core: update insta after ndstrides

New type vars are introduced when programming new ndarray functions.

.clang-format

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

.pre-commit-config.yaml

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

Cargo.toml

@@ -4,7 +4,6 @@ members = [
   "nac3ast",
   "nac3parser",
   "nac3core",
-  "nac3core/nac3core_derive",
   "nac3standalone",
   "nac3artiq",
   "runkernel",

flake.lock

@@ -2,11 +2,11 @@
   "nodes": {
     "nixpkgs": {
       "locked": {
-        "lastModified": 1733940404,
-        "narHash": "sha256-Pj39hSoUA86ZePPF/UXiYHHM7hMIkios8TYG29kQT4g=",
+        "lastModified": 1723637854,
+        "narHash": "sha256-med8+5DSWa2UnOqtdICndjDAEjxr5D7zaIiK4pn0Q7c=",
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "5d67ea6b4b63378b9c13be21e2ec9d1afc921713",
+        "rev": "c3aa7b8938b17aebd2deecf7be0636000d62a2b9",
         "type": "github"
       },
       "original": {

flake.nix

@@ -107,18 +107,18 @@
             (pkgs.fetchFromGitHub {
               owner = "m-labs";
               repo = "sipyco";
-              rev = "094a6cd63ffa980ef63698920170e50dc9ba77fd";
-              sha256 = "sha256-PPnAyDedUQ7Og/Cby9x5OT9wMkNGTP8GS53V6N/dk4w=";
+              rev = "939f84f9b5eef7efbf7423c735d1834783b6140e";
+              sha256 = "sha256-15Nun4EY35j+6SPZkjzZtyH/ncxLS60KuGJjFh5kSTc=";
             })
             (pkgs.fetchFromGitHub {
               owner = "m-labs";
               repo = "artiq";
-              rev = "28c9de3e251daa89a8c9fd79d5ab64a3ec03bac6";
-              sha256 = "sha256-vAvpbHc5B+1wtG8zqN7j9dQE1ON+i22v+uqA+tw6Gak=";
+              rev = "923ca3377d42c815f979983134ec549dc39d3ca0";
+              sha256 = "sha256-oJoEeNEeNFSUyh6jXG8Tzp6qHVikeHS0CzfE+mODPgw=";
             })
           ];
           buildInputs = [
-            (python3-mimalloc.withPackages(ps: [ ps.numpy ps.scipy ps.jsonschema ps.lmdb ps.platformdirs nac3artiq-instrumented ]))
+            (python3-mimalloc.withPackages(ps: [ ps.numpy ps.scipy ps.jsonschema ps.lmdb nac3artiq-instrumented ]))
             pkgs.llvmPackages_14.llvm.out
           ];
           phases = [ "buildPhase" "installPhase" ];

nac3artiq/Cargo.toml

@@ -12,10 +12,16 @@ crate-type = ["cdylib"]
 itertools = "0.13"
 pyo3 = { version = "0.21", features = ["extension-module", "gil-refs"] }
 parking_lot = "0.12"
-tempfile = "3.13"
+tempfile = "3.10"
+nac3parser = { path = "../nac3parser" }
 nac3core = { path = "../nac3core" }
 nac3ld = { path = "../nac3ld" }
 
+[dependencies.inkwell]
+version = "0.4"
+default-features = false
+features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
+
 [features]
 init-llvm-profile = []
 no-escape-analysis = ["nac3core/no-escape-analysis"]

nac3artiq/demo/embedding_map.py

@@ -7,6 +7,33 @@ class EmbeddingMap:
         self.function_map = {}
         self.attributes_writeback = []
 
+        # preallocate exception names
+        self.preallocate_runtime_exception_names(["RuntimeError",
+                                          "RTIOUnderflow",
+                                          "RTIOOverflow",
+                                          "RTIODestinationUnreachable",
+                                          "DMAError",
+                                          "I2CError",
+                                          "CacheError",
+                                          "SPIError",
+                                          "0:ZeroDivisionError",
+                                          "0:IndexError",
+                                          "0:ValueError",
+                                          "0:RuntimeError",
+                                          "0:AssertionError",
+                                          "0:KeyError",
+                                          "0:NotImplementedError",
+                                          "0:OverflowError",
+                                          "0:IOError",
+                                          "0:UnwrapNoneError"])
+
+    def preallocate_runtime_exception_names(self, names):
+        for i, name in enumerate(names):
+            if ":" not in name:
+                name = "0:artiq.coredevice.exceptions." + name
+            exn_id = self.store_str(name)
+            assert exn_id == i
+
     def store_function(self, key, fun):
         self.function_map[key] = fun
         return key

nac3artiq/demo/min_artiq.py

@@ -112,15 +112,10 @@ def extern(function):
     register_function(function)
     return function
 
