGenTask {
 fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
     ctx: &'ctx Context,
     module: &Module<'ctx>,
-    generator: &mut G,
+    generator: &G,
     unifier: &mut Unifier,
     top_level: &TopLevelContext,
     type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
@@ -471,12 +498,7 @@ 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);
+                            Ptr(Struct(NDArray)).get_type(generator, ctx).as_basic_type_enum()
-                            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()
                         }
 
                         _ => unreachable!(
@@ -520,8 +542,10 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
                 };
                 return ty;
             }
-            TTuple { ty } => {
+            TTuple { ty, is_vararg_ctx } => {
                 // a struct with fields in the order present in the tuple
+                assert!(!is_vararg_ctx, "Tuples in vararg context must be instantiated with the correct number of arguments before calling get_llvm_type");
+
                 let fields = ty
                     .iter()
                     .map(|ty| {
@@ -551,7 +575,7 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
 fn get_llvm_abi_type<'ctx, G: CodeGenerator + ?Sized>(
     ctx: &'ctx Context,
     module: &Module<'ctx>,
-    generator: &mut G,
+    generator: &G,
     unifier: &mut Unifier,
     top_level: &TopLevelContext,
     type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
@@ -560,11 +584,11 @@ fn get_llvm_abi_type<'ctx, G: CodeGenerator + ?Sized>(
 ) -> BasicTypeEnum<'ctx> {
     // If the type is used in the definition of a function, return `i1` instead of `i8` for ABI
     // consistency.
-    return if unifier.unioned(ty, primitives.bool) {
+    if unifier.unioned(ty, primitives.bool) {
         ctx.bool_type().into()
     } else {
         get_llvm_type(ctx, module, generator, unifier, top_level, type_cache, ty)
-    };
+    }
 }
 
 /// Whether `sret` is needed for a return value with type `ty`.
@@ -589,6 +613,40 @@ fn need_sret(ty: BasicTypeEnum) -> bool {
     need_sret_impl(ty, true)
 }
 
+/// Returns the [`BasicTypeEnum`] representing a `va_list` struct for variadic arguments.
+fn get_llvm_valist_type<'ctx>(ctx: &'ctx Context, triple: &TargetTriple) -> BasicTypeEnum<'ctx> {
+    let triple = TargetMachine::normalize_triple(triple);
+    let triple = triple.as_str().to_str().unwrap();
+    let arch = triple.split('-').next().unwrap();
+
+    let llvm_pi8 = ctx.i8_type().ptr_type(AddressSpace::default());
+
+    // Referenced from parseArch() in llvm/lib/Support/Triple.cpp
+    match arch {
+        "i386" | "i486" | "i586" | "i686" | "riscv32" => {
+            ctx.i8_type().ptr_type(AddressSpace::default()).into()
+        }
+        "amd64" | "x86_64" | "x86_64h" => {
+            let llvm_i32 = ctx.i32_type();
+
+            let va_list_tag = ctx.opaque_struct_type("struct.__va_list_tag");
+            va_list_tag.set_body(
+                &[llvm_i32.into(), llvm_i32.into(), llvm_pi8.into(), llvm_pi8.into()],
+                false,
+            );
+            va_list_tag.into()
+        }
+        "armv7" => {
+            let va_list = ctx.opaque_struct_type("struct.__va_list");
+            va_list.set_body(&[llvm_pi8.into()], false);
+            va_list.into()
+        }
+        triple => {
+            todo!("Unsupported platform for varargs: {triple}")
+        }
+    }
+}
+
 /// Implementation for generating LLVM IR for a function.
 pub fn gen_func_impl<
     'ctx,
@@ -653,36 +711,9 @@ pub fn gen_func_impl<
         (primitives.uint64, context.i64_type().into()),
         (primitives.float, context.f64_type().into()),
         (primitives.bool, context.i8_type().into()),
-        (primitives.str, {
+        (primitives.str, str_model().get_type(generator, context).into()),
-            let name = "str";
+        (primitives.range, Ptr(range_model()).get_type(generator, context).into()),
-            match module.get_struct_type(name) {
+        (primitives.exception, { Ptr(Struct(Exception)).get_type(generator, context).into() }),
-                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.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()
-            }
-        }),
     ]
     .iter()
     .copied()
@@ -700,6 +731,7 @@ pub fn gen_func_impl<
                 name: arg.name,
                 ty: task.store.to_unifier_type(&mut unifier, &primitives, arg.ty, &mut cache),
                 default_value: arg.default_value.clone(),
+                is_vararg: arg.is_vararg,
             })
             .collect_vec(),
         task.store.to_unifier_type(&mut unifier, &primitives, *ret, &mut cache),
@@ -722,7 +754,10 @@ pub fn gen_func_impl<
     let has_sret = ret_type.map_or(false, |ty| need_sret(ty));
     let mut params = args
         .iter()
+        .filter(|arg| !arg.is_vararg)
         .map(|arg| {
+            debug_assert!(!arg.is_vararg);
+
             get_llvm_abi_type(
                 context,
                 &module,
@@ -741,9 +776,12 @@ pub fn gen_func_impl<
         params.insert(0, ret_type.unwrap().ptr_type(AddressSpace::default()).into());
     }
 
+    debug_assert!(matches!(args.iter().filter(|arg| arg.is_vararg).count(), 0..=1));
+    let vararg_arg = args.iter().find(|arg| arg.is_vararg);
+
     let fn_type = match ret_type {
-        Some(ret_type) if !has_sret => ret_type.fn_type(&params, false),
+        Some(ret_type) if !has_sret => ret_type.fn_type(&params, vararg_arg.is_some()),
-        _ => context.void_type().fn_type(&params, false),
+        _ => context.void_type().fn_type(&params, vararg_arg.is_some()),
     };
 
     let symbol = &task.symbol_name;
@@ -773,7 +811,9 @@ pub fn gen_func_impl<
 
     let mut var_assignment = HashMap::new();
     let offset = u32::from(has_sret);
-    for (n, arg) in args.iter().enumerate() {
+
+    // Store non-vararg argument values into local variables
+    for (n, arg) in args.iter().enumerate().filter(|(_, arg)| !arg.is_vararg) {
         let param = fn_val.get_nth_param((n as u32) + offset).unwrap();
         let local_type = get_llvm_type(
             context,
@@ -806,6 +846,8 @@ pub fn gen_func_impl<
         var_assignment.insert(arg.name, (alloca, None, 0));
     }
 
+    // TODO: Save vararg parameters as list
+
     let return_buffer = if has_sret {
         Some(fn_val.get_nth_param(0).unwrap().into_pointer_value())
     } else {
@@ -1028,3 +1070,9 @@ fn gen_in_range_check<'ctx>(
 
     ctx.builder.build_int_compare(IntPredicate::SLT, lo, hi, "cmp").unwrap()
 }
+
+/// Returns the internal name for the `va_count` argument, used to indicate the number of arguments
+/// passed to the variadic function.
+fn get_va_count_arg_name(arg_name: StrRef) -> StrRef {
+    format!("__{}_va_count", &arg_name).into()
+}

nac3core/src/codegen/model/any.rs

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

nac3core/src/codegen/model/array.rs

@@ -0,0 +1,141 @@
+use std::fmt;
+
+use inkwell::{
+    context::Context,
+    types::{ArrayType, BasicType, BasicTypeEnum},
+    values::{ArrayValue, IntValue},
+};
+
+use crate::codegen::{CodeGenContext, CodeGenerator};
+
+use super::*;
+
+/// Traits for a Rust struct that describes a length value for [`Array`].
+pub trait LenKind: fmt::Debug + Clone + Copy {
+    fn get_length(&self) -> u32;
+}
+
+/// A statically known length.
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Len<const N: u32>;
+
+/// A dynamically known length.
+#[derive(Debug, Clone, Copy)]
+pub struct AnyLen(pub u32);
+
+impl<const N: u32> LenKind for Len<N> {
+    fn get_length(&self) -> u32 {
+        N
+    }
+}
+
+impl LenKind for AnyLen {
+    fn get_length(&self) -> u32 {
+        self.0
+    }
+}
+
+/// A Model for an [`ArrayType`].
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Array<Len, Item> {
+    /// Length of this array.
+    pub len: Len,
+    /// [`Model`] of an array item.
+    pub item: Item,
+}
+
+impl<'ctx, Len: LenKind, Item: Model<'ctx>> Model<'ctx> for Array<Len, Item> {
+    type Value = ArrayValue<'ctx>;
+    type Type = ArrayType<'ctx>;
+
+    fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
+        self.item.get_type(generator, ctx).array_type(self.len.get_length())
+    }
+
+    fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        ty: T,
+    ) -> Result<(), ModelError> {
+        let ty = ty.as_basic_type_enum();
+        let BasicTypeEnum::ArrayType(ty) = ty else {
+            return Err(ModelError(format!("Expecting ArrayType, but got {ty:?}")));
+        };
+
+        if ty.len() != self.len.get_length() {
+            return Err(ModelError(format!(
+                "Expecting ArrayType with size {}, but got an ArrayType with size {}",
+                ty.len(),
+                self.len.get_length()
+            )));
+        }
+
+        self.item
+            .check_type(generator, ctx, ty.get_element_type())
+            .map_err(|err| err.under_context("an ArrayType"))?;
+
+        Ok(())
+    }
+}
+
+impl<'ctx, Len: LenKind, Item: Model<'ctx>> Instance<'ctx, Ptr<Array<Len, Item>>> {
+    /// Get the pointer to the `i`-th (0-based) array element.
+    pub fn gep(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        i: IntValue<'ctx>,
+    ) -> Instance<'ctx, Ptr<Item>> {
+        let zero = ctx.ctx.i32_type().const_zero();
+        let ptr = unsafe { ctx.builder.build_in_bounds_gep(self.value, &[zero, i], "").unwrap() };
+
+        Ptr(self.model.0.item).believe_value(ptr)
+    }
+
+    /// Like `gep` but `i` is a constant.
+    pub fn gep_const(&self, ctx: &CodeGenContext<'ctx, '_>, i: u64) -> Instance<'ctx, Ptr<Item>> {
+        assert!(
+            i < u64::from(self.model.0.len.get_length()),
+            "Index {i} is out of bounds. Array length = {}",
+            self.model.0.len.get_length()
+        );
+
+        let i = ctx.ctx.i32_type().const_int(i, false);
+        self.gep(ctx, i)
+    }
+
+    /// Convenience function equivalent to `.gep(...).load(...)`.
+    pub fn get<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        i: IntValue<'ctx>,
+    ) -> Instance<'ctx, Item> {
+        self.gep(ctx, i).load(generator, ctx)
+    }
+
+    /// Like `get` but `i` is a constant.
+    pub fn get_const<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        i: u64,
+    ) -> Instance<'ctx, Item> {
+        self.gep_const(ctx, i).load(generator, ctx)
+    }
+
+    /// Convenience function equivalent to `.gep(...).store(...)`.
+    pub fn set(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        i: IntValue<'ctx>,
+        value: Instance<'ctx, Item>,
+    ) {
+        self.gep(ctx, i).store(ctx, value);
+    }
+
+    /// Like `set` but `i` is a constant.
+    pub fn set_const(&self, ctx: &CodeGenContext<'ctx, '_>, i: u64, value: Instance<'ctx, Item>) {
+        self.gep_const(ctx, i).store(ctx, value);
+    }
+}