-
-def rpc(arg=None, flags={}):
-    """Decorates a function or method to be executed on the host interpreter."""
-    if arg is None:
-        def inner_decorator(function):
-            return rpc(function, flags)
-        return inner_decorator
-    register_function(arg)
-    return arg
+def rpc(function):
+    """Decorates a function declaration defined by the core device runtime."""
+    register_function(function)
+    return function
 
 def kernel(function_or_method):
     """Decorates a function or method to be executed on the core device."""
@@ -206,7 +201,7 @@ class Core:
         embedding = EmbeddingMap()
 
         if allow_registration:
-            compiler.analyze(registered_functions, registered_classes, set())
+            compiler.analyze(registered_functions, registered_classes)
             allow_registration = False
 
         if hasattr(method, "__self__"):

nac3artiq/src/codegen.rs

@@ -1,3 +1,36 @@
+use nac3core::{
+    codegen::{
+        classes::{ListValue, RangeValue, UntypedArrayLikeAccessor},
+        expr::{destructure_range, gen_call},
+        llvm_intrinsics::{call_int_smax, call_stackrestore, call_stacksave},
+        model::*,
+        object::{any::AnyObject, ndarray::NDArrayObject},
+        stmt::{gen_block, gen_for_callback_incrementing, gen_if_callback, gen_with},
+        CodeGenContext, CodeGenerator,
+    },
+    symbol_resolver::ValueEnum,
+    toplevel::{helper::PrimDef, numpy::unpack_ndarray_var_tys, DefinitionId, GenCall},
+    typecheck::typedef::{iter_type_vars, FunSignature, FuncArg, Type, TypeEnum, VarMap},
+};
+
+use nac3parser::ast::{Expr, ExprKind, Located, Stmt, StmtKind, StrRef};
+
+use inkwell::{
+    context::Context,
+    module::Linkage,
+    types::IntType,
+    values::{BasicValueEnum, PointerValue, StructValue},
+    AddressSpace, IntPredicate,
+};
+
+use pyo3::{
+    types::{PyDict, PyList},
+    PyObject, PyResult, Python,
+};
+
+use crate::{symbol_resolver::InnerResolver, timeline::TimeFns};
+
+use itertools::Itertools;
 use std::{
     collections::{hash_map::DefaultHasher, HashMap},
     hash::{Hash, Hasher},
@@ -6,43 +39,6 @@ use std::{
     sync::Arc,
 };
 
-use itertools::Itertools;
-use pyo3::{
-    types::{PyDict, PyList},
-    PyObject, PyResult, Python,
-};
-
-use super::{symbol_resolver::InnerResolver, timeline::TimeFns};
-use nac3core::{
-    codegen::{
-        expr::{destructure_range, gen_call},
-        llvm_intrinsics::{call_int_smax, call_memcpy, call_stackrestore, call_stacksave},
-        stmt::{gen_block, gen_for_callback_incrementing, gen_if_callback, gen_with},
-        type_aligned_alloca,
-        types::ndarray::NDArrayType,
-        values::{
-            ArrayLikeIndexer, ArrayLikeValue, ArraySliceValue, ListValue, ProxyValue, RangeValue,
-            UntypedArrayLikeAccessor,
-        },
-        CodeGenContext, CodeGenerator,
-    },
-    inkwell::{
-        context::Context,
-        module::Linkage,
-        types::{BasicType, IntType},
-        values::{BasicValueEnum, IntValue, PointerValue, StructValue},
-        AddressSpace, IntPredicate, OptimizationLevel,
-    },
-    nac3parser::ast::{Expr, ExprKind, Located, Stmt, StmtKind, StrRef},
-    symbol_resolver::ValueEnum,
-    toplevel::{
-        helper::{extract_ndims, PrimDef},
-        numpy::unpack_ndarray_var_tys,
-        DefinitionId, GenCall,
-    },
-    typecheck::typedef::{iter_type_vars, FunSignature, FuncArg, Type, TypeEnum, VarMap},
-};
-
 /// The parallelism mode within a block.
 #[derive(Copy, Clone, Eq, PartialEq)]
 enum ParallelMode {
@@ -458,52 +454,41 @@ fn format_rpc_arg<'ctx>(
             // NAC3: NDArray = { usize, usize*, T* }
             // libproto_artiq: NDArray = [data[..], dim_sz[..]]
 
-            let llvm_i1 = ctx.ctx.bool_type();
-            let llvm_usize = generator.get_size_type(ctx.ctx);
+            let ndarray = AnyObject { ty: arg_ty, value: arg };
+            let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
 
-            let (elem_ty, ndims) = unpack_ndarray_var_tys(&mut ctx.unifier, arg_ty);
-            let ndims = extract_ndims(&ctx.unifier, ndims);
-            let dtype = ctx.get_llvm_type(generator, elem_ty);
-            let ndarray = NDArrayType::new(generator, ctx.ctx, dtype, Some(ndims))
-                .map_value(arg.into_pointer_value(), None);
-
-            let ndims = llvm_usize.const_int(ndims, false);
+            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.
-            // Turning it into a ContiguousNDArray to get a `data` that is contiguous.
-            let carray = ndarray.make_contiguous_ndarray(generator, ctx);
+            let carray = ndarray.make_contiguous_ndarray(generator, ctx, Any(dtype));
 
-            let sizeof_usize = llvm_usize.size_of();
-            let sizeof_usize =
-                ctx.builder.build_int_z_extend_or_bit_cast(sizeof_usize, llvm_usize, "").unwrap();
+            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 = dtype.ptr_type(AddressSpace::default()).size_of();
-            let sizeof_pdata =
-                ctx.builder.build_int_z_extend_or_bit_cast(sizeof_pdata, llvm_usize, "").unwrap();
+            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 = ctx.builder.build_int_mul(sizeof_usize, ndims, "").unwrap();
-            let sizeof_buf = ctx.builder.build_int_add(sizeof_buf_shape, sizeof_pdata, "").unwrap();
+            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 = ctx.builder.build_array_alloca(llvm_i8, sizeof_buf, "rpc.arg").unwrap();
-            let buf = ArraySliceValue::from_ptr_val(buf, sizeof_buf, Some("rpc.arg"));
-            let buf_data = buf.base_ptr(ctx, generator);
-            let buf_shape =
-                unsafe { buf.ptr_offset_unchecked(ctx, generator, &sizeof_pdata, None) };
+            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.load_data(ctx);
-            let carray_data = ctx.builder.build_pointer_cast(carray_data, llvm_pi8, "").unwrap();
-            call_memcpy(ctx, buf_data, carray_data, sizeof_pdata, llvm_i1.const_zero());
+            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.shape().base_ptr(ctx, generator);
-            let carray_shape_i8 =
-                ctx.builder.build_pointer_cast(carray_shape, llvm_pi8, "").unwrap();
-            call_memcpy(ctx, buf_shape, carray_shape_i8, sizeof_buf_shape, llvm_i1.const_zero());
+            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.base_ptr(ctx, generator)
+            buf.value
         }
 
         _ => {
@@ -513,7 +498,7 @@ fn format_rpc_arg<'ctx>(
             ctx.builder.build_store(arg_slot, arg).unwrap();
 
             ctx.builder
-                .build_bit_cast(arg_slot, llvm_pi8, "rpc.arg")
+                .build_bitcast(arg_slot, llvm_pi8, "rpc.arg")
                 .map(BasicValueEnum::into_pointer_value)
                 .unwrap()
         }
@@ -524,276 +509,12 @@ fn format_rpc_arg<'ctx>(
     arg_slot
 }
 