nac3core/src/codegen/model/core.rs

@@ -0,0 +1,212 @@
+use std::fmt;
+
+use inkwell::{context::Context, types::*, values::*};
+use itertools::Itertools;
+
+use super::*;
+use crate::codegen::{CodeGenContext, CodeGenerator};
+
+/// A error type for reporting any [`Model`]-related error (e.g., a [`BasicType`] mismatch).
+#[derive(Debug, Clone)]
+pub struct ModelError(pub String);
+
+impl ModelError {
+    // Append a context message to the error.
+    pub(super) fn under_context(mut self, context: &str) -> Self {
+        self.0.push_str(" ... in ");
+        self.0.push_str(context);
+        self
+    }
+}
+
+/// Trait for Rust structs identifying [`BasicType`]s in the context of a known [`CodeGenerator`] and [`CodeGenContext`].
+///
+/// For instance,
+/// - [`Int<Int32>`] identifies an [`IntType`] with 32-bits.
+/// - [`Int<SizeT>`] identifies an [`IntType`] with bit-width [`CodeGenerator::get_size_type`].
+/// - [`Ptr<Int<SizeT>>`] identifies a [`PointerType`] that points to an [`IntType`] with bit-width [`CodeGenerator::get_size_type`].
+/// - [`Int<AnyInt>`] identifies an [`IntType`] with bit-width of whatever is set in the [`AnyInt`] object.
+/// - [`Any`] identifies a [`BasicType`] set in the [`Any`] object itself.
+///
+/// You can get the [`BasicType`] out of a model with [`Model::get_type`].
+///
+/// Furthermore, [`Instance<'ctx, M>`] is a simple structure that carries a [`BasicValue`] with a [`BasicType`] identified by model `M`.
+///
+/// The main purpose of this abstraction is to have a more Rust type-safe way to use Inkwell and give type-hints
+/// for programmers.
+///
+/// ### Notes on `Default` trait
+///
+/// For some models like [`Int<Int32>`] or [`Int<SizeT>`], they have a [`Default`] trait since just by looking at the type, it is possible
+/// to tell which [`BasicType`] they are identifying.
+///
+/// This can be used to create strongly-typed interfaces accepting only values of a specific [`BasicType`] without having to worry about
+/// writing debug assertions to check if the programmer has passed in an [`IntValue`] with the wrong bit-width.
+/// ```ignore
+/// fn give_me_i32_and_get_a_size_t_back<'ctx>(i32: Instance<'ctx, Int<Int32>>) -> Instance<'ctx, Int<SizeT>> {
+///     // code...
+/// }
+/// ```
+///
+/// ### Notes on converting between Inkwell and model.
+///
+/// Suppose you have an [`IntValue`], and you want to pass it into a function that takes a [`Instance<'ctx, Int<Int32>>`]. You can do use
+/// [`Model::check_value`] or [`Model::believe_value`].
+/// ```ignore
+/// let my_value: IntValue<'ctx>;
+///
+/// let my_value = Int(Int32).check_value(my_value).unwrap(); // Panics if `my_value` is not 32-bit with a descriptive error message.
+///
+/// // or, if you are absolutely certain that `my_value` is 32-bit and doing extra checks is a waste of time:
+/// let my_value = Int(Int32).believe_value(my_value);
+/// ```
+pub trait Model<'ctx>: fmt::Debug + Clone + Copy {
+    /// The [`BasicType`] *variant* this model is identifying.
+    ///
+    /// For [`Int<Int32>`], [`Int<SizeT>`], [`Int<Any>`], etc, this is [`IntValue`];
+    ///
+    /// For [`Ptr<???>`], etc, this is [`PointerValue`];
+    ///
+    /// For [`Any`], this is just [`BasicValueEnum`];
+    ///
+    /// and so on.
+    type Type: BasicType<'ctx>;
+
+    /// The [`BasicValue`] type of the [`BasicType`] of this model.
+    type Value: BasicValue<'ctx> + TryFrom<BasicValueEnum<'ctx>>;
+
+    /// Return the [`BasicType`] of this model.
+    #[must_use]
+    fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type;
+
+    /// Get the number of bytes of the [`BasicType`] of this model.
+    fn sizeof<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+    ) -> IntValue<'ctx> {
+        self.get_type(generator, ctx).size_of().unwrap()
+    }
+
+    /// Check if a [`BasicType`] matches the [`BasicType`] of this model.
+    fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        ty: T,
+    ) -> Result<(), ModelError>;
+
+    /// Create an instance from a value with [`Instance::model`] being this model.
+    ///
+    /// Caller must make sure the type of `value` and the type of this `model` are equivalent.
+    #[must_use]
+    fn believe_value(&self, value: Self::Value) -> Instance<'ctx, Self> {
+        Instance { model: *self, value }
+    }
+
+    /// Check if a [`BasicValue`]'s type is equivalent to the type of this model.
+    /// Wrap it into an [`Instance`] if it is.
+    fn check_value<V: BasicValue<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        value: V,
+    ) -> Result<Instance<'ctx, Self>, ModelError> {
+        let value = value.as_basic_value_enum();
+        self.check_type(generator, ctx, value.get_type())
+            .map_err(|err| err.under_context(format!("the value {value:?}").as_str()))?;
+
+        let Ok(value) = Self::Value::try_from(value) else {
+            unreachable!("check_type() has bad implementation")
+        };
+        Ok(self.believe_value(value))
+    }
+
+    // Allocate a value on the stack and return its pointer.
+    fn alloca<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Ptr<Self>> {
+        let p = ctx.builder.build_alloca(self.get_type(generator, ctx.ctx), "").unwrap();
+        Ptr(*self).believe_value(p)
+    }
+
+    // Allocate an array on the stack and return its pointer.
+    fn array_alloca<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        len: IntValue<'ctx>,
+    ) -> Instance<'ctx, Ptr<Self>> {
+        let p = ctx.builder.build_array_alloca(self.get_type(generator, ctx.ctx), len, "").unwrap();
+        Ptr(*self).believe_value(p)
+    }
+
+    fn var_alloca<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        name: Option<&str>,
+    ) -> Result<Instance<'ctx, Ptr<Self>>, String> {
+        let ty = self.get_type(generator, ctx.ctx).as_basic_type_enum();
+        let p = generator.gen_var_alloc(ctx, ty, name)?;
+        Ok(Ptr(*self).believe_value(p))
+    }
+
+    fn array_var_alloca<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        len: IntValue<'ctx>,
+        name: Option<&'ctx str>,
+    ) -> Result<Instance<'ctx, Ptr<Self>>, String> {
+        // TODO: Remove ArraySliceValue
+        let ty = self.get_type(generator, ctx.ctx).as_basic_type_enum();
+        let p = generator.gen_array_var_alloc(ctx, ty, len, name)?;
+        Ok(Ptr(*self).believe_value(PointerValue::from(p)))
+    }
+
+    /// Allocate a constant array.
+    fn const_array<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        values: &[Instance<'ctx, Self>],
+    ) -> Instance<'ctx, Array<AnyLen, Self>> {
+        macro_rules! make {
+            ($t:expr, $into_value:expr) => {
+                $t.const_array(
+                    &values
+                        .iter()
+                        .map(|x| $into_value(x.value.as_basic_value_enum()))
+                        .collect_vec(),
+                )
+            };
+        }
+
+        let value = match self.get_type(generator, ctx).as_basic_type_enum() {
+            BasicTypeEnum::ArrayType(t) => make!(t, BasicValueEnum::into_array_value),
+            BasicTypeEnum::IntType(t) => make!(t, BasicValueEnum::into_int_value),
+            BasicTypeEnum::FloatType(t) => make!(t, BasicValueEnum::into_float_value),
+            BasicTypeEnum::PointerType(t) => make!(t, BasicValueEnum::into_pointer_value),
+            BasicTypeEnum::StructType(t) => make!(t, BasicValueEnum::into_struct_value),
+            BasicTypeEnum::VectorType(t) => make!(t, BasicValueEnum::into_vector_value),
+        };
+
+        Array { len: AnyLen(values.len() as u32), item: *self }
+            .check_value(generator, ctx, value)
+            .unwrap()
+    }
+}
+
+#[derive(Debug, Clone, Copy)]
+pub struct Instance<'ctx, M: Model<'ctx>> {
+    /// The model of this instance.
+    pub model: M,
+    /// The value of this instance.
+    ///
+    /// Caller must make sure the type of `value` and the type of this `model` are equivalent,
+    /// down to having the same [`IntType::get_bit_width`] in case of [`IntType`] for example.
+    pub value: M::Value,
+}

nac3core/src/codegen/model/float.rs

@@ -0,0 +1,86 @@
+use std::fmt;
+
+use inkwell::{context::Context, types::FloatType, values::FloatValue};
+
+use crate::codegen::CodeGenerator;
+
+use super::*;
+
+pub trait FloatKind<'ctx>: fmt::Debug + Clone + Copy {
+    fn get_float_type<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &G,
+        ctx: &'ctx Context,
+    ) -> FloatType<'ctx>;
+}
+
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Float32;
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Float64;
+
+impl<'ctx> FloatKind<'ctx> for Float32 {
+    fn get_float_type<G: CodeGenerator + ?Sized>(
+        &self,
+        _generator: &G,
+        ctx: &'ctx Context,
+    ) -> FloatType<'ctx> {
+        ctx.f32_type()
+    }
+}
+
+impl<'ctx> FloatKind<'ctx> for Float64 {
+    fn get_float_type<G: CodeGenerator + ?Sized>(
+        &self,
+        _generator: &G,
+        ctx: &'ctx Context,
+    ) -> FloatType<'ctx> {
+        ctx.f64_type()
+    }
+}
+
+#[derive(Debug, Clone, Copy)]
+pub struct AnyFloat<'ctx>(FloatType<'ctx>);
+
+impl<'ctx> FloatKind<'ctx> for AnyFloat<'ctx> {
+    fn get_float_type<G: CodeGenerator + ?Sized>(
+        &self,
+        _generator: &G,
+        _ctx: &'ctx Context,
+    ) -> FloatType<'ctx> {
+        self.0
+    }
+}
+
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Float<N>(pub N);
+
+impl<'ctx, N: FloatKind<'ctx>> Model<'ctx> for Float<N> {
+    type Value = FloatValue<'ctx>;
+    type Type = FloatType<'ctx>;
+
+    fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
+        self.0.get_float_type(generator, ctx)
+    }
+
+    fn check_type<T: inkwell::types::BasicType<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        ty: T,
+    ) -> Result<(), ModelError> {
+        let ty = ty.as_basic_type_enum();
+        let Ok(ty) = FloatType::try_from(ty) else {
+            return Err(ModelError(format!("Expecting FloatType, but got {ty:?}")));
+        };
+
+        let exp_ty = self.0.get_float_type(generator, ctx);
+
+        // TODO: Inkwell does not have get_bit_width for FloatType?
+        if ty != exp_ty {
+            return Err(ModelError(format!("Expecting {exp_ty:?}, but got {ty:?}")));
+        }
+
+        Ok(())
+    }
+}

nac3core/src/codegen/model/function.rs

@@ -0,0 +1,103 @@
+use inkwell::{
+    attributes::{Attribute, AttributeLoc},
+    types::{BasicMetadataTypeEnum, BasicType, FunctionType},
+    values::{AnyValue, BasicMetadataValueEnum, BasicValue, BasicValueEnum, CallSiteValue},
+};
+use itertools::Itertools;
+
+use crate::codegen::{CodeGenContext, CodeGenerator};
+
+use super::*;
+
+#[derive(Debug, Clone, Copy)]
+struct Arg<'ctx> {
+    ty: BasicMetadataTypeEnum<'ctx>,
+    val: BasicMetadataValueEnum<'ctx>,
+}
+
+/// A convenience structure to construct & call an LLVM function.
+pub struct CallFunction<'ctx, 'a, 'b, 'c, 'd, G: CodeGenerator + ?Sized> {
+    generator: &'d mut G,
+    ctx: &'b CodeGenContext<'ctx, 'a>,
+    /// Function name
+    name: &'c str,
+    /// Call arguments
+    args: Vec<Arg<'ctx>>,
+    /// LLVM function Attributes
+    attrs: Vec<&'static str>,
+}
+
+impl<'ctx, 'a, 'b, 'c, 'd, G: CodeGenerator + ?Sized> CallFunction<'ctx, 'a, 'b, 'c, 'd, G> {
+    pub fn begin(generator: &'d mut G, ctx: &'b CodeGenContext<'ctx, 'a>, name: &'c str) -> Self {
+        CallFunction { generator, ctx, name, args: Vec::new(), attrs: Vec::new() }
+    }
+
+    /// Push a list of LLVM function attributes to the function declaration.
+    #[must_use]
+    pub fn attrs(mut self, attrs: Vec<&'static str>) -> Self {
+        self.attrs = attrs;
+        self
+    }
+
+    /// Push a call argument to the function call.
+    #[allow(clippy::needless_pass_by_value)]
+    #[must_use]
+    pub fn arg<M: Model<'ctx>>(mut self, arg: Instance<'ctx, M>) -> Self {
+        let arg = Arg {
+            ty: arg.model.get_type(self.generator, self.ctx.ctx).as_basic_type_enum().into(),
+            val: arg.value.as_basic_value_enum().into(),
+        };
+        self.args.push(arg);
+        self
+    }
+
+    /// Call the function and expect the function to return a value of type of `return_model`.
+    #[must_use]
+    pub fn returning<M: Model<'ctx>>(self, name: &str, return_model: M) -> Instance<'ctx, M> {
+        let ret_ty = return_model.get_type(self.generator, self.ctx.ctx);
+
+        let ret = self.call(|tys| ret_ty.fn_type(tys, false), name);
+        let ret = BasicValueEnum::try_from(ret.as_any_value_enum()).unwrap(); // Must work
+        let ret = return_model.check_value(self.generator, self.ctx.ctx, ret).unwrap(); // Must work
+        ret
+    }
+
+    /// Like [`CallFunction::returning_`] but `return_model` is automatically inferred.
+    #[must_use]
+    pub fn returning_auto<M: Model<'ctx> + Default>(self, name: &str) -> Instance<'ctx, M> {
+        self.returning(name, M::default())
+    }
+
+    /// Call the function and expect the function to return a void-type.
+    pub fn returning_void(self) {
+        let ret_ty = self.ctx.ctx.void_type();
+
+        let _ = self.call(|tys| ret_ty.fn_type(tys, false), "");
+    }
+
+    fn call<F>(&self, make_fn_type: F, return_value_name: &str) -> CallSiteValue<'ctx>
+    where
+        F: FnOnce(&[BasicMetadataTypeEnum<'ctx>]) -> FunctionType<'ctx>,
+    {
+        // Get the LLVM function.
+        let func = self.ctx.module.get_function(self.name).unwrap_or_else(|| {
+            // Declare the function if it doesn't exist.
+            let tys = self.args.iter().map(|arg| arg.ty).collect_vec();
+
+            let func_type = make_fn_type(&tys);
+            let func = self.ctx.module.add_function(self.name, func_type, None);
+
+            for attr in &self.attrs {
+                func.add_attribute(
+                    AttributeLoc::Function,
+                    self.ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
+                );
+            }
+
+            func
+        });
+
+        let vals = self.args.iter().map(|arg| arg.val).collect_vec();
+        self.ctx.builder.build_call(func, &vals, return_value_name).unwrap()
+    }
+}