-/// Formats an RPC return value to conform to the expected format required by NAC3.
-fn format_rpc_ret<'ctx>(
-    generator: &mut dyn CodeGenerator,
-    ctx: &mut CodeGenContext<'ctx, '_>,
-    ret_ty: Type,
-) -> Option<BasicValueEnum<'ctx>> {
-    // -- receive value:
-    // T result = {
-    //   void *ret_ptr = alloca(sizeof(T));
-    //   void *ptr = ret_ptr;
-    //   loop: int size = rpc_recv(ptr);
-    //   // Non-zero: Provide `size` bytes of extra storage for variable-length data.
-    //   if(size) { ptr = alloca(size); goto loop; }
-    //   else *(T*)ret_ptr
-    // }
-
-    let llvm_i8 = ctx.ctx.i8_type();
-    let llvm_i32 = ctx.ctx.i32_type();
-    let llvm_i8_8 = ctx.ctx.struct_type(&[llvm_i8.array_type(8).into()], false);
-    let llvm_pi8 = llvm_i8.ptr_type(AddressSpace::default());
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    let rpc_recv = ctx.module.get_function("rpc_recv").unwrap_or_else(|| {
-        ctx.module.add_function("rpc_recv", llvm_i32.fn_type(&[llvm_pi8.into()], false), None)
-    });
-
-    if ctx.unifier.unioned(ret_ty, ctx.primitives.none) {
-        ctx.build_call_or_invoke(rpc_recv, &[llvm_pi8.const_null().into()], "rpc_recv");
-        return None;
-    }
-
-    let prehead_bb = ctx.builder.get_insert_block().unwrap();
-    let current_function = prehead_bb.get_parent().unwrap();
-    let head_bb = ctx.ctx.append_basic_block(current_function, "rpc.head");
-    let alloc_bb = ctx.ctx.append_basic_block(current_function, "rpc.continue");
-    let tail_bb = ctx.ctx.append_basic_block(current_function, "rpc.tail");
-
-    let llvm_ret_ty = ctx.get_llvm_abi_type(generator, ret_ty);
-
-    let result = match &*ctx.unifier.get_ty_immutable(ret_ty) {
-        TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
-            let num_0 = llvm_usize.const_zero();
-
-            // Round `val` up to its modulo `power_of_two`
-            let round_up = |ctx: &mut CodeGenContext<'ctx, '_>,
-                            val: IntValue<'ctx>,
-                            power_of_two: IntValue<'ctx>| {
-                debug_assert_eq!(
-                    val.get_type().get_bit_width(),
-                    power_of_two.get_type().get_bit_width()
-                );
-
-                let llvm_val_t = val.get_type();
-
-                let max_rem = ctx
-                    .builder
-                    .build_int_sub(power_of_two, llvm_val_t.const_int(1, false), "")
-                    .unwrap();
-                ctx.builder
-                    .build_and(
-                        ctx.builder.build_int_add(val, max_rem, "").unwrap(),
-                        ctx.builder.build_not(max_rem, "").unwrap(),
-                        "",
-                    )
-                    .unwrap()
-            };
-
-            // Allocate the resulting ndarray
-            // A condition after format_rpc_ret ensures this will not be popped this off.
-            let (dtype, ndims) = unpack_ndarray_var_tys(&mut ctx.unifier, ret_ty);
-            let dtype_llvm = ctx.get_llvm_type(generator, dtype);
-            let ndims = extract_ndims(&ctx.unifier, ndims);
-            let ndarray = NDArrayType::new(generator, ctx.ctx, dtype_llvm, Some(ndims))
-                .construct_uninitialized(generator, ctx, None);
-
-            // NOTE: Current content of `ndarray`:
-            //   - * `data` - **NOT YET** allocated.
-            //   - * `itemsize` - initialized to be size_of(dtype).
-            //   - * `ndims` - initialized.
-            //   - * `shape` - allocated; has uninitialized values.
-            //   - * `strides` - allocated; has uninitialized values.
-
-            let itemsize = ndarray.load_itemsize(ctx); // Same as doing a `ctx.get_llvm_type` on `dtype` and get its `size_of()`.
-
-            // Allocates a buffer for the initial RPC'ed object, which is guaranteed to be
-            // (4 + 4 * ndims) bytes with 8-byte alignment
-            let sizeof_usize = llvm_usize.size_of();
-            let sizeof_usize =
-                ctx.builder.build_int_truncate_or_bit_cast(sizeof_usize, llvm_usize, "").unwrap();
-
-            let sizeof_ptr = llvm_i8.ptr_type(AddressSpace::default()).size_of();
-            let sizeof_ptr =
-                ctx.builder.build_int_z_extend_or_bit_cast(sizeof_ptr, llvm_usize, "").unwrap();
-
-            let sizeof_shape =
-                ctx.builder.build_int_mul(ndarray.load_ndims(ctx), sizeof_usize, "").unwrap();
-
-            // Size of the buffer for the initial `rpc_recv()`.
-            let unaligned_buffer_size =
-                ctx.builder.build_int_add(sizeof_ptr, sizeof_shape, "").unwrap();
-
-            let stackptr = call_stacksave(ctx, None);
-            let buffer = type_aligned_alloca(
-                generator,
-                ctx,
-                llvm_i8_8,
-                unaligned_buffer_size,
-                Some("rpc.buffer"),
-            );
-            let buffer = ArraySliceValue::from_ptr_val(buffer, unaligned_buffer_size, None);
-
-            // The first call to `rpc_recv` reads the top-level ndarray object: [pdata, shape]
-            //
-            // The returned value is the number of bytes for `ndarray.data`.
-            let ndarray_nbytes = ctx
-                .build_call_or_invoke(
-                    rpc_recv,
-                    &[buffer.base_ptr(ctx, generator).into()], // Reads [usize; ndims]
-                    "rpc.size.next",
-                )
-                .map(BasicValueEnum::into_int_value)
-                .unwrap();
-
-            // debug_assert(ndarray_nbytes > 0)
-            if ctx.registry.llvm_options.opt_level == OptimizationLevel::None {
-                let cmp = ctx
-                    .builder
-                    .build_int_compare(IntPredicate::UGT, ndarray_nbytes, num_0, "")
-                    .unwrap();
-
-                ctx.make_assert(
-                    generator,
-                    cmp,
-                    "0:AssertionError",
-                    "Unexpected RPC termination for ndarray - Expected data buffer next",
-                    [None, None, None],
-                    ctx.current_loc,
-                );
-            }
-
-            // Copy shape from the buffer to `ndarray.shape`.
-            // We need to skip the first `sizeof(uint8_t*)` bytes to skip the `pdata` in `[pdata, shape]`.
-            let pbuffer_shape =
-                unsafe { buffer.ptr_offset_unchecked(ctx, generator, &sizeof_ptr, None) };
-            let pbuffer_shape =
-                ctx.builder.build_pointer_cast(pbuffer_shape, llvm_pusize, "").unwrap();
-
-            // Copy shape from buffer to `ndarray.shape`
-            ndarray.copy_shape_from_array(generator, ctx, pbuffer_shape);
-
-            // Restore stack from before allocation of buffer
-            call_stackrestore(ctx, stackptr);
-
-            // Allocate `ndarray.data`.
-            // `ndarray.shape` must be initialized beforehand in this implementation
-            //   (for ndarray.create_data() to know how many elements to allocate)
-            unsafe { ndarray.create_data(generator, ctx) }; // NOTE: the strides of `ndarray` has also been set to contiguous in `create_data`.
-
-            // debug_assert(nelems * sizeof(T) >= ndarray_nbytes)
-            if ctx.registry.llvm_options.opt_level == OptimizationLevel::None {
-                let num_elements = ndarray.size(generator, ctx);
-
-                let expected_ndarray_nbytes =
-                    ctx.builder.build_int_mul(num_elements, itemsize, "").unwrap();
-                let cmp = ctx
-                    .builder
-                    .build_int_compare(
-                        IntPredicate::UGE,
-                        expected_ndarray_nbytes,
-                        ndarray_nbytes,
-                        "",
-                    )
-                    .unwrap();
-
-                ctx.make_assert(
-                    generator,
-                    cmp,
-                    "0:AssertionError",
-                    "Unexpected allocation size request for ndarray data - Expected up to {0} bytes, got {1} bytes",
-                    [Some(expected_ndarray_nbytes), Some(ndarray_nbytes), None],
-                    ctx.current_loc,
-                );
-            }
-
-            let ndarray_data = ndarray.data().base_ptr(ctx, generator);
-
-            // NOTE: Currently on `prehead_bb`
-            ctx.builder.build_unconditional_branch(head_bb).unwrap();
-
-            // Inserting into `head_bb`. Do `rpc_recv` for `data` recursively.
-            ctx.builder.position_at_end(head_bb);
-
-            let phi = ctx.builder.build_phi(llvm_pi8, "rpc.ptr").unwrap();
-            phi.add_incoming(&[(&ndarray_data, prehead_bb)]);
-
-            let alloc_size = ctx
-                .build_call_or_invoke(rpc_recv, &[phi.as_basic_value()], "rpc.size.next")
-                .map(BasicValueEnum::into_int_value)
-                .unwrap();
-
-            let is_done = ctx
-                .builder
-                .build_int_compare(IntPredicate::EQ, llvm_i32.const_zero(), alloc_size, "rpc.done")
-                .unwrap();
-            ctx.builder.build_conditional_branch(is_done, tail_bb, alloc_bb).unwrap();
-
-            ctx.builder.position_at_end(alloc_bb);
-            // Align the allocation to sizeof(T)
-            let alloc_size = round_up(ctx, alloc_size, itemsize);
-            // TODO(Derppening): Candidate for refactor into type_aligned_alloca
-            let alloc_ptr = ctx
-                .builder
-                .build_array_alloca(
-                    dtype_llvm,
-                    ctx.builder.build_int_unsigned_div(alloc_size, itemsize, "").unwrap(),
-                    "rpc.alloc",
-                )
-                .unwrap();
-            let alloc_ptr =
-                ctx.builder.build_pointer_cast(alloc_ptr, llvm_pi8, "rpc.alloc.ptr").unwrap();
-            phi.add_incoming(&[(&alloc_ptr, alloc_bb)]);
-            ctx.builder.build_unconditional_branch(head_bb).unwrap();
-
-            ctx.builder.position_at_end(tail_bb);
-            ndarray.as_base_value().into()
-        }
-
-        _ => {
-            let slot = ctx.builder.build_alloca(llvm_ret_ty, "rpc.ret.slot").unwrap();
-            let slotgen = ctx.builder.build_bit_cast(slot, llvm_pi8, "rpc.ret.ptr").unwrap();
-            ctx.builder.build_unconditional_branch(head_bb).unwrap();
-            ctx.builder.position_at_end(head_bb);
-
-            let phi = ctx.builder.build_phi(llvm_pi8, "rpc.ptr").unwrap();
-            phi.add_incoming(&[(&slotgen, prehead_bb)]);
-            let alloc_size = ctx
-                .build_call_or_invoke(rpc_recv, &[phi.as_basic_value()], "rpc.size.next")
-                .unwrap()
-                .into_int_value();
-            let is_done = ctx
-                .builder
-                .build_int_compare(IntPredicate::EQ, llvm_i32.const_zero(), alloc_size, "rpc.done")
-                .unwrap();
-
-            ctx.builder.build_conditional_branch(is_done, tail_bb, alloc_bb).unwrap();
-            ctx.builder.position_at_end(alloc_bb);
-
-            let alloc_ptr =
-                ctx.builder.build_array_alloca(llvm_pi8, alloc_size, "rpc.alloc").unwrap();
-            let alloc_ptr =
-                ctx.builder.build_bit_cast(alloc_ptr, llvm_pi8, "rpc.alloc.ptr").unwrap();
-            phi.add_incoming(&[(&alloc_ptr, alloc_bb)]);
-            ctx.builder.build_unconditional_branch(head_bb).unwrap();
-
-            ctx.builder.position_at_end(tail_bb);
-            ctx.builder.build_load(slot, "rpc.result").unwrap()
-        }
-    };
-
-    Some(result)
-}
-
 fn rpc_codegen_callback_fn<'ctx>(
     ctx: &mut CodeGenContext<'ctx, '_>,
     obj: Option<(Type, ValueEnum<'ctx>)>,
     fun: (&FunSignature, DefinitionId),
     args: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
     generator: &mut dyn CodeGenerator,
-    is_async: bool,
 ) -> Result<Option<BasicValueEnum<'ctx>>, String> {
     let int8 = ctx.ctx.i8_type();
     let int32 = ctx.ctx.i32_type();
@@ -902,72 +623,91 @@ fn rpc_codegen_callback_fn<'ctx>(
     }
 