nac3core/src/codegen/model/int.rs

@@ -0,0 +1,417 @@
+use std::{cmp::Ordering, fmt};
+
+use inkwell::{
+    context::Context,
+    types::{BasicType, IntType},
+    values::IntValue,
+    IntPredicate,
+};
+
+use crate::codegen::{CodeGenContext, CodeGenerator};
+
+use super::*;
+
+pub trait IntKind<'ctx>: fmt::Debug + Clone + Copy {
+    fn get_int_type<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &G,
+        ctx: &'ctx Context,
+    ) -> IntType<'ctx>;
+}
+
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Bool;
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Byte;
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Int32;
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Int64;
+#[derive(Debug, Clone, Copy, Default)]
+pub struct SizeT;
+
+impl<'ctx> IntKind<'ctx> for Bool {
+    fn get_int_type<G: CodeGenerator + ?Sized>(
+        &self,
+        _generator: &G,
+        ctx: &'ctx Context,
+    ) -> IntType<'ctx> {
+        ctx.bool_type()
+    }
+}
+
+impl<'ctx> IntKind<'ctx> for Byte {
+    fn get_int_type<G: CodeGenerator + ?Sized>(
+        &self,
+        _generator: &G,
+        ctx: &'ctx Context,
+    ) -> IntType<'ctx> {
+        ctx.i8_type()
+    }
+}
+
+impl<'ctx> IntKind<'ctx> for Int32 {
+    fn get_int_type<G: CodeGenerator + ?Sized>(
+        &self,
+        _generator: &G,
+        ctx: &'ctx Context,
+    ) -> IntType<'ctx> {
+        ctx.i32_type()
+    }
+}
+
+impl<'ctx> IntKind<'ctx> for Int64 {
+    fn get_int_type<G: CodeGenerator + ?Sized>(
+        &self,
+        _generator: &G,
+        ctx: &'ctx Context,
+    ) -> IntType<'ctx> {
+        ctx.i64_type()
+    }
+}
+
+impl<'ctx> IntKind<'ctx> for SizeT {
+    fn get_int_type<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &G,
+        ctx: &'ctx Context,
+    ) -> IntType<'ctx> {
+        generator.get_size_type(ctx)
+    }
+}
+
+#[derive(Debug, Clone, Copy)]
+pub struct AnyInt<'ctx>(pub IntType<'ctx>);
+
+impl<'ctx> IntKind<'ctx> for AnyInt<'ctx> {
+    fn get_int_type<G: CodeGenerator + ?Sized>(
+        &self,
+        _generator: &G,
+        _ctx: &'ctx Context,
+    ) -> IntType<'ctx> {
+        self.0
+    }
+}
+
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Int<N>(pub N);
+
+impl<'ctx, N: IntKind<'ctx>> Model<'ctx> for Int<N> {
+    type Value = IntValue<'ctx>;
+    type Type = IntType<'ctx>;
+
+    fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
+        self.0.get_int_type(generator, ctx)
+    }
+
+    fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        ty: T,
+    ) -> Result<(), ModelError> {
+        let ty = ty.as_basic_type_enum();
+        let Ok(ty) = IntType::try_from(ty) else {
+            return Err(ModelError(format!("Expecting IntType, but got {ty:?}")));
+        };
+
+        let exp_ty = self.0.get_int_type(generator, ctx);
+        if ty.get_bit_width() != exp_ty.get_bit_width() {
+            return Err(ModelError(format!(
+                "Expecting IntType to have {} bit(s), but got {} bit(s)",
+                exp_ty.get_bit_width(),
+                ty.get_bit_width()
+            )));
+        }
+
+        Ok(())
+    }
+}
+
+impl<'ctx, N: IntKind<'ctx>> Int<N> {
+    pub fn const_int<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        value: u64,
+    ) -> Instance<'ctx, Self> {
+        let value = self.get_type(generator, ctx).const_int(value, false);
+        self.believe_value(value)
+    }
+
+    pub fn const_0<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+    ) -> Instance<'ctx, Self> {
+        let value = self.get_type(generator, ctx).const_zero();
+        self.believe_value(value)
+    }
+
+    pub fn const_1<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+    ) -> Instance<'ctx, Self> {
+        self.const_int(generator, ctx, 1)
+    }
+
+    pub fn const_all_ones<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+    ) -> Instance<'ctx, Self> {
+        let value = self.get_type(generator, ctx).const_all_ones();
+        self.believe_value(value)
+    }
+
+    pub fn s_extend_or_bit_cast<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> Instance<'ctx, Self> {
+        assert!(
+            value.get_type().get_bit_width()
+                <= self.0.get_int_type(generator, ctx.ctx).get_bit_width()
+        );
+        let value = ctx
+            .builder
+            .build_int_s_extend_or_bit_cast(value, self.get_type(generator, ctx.ctx), "")
+            .unwrap();
+        self.believe_value(value)
+    }
+
+    pub fn s_extend<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> Instance<'ctx, Self> {
+        assert!(
+            value.get_type().get_bit_width()
+                < self.0.get_int_type(generator, ctx.ctx).get_bit_width()
+        );
+        let value =
+            ctx.builder.build_int_s_extend(value, self.get_type(generator, ctx.ctx), "").unwrap();
+        self.believe_value(value)
+    }
+
+    pub fn z_extend_or_bit_cast<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> Instance<'ctx, Self> {
+        assert!(
+            value.get_type().get_bit_width()
+                <= self.0.get_int_type(generator, ctx.ctx).get_bit_width()
+        );
+        let value = ctx
+            .builder
+            .build_int_z_extend_or_bit_cast(value, self.get_type(generator, ctx.ctx), "")
+            .unwrap();
+        self.believe_value(value)
+    }
+
+    pub fn z_extend<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> Instance<'ctx, Self> {
+        assert!(
+            value.get_type().get_bit_width()
+                < self.0.get_int_type(generator, ctx.ctx).get_bit_width()
+        );
+        let value =
+            ctx.builder.build_int_z_extend(value, self.get_type(generator, ctx.ctx), "").unwrap();
+        self.believe_value(value)
+    }
+
+    pub fn truncate_or_bit_cast<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> Instance<'ctx, Self> {
+        assert!(
+            value.get_type().get_bit_width()
+                >= self.0.get_int_type(generator, ctx.ctx).get_bit_width()
+        );
+        let value = ctx
+            .builder
+            .build_int_truncate_or_bit_cast(value, self.get_type(generator, ctx.ctx), "")
+            .unwrap();
+        self.believe_value(value)
+    }
+
+    pub fn truncate<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> Instance<'ctx, Self> {
+        assert!(
+            value.get_type().get_bit_width()
+                > self.0.get_int_type(generator, ctx.ctx).get_bit_width()
+        );
+        let value =
+            ctx.builder.build_int_truncate(value, self.get_type(generator, ctx.ctx), "").unwrap();
+        self.believe_value(value)
+    }
+
+    /// `sext` or `trunc` an int to this model's int type. Does nothing if equal bit-widths.
+    pub fn s_extend_or_truncate<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> Instance<'ctx, Self> {
+        let their_width = value.get_type().get_bit_width();
+        let our_width = self.0.get_int_type(generator, ctx.ctx).get_bit_width();
+        match their_width.cmp(&our_width) {
+            Ordering::Less => self.s_extend(generator, ctx, value),
+            Ordering::Equal => self.believe_value(value),
+            Ordering::Greater => self.truncate(generator, ctx, value),
+        }
+    }
+
+    /// `zext` or `trunc` an int to this model's int type. Does nothing if equal bit-widths.
+    pub fn z_extend_or_truncate<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> Instance<'ctx, Self> {
+        let their_width = value.get_type().get_bit_width();
+        let our_width = self.0.get_int_type(generator, ctx.ctx).get_bit_width();
+        match their_width.cmp(&our_width) {
+            Ordering::Less => self.z_extend(generator, ctx, value),
+            Ordering::Equal => self.believe_value(value),
+            Ordering::Greater => self.truncate(generator, ctx, value),
+        }
+    }
+}
+
+impl Int<Bool> {
+    #[must_use]
+    pub fn const_false<'ctx, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+    ) -> Instance<'ctx, Self> {
+        self.const_int(generator, ctx, 0)
+    }
+
+    #[must_use]
+    pub fn const_true<'ctx, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+    ) -> Instance<'ctx, Self> {
+        self.const_int(generator, ctx, 1)
+    }
+}
+
+impl<'ctx, N: IntKind<'ctx>> Instance<'ctx, Int<N>> {
+    pub fn s_extend_or_bit_cast<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        to_int_kind: NewN,
+    ) -> Instance<'ctx, Int<NewN>> {
+        Int(to_int_kind).s_extend_or_bit_cast(generator, ctx, self.value)
+    }
+
+    pub fn s_extend<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        to_int_kind: NewN,
+    ) -> Instance<'ctx, Int<NewN>> {
+        Int(to_int_kind).s_extend(generator, ctx, self.value)
+    }
+
+    pub fn z_extend_or_bit_cast<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        to_int_kind: NewN,
+    ) -> Instance<'ctx, Int<NewN>> {
+        Int(to_int_kind).z_extend_or_bit_cast(generator, ctx, self.value)
+    }
+
+    pub fn z_extend<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        to_int_kind: NewN,
+    ) -> Instance<'ctx, Int<NewN>> {
+        Int(to_int_kind).z_extend(generator, ctx, self.value)
+    }
+
+    pub fn truncate_or_bit_cast<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        to_int_kind: NewN,
+    ) -> Instance<'ctx, Int<NewN>> {
+        Int(to_int_kind).truncate_or_bit_cast(generator, ctx, self.value)
+    }
+
+    pub fn truncate<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        to_int_kind: NewN,
+    ) -> Instance<'ctx, Int<NewN>> {
+        Int(to_int_kind).truncate(generator, ctx, self.value)
+    }
+
+    pub fn s_extend_or_truncate<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        to_int_kind: NewN,
+    ) -> Instance<'ctx, Int<NewN>> {
+        Int(to_int_kind).s_extend_or_truncate(generator, ctx, self.value)
+    }
+
+    pub fn z_extend_or_truncate<NewN: IntKind<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        to_int_kind: NewN,
+    ) -> Instance<'ctx, Int<NewN>> {
+        Int(to_int_kind).z_extend_or_truncate(generator, ctx, self.value)
+    }
+
+    #[must_use]
+    pub fn add(&self, ctx: &CodeGenContext<'ctx, '_>, other: Self) -> Self {
+        let value = ctx.builder.build_int_add(self.value, other.value, "").unwrap();
+        self.model.believe_value(value)
+    }
+
+    #[must_use]
+    pub fn sub(&self, ctx: &CodeGenContext<'ctx, '_>, other: Self) -> Self {
+        let value = ctx.builder.build_int_sub(self.value, other.value, "").unwrap();
+        self.model.believe_value(value)
+    }
+
+    #[must_use]
+    pub fn mul(&self, ctx: &CodeGenContext<'ctx, '_>, other: Self) -> Self {
+        let value = ctx.builder.build_int_mul(self.value, other.value, "").unwrap();
+        self.model.believe_value(value)
+    }
+
+    pub fn compare(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        op: IntPredicate,
+        other: Self,
+    ) -> Instance<'ctx, Int<Bool>> {
+        let value = ctx.builder.build_int_compare(op, self.value, other.value, "").unwrap();
+        Int(Bool).believe_value(value)
+    }
+}

nac3core/src/codegen/model/mod.rs

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

nac3core/src/codegen/model/ptr.rs

@@ -0,0 +1,207 @@
+use inkwell::{
+    context::Context,
+    types::{BasicType, BasicTypeEnum, PointerType},
+    values::{IntValue, PointerValue},
+    AddressSpace,
+};
+
+use crate::codegen::{llvm_intrinsics::call_memcpy_generic, CodeGenContext, CodeGenerator};
+
+use super::*;
+
+/// A model for [`PointerType`].
+// TODO: LLVM 15: `Item` is a Rust type-hint for the LLVM type of value the `.store()/.load()` family
+// of functions return. If a truly opaque pointer is needed, tell the programmer to use `OpaquePtr`.
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Ptr<Item>(pub Item);
+
+/// An opaque pointer. Like [`Ptr`] but without any Rust type-hints about its element type.
+///
+/// `.load()/.store()` is not available for [`Instance`]s of opaque pointers.
+pub type OpaquePtr = Ptr<()>;
+
+// TODO: LLVM 15: `Item: Model<'ctx>` don't even need to be a model anymore. It will only be
+// a type-hint for the `.load()/.store()` functions for the `pointee_ty`.
+//
+// See https://thedan64.github.io/inkwell/inkwell/builder/struct.Builder.html#method.build_load.
+impl<'ctx, Item: Model<'ctx>> Model<'ctx> for Ptr<Item> {
+    type Value = PointerValue<'ctx>;
+    type Type = PointerType<'ctx>;
+
+    fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
+        // TODO: LLVM 15: ctx.ptr_type(AddressSpace::default())
+        self.0.get_type(generator, ctx).ptr_type(AddressSpace::default())
+    }
+
+    fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        ty: T,
+    ) -> Result<(), ModelError> {
+        let ty = ty.as_basic_type_enum();
+        let Ok(ty) = PointerType::try_from(ty) else {
+            return Err(ModelError(format!("Expecting PointerType, but got {ty:?}")));
+        };
+
+        let elem_ty = ty.get_element_type();
+        let Ok(elem_ty) = BasicTypeEnum::try_from(elem_ty) else {
+            return Err(ModelError(format!(
+                "Expecting pointer element type to be a BasicTypeEnum, but got {elem_ty:?}"
+            )));
+        };
+
+        // TODO: inkwell `get_element_type()` will be deprecated.
+        // Remove the check for `get_element_type()` when the time comes.
+        self.0
+            .check_type(generator, ctx, elem_ty)
+            .map_err(|err| err.under_context("a PointerType"))?;
+
+        Ok(())
+    }
+}
+
+impl<'ctx, Item: Model<'ctx>> Ptr<Item> {
+    /// Return a ***constant*** nullptr.
+    pub fn nullptr<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+    ) -> Instance<'ctx, Ptr<Item>> {
+        let ptr = self.get_type(generator, ctx).const_null();
+        self.believe_value(ptr)
+    }
+
+    /// Cast a pointer into this model with [`inkwell::builder::Builder::build_pointer_cast`]
+    pub fn pointer_cast<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        ptr: PointerValue<'ctx>,
+    ) -> Instance<'ctx, Ptr<Item>> {
+        // TODO: LLVM 15: Write in an impl where `Item` does not have to be `Model<'ctx>`.
+        // TODO: LLVM 15: This function will only have to be:
+        // ```
+        // return self.believe_value(ptr);
+        // ```
+        let t = self.get_type(generator, ctx.ctx);
+        let ptr = ctx.builder.build_pointer_cast(ptr, t, "").unwrap();
+        self.believe_value(ptr)
+    }
+}
+
+impl<'ctx, Item: Model<'ctx>> Instance<'ctx, Ptr<Item>> {
+    /// Offset the pointer by [`inkwell::builder::Builder::build_in_bounds_gep`].
+    #[must_use]
+    pub fn offset(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        offset: IntValue<'ctx>,
+    ) -> Instance<'ctx, Ptr<Item>> {
+        let p = unsafe { ctx.builder.build_in_bounds_gep(self.value, &[offset], "").unwrap() };
+        self.model.believe_value(p)
+    }
+
+    /// Offset the pointer by [`inkwell::builder::Builder::build_in_bounds_gep`] by a constant offset.
+    #[must_use]
+    pub fn offset_const(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        offset: u64,
+    ) -> Instance<'ctx, Ptr<Item>> {
+        let offset = ctx.ctx.i32_type().const_int(offset, false);
+        self.offset(ctx, offset)
+    }
+
+    pub fn set_index(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        index: IntValue<'ctx>,
+        value: Instance<'ctx, Item>,
+    ) {
+        self.offset(ctx, index).store(ctx, value);
+    }
+
+    pub fn set_index_const(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        index: u64,
+        value: Instance<'ctx, Item>,
+    ) {
+        self.offset_const(ctx, index).store(ctx, value);
+    }
+
+    pub fn get_index<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        index: IntValue<'ctx>,
+    ) -> Instance<'ctx, Item> {
+        self.offset(ctx, index).load(generator, ctx)
+    }
+
+    pub fn get_index_const<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        index: u64,
+    ) -> Instance<'ctx, Item> {
+        self.offset_const(ctx, index).load(generator, ctx)
+    }
+
+    /// Load the value with [`inkwell::builder::Builder::build_load`].
+    pub fn load<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Item> {
+        let value = ctx.builder.build_load(self.value, "").unwrap();
+        self.model.0.check_value(generator, ctx.ctx, value).unwrap() // If unwrap() panics, there is a logic error.
+    }
+
+    /// Store a value with [`inkwell::builder::Builder::build_store`].
+    pub fn store(&self, ctx: &CodeGenContext<'ctx, '_>, value: Instance<'ctx, Item>) {
+        ctx.builder.build_store(self.value, value.value).unwrap();
+    }
+
+    /// Return a casted pointer of element type `NewElement` with [`inkwell::builder::Builder::build_pointer_cast`].
+    pub fn pointer_cast<NewItem: Model<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        new_item: NewItem,
+    ) -> Instance<'ctx, Ptr<NewItem>> {
+        // TODO: LLVM 15: Write in an impl where `Item` does not have to be `Model<'ctx>`.
+        Ptr(new_item).pointer_cast(generator, ctx, self.value)
+    }
+
+    /// Check if the pointer is null with [`inkwell::builder::Builder::build_is_null`].
+    pub fn is_null(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Int<Bool>> {
+        let value = ctx.builder.build_is_null(self.value, "").unwrap();
+        Int(Bool).believe_value(value)
+    }
+
+    /// Check if the pointer is not null with [`inkwell::builder::Builder::build_is_not_null`].
+    pub fn is_not_null(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Int<Bool>> {
+        let value = ctx.builder.build_is_not_null(self.value, "").unwrap();
+        Int(Bool).believe_value(value)
+    }
+
+    /// `memcpy` from another pointer.
+    pub fn copy_from<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        source: Self,
+        num_items: IntValue<'ctx>,
+    ) {
+        // Force extend `num_items` and `itemsize` to `i64` so their types would match.
+        let itemsize = self.model.sizeof(generator, ctx.ctx);
+        let itemsize = Int(Int64).z_extend_or_truncate(generator, ctx, itemsize);
+        let num_items = Int(Int64).z_extend_or_truncate(generator, ctx, num_items);
+        let totalsize = itemsize.mul(ctx, num_items);
+
+        let is_volatile = ctx.ctx.bool_type().const_zero(); // is_volatile = false
+        call_memcpy_generic(ctx, self.value, source.value, totalsize.value, is_volatile);
+    }
+}

nac3core/src/codegen/model/structure.rs

@@ -0,0 +1,345 @@
+use std::fmt;
+
+use inkwell::{
+    context::Context,
+    types::{BasicType, BasicTypeEnum, StructType},
+    values::{BasicValueEnum, StructValue},
+};
+
+use crate::codegen::{CodeGenContext, CodeGenerator};
+
+use super::*;
+
+/// A traveral that traverses a Rust `struct` that is used to declare an LLVM's struct's field types.
+pub trait FieldTraversal<'ctx> {
+    /// Output type of [`FieldTraversal::add`].
+    type Out<M>;
+
+    /// Traverse through the type of a declared field and do something with it.
+    ///
+    /// * `name` - The cosmetic name of the LLVM field. Used for debugging.
+    /// * `model` - The [`Model`] representing the LLVM type of this field.
+    fn add<M: Model<'ctx>>(&mut self, name: &'static str, model: M) -> Self::Out<M>;
+
+    /// Like [`FieldTraversal::add`] but [`Model`] is automatically inferred from its [`Default`] trait.
+    fn add_auto<M: Model<'ctx> + Default>(&mut self, name: &'static str) -> Self::Out<M> {
+        self.add(name, M::default())
+    }
+}
+
+/// Descriptor of an LLVM struct field.
+#[derive(Debug, Clone, Copy)]
+pub struct GepField<M> {
+    /// The GEP index of this field. This is the index to use with `build_gep`.
+    pub gep_index: u64,
+    /// The cosmetic name of this field.
+    pub name: &'static str,
+    /// The [`Model`] of this field's type.
+    pub model: M,
+}
+
+/// A traversal to get the GEP index of fields.
+pub struct GepFieldTraversal {
+    /// The current GEP index.
+    gep_index_counter: u64,
+}
+
+impl<'ctx> FieldTraversal<'ctx> for GepFieldTraversal {
+    type Out<M> = GepField<M>;
+
+    fn add<M: Model<'ctx>>(&mut self, name: &'static str, model: M) -> Self::Out<M> {
+        let gep_index = self.gep_index_counter;
+        self.gep_index_counter += 1;
+        Self::Out { gep_index, name, model }
+    }
+}
+
+/// A traversal to collect the field types of a struct.
+///
+/// This is used to collect the field types for [`Context::struct_type`].
+struct TypeFieldTraversal<'ctx, 'a, G: CodeGenerator + ?Sized> {
+    generator: &'a G,
+    ctx: &'ctx Context,
+    /// The collected field types so far, in order.
+    field_types: Vec<BasicTypeEnum<'ctx>>,
+}
+
+impl<'ctx, 'a, G: CodeGenerator + ?Sized> FieldTraversal<'ctx> for TypeFieldTraversal<'ctx, 'a, G> {
+    type Out<M> = (); // Checking types return nothing.
+
+    fn add<M: Model<'ctx>>(&mut self, _name: &'static str, model: M) -> Self::Out<M> {
+        let t = model.get_type(self.generator, self.ctx).as_basic_type_enum();
+        self.field_types.push(t);
+    }
+}
+
+/// A traversal to check the field types of a [`StructType`].
+struct CheckTypeFieldTraversal<'ctx, 'a, G: CodeGenerator + ?Sized> {
+    generator: &'a mut G,
+    ctx: &'ctx Context,
+    /// The current GEP index, so we can tell the index of the field we are checking
+    /// and report the GEP index.
+    index: u32,
+    /// The [`StructType`] to check.
+    scrutinee: StructType<'ctx>,
+    /// The list of collected errors so far.
+    errors: Vec<ModelError>,
+}
+
+impl<'ctx, 'a, G: CodeGenerator + ?Sized> FieldTraversal<'ctx>
+    for CheckTypeFieldTraversal<'ctx, 'a, G>
+{
+    type Out<M> = (); // Checking types return nothing.
+
+    fn add<M: Model<'ctx>>(&mut self, name: &'static str, model: M) -> Self::Out<M> {
+        let i = self.index;
+        self.index += 1;
+
+        if let Some(t) = self.scrutinee.get_field_type_at_index(i) {
+            if let Err(err) = model.check_type(self.generator, self.ctx, t) {
+                self.errors.push(err.under_context(format!("field #{i} '{name}'").as_str()));
+            }
+        } // Otherwise, it will be caught by Struct's `check_type`.
+    }
+}
+
+/// A trait for Rust structs identifying LLVM structures.
+///
+/// ### Example
+///
+/// Suppose you want to define this structure:
+/// ```c
+/// template <typename T>
+/// struct ContiguousNDArray {
+///     size_t ndims;
+///     size_t* shape;
+///     T* data;
+/// }
+/// ```
+///
+/// This is how it should be done:
+/// ```ignore
+/// pub struct ContiguousNDArrayFields<'ctx, F: FieldTraversal<'ctx>, Item: Model<'ctx>> {
+///     pub ndims: F::Out<Int<SizeT>>,
+///     pub shape: F::Out<Ptr<Int<SizeT>>>,
+///     pub data: F::Out<Ptr<Item>>,
+/// }
+///
+/// /// An ndarray without strides and non-opaque `data` field in NAC3.
+/// #[derive(Debug, Clone, Copy)]
+/// pub struct ContiguousNDArray<M> {
+///     /// [`Model`] of the items.
+///     pub item: M,
+/// }
+///
+/// impl<'ctx, Item: Model<'ctx>> StructKind<'ctx> for ContiguousNDArray<Item> {
+///     type Fields<F: FieldTraversal<'ctx>> = ContiguousNDArrayFields<'ctx, F, Item>;
+///
+///     fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
+///         // The order of `traversal.add*` is important
+///         Self::Fields {
+///             ndims: traversal.add_auto("ndims"),
+///             shape: traversal.add_auto("shape"),
+///             data: traversal.add("data", Ptr(self.item)),
+///         }
+///     }
+/// }
+/// ```
+///
+/// The [`FieldTraversal`] here is a mechanism to allow the fields of `ContiguousNDArrayFields` to be
+/// traversed to do useful work such as:
+///
+/// - To create the [`StructType`] of `ContiguousNDArray` by collecting [`BasicType`]s of the fields.
+/// - To enable the `.gep(ctx, |f| f.ndims).store(ctx, ...)` syntax.
+///
+/// Suppose now that you have defined `ContiguousNDArray` and you want to allocate a `ContiguousNDArray`
+/// with dtype `float64` in LLVM, this is how you do it:
+/// ```rust
+/// type F64NDArray = ContiguousNDArray<Float<Float64>>; // Type alias for leaner documentation
+/// let model: Struct<F64NDArray> = Struct(ContigousNDArray { item: Float(Float64) });
+/// // In fact you may even do `let model = Struct<F64NDArray>::default()`.
+/// let ndarray: Instance<'ctx, Ptr<F64NDArray>> = model.alloca(generator, ctx);
+/// ```
+///
+/// ...and here is how you may manipulate/access `ndarray`:
+///
+/// (NOTE: some arguments have been omitted)
+///
+/// ```rust
+/// // Get `&ndarray->data`
+/// ndarray.gep(|f| f.data); // type: Instance<'ctx, Ptr<Float<Float64>>>
+///
+/// // Get `ndarray->ndims`
+/// ndarray.get(|f| f.ndims); // type: Instance<'ctx, Int<SizeT>>
+///
+/// // Get `&ndarray->ndims`
+/// ndarray.gep(|f| f.ndims); // type: Instance<'ctx, Ptr<Int<SizeT>>>
+///
+/// // Get `ndarray->shape[0]`
+/// ndarray.get(|f| f.shape).get_index_const(0); // Instance<'ctx, Int<SizeT>>
+///
+/// // Get `&ndarray->shape[2]`
+/// ndarray.get(|f| f.shape).offset_const(2); // Instance<'ctx, Ptr<Int<SizeT>>>
+///
+/// // Do `ndarray->ndims = 3;`
+/// let num_3 = Int(SizeT).const_int(3);
+/// ndarray.set(|f| f.ndims, num_3);
+/// ```
+pub trait StructKind<'ctx>: fmt::Debug + Clone + Copy {
+    /// The associated fields of this struct.
+    type Fields<F: FieldTraversal<'ctx>>;
+
+    /// Traverse map through all fields of this [`StructKind`].
+    fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F>;
+
+    /// Get a convenience structure to get a struct field's GEP index through its corresponding Rust field.
+    fn fields(&self) -> Self::Fields<GepFieldTraversal> {
+        self.traverse_fields(&mut GepFieldTraversal { gep_index_counter: 0 })
+    }
+
+    /// Get the LLVM [`StructType`] of this [`StructKind`].
+    fn get_struct_type<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &G,
+        ctx: &'ctx Context,
+    ) -> StructType<'ctx> {
+        let mut traversal = TypeFieldTraversal { generator, ctx, field_types: Vec::new() };
+        self.traverse_fields(&mut traversal);
+
+        ctx.struct_type(&traversal.field_types, false)
+    }
+}
+
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Struct<S>(pub S);
+
+impl<'ctx, S: StructKind<'ctx>> Struct<S> {
+    /// Create a constant struct value.
+    ///
+    /// This function also validates `fields` and panics types don't match.
+    pub fn const_struct<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        fields: &[BasicValueEnum<'ctx>],
+    ) -> Instance<'ctx, Self> {
+        // NOTE: There *could* have been a functor `F<M> = Instance<'ctx, M>` for `S::Fields<F>`
+        // to create a more user-friendly interface, but Rust's type system is not sophisticated enough
+        // and if you try doing that Rust would force you put lifetimes everywhere.
+        let val = ctx.const_struct(fields, false);
+        self.check_value(generator, ctx, val).unwrap()
+    }
+}
+
+impl<'ctx, S: StructKind<'ctx>> Model<'ctx> for Struct<S> {
+    type Value = StructValue<'ctx>;
+    type Type = StructType<'ctx>;
+
+    fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
+        self.0.get_struct_type(generator, ctx)
+    }
+
+    fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        ty: T,
+    ) -> Result<(), ModelError> {
+        let ty = ty.as_basic_type_enum();
+        let Ok(ty) = StructType::try_from(ty) else {
+            return Err(ModelError(format!("Expecting StructType, but got {ty:?}")));
+        };
+
+        let mut traversal =
+            CheckTypeFieldTraversal { generator, ctx, index: 0, errors: Vec::new(), scrutinee: ty };
+        self.0.traverse_fields(&mut traversal);
+
+        let exp_num_fields = traversal.index;
+        let got_num_fields = u32::try_from(ty.get_field_types().len()).unwrap();
+        if exp_num_fields != got_num_fields {
+            return Err(ModelError(format!(
+                "Expecting StructType with {exp_num_fields} field(s), but got {got_num_fields}"
+            )));
+        }
+
+        if !traversal.errors.is_empty() {
+            // Currently, only the first error is reported.
+            return Err(traversal.errors[0].clone());
+        }
+
+        Ok(())
+    }
+}
+
+impl<'ctx, S: StructKind<'ctx>> Instance<'ctx, Struct<S>> {
+    /// Get a field with [`StructValue::get_field_at_index`].
+    pub fn get_field<G: CodeGenerator + ?Sized, M, GetField>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        get_field: GetField,
+    ) -> Instance<'ctx, M>
+    where
+        M: Model<'ctx>,
+        GetField: FnOnce(S::Fields<GepFieldTraversal>) -> GepField<M>,
+    {
+        let field = get_field(self.model.0.fields());
+        let val = self.value.get_field_at_index(field.gep_index as u32).unwrap();
+        field.model.check_value(generator, ctx, val).unwrap()
+    }
+}
+
+impl<'ctx, S: StructKind<'ctx>> Instance<'ctx, Ptr<Struct<S>>> {
+    /// Get a pointer to a field with [`Builder::build_in_bounds_gep`].
+    pub fn gep<M, GetField>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        get_field: GetField,
+    ) -> Instance<'ctx, Ptr<M>>
+    where
+        M: Model<'ctx>,
+        GetField: FnOnce(S::Fields<GepFieldTraversal>) -> GepField<M>,
+    {
+        let field = get_field(self.model.0 .0.fields());
+        let llvm_i32 = ctx.ctx.i32_type();
+
+        let ptr = unsafe {
+            ctx.builder
+                .build_in_bounds_gep(
+                    self.value,
+                    &[llvm_i32.const_zero(), llvm_i32.const_int(field.gep_index, false)],
+                    field.name,
+                )
+                .unwrap()
+        };
+
+        Ptr(field.model).believe_value(ptr)
+    }
+
+    /// Convenience function equivalent to `.gep(...).load(...)`.
+    pub fn get<M, GetField, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        get_field: GetField,
+    ) -> Instance<'ctx, M>
+    where
+        M: Model<'ctx>,
+        GetField: FnOnce(S::Fields<GepFieldTraversal>) -> GepField<M>,
+    {
+        self.gep(ctx, get_field).load(generator, ctx)
+    }
+
+    /// Convenience function equivalent to `.gep(...).store(...)`.
+    pub fn set<M, GetField>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        get_field: GetField,
+        value: Instance<'ctx, M>,
+    ) where
+        M: Model<'ctx>,
+        GetField: FnOnce(S::Fields<GepFieldTraversal>) -> GepField<M>,
+    {
+        self.gep(ctx, get_field).store(ctx, value);
+    }
+}

nac3core/src/codegen/model/util.rs

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

nac3core/src/codegen/object/any.rs

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

nac3core/src/codegen/object/cslice.rs

@@ -0,0 +1,24 @@
+use crate::codegen::model::*;
+
+/// Fields of [`CSlice`]
+pub struct CSliceFields<'ctx, F: FieldTraversal<'ctx>, Item: Model<'ctx>> {
+    /// Pointer to items
+    pub base: F::Out<Ptr<Item>>,
+    /// Number of items (not bytes)
+    pub len: F::Out<Int<SizeT>>,
+}
+
+/// See <https://docs.rs/cslice/0.3.0/cslice/struct.CSlice.html>.
+///
+/// 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<Item>(pub Item);
+
+impl<'ctx, Item: Model<'ctx>> StructKind<'ctx> for CSlice<Item> {
+    type Fields<F: FieldTraversal<'ctx>> = CSliceFields<'ctx, F, Item>;
+
+    fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
+        CSliceFields { base: traversal.add("base", Ptr(self.0)), len: traversal.add_auto("len") }
+    }
+}

nac3core/src/codegen/object/exception.rs

@@ -0,0 +1,41 @@
+use crate::codegen::model::*;
+
+use super::str::Str;
+
+/// 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<'ctx, F: FieldTraversal<'ctx>> {
+    pub id: F::Out<Int<Int32>>,
+    pub filename: F::Out<Str>,
+    pub line: F::Out<Int<Int32>>,
+    pub column: F::Out<Int<Int32>>,
+    pub function: F::Out<Str>,
+    pub msg: F::Out<Str>,
+    pub params: [F::Out<Int<Int64>>; 3],
+}
+
+/// nac3core & ARTIQ's Exception
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Exception;
+
+impl<'ctx> StructKind<'ctx> for Exception {
+    type Fields<F: FieldTraversal<'ctx>> = ExceptionFields<'ctx, F>;
+
+    fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
+        Self::Fields {
+            id: traversal.add_auto("id"),
+            filename: traversal.add_auto("filename"),
+            line: traversal.add_auto("line"),
+            column: traversal.add_auto("column"),
+            function: traversal.add_auto("function"),
+            msg: traversal.add_auto("msg"),
+            params: [
+                traversal.add_auto("params[0]"),
+                traversal.add_auto("params[1]"),
+                traversal.add_auto("params[2]"),
+            ],
+        }
+    }
+}