     // call
-    if is_async {
-        let rpc_send_async = ctx.module.get_function("rpc_send_async").unwrap_or_else(|| {
-            ctx.module.add_function(
-                "rpc_send_async",
-                ctx.ctx.void_type().fn_type(
-                    &[
-                        int32.into(),
-                        tag_ptr_type.ptr_type(AddressSpace::default()).into(),
-                        ptr_type.ptr_type(AddressSpace::default()).into(),
-                    ],
-                    false,
-                ),
-                None,
-            )
-        });
-        ctx.builder
-            .build_call(
-                rpc_send_async,
-                &[service_id.into(), tag_ptr.into(), args_ptr.into()],
-                "rpc.send",
-            )
-            .unwrap();
-    } else {
-        let rpc_send = ctx.module.get_function("rpc_send").unwrap_or_else(|| {
-            ctx.module.add_function(
-                "rpc_send",
-                ctx.ctx.void_type().fn_type(
-                    &[
-                        int32.into(),
-                        tag_ptr_type.ptr_type(AddressSpace::default()).into(),
-                        ptr_type.ptr_type(AddressSpace::default()).into(),
-                    ],
-                    false,
-                ),
-                None,
-            )
-        });
-        ctx.builder
-            .build_call(rpc_send, &[service_id.into(), tag_ptr.into(), args_ptr.into()], "rpc.send")
-            .unwrap();
-    }
+    let rpc_send = ctx.module.get_function("rpc_send").unwrap_or_else(|| {
+        ctx.module.add_function(
+            "rpc_send",
+            ctx.ctx.void_type().fn_type(
+                &[
+                    int32.into(),
+                    tag_ptr_type.ptr_type(AddressSpace::default()).into(),
+                    ptr_type.ptr_type(AddressSpace::default()).into(),
+                ],
+                false,
+            ),
+            None,
+        )
+    });
+    ctx.builder
+        .build_call(rpc_send, &[service_id.into(), tag_ptr.into(), args_ptr.into()], "rpc.send")
+        .unwrap();
 