nac3core/src/codegen/object/list.rs

@@ -0,0 +1,107 @@
+use crate::{
+    codegen::{model::*, CodeGenContext, CodeGenerator},
+    typecheck::typedef::{iter_type_vars, Type, TypeEnum},
+};
+
+use super::any::AnyObject;
+
+/// Fields of [`List`]
+pub struct ListFields<'ctx, F: FieldTraversal<'ctx>, Item: Model<'ctx>> {
+    /// Array pointer to content
+    pub items: F::Out<Ptr<Item>>,
+    /// Number of items in the array
+    pub len: F::Out<Int<SizeT>>,
+}
+
+/// A list in NAC3.
+#[derive(Debug, Clone, Copy, Default)]
+pub struct List<Item> {
+    /// Model of the list items
+    pub item: Item,
+}
+
+impl<'ctx, Item: Model<'ctx>> StructKind<'ctx> for List<Item> {
+    type Fields<F: FieldTraversal<'ctx>> = ListFields<'ctx, F, Item>;
+
+    fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
+        Self::Fields {
+            items: traversal.add("items", Ptr(self.item)),
+            len: traversal.add_auto("len"),
+        }
+    }
+}
+
+/// A NAC3 Python List object.
+#[derive(Debug, Clone, Copy)]
+pub struct ListObject<'ctx> {
+    /// Typechecker type of the list items
+    pub item_type: Type,
+    pub instance: Instance<'ctx, Ptr<Struct<List<Any<'ctx>>>>>,
+}
+
+impl<'ctx> ListObject<'ctx> {
+    /// Create a [`ListObject`] from an LLVM value and its typechecker [`Type`].
+    pub fn from_object<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        object: AnyObject<'ctx>,
+    ) -> Self {
+        // Check typechecker type and extract `item_type`
+        let item_type = match &*ctx.unifier.get_ty(object.ty) {
+            TypeEnum::TObj { obj_id, params, .. }
+                if *obj_id == ctx.primitives.list.obj_id(&ctx.unifier).unwrap() =>
+            {
+                iter_type_vars(params).next().unwrap().ty // Extract `item_type`
+            }
+            _ => {
+                panic!("Expecting type to be a list, but got {}", ctx.unifier.stringify(object.ty))
+            }
+        };
+
+        let plist = Ptr(Struct(List { item: Any(ctx.get_llvm_type(generator, item_type)) }));
+
+        // Create object
+        let value = plist.check_value(generator, ctx.ctx, object.value).unwrap();
+        ListObject { item_type, instance: value }
+    }
+
+    /// Get the `len()` of this list.
+    pub fn len<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Int<SizeT>> {
+        self.instance.get(generator, ctx, |f| f.len)
+    }
+
+    /// Get the `items` field as an opaque pointer.
+    pub fn get_opaque_items_ptr<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Ptr<Int<Byte>>> {
+        self.instance.get(generator, ctx, |f| f.items).pointer_cast(generator, ctx, Int(Byte))
+    }
+
+    /// Get the value of this [`ListObject`] as a list with opaque items.
+    ///
+    /// This function allocates on the stack to create the list, but the
+    /// reference to the `items` are preserved.
+    pub fn get_opaque_list_ptr<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>> {
+        let opaque_list = Struct(List { item: Int(Byte) }).alloca(generator, ctx);
+
+        // Copy items pointer
+        let items = self.get_opaque_items_ptr(generator, ctx);
+        opaque_list.set(ctx, |f| f.items, items);
+
+        // Copy len
+        let len = self.instance.get(generator, ctx, |f| f.len);
+        opaque_list.set(ctx, |f| f.len, len);
+
+        opaque_list
+    }
+}

nac3core/src/codegen/object/mod.rs

@@ -0,0 +1,9 @@
+pub mod any;
+pub mod cslice;
+pub mod exception;
+pub mod list;
+pub mod ndarray;
+pub mod range;
+pub mod slice;
+pub mod str;
+pub mod tuple;

nac3core/src/codegen/object/ndarray/array.rs

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

nac3core/src/codegen/object/ndarray/broadcast.rs

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

nac3core/src/codegen/object/ndarray/contiguous.rs

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

nac3core/src/codegen/object/ndarray/factory.rs

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

nac3core/src/codegen/object/ndarray/indexing.rs

@@ -0,0 +1,223 @@
+use crate::codegen::{
+    irrt::call_nac3_ndarray_index,
+    model::*,
+    object::slice::{RustSlice, Slice},
+    CodeGenContext, CodeGenerator,
+};
+
+use super::NDArrayObject;
+
+pub type NDIndexType = Byte;
+
+/// Fields of [`NDIndex`]
+#[derive(Debug, Clone, Copy)]
+pub struct NDIndexFields<'ctx, F: FieldTraversal<'ctx>> {
+    pub type_: F::Out<Int<NDIndexType>>, // Defined to be uint8_t in IRRT
+    pub data: F::Out<Ptr<Int<Byte>>>,
+}
+
+/// An IRRT representation of an ndarray subscript index.
+#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
+pub struct NDIndex;
+
+impl<'ctx> StructKind<'ctx> for NDIndex {
+    type Fields<F: FieldTraversal<'ctx>> = NDIndexFields<'ctx, F>;
+
+    fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
+        Self::Fields { type_: traversal.add_auto("type"), data: traversal.add_auto("data") }
+    }
+}
+
+// A convenience enum to prepare an [`NDIndex`].
+#[derive(Debug, Clone)]
+pub enum RustNDIndex<'ctx> {
+    SingleElement(Instance<'ctx, Int<Int32>>), // TODO: To be SizeT
+    Slice(RustSlice<'ctx, Int32>),             // TODO: To be SizeT
+    NewAxis,
+    Ellipsis,
+}
+
+impl<'ctx> RustNDIndex<'ctx> {
+    /// Get the value to set `NDIndex::type` for this variant.
+    fn get_type_id(&self) -> u64 {
+        // Defined in IRRT, must be in sync
+        match self {
+            RustNDIndex::SingleElement(_) => 0,
+            RustNDIndex::Slice(_) => 1,
+            RustNDIndex::NewAxis => 2,
+            RustNDIndex::Ellipsis => 3,
+        }
+    }
+
+    /// Write the contents to an LLVM [`NDIndex`].
+    fn write_to_ndindex<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        dst_ndindex_ptr: Instance<'ctx, Ptr<Struct<NDIndex>>>,
+    ) {
+        // Set `dst_ndindex_ptr->type`
+        dst_ndindex_ptr.gep(ctx, |f| f.type_).store(
+            ctx,
+            Int(NDIndexType::default()).const_int(generator, ctx.ctx, self.get_type_id()),
+        );
+
+        // Set `dst_ndindex_ptr->data`
+        match self {
+            RustNDIndex::SingleElement(in_index) => {
+                let index_ptr = Int(Int32).alloca(generator, ctx);
+                index_ptr.store(ctx, *in_index);
+
+                dst_ndindex_ptr
+                    .gep(ctx, |f| f.data)
+                    .store(ctx, index_ptr.pointer_cast(generator, ctx, Int(Byte)));
+            }
+            RustNDIndex::Slice(in_rust_slice) => {
+                let user_slice_ptr = Struct(Slice(Int32)).alloca(generator, ctx);
+                in_rust_slice.write_to_slice(generator, ctx, user_slice_ptr);
+
+                dst_ndindex_ptr
+                    .gep(ctx, |f| f.data)
+                    .store(ctx, user_slice_ptr.pointer_cast(generator, ctx, Int(Byte)));
+            }
+            RustNDIndex::NewAxis | RustNDIndex::Ellipsis => {}
+        }
+    }
+
+    /// Allocate an array of `NDIndex`es on the stack and return its stack pointer.
+    pub fn alloca_ndindices<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        in_ndindices: &[RustNDIndex<'ctx>],
+    ) -> (Instance<'ctx, Int<SizeT>>, Instance<'ctx, Ptr<Struct<NDIndex>>>) {
+        let ndindex_model = Struct(NDIndex);
+
+        let num_ndindices = Int(SizeT).const_int(generator, ctx.ctx, in_ndindices.len() as u64);
+        let ndindices = ndindex_model.array_alloca(generator, ctx, num_ndindices.value);
+        for (i, in_ndindex) in in_ndindices.iter().enumerate() {
+            let pndindex = ndindices.offset_const(ctx, i as u64);
+            in_ndindex.write_to_ndindex(generator, ctx, pndindex);
+        }
+
+        (num_ndindices, ndindices)
+    }
+}
+
+impl<'ctx> NDArrayObject<'ctx> {
+    /// Get the ndims [`Type`] after indexing with a given slice.
+    #[must_use]
+    pub fn deduce_ndims_after_indexing_with(&self, indices: &[RustNDIndex<'ctx>]) -> u64 {
+        let mut ndims = self.ndims;
+        for index in indices {
+            match index {
+                RustNDIndex::SingleElement(_) => {
+                    ndims -= 1; // Single elements decrements ndims
+                }
+                RustNDIndex::NewAxis => {
+                    ndims += 1; // `np.newaxis` / `none` adds a new axis
+                }
+                RustNDIndex::Ellipsis | RustNDIndex::Slice(_) => {}
+            }
+        }
+        ndims
+    }
+
+    /// Index into the ndarray, and return a newly-allocated view on this ndarray.
+    ///
+    /// This function behaves like NumPy's ndarray indexing, but if the indices index
+    /// into a single element, an unsized ndarray is returned.
+    #[must_use]
+    pub fn index<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        indices: &[RustNDIndex<'ctx>],
+    ) -> Self {
+        let dst_ndims = self.deduce_ndims_after_indexing_with(indices);
+        let dst_ndarray = NDArrayObject::alloca(generator, ctx, self.dtype, dst_ndims);
+
+        let (num_indices, indices) = RustNDIndex::alloca_ndindices(generator, ctx, indices);
+        call_nac3_ndarray_index(
+            generator,
+            ctx,
+            num_indices,
+            indices,
+            self.instance,
+            dst_ndarray.instance,
+        );
+
+        dst_ndarray
+    }
+}
+
+pub mod util {
+    use itertools::Itertools;
+    use nac3parser::ast::{Expr, ExprKind};
+
+    use crate::{
+        codegen::{model::*, object::slice::util::gen_slice, CodeGenContext, CodeGenerator},
+        typecheck::typedef::Type,
+    };
+
+    use super::RustNDIndex;
+
+    /// Generate LLVM code to transform an ndarray subscript expression to
+    /// its list of [`RustNDIndex`]
+    ///
+    /// i.e.,
+    /// ```python
+    /// my_ndarray[::3, 1, :2:]
+    ///            ^^^^^^^^^^^ Then these into a three `RustNDIndex`es
+    /// ```
+    pub fn gen_ndarray_subscript_ndindices<'ctx, G: CodeGenerator>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        subscript: &Expr<Option<Type>>,
+    ) -> Result<Vec<RustNDIndex<'ctx>>, String> {
+        // TODO: Support https://numpy.org/doc/stable/user/basics.indexing.html#dimensional-indexing-tools
+
+        // Annoying notes about `slice`
+        //  - `my_array[5]`
+        //    - slice is a `Constant`
+        //  - `my_array[:5]`
+        //    - slice is a `Slice`
+        //  - `my_array[:]`
+        //    - slice is a `Slice`, but lower upper step would all be `Option::None`
+        //  - `my_array[:, :]`
+        //    - slice is now a `Tuple` of two `Slice`-s
+        //
+        // In summary:
+        //  - when there is a comma "," within [], `slice` will be a `Tuple` of the entries.
+        //  - when there is not comma "," within [] (i.e., just a single entry), `slice` will be that entry itself.
+        //
+        // So we first "flatten" out the slice expression
+        let index_exprs = match &subscript.node {
+            ExprKind::Tuple { elts, .. } => elts.iter().collect_vec(),
+            _ => vec![subscript],
+        };
+
+        // Process all index expressions
+        let mut rust_ndindices: Vec<RustNDIndex> = Vec::with_capacity(index_exprs.len()); // Not using iterators here because `?` is used here.
+        for index_expr in index_exprs {
+            // NOTE: Currently nac3core's slices do not have an object representation,
+            // so the code/implementation looks awkward - we have to do pattern matching on the expression
+            let ndindex = if let ExprKind::Slice { lower, upper, step } = &index_expr.node {
+                // Handle slices
+                let slice = gen_slice(generator, ctx, lower, upper, step)?;
+                RustNDIndex::Slice(slice)
+            } else {
+                // Treat and handle everything else as a single element index.
+                let index = generator.gen_expr(ctx, index_expr)?.unwrap().to_basic_value_enum(
+                    ctx,
+                    generator,
+                    ctx.primitives.int32, // Must be int32, this checks for illegal values
+                )?;
+                let index = Int(Int32).check_value(generator, ctx.ctx, index).unwrap();
+
+                RustNDIndex::SingleElement(index)
+            };
+            rust_ndindices.push(ndindex);
+        }
+        Ok(rust_ndindices)
+    }
+}