     // reclaim stack space used by arguments
     call_stackrestore(ctx, stackptr);
 
-    if is_async {
-        // async RPCs do not return any values
-        Ok(None)
-    } else {
-        let result = format_rpc_ret(generator, ctx, fun.0.ret);
+    // -- receive value:
+    // T result = {
+    //   void *ret_ptr = alloca(sizeof(T));
+    //   void *ptr = ret_ptr;
+    //   loop: int size = rpc_recv(ptr);
+    //   // Non-zero: Provide `size` bytes of extra storage for variable-length data.
+    //   if(size) { ptr = alloca(size); goto loop; }
+    //   else *(T*)ret_ptr
+    // }
+    let rpc_recv = ctx.module.get_function("rpc_recv").unwrap_or_else(|| {
+        ctx.module.add_function("rpc_recv", int32.fn_type(&[ptr_type.into()], false), None)
+    });
 
-        if !result.is_some_and(|res| res.get_type().is_pointer_type()) {
-            // An RPC returning an NDArray would not touch here.
-            call_stackrestore(ctx, stackptr);
-        }
-
-        Ok(result)
+    if ctx.unifier.unioned(fun.0.ret, ctx.primitives.none) {
+        ctx.build_call_or_invoke(rpc_recv, &[ptr_type.const_null().into()], "rpc_recv");
+        return Ok(None);
     }
+
+    let prehead_bb = ctx.builder.get_insert_block().unwrap();
+    let current_function = prehead_bb.get_parent().unwrap();
+    let head_bb = ctx.ctx.append_basic_block(current_function, "rpc.head");
+    let alloc_bb = ctx.ctx.append_basic_block(current_function, "rpc.continue");
+    let tail_bb = ctx.ctx.append_basic_block(current_function, "rpc.tail");
+
+    let ret_ty = ctx.get_llvm_abi_type(generator, fun.0.ret);
+    let need_load = !ret_ty.is_pointer_type();
+    let slot = ctx.builder.build_alloca(ret_ty, "rpc.ret.slot").unwrap();
+    let slotgen = ctx.builder.build_bitcast(slot, ptr_type, "rpc.ret.ptr").unwrap();
+    ctx.builder.build_unconditional_branch(head_bb).unwrap();
+    ctx.builder.position_at_end(head_bb);
+
+    let phi = ctx.builder.build_phi(ptr_type, "rpc.ptr").unwrap();
+    phi.add_incoming(&[(&slotgen, prehead_bb)]);
+    let alloc_size = ctx
+        .build_call_or_invoke(rpc_recv, &[phi.as_basic_value()], "rpc.size.next")
+        .unwrap()
+        .into_int_value();
+    let is_done = ctx
+        .builder
+        .build_int_compare(inkwell::IntPredicate::EQ, int32.const_zero(), alloc_size, "rpc.done")
+        .unwrap();
+
+    ctx.builder.build_conditional_branch(is_done, tail_bb, alloc_bb).unwrap();
+    ctx.builder.position_at_end(alloc_bb);
+
+    let alloc_ptr = ctx.builder.build_array_alloca(ptr_type, alloc_size, "rpc.alloc").unwrap();
+    let alloc_ptr = ctx.builder.build_bitcast(alloc_ptr, ptr_type, "rpc.alloc.ptr").unwrap();
+    phi.add_incoming(&[(&alloc_ptr, alloc_bb)]);
+    ctx.builder.build_unconditional_branch(head_bb).unwrap();
+
+    ctx.builder.position_at_end(tail_bb);
+
+    let result = ctx.builder.build_load(slot, "rpc.result").unwrap();
+    if need_load {
+        call_stackrestore(ctx, stackptr);
+    }
+    Ok(Some(result))
 }
 