nac3core/src/codegen/object/ndarray/map.rs

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

nac3core/src/codegen/object/ndarray/matmul.rs

@@ -0,0 +1,218 @@
+use std::cmp::max;
+
+use nac3parser::ast::Operator;
+use util::gen_for_model;
+
+use crate::{
+    codegen::{
+        expr::gen_binop_expr_with_values, irrt::call_nac3_ndarray_matmul_calculate_shapes,
+        model::*, object::ndarray::indexing::RustNDIndex, CodeGenContext, CodeGenerator,
+    },
+    typecheck::{magic_methods::Binop, typedef::Type},
+};
+
+use super::{NDArrayObject, NDArrayOut};
+
+/// Perform `np.einsum("...ij,...jk->...ik", in_a, in_b)`.
+///
+/// `dst_dtype` defines the dtype of the returned ndarray.
+fn matmul_at_least_2d<'ctx, G: CodeGenerator>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    dst_dtype: Type,
+    in_a: NDArrayObject<'ctx>,
+    in_b: NDArrayObject<'ctx>,
+) -> NDArrayObject<'ctx> {
+    assert!(in_a.ndims >= 2);
+    assert!(in_b.ndims >= 2);
+
+    // Deduce ndims of the result of matmul.
+    let ndims_int = max(in_a.ndims, in_b.ndims);
+    let ndims = Int(SizeT).const_int(generator, ctx.ctx, ndims_int);
+
+    let num_0 = Int(SizeT).const_int(generator, ctx.ctx, 0);
+    let num_1 = Int(SizeT).const_int(generator, ctx.ctx, 1);
+
+    // Broadcasts `in_a.shape[:-2]` and `in_b.shape[:-2]` together and allocate the
+    // destination ndarray to store the result of matmul.
+    let (lhs, rhs, dst) = {
+        let in_lhs_ndims = in_a.ndims_llvm(generator, ctx.ctx);
+        let in_lhs_shape = in_a.instance.get(generator, ctx, |f| f.shape);
+        let in_rhs_ndims = in_b.ndims_llvm(generator, ctx.ctx);
+        let in_rhs_shape = in_b.instance.get(generator, ctx, |f| f.shape);
+        let lhs_shape = Int(SizeT).array_alloca(generator, ctx, ndims.value);
+        let rhs_shape = Int(SizeT).array_alloca(generator, ctx, ndims.value);
+        let dst_shape = Int(SizeT).array_alloca(generator, ctx, ndims.value);
+
+        // Matmul dimension compatibility is checked here.
+        call_nac3_ndarray_matmul_calculate_shapes(
+            generator,
+            ctx,
+            in_lhs_ndims,
+            in_lhs_shape,
+            in_rhs_ndims,
+            in_rhs_shape,
+            ndims,
+            lhs_shape,
+            rhs_shape,
+            dst_shape,
+        );
+
+        let lhs = in_a.broadcast_to(generator, ctx, ndims_int, lhs_shape);
+        let rhs = in_b.broadcast_to(generator, ctx, ndims_int, rhs_shape);
+
+        let dst = NDArrayObject::alloca(generator, ctx, dst_dtype, ndims_int);
+        dst.copy_shape_from_array(generator, ctx, dst_shape);
+        dst.create_data(generator, ctx);
+
+        (lhs, rhs, dst)
+    };
+
+    let len = lhs.instance.get(generator, ctx, |f| f.shape).get_index_const(
+        generator,
+        ctx,
+        ndims_int - 1,
+    );
+
+    let at_row = ndims_int - 2;
+    let at_col = ndims_int - 1;
+
+    let dst_dtype_llvm = ctx.get_llvm_type(generator, dst_dtype);
+    let dst_zero = dst_dtype_llvm.const_zero();
+
+    dst.foreach(generator, ctx, |generator, ctx, _, hdl| {
+        let pdst_ij = hdl.get_pointer(generator, ctx);
+
+        ctx.builder.build_store(pdst_ij, dst_zero).unwrap();
+
+        let indices = hdl.get_indices();
+        let i = indices.get_index_const(generator, ctx, at_row);
+        let j = indices.get_index_const(generator, ctx, at_col);
+
+        gen_for_model(generator, ctx, num_0, len, num_1, |generator, ctx, _, k| {
+            // `indices` is modified to index into `a` and `b`, and restored.
+            indices.set_index_const(ctx, at_row, i);
+            indices.set_index_const(ctx, at_col, k);
+            let a_ik = lhs.get_scalar_by_indices(generator, ctx, indices);
+
+            indices.set_index_const(ctx, at_row, k);
+            indices.set_index_const(ctx, at_col, j);
+            let b_kj = rhs.get_scalar_by_indices(generator, ctx, indices);
+
+            // Restore `indices`.
+            indices.set_index_const(ctx, at_row, i);
+            indices.set_index_const(ctx, at_col, j);
+
+            // x = a_[...]ik * b_[...]kj
+            let x = gen_binop_expr_with_values(
+                generator,
+                ctx,
+                (&Some(lhs.dtype), a_ik.value),
+                Binop::normal(Operator::Mult),
+                (&Some(rhs.dtype), b_kj.value),
+                ctx.current_loc,
+            )?
+            .unwrap()
+            .to_basic_value_enum(ctx, generator, dst_dtype)?;
+
+            // dst_[...]ij += x
+            let dst_ij = ctx.builder.build_load(pdst_ij, "").unwrap();
+            let dst_ij = gen_binop_expr_with_values(
+                generator,
+                ctx,
+                (&Some(dst_dtype), dst_ij),
+                Binop::normal(Operator::Add),
+                (&Some(dst_dtype), x),
+                ctx.current_loc,
+            )?
+            .unwrap()
+            .to_basic_value_enum(ctx, generator, dst_dtype)?;
+            ctx.builder.build_store(pdst_ij, dst_ij).unwrap();
+
+            Ok(())
+        })
+    })
+    .unwrap();
+
+    dst
+}
+
+impl<'ctx> NDArrayObject<'ctx> {
+    /// Perform `np.matmul` according to the rules in
+    /// <https://numpy.org/doc/stable/reference/generated/numpy.matmul.html>.
+    ///
+    /// This function always return an [`NDArrayObject`]. You may want to use [`NDArrayObject::split_unsized`]
+    /// to handle when the output could be a scalar.
+    ///
+    /// `dst_dtype` defines the dtype of the returned ndarray.
+    pub fn matmul<G: CodeGenerator>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        a: Self,
+        b: Self,
+        out: NDArrayOut<'ctx>,
+    ) -> Self {
+        // Sanity check, but type inference should prevent this.
+        assert!(a.ndims > 0 && b.ndims > 0, "np.matmul disallows scalar input");
+
+        /*
+            If both arguments are 2-D they are multiplied like conventional matrices.
+            If either argument is N-D, N > 2, it is treated as a stack of matrices residing in the last two indices and broadcast accordingly.
+            If the first argument is 1-D, it is promoted to a matrix by prepending a 1 to its dimensions. After matrix multiplication the prepended 1 is removed.
+            If the second argument is 1-D, it is promoted to a matrix by appending a 1 to its dimensions. After matrix multiplication the appended 1 is removed.
+        */
+
+        let new_a = if a.ndims == 1 {
+            // Prepend 1 to its dimensions
+            a.index(generator, ctx, &[RustNDIndex::NewAxis, RustNDIndex::Ellipsis])
+        } else {
+            a
+        };
+
+        let new_b = if b.ndims == 1 {
+            // Append 1 to its dimensions
+            b.index(generator, ctx, &[RustNDIndex::Ellipsis, RustNDIndex::NewAxis])
+        } else {
+            b
+        };
+
+        // NOTE: `result` will always be a newly allocated ndarray.
+        // Current implementation cannot do in-place matrix muliplication.
+        let mut result = matmul_at_least_2d(generator, ctx, out.get_dtype(), new_a, new_b);
+
+        // Postprocessing on the result to remove prepended/appended axes.
+        let mut postindices = vec![];
+        let zero = Int(Int32).const_0(generator, ctx.ctx);
+
+        if a.ndims == 1 {
+            // Remove the prepended 1
+            postindices.push(RustNDIndex::SingleElement(zero));
+        }
+
+        if b.ndims == 1 {
+            // Remove the appended 1
+            postindices.push(RustNDIndex::Ellipsis);
+            postindices.push(RustNDIndex::SingleElement(zero));
+        }
+
+        if !postindices.is_empty() {
+            result = result.index(generator, ctx, &postindices);
+        }
+
+        match out {
+            NDArrayOut::NewNDArray { .. } => result,
+            NDArrayOut::WriteToNDArray { ndarray: out_ndarray } => {
+                let result_shape = result.instance.get(generator, ctx, |f| f.shape);
+                out_ndarray.assert_can_be_written_by_out(
+                    generator,
+                    ctx,
+                    result.ndims,
+                    result_shape,
+                );
+
+                out_ndarray.copy_data_from(generator, ctx, result);
+                out_ndarray
+            }
+        }
+    }
+}

nac3core/src/codegen/object/ndarray/mod.rs

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

nac3core/src/codegen/object/ndarray/nditer.rs

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

nac3core/src/codegen/object/ndarray/shape_util.rs

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

nac3core/src/codegen/object/ndarray/str.rs

@@ -0,0 +1,3 @@
+
+pub fn str_type() {
+}

nac3core/src/codegen/object/ndarray/view.rs

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

nac3core/src/codegen/object/range.rs

@@ -0,0 +1,139 @@
+use inkwell::{values::IntValue, IntPredicate};
+
+use crate::codegen::{irrt::call_nac3_range_len, model::*, CodeGenContext, CodeGenerator};
+
+use super::any::AnyObject;
+
+/// A range in NAC3.
+pub type Range<N> = Array<Len<3>, Int<N>>;
+
+/// An alias for `Range::<Int32>::default()`
+#[must_use]
+pub fn range_model() -> Range<Int32> {
+    Array::default()
+}
+
+impl<'ctx, N: IntKind<'ctx>> Instance<'ctx, Ptr<Range<N>>> {
+    /// Get GEP to `range.start`.
+    pub fn start(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Ptr<Int<N>>> {
+        self.gep_const(ctx, 0)
+    }
+
+    /// Get GEP to `range.stop`.
+    pub fn stop(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Ptr<Int<N>>> {
+        self.gep_const(ctx, 1)
+    }
+
+    /// Get GEP to `range.step`.
+    pub fn step(&self, ctx: &CodeGenContext<'ctx, '_>) -> Instance<'ctx, Ptr<Int<N>>> {
+        self.gep_const(ctx, 2)
+    }
+
+    /// Convenience function to load the `(start, stop, step)` of this range.
+    #[allow(clippy::type_complexity)]
+    pub fn destructure<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+    ) -> (Instance<'ctx, Int<N>>, Instance<'ctx, Int<N>>, Instance<'ctx, Int<N>>) {
+        let start = self.start(ctx).load(generator, ctx);
+        let stop = self.stop(ctx).load(generator, ctx);
+        let step = self.step(ctx).load(generator, ctx);
+        (start, stop, step)
+    }
+}
+
+/// Generate LLVM IR to check that a range's `step` is not zero.
+/// Throws "range step must not be zero" if it is the case.
+pub fn assert_range_step_non_zero<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    step: IntValue<'ctx>,
+) {
+    let int32 = ctx.ctx.i32_type();
+    let rangenez =
+        ctx.builder.build_int_compare(IntPredicate::NE, step, int32.const_zero(), "").unwrap();
+    ctx.make_assert(
+        generator,
+        rangenez,
+        "0:ValueError",
+        "range step must not be zero",
+        [None, None, None],
+        ctx.current_loc,
+    );
+}
+
+/// A Rust structure that has [`Range`] utilities and looks like a [`Range`] but
+/// `start`, `stop` and `step` are held by LLVM registers only.
+///
+/// This structure exists because many implementations use [`Range`] utilities but
+/// it might not be good to alloca an actual [`Range`] value on the stack in order
+/// to perform calculations.
+pub struct RustRange<'ctx, N: IntKind<'ctx>> {
+    pub start: Instance<'ctx, Int<N>>,
+    pub stop: Instance<'ctx, Int<N>>,
+    pub step: Instance<'ctx, Int<N>>,
+}
+
+impl<'ctx, N: IntKind<'ctx>> RustRange<'ctx, N> {
+    pub fn assert_step_non_zero<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) {
+        assert_range_step_non_zero(generator, ctx, self.step.value);
+    }
+
+    /// Calculate the `len()` of this range.
+    pub fn len<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Int<N>> {
+        let int_kind = self.start.model.0;
+        call_nac3_range_len(generator, ctx, int_kind, self.start, self.stop, self.step)
+    }
+}
+
+// TODO: `RangeObject` in the future will have range32, range64
+
+/// A NAC3 Python range object.
+#[derive(Debug, Clone, Copy)]
+pub struct RangeObject<'ctx> {
+    pub instance: Instance<'ctx, Ptr<Range<Int32>>>,
+}
+
+impl<'ctx> RangeObject<'ctx> {
+    /// Attempt to convert an [`AnyObject`] into a [`RangeObject`].
+    pub fn from_object<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        object: AnyObject<'ctx>,
+    ) -> RangeObject<'ctx> {
+        assert!(ctx.unifier.unioned(object.ty, ctx.primitives.range));
+
+        let instance = Ptr(Range::default()).check_value(generator, ctx.ctx, object.value).unwrap();
+        RangeObject { instance }
+    }
+
+    /// Convert into a [`RustRange`].
+    pub fn as_rust_range<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> RustRange<'ctx, Int32> {
+        let (start, stop, step) = self.instance.destructure(generator, ctx);
+        RustRange { start, stop, step }
+    }
+
+    /// Get the `len()` of this range.
+    pub fn len<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Int<Int32>> {
+        let range = self.as_rust_range(generator, ctx);
+        range.assert_step_non_zero(generator, ctx);
+        range.len(generator, ctx)
+    }
+}

nac3core/src/codegen/object/slice.rs

@@ -0,0 +1,185 @@
+use crate::codegen::{irrt::call_nac3_slice_indices, model::*, CodeGenContext, CodeGenerator};
+
+use super::range::RustRange;
+
+/// Fields of [`Slice`]
+#[derive(Debug, Clone)]
+pub struct SliceFields<'ctx, F: FieldTraversal<'ctx>, N: IntKind<'ctx>> {
+    pub start_defined: F::Out<Int<Bool>>,
+    pub start: F::Out<Int<N>>,
+    pub stop_defined: F::Out<Int<Bool>>,
+    pub stop: F::Out<Int<N>>,
+    pub step_defined: F::Out<Int<Bool>>,
+    pub step: F::Out<Int<N>>,
+}
+
+/// An IRRT representation of an (unresolved) slice.
+#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
+pub struct Slice<N>(pub N);
+
+impl<'ctx, N: IntKind<'ctx>> StructKind<'ctx> for Slice<N> {
+    type Fields<F: FieldTraversal<'ctx>> = SliceFields<'ctx, F, N>;
+
+    fn traverse_fields<F: FieldTraversal<'ctx>>(&self, traversal: &mut F) -> Self::Fields<F> {
+        Self::Fields {
+            start_defined: traversal.add_auto("start_defined"),
+            start: traversal.add("start", Int(self.0)),
+            stop_defined: traversal.add_auto("stop_defined"),
+            stop: traversal.add("stop", Int(self.0)),
+            step_defined: traversal.add_auto("step_defined"),
+            step: traversal.add("step", Int(self.0)),
+        }
+    }
+}
+
+/// A Rust structure that has [`Slice`] utilities and looks like a [`Slice`] but
+/// `start`, `stop` and `step` are held by LLVM registers only and possibly
+/// [`Option::None`] if unspecified.
+#[derive(Debug, Clone)]
+pub struct RustSlice<'ctx, N: IntKind<'ctx>> {
+    // It is possible that `start`, `stop`, and `step` are all `None`.
+    // We need to know the `int_kind` even when that is the case.
+    pub int_kind: N,
+    pub start: Option<Instance<'ctx, Int<N>>>,
+    pub stop: Option<Instance<'ctx, Int<N>>>,
+    pub step: Option<Instance<'ctx, Int<N>>>,
+}
+
+impl<'ctx, N: IntKind<'ctx>> RustSlice<'ctx, N> {
+    /// Write the contents to an LLVM [`Slice`].
+    pub fn write_to_slice<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        dst_slice_ptr: Instance<'ctx, Ptr<Struct<Slice<N>>>>,
+    ) {
+        let false_ = Int(Bool).const_false(generator, ctx.ctx);
+        let true_ = Int(Bool).const_true(generator, ctx.ctx);
+
+        match self.start {
+            Some(start) => {
+                dst_slice_ptr.gep(ctx, |f| f.start_defined).store(ctx, true_);
+                dst_slice_ptr.gep(ctx, |f| f.start).store(ctx, start);
+            }
+            None => dst_slice_ptr.gep(ctx, |f| f.start_defined).store(ctx, false_),
+        }
+
+        match self.stop {
+            Some(stop) => {
+                dst_slice_ptr.gep(ctx, |f| f.stop_defined).store(ctx, true_);
+                dst_slice_ptr.gep(ctx, |f| f.stop).store(ctx, stop);
+            }
+            None => dst_slice_ptr.gep(ctx, |f| f.stop_defined).store(ctx, false_),
+        }
+
+        match self.step {
+            Some(step) => {
+                dst_slice_ptr.gep(ctx, |f| f.step_defined).store(ctx, true_);
+                dst_slice_ptr.gep(ctx, |f| f.step).store(ctx, step);
+            }
+            None => dst_slice_ptr.gep(ctx, |f| f.step_defined).store(ctx, false_),
+        }
+    }
+
+    /// Resolve this [`RustSlice`] into a [`RustRange`] like `slice.indices` in Python.
+    ///
+    /// NOTE: This function does stack allocation.
+    pub fn indices<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        length: Instance<'ctx, Int<N>>,
+    ) -> RustRange<'ctx, N> {
+        let mut is_defined = |value: Option<_>| -> Instance<'ctx, Int<Bool>> {
+            Int(Bool).const_int(generator, ctx.ctx, u64::from(value.is_some()))
+        };
+
+        let start_defined = is_defined(self.start);
+        let stop_defined = is_defined(self.stop);
+        let step_defined = is_defined(self.step);
+
+        let mut defined_or_zero = |value: Option<_>| -> Instance<'ctx, Int<N>> {
+            if let Some(value) = value {
+                value
+            } else {
+                // If undefined, return 0 as a placeholder.
+                Int(self.int_kind).const_0(generator, ctx.ctx)
+            }
+        };
+
+        let start = defined_or_zero(self.start);
+        let stop = defined_or_zero(self.stop);
+        let step = defined_or_zero(self.step);
+
+        // Stack allocation here.
+        let range_start = Int(self.int_kind).alloca(generator, ctx);
+        let range_stop = Int(self.int_kind).alloca(generator, ctx);
+        let range_step = Int(self.int_kind).alloca(generator, ctx);
+
+        call_nac3_slice_indices(
+            generator,
+            ctx,
+            self.int_kind,
+            start_defined,
+            start,
+            stop_defined,
+            stop,
+            step_defined,
+            step,
+            length,
+            range_start,
+            range_stop,
+            range_step,
+        );
+
+        let start = range_start.load(generator, ctx);
+        let stop = range_stop.load(generator, ctx);
+        let step = range_step.load(generator, ctx);
+
+        RustRange { start, stop, step }
+    }
+}
+
+pub mod util {
+    use nac3parser::ast::Expr;
+
+    use crate::{
+        codegen::{model::*, CodeGenContext, CodeGenerator},
+        typecheck::typedef::Type,
+    };
+
+    use super::RustSlice;
+
+    /// Generate LLVM IR for an [`ExprKind::Slice`] and convert it into a [`RustSlice`].
+    #[allow(clippy::type_complexity)]
+    pub fn gen_slice<'ctx, G: CodeGenerator>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        lower: &Option<Box<Expr<Option<Type>>>>,
+        upper: &Option<Box<Expr<Option<Type>>>>,
+        step: &Option<Box<Expr<Option<Type>>>>,
+    ) -> Result<RustSlice<'ctx, Int32>, String> {
+        let mut help = |value_expr: &Option<Box<Expr<Option<Type>>>>| -> Result<_, String> {
+            Ok(match value_expr {
+                None => None,
+                Some(value_expr) => {
+                    let value_expr = generator
+                        .gen_expr(ctx, value_expr)?
+                        .unwrap()
+                        .to_basic_value_enum(ctx, generator, ctx.primitives.int32)?;
+
+                    let value_expr =
+                        Int(Int32).check_value(generator, ctx.ctx, value_expr).unwrap();
+
+                    Some(value_expr)
+                }
+            })
+        };
+
+        let start = help(lower)?;
+        let stop = help(upper)?;
+        let step = help(step)?;
+
+        Ok(RustSlice { int_kind: Int32, start, stop, step })
+    }
+}

nac3core/src/codegen/object/str.rs

@@ -0,0 +1,11 @@
+use super::cslice::CSlice;
+use crate::codegen::model::*;
+
+/// A string in NAC3.
+pub type Str = Struct<CSlice<Int<Byte>>>;
+
+/// An alias for `Str::default()`
+#[must_use]
+pub fn str_model() -> Str {
+    Str::default()
+}

nac3core/src/codegen/object/tuple.rs

@@ -0,0 +1,99 @@
+use inkwell::values::StructValue;
+use itertools::Itertools;
+
+use crate::{
+    codegen::{model::*, CodeGenContext, CodeGenerator},
+    typecheck::typedef::{Type, TypeEnum},
+};
+
+use super::any::AnyObject;
+
+/// A NAC3 tuple object.
+///
+/// NOTE: This struct has no copy trait.
+#[derive(Debug, Clone)]
+pub struct TupleObject<'ctx> {
+    /// The type of the tuple.
+    pub tys: Vec<Type>,
+    /// The underlying LLVM struct value of this tuple.
+    pub value: StructValue<'ctx>,
+}
+
+impl<'ctx> TupleObject<'ctx> {
+    pub fn from_object(ctx: &mut CodeGenContext<'ctx, '_>, object: AnyObject<'ctx>) -> Self {
+        // TODO: Keep `is_vararg_ctx` from TTuple?
+
+        // Sanity check on object type.
+        let TypeEnum::TTuple { ty: tys, .. } = &*ctx.unifier.get_ty(object.ty) else {
+            panic!(
+                "Expected type to be a TypeEnum::TTuple, got {}",
+                ctx.unifier.stringify(object.ty)
+            );
+        };
+
+        // Check number of fields
+        let value = object.value.into_struct_value();
+        let value_num_fields = value.get_type().count_fields() as usize;
+        assert!(
+            value_num_fields == tys.len(),
+            "Tuple type has {} item(s), but the LLVM struct value has {} field(s)",
+            tys.len(),
+            value_num_fields
+        );
+
+        TupleObject { tys: tys.clone(), value }
+    }
+
+    /// Convenience function. Create a [`TupleObject`] from an iterator of objects.
+    pub fn from_objects<I, G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        objects: I,
+    ) -> Self
+    where
+        I: IntoIterator<Item = AnyObject<'ctx>>,
+    {
+        let (values, tys): (Vec<_>, Vec<_>) =
+            objects.into_iter().map(|object| (object.value, object.ty)).unzip();
+
+        let llvm_tys = tys.iter().map(|ty| ctx.get_llvm_type(generator, *ty)).collect_vec();
+        let llvm_tuple_ty = ctx.ctx.struct_type(&llvm_tys, false);
+
+        let pllvm_tuple = ctx.builder.build_alloca(llvm_tuple_ty, "tuple").unwrap();
+        for (i, val) in values.into_iter().enumerate() {
+            let pval = ctx.builder.build_struct_gep(pllvm_tuple, i as u32, "value").unwrap();
+            ctx.builder.build_store(pval, val).unwrap();
+        }
+
+        let value = ctx.builder.build_load(pllvm_tuple, "").unwrap().into_struct_value();
+        TupleObject { tys, value }
+    }
+
+    #[must_use]
+    pub fn num_elements(&self) -> usize {
+        self.tys.len()
+    }
+
+    /// Get the `len()` of this tuple.
+    #[must_use]
+    pub fn len<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Int<SizeT>> {
+        Int(SizeT).const_int(generator, ctx.ctx, self.num_elements() as u64)
+    }
+
+    /// Get the `i`-th (0-based) object in this tuple.
+    pub fn index(&self, ctx: &mut CodeGenContext<'ctx, '_>, i: usize) -> AnyObject<'ctx> {
+        assert!(
+            i < self.num_elements(),
+            "Tuple object with length {} have index {i}",
+            self.num_elements()
+        );
+
+        let value = ctx.builder.build_extract_value(self.value, i as u32, "tuple[{i}]").unwrap();
+        let ty = self.tys[i];
+        AnyObject { ty, value }
+    }
+}

nac3core/src/codegen/test.rs

@@ -1,6 +1,6 @@
 use crate::{
     codegen::{
-        classes::{ListType, NDArrayType, ProxyType, RangeType},
+        classes::{ListType, ProxyType},
         concrete_type::ConcreteTypeStore,
         CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask,
         CodeGenerator, DefaultCodeGenerator, WithCall, WorkerRegistry,
@@ -94,7 +94,7 @@ fn test_primitives() {
         "};
     let statements = parse_program(source, FileName::default()).unwrap();
 
-    let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
+    let composer = TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 32).0;
     let mut unifier = composer.unifier.clone();
     let primitives = composer.primitives_ty;
     let top_level = Arc::new(composer.make_top_level_context());
@@ -109,8 +109,18 @@ fn test_primitives() {
     let threads = vec![DefaultCodeGenerator::new("test".into(), 32).into()];
     let signature = FunSignature {
         args: vec![
-            FuncArg { name: "a".into(), ty: primitives.int32, default_value: None },
+            FuncArg {
-            FuncArg { name: "b".into(), ty: primitives.int32, default_value: None },
+                name: "a".into(),
+                ty: primitives.int32,
+                default_value: None,
+                is_vararg: false,
+            },
+            FuncArg {
+                name: "b".into(),
+                ty: primitives.int32,
+                default_value: None,
+                is_vararg: false,
+            },
         ],
         ret: primitives.int32,
         vars: VarMap::new(),
@@ -189,6 +199,8 @@ fn test_primitives() {
         let expected = indoc! {"
             ; ModuleID = 'test'
             source_filename = \"test\"
+            target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"
+            target triple = \"x86_64-unknown-linux-gnu\"
 
             ; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
             define i32 @testing(i32 %0, i32 %1) local_unnamed_addr #0 !dbg !4 {
@@ -246,14 +258,19 @@ fn test_simple_call() {
         "};
     let statements_2 = parse_program(source_2, FileName::default()).unwrap();
 
-    let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
+    let composer = TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 32).0;
     let mut unifier = composer.unifier.clone();
     let primitives = composer.primitives_ty;
     let top_level = Arc::new(composer.make_top_level_context());
     unifier.top_level = Some(top_level.clone());
 
     let signature = FunSignature {
-        args: vec![FuncArg { name: "a".into(), ty: primitives.int32, default_value: None }],
+        args: vec![FuncArg {
+            name: "a".into(),
+            ty: primitives.int32,
+            default_value: None,
+            is_vararg: false,
+        }],
         ret: primitives.int32,
         vars: VarMap::new(),
     };
@@ -368,6 +385,8 @@ fn test_simple_call() {
         let expected = indoc! {"
             ; ModuleID = 'test'
             source_filename = \"test\"
+            target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"
+            target triple = \"x86_64-unknown-linux-gnu\"
 
             ; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
             define i32 @testing(i32 %0) local_unnamed_addr #0 !dbg !5 {
@@ -429,23 +448,3 @@ fn test_classes_list_type_new() {
     let llvm_list = ListType::new(&generator, &ctx, llvm_i32.into());
     assert!(ListType::is_type(llvm_list.as_base_type(), llvm_usize).is_ok());
 }
-
-#[test]
-fn test_classes_range_type_new() {
-    let ctx = inkwell::context::Context::create();
-
-    let llvm_range = RangeType::new(&ctx);
-    assert!(RangeType::is_type(llvm_range.as_base_type()).is_ok());
-}
-
-#[test]
-fn test_classes_ndarray_type_new() {
-    let ctx = inkwell::context::Context::create();
-    let generator = DefaultCodeGenerator::new(String::new(), 64);
-
-    let llvm_i32 = ctx.i32_type();
-    let llvm_usize = generator.get_size_type(&ctx);
-
-    let llvm_ndarray = NDArrayType::new(&generator, &ctx, llvm_i32.into());
-    assert!(NDArrayType::is_type(llvm_ndarray.as_base_type(), llvm_usize).is_ok());
-}

nac3core/src/lib.rs

@@ -23,4 +23,3 @@ pub mod codegen;
 pub mod symbol_resolver;
 pub mod toplevel;
 pub mod typecheck;
-pub mod util;

nac3core/src/symbol_resolver.rs

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

nac3core/src/toplevel/composer.rs

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

nac3core/src/toplevel/helper.rs

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

nac3core/src/toplevel/mod.rs

@@ -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
+        ///   order, including type variables associated with the class. Excluding rigid type
-        /// variables.
+        ///   variables.
         ///
         /// Rigid type variables that would be substituted when the function is instantiated.
         instance_to_stmt: HashMap<String, FunInstance>,

nac3core/src/toplevel/numpy.rs

@@ -10,9 +10,9 @@ 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,
@@ -25,9 +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,

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__generic_class.snap

@@ -5,7 +5,7 @@ expression: res_vec
 [
     "Class {\nname: \"Generic_A\",\nancestors: [\"Generic_A[V]\", \"B\"],\nfields: [\"aa\", \"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"foo\", \"fn[[b:T], none]\"), (\"fun\", \"fn[[a:int32], V]\")],\ntype_vars: [\"V\"]\n}\n",
     "Function {\nname: \"Generic_A.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
-    "Function {\nname: \"Generic_A.fun\",\nsig: \"fn[[a:int32], V]\",\nvar_id: [TypeVarId(245)]\n}\n",
+    "Function {\nname: \"Generic_A.fun\",\nsig: \"fn[[a:int32], V]\",\nvar_id: [TypeVarId(257)]\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",

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__inheritance_override.snap

@@ -7,7 +7,7 @@ expression: res_vec
     "Function {\nname: \"A.__init__\",\nsig: \"fn[[t:T], none]\",\nvar_id: []\n}\n",
     "Function {\nname: \"A.fun\",\nsig: \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\",\nvar_id: []\n}\n",
     "Function {\nname: \"A.foo\",\nsig: \"fn[[c:C], none]\",\nvar_id: []\n}\n",
-    "Class {\nname: \"B\",\nancestors: [\"B[typevar234]\", \"A[float]\"],\nfields: [\"a\", \"b\", \"c\", \"d\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: [\"typevar234\"]\n}\n",
+    "Class {\nname: \"B\",\nancestors: [\"B[typevar246]\", \"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: [\"typevar246\"]\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",

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__list_tuple_generic.snap

@@ -5,8 +5,8 @@ expression: res_vec
 [
     "Function {\nname: \"foo\",\nsig: \"fn[[a:list[int32], b:tuple[T, float]], A[B, bool]]\",\nvar_id: []\n}\n",
     "Class {\nname: \"A\",\nancestors: [\"A[T, V]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[v:V], none]\"), (\"fun\", \"fn[[a:T], V]\")],\ntype_vars: [\"T\", \"V\"]\n}\n",
-    "Function {\nname: \"A.__init__\",\nsig: \"fn[[v:V], none]\",\nvar_id: [TypeVarId(247)]\n}\n",
+    "Function {\nname: \"A.__init__\",\nsig: \"fn[[v:V], none]\",\nvar_id: [TypeVarId(259)]\n}\n",
-    "Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(252)]\n}\n",
+    "Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(264)]\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",

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__self1.snap

@@ -3,7 +3,7 @@ source: nac3core/src/toplevel/test.rs
 expression: res_vec
 ---
 [
-    "Class {\nname: \"A\",\nancestors: [\"A[typevar233, typevar234]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[float, bool], b:B], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\")],\ntype_vars: [\"typevar233\", \"typevar234\"]\n}\n",
+    "Class {\nname: \"A\",\nancestors: [\"A[typevar245, typevar246]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[float, bool], b:B], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\")],\ntype_vars: [\"typevar245\", \"typevar246\"]\n}\n",
     "Function {\nname: \"A.__init__\",\nsig: \"fn[[a:A[float, bool], b:B], none]\",\nvar_id: []\n}\n",
     "Function {\nname: \"A.fun\",\nsig: \"fn[[a:A[float, bool]], A[bool, int32]]\",\nvar_id: []\n}\n",
     "Class {\nname: \"B\",\nancestors: [\"B\", \"A[int64, bool]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\"), (\"foo\", \"fn[[b:B], B]\"), (\"bar\", \"fn[[a:A[list[B], int32]], tuple[A[virtual[A[B, int32]], bool], B]]\")],\ntype_vars: []\n}\n",