-pub fn attributes_writeback<'ctx>(
-    ctx: &mut CodeGenContext<'ctx, '_>,
+pub fn attributes_writeback(
+    ctx: &mut CodeGenContext<'_, '_>,
     generator: &mut dyn CodeGenerator,
     inner_resolver: &InnerResolver,
     host_attributes: &PyObject,
-    return_obj: Option<(Type, ValueEnum<'ctx>)>,
 ) -> Result<(), String> {
     Python::with_gil(|py| -> PyResult<Result<(), String>> {
         let host_attributes: &PyList = host_attributes.downcast(py)?;
@@ -977,11 +717,6 @@ pub fn attributes_writeback<'ctx>(
         let zero = int32.const_zero();
         let mut values = Vec::new();
         let mut scratch_buffer = Vec::new();
-
-        if let Some((ty, obj)) = return_obj {
-            values.push((ty, obj.to_basic_value_enum(ctx, generator, ty).unwrap()));
-        }
-
         for val in (*globals).values() {
             let val = val.as_ref(py);
             let ty = inner_resolver.get_obj_type(
@@ -1060,7 +795,7 @@ pub fn attributes_writeback<'ctx>(
         let args: Vec<_> =
             values.into_iter().map(|(_, val)| (None, ValueEnum::Dynamic(val))).collect();
         if let Err(e) =
-            rpc_codegen_callback_fn(ctx, None, (&fun, PrimDef::Int32.id()), args, generator, true)
+            rpc_codegen_callback_fn(ctx, None, (&fun, PrimDef::Int32.id()), args, generator)
         {
             return Ok(Err(e));
         }
@@ -1070,9 +805,9 @@ pub fn attributes_writeback<'ctx>(
     Ok(())
 }
 
-pub fn rpc_codegen_callback(is_async: bool) -> Arc<GenCall> {
-    Arc::new(GenCall::new(Box::new(move |ctx, obj, fun, args, generator| {
-        rpc_codegen_callback_fn(ctx, obj, fun, args, generator, is_async)
+pub fn rpc_codegen_callback() -> Arc<GenCall> {
+    Arc::new(GenCall::new(Box::new(|ctx, obj, fun, args, generator| {
+        rpc_codegen_callback_fn(ctx, obj, fun, args, generator)
     })))
 }
 
@@ -1286,8 +1021,7 @@ fn polymorphic_print<'ctx>(
                 fmt.push('[');
                 flush(ctx, generator, &mut fmt, &mut args);
 
-                let val =
-                    ListValue::from_pointer_value(value.into_pointer_value(), llvm_usize, None);
+                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();
@@ -1341,27 +1075,23 @@ fn polymorphic_print<'ctx>(
                 fmt.push_str("array([");
                 flush(ctx, generator, &mut fmt, &mut args);
 
-                let (dtype, _) = unpack_ndarray_var_tys(&mut ctx.unifier, ty);
-                let ndarray = NDArrayType::from_unifier_type(generator, ctx, ty)
-                    .map_value(value.into_pointer_value(), None);
+                let ndarray = AnyObject { ty, value };
+                let ndarray = NDArrayObject::from_object(generator, ctx, ndarray);
 
-                let num_0 = llvm_usize.const_zero();
+                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(ctx);
-                    let scalar = hdl.get_scalar(ctx);
+                    let i = hdl.get_index(generator, ctx);
+                    let scalar = hdl.get_scalar(generator, ctx);
 