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__simple_class_compose.snap

@@ -6,12 +6,12 @@ expression: res_vec
     "Class {\nname: \"A\",\nancestors: [\"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
     "Function {\nname: \"A.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
     "Function {\nname: \"A.fun\",\nsig: \"fn[[b:B], none]\",\nvar_id: []\n}\n",
-    "Function {\nname: \"A.foo\",\nsig: \"fn[[a:T, b:V], none]\",\nvar_id: [TypeVarId(253)]\n}\n",
+    "Function {\nname: \"A.foo\",\nsig: \"fn[[a:T, b:V], none]\",\nvar_id: [TypeVarId(265)]\n}\n",
     "Class {\nname: \"B\",\nancestors: [\"B\", \"C\", \"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
     "Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
     "Class {\nname: \"C\",\nancestors: [\"C\", \"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
     "Function {\nname: \"C.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
     "Function {\nname: \"C.fun\",\nsig: \"fn[[b:B], none]\",\nvar_id: []\n}\n",
     "Function {\nname: \"foo\",\nsig: \"fn[[a:A], none]\",\nvar_id: []\n}\n",
-    "Function {\nname: \"ff\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(261)]\n}\n",
+    "Function {\nname: \"ff\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(273)]\n}\n",
 ]

nac3core/src/toplevel/test.rs

@@ -117,7 +117,8 @@ impl SymbolResolver for Resolver {
     "register"
 )]
 fn test_simple_register(source: Vec<&str>) {
-    let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
 
     for s in source {
         let ast = parse_program(s, FileName::default()).unwrap();
@@ -137,7 +138,8 @@ fn test_simple_register(source: Vec<&str>) {
     "register"
 )]
 fn test_simple_register_without_constructor(source: &str) {
-    let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
     let ast = parse_program(source, FileName::default()).unwrap();
     let ast = ast[0].clone();
     composer.register_top_level(ast, None, "", true).unwrap();
@@ -171,7 +173,8 @@ fn test_simple_register_without_constructor(source: &str) {
     "function compose"
 )]
 fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
-    let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
 
     let internal_resolver = Arc::new(ResolverInternal {
         id_to_def: Mutex::default(),
@@ -519,7 +522,8 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
 )]
 fn test_analyze(source: &[&str], res: &[&str]) {
     let print = false;
-    let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
 
     let internal_resolver = make_internal_resolver_with_tvar(
         vec![
@@ -696,7 +700,8 @@ fn test_analyze(source: &[&str], res: &[&str]) {
 )]
 fn test_inference(source: Vec<&str>, res: &[&str]) {
     let print = true;
-    let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
 
     let internal_resolver = make_internal_resolver_with_tvar(
         vec![

nac3core/src/toplevel/type_annotation.rs

@@ -1,8 +1,9 @@
 use super::*;
 use crate::symbol_resolver::SymbolValue;
-use crate::toplevel::helper::PrimDef;
+use crate::toplevel::helper::{PrimDef, PrimDefDetails};
 use crate::typecheck::typedef::VarMap;
 use nac3parser::ast::Constant;
+use strum::IntoEnumIterator;
 
 #[derive(Clone, Debug)]
 pub enum TypeAnnotation {
@@ -63,9 +64,9 @@ impl TypeAnnotation {
 /// Parses an AST expression `expr` into a [`TypeAnnotation`].
 ///
 /// * `locked` - A [`HashMap`] containing the IDs of known definitions, mapped to a [`Vec`] of all
-/// generic variables associated with the definition.
+///   generic variables associated with the definition.
 /// * `type_var` - The type variable associated with the type argument currently being parsed. Pass
-/// [`None`] when this function is invoked externally.
+///   [`None`] when this function is invoked externally.
 pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
     resolver: &(dyn SymbolResolver + Send + Sync),
     top_level_defs: &[Arc<RwLock<TopLevelDef>>],
@@ -357,6 +358,7 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
 pub fn get_type_from_type_annotation_kinds(
     top_level_defs: &[Arc<RwLock<TopLevelDef>>],
     unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
     ann: &TypeAnnotation,
     subst_list: &mut Option<Vec<Type>>,
 ) -> Result<Type, HashSet<String>> {
@@ -379,100 +381,141 @@ pub fn get_type_from_type_annotation_kinds(
             let param_ty = params
                 .iter()
                 .map(|x| {
-                    get_type_from_type_annotation_kinds(top_level_defs, unifier, x, subst_list)
+                    get_type_from_type_annotation_kinds(
+                        top_level_defs,
+                        unifier,
+                        primitives,
+                        x,
+                        subst_list,
+                    )
                 })
                 .collect::<Result<Vec<_>, _>>()?;
 
-            let subst = {
+            let ty = if let Some(prim_def) = PrimDef::iter().find(|prim| prim.id() == *obj_id) {
-                // check for compatible range
+                // Primitive TopLevelDefs do not contain all fields that are present in their Type
-                // TODO: if allow type var to be applied(now this disallowed in the parse_to_type_annotation), need more check
+                // counterparts, so directly perform subst on the Type instead.
-                let mut result = VarMap::new();
-                for (tvar, p) in type_vars.iter().zip(param_ty) {
-                    match unifier.get_ty(*tvar).as_ref() {
-                        TypeEnum::TVar {
-                            id,
-                            range,
-                            fields: None,
-                            name,
-                            loc,
-                            is_const_generic: false,
-                        } => {
-                            let ok: bool = {
-                                // create a temp type var and unify to check compatibility
-                                p == *tvar || {
-                                    let temp = unifier.get_fresh_var_with_range(
-                                        range.as_slice(),
-                                        *name,
-                                        *loc,
-                                    );
-                                    unifier.unify(temp.ty, p).is_ok()
-                                }
-                            };
-                            if ok {
-                                result.insert(*id, p);
-                            } else {
-                                return Err(HashSet::from([format!(
-                                    "cannot apply type {} to type variable with id {:?}",
-                                    unifier.internal_stringify(
-                                        p,
-                                        &mut |id| format!("class{id}"),
-                                        &mut |id| format!("typevar{id}"),
-                                        &mut None
-                                    ),
-                                    *id
-                                )]));
-                            }
-                        }
 
-                        TypeEnum::TVar { id, range, name, loc, is_const_generic: true, .. } => {
+                let PrimDefDetails::PrimClass { get_ty_fn, .. } = prim_def.details() else {
-                            let ty = range[0];
+                    unreachable!()
-                            let ok: bool = {
+                };
-                                // create a temp type var and unify to check compatibility
-                                p == *tvar || {
-                                    let temp = unifier.get_fresh_const_generic_var(ty, *name, *loc);
-                                    unifier.unify(temp.ty, p).is_ok()
-                                }
-                            };
-                            if ok {
-                                result.insert(*id, p);
-                            } else {
-                                return Err(HashSet::from([format!(
-                                    "cannot apply type {} to type variable {}",
-                                    unifier.stringify(p),
-                                    name.unwrap_or_else(|| format!("typevar{id}").into()),
-                                )]));
-                            }
-                        }
 
-                        _ => unreachable!("must be generic type var"),
+                let base_ty = get_ty_fn(primitives);
+                let params =
+                    if let TypeEnum::TObj { params, .. } = &*unifier.get_ty_immutable(base_ty) {
+                        params.clone()
+                    } else {
+                        unreachable!()
+                    };
+
+                unifier
+                    .subst(
+                        get_ty_fn(primitives),
+                        &params
+                            .iter()
+                            .zip(param_ty)
+                            .map(|(obj_tv, param)| (*obj_tv.0, param))
+                            .collect(),
+                    )
+                    .unwrap_or(base_ty)
+            } else {
+                let subst = {
+                    // check for compatible range
+                    // TODO: if allow type var to be applied(now this disallowed in the parse_to_type_annotation), need more check
+                    let mut result = VarMap::new();
+                    for (tvar, p) in type_vars.iter().zip(param_ty) {
+                        match unifier.get_ty(*tvar).as_ref() {
+                            TypeEnum::TVar {
+                                id,
+                                range,
+                                fields: None,
+                                name,
+                                loc,
+                                is_const_generic: false,
+                            } => {
+                                let ok: bool = {
+                                    // create a temp type var and unify to check compatibility
+                                    p == *tvar || {
+                                        let temp = unifier.get_fresh_var_with_range(
+                                            range.as_slice(),
+                                            *name,
+                                            *loc,
+                                        );
+                                        unifier.unify(temp.ty, p).is_ok()
+                                    }
+                                };
+                                if ok {
+                                    result.insert(*id, p);
+                                } else {
+                                    return Err(HashSet::from([format!(
+                                        "cannot apply type {} to type variable with id {:?}",
+                                        unifier.internal_stringify(
+                                            p,
+                                            &mut |id| format!("class{id}"),
+                                            &mut |id| format!("typevar{id}"),
+                                            &mut None
+                                        ),
+                                        *id
+                                    )]));
+                                }
+                            }
+
+                            TypeEnum::TVar {
+                                id, range, name, loc, is_const_generic: true, ..
+                            } => {
+                                let ty = range[0];
+                                let ok: bool = {
+                                    // create a temp type var and unify to check compatibility
+                                    p == *tvar || {
+                                        let temp =
+                                            unifier.get_fresh_const_generic_var(ty, *name, *loc);
+                                        unifier.unify(temp.ty, p).is_ok()
+                                    }
+                                };
+                                if ok {
+                                    result.insert(*id, p);
+                                } else {
+                                    return Err(HashSet::from([format!(
+                                        "cannot apply type {} to type variable {}",
+                                        unifier.stringify(p),
+                                        name.unwrap_or_else(|| format!("typevar{id}").into()),
+                                    )]));
+                                }
+                            }
+
+                            _ => unreachable!("must be generic type var"),
+                        }
+                    }
+                    result
+                };
+                // Class Attributes keep a copy with Class Definition and are not added to objects
+                let mut tobj_fields = methods
+                    .iter()
+                    .map(|(name, ty, _)| {
+                        let subst_ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
+                        // methods are immutable
+                        (*name, (subst_ty, false))
+                    })
+                    .collect::<HashMap<_, _>>();
+                tobj_fields.extend(fields.iter().map(|(name, ty, mutability)| {
+                    let subst_ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
+                    (*name, (subst_ty, *mutability))
+                }));
+                let need_subst = !subst.is_empty();
+                let ty = unifier.add_ty(TypeEnum::TObj {
+                    obj_id: *obj_id,
+                    fields: tobj_fields,
+                    params: subst,
+                });
+
+                if need_subst {
+                    if let Some(wl) = subst_list.as_mut() {
+                        wl.push(ty);
                     }
                 }
-                result
+
+                ty
             };
-            // Class Attributes keep a copy with Class Definition and are not added to objects
+
-            let mut tobj_fields = methods
-                .iter()
-                .map(|(name, ty, _)| {
-                    let subst_ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
-                    // methods are immutable
-                    (*name, (subst_ty, false))
-                })
-                .collect::<HashMap<_, _>>();
-            tobj_fields.extend(fields.iter().map(|(name, ty, mutability)| {
-                let subst_ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
-                (*name, (subst_ty, *mutability))
-            }));
-            let need_subst = !subst.is_empty();
-            let ty = unifier.add_ty(TypeEnum::TObj {
-                obj_id: *obj_id,
-                fields: tobj_fields,
-                params: subst,
-            });
-            if need_subst {
-                if let Some(wl) = subst_list.as_mut() {
-                    wl.push(ty);
-                }
-            }
             Ok(ty)
         }
         TypeAnnotation::Primitive(ty) | TypeAnnotation::TypeVar(ty) => Ok(*ty),
@@ -490,6 +533,7 @@ pub fn get_type_from_type_annotation_kinds(
             let ty = get_type_from_type_annotation_kinds(
                 top_level_defs,
                 unifier,
+                primitives,
                 ty.as_ref(),
                 subst_list,
             )?;
@@ -499,10 +543,16 @@ pub fn get_type_from_type_annotation_kinds(
             let tys = tys
                 .iter()
                 .map(|x| {
-                    get_type_from_type_annotation_kinds(top_level_defs, unifier, x, subst_list)
+                    get_type_from_type_annotation_kinds(
+                        top_level_defs,
+                        unifier,
+                        primitives,
+                        x,
+                        subst_list,
+                    )
                 })
                 .collect::<Result<Vec<_>, _>>()?;
-            Ok(unifier.add_ty(TypeEnum::TTuple { ty: tys }))
+            Ok(unifier.add_ty(TypeEnum::TTuple { ty: tys, is_vararg_ctx: false }))
         }
     }
 }

nac3core/src/typecheck/function_check.rs

@@ -34,13 +34,18 @@ impl<'a> Inferencer<'a> {
                 self.should_have_value(pattern)?;
                 Ok(())
             }
-            ExprKind::Tuple { elts, .. } => {
+            ExprKind::List { elts, .. } | ExprKind::Tuple { elts, .. } => {
                 for elt in elts {
                     self.check_pattern(elt, defined_identifiers)?;
                     self.should_have_value(elt)?;
                 }
                 Ok(())
             }
+            ExprKind::Starred { value, .. } => {
+                self.check_pattern(value, defined_identifiers)?;
+                self.should_have_value(value)?;
+                Ok(())
+            }
             ExprKind::Subscript { value, slice, .. } => {
                 self.check_expr(value, defined_identifiers)?;
                 self.should_have_value(value)?;
@@ -207,19 +212,23 @@ impl<'a> Inferencer<'a> {
     /// This is a workaround preventing the caller from using a variable `alloca`-ed in the body, which
     /// is freed when the function returns.
     fn check_return_value_ty(&mut self, ret_ty: Type) -> bool {
-        match &*self.unifier.get_ty_immutable(ret_ty) {
+        if cfg!(feature = "no-escape-analysis") {
-            TypeEnum::TObj { .. } => [
+            true
-                self.primitives.int32,
+        } else {
-                self.primitives.int64,
+            match &*self.unifier.get_ty_immutable(ret_ty) {
-                self.primitives.uint32,
+                TypeEnum::TObj { .. } => [
-                self.primitives.uint64,
+                    self.primitives.int32,
-                self.primitives.float,
+                    self.primitives.int64,
-                self.primitives.bool,
+                    self.primitives.uint32,
-            ]
+                    self.primitives.uint64,
-            .iter()
+                    self.primitives.float,
-            .any(|allowed_ty| self.unifier.unioned(ret_ty, *allowed_ty)),
+                    self.primitives.bool,
-            TypeEnum::TTuple { ty } => ty.iter().all(|t| self.check_return_value_ty(*t)),
+                ]
-            _ => false,
+                .iter()
+                .any(|allowed_ty| self.unifier.unioned(ret_ty, *allowed_ty)),
+                TypeEnum::TTuple { ty, .. } => ty.iter().all(|t| self.check_return_value_ty(*t)),
+                _ => false,
+            }
         }
     }