-                    // if (i != 0) puts(", ");
+                    // if (i != 0) { puts(", "); }
                     gen_if_callback(
                         generator,
                         ctx,
                         |_, ctx| {
-                            let not_first = ctx
-                                .builder
-                                .build_int_compare(IntPredicate::NE, i, num_0, "")
-                                .unwrap();
-                            Ok(not_first)
+                            let not_first = i.compare(ctx, IntPredicate::NE, num_0);
+                            Ok(not_first.value)
                         },
                         |generator, ctx| {
                             printf(ctx, generator, ", \0".into(), Vec::default());
@@ -1374,7 +1104,7 @@ fn polymorphic_print<'ctx>(
                     polymorphic_print(
                         ctx,
                         generator,
-                        &[(dtype, scalar.into())],
+                        &[(scalar.ty, scalar.value.into())],
                         "",
                         None,
                         true,
@@ -1391,7 +1121,7 @@ fn polymorphic_print<'ctx>(
                 fmt.push_str("range(");
                 flush(ctx, generator, &mut fmt, &mut args);
 
-                let val = RangeValue::from_pointer_value(value.into_pointer_value(), None);
+                let val = RangeValue::from_ptr_val(value.into_pointer_value(), None);
 
                 let (start, stop, step) = destructure_range(ctx, val);
 

nac3artiq/src/lib.rs

@@ -1,4 +1,10 @@
-#![deny(future_incompatible, let_underscore, nonstandard_style, clippy::all)]
+#![deny(
+    future_incompatible,
+    let_underscore,
+    nonstandard_style,
+    rust_2024_compatibility,
+    clippy::all
+)]
 #![warn(clippy::pedantic)]
 #![allow(
     unsafe_op_in_unsafe_fn,
@@ -10,65 +16,65 @@
     clippy::wildcard_imports
 )]
 
-use std::{
-    collections::{HashMap, HashSet},
-    fs,
-    io::Write,
-    process::Command,
-    rc::Rc,
-    sync::Arc,
-};
+use std::collections::{HashMap, HashSet};
+use std::fs;
+use std::io::Write;
+use std::process::Command;
+use std::rc::Rc;
+use std::sync::Arc;
 
-use itertools::Itertools;
-use parking_lot::{Mutex, RwLock};
-use pyo3::{
-    create_exception, exceptions,
-    prelude::*,
-    types::{PyBytes, PyDict, PyNone, PySet},
+use inkwell::{
+    context::Context,
+    memory_buffer::MemoryBuffer,
+    module::{Linkage, Module},
+    passes::PassBuilderOptions,
+    support::is_multithreaded,
+    targets::*,
+    OptimizationLevel,
 };
-use tempfile::{self, TempDir};
+use itertools::Itertools;
+use nac3core::codegen::irrt::setup_irrt_exceptions;
+use nac3core::codegen::{gen_func_impl, CodeGenLLVMOptions, CodeGenTargetMachineOptions};
+use nac3core::toplevel::builtins::get_exn_constructor;
+use nac3core::typecheck::typedef::{into_var_map, TypeEnum, Unifier, VarMap};
+use nac3parser::{
+    ast::{ExprKind, Stmt, StmtKind, StrRef},
+    parser::parse_program,
+};
+use pyo3::create_exception;
+use pyo3::prelude::*;
+use pyo3::{exceptions, types::PyBytes, types::PyDict, types::PySet};
+
+use parking_lot::{Mutex, RwLock};
 
 use nac3core::{
-    codegen::{
-        concrete_type::ConcreteTypeStore, gen_func_impl, irrt::load_irrt, CodeGenLLVMOptions,
-        CodeGenTargetMachineOptions, CodeGenTask, CodeGenerator, WithCall, WorkerRegistry,
-    },
-    inkwell::{
-        context::Context,
-        memory_buffer::MemoryBuffer,
-        module::{FlagBehavior, Linkage, Module},
-        passes::PassBuilderOptions,
-        support::is_multithreaded,
-        targets::*,
-        OptimizationLevel,
-    },
-    nac3parser::{
-        ast::{Constant, ExprKind, Located, Stmt, StmtKind, StrRef},
-        parser::parse_program,
-    },
+    codegen::irrt::load_irrt,
+    codegen::{concrete_type::ConcreteTypeStore, CodeGenTask, WithCall, WorkerRegistry},
     symbol_resolver::SymbolResolver,
     toplevel::{
-        builtins::get_exn_constructor,
         composer::{BuiltinFuncCreator, BuiltinFuncSpec, ComposerConfig, TopLevelComposer},
         DefinitionId, GenCall, TopLevelDef,
     },
-    typecheck::{
-        type_inferencer::PrimitiveStore,
-        typedef::{into_var_map, FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
-    },
+    typecheck::typedef::{FunSignature, FuncArg},
+    typecheck::{type_inferencer::PrimitiveStore, typedef::Type},
 };
+
 use nac3ld::Linker;
 
-use codegen::{
-    attributes_writeback, gen_core_log, gen_rtio_log, rpc_codegen_callback, ArtiqCodeGenerator,
+use crate::{
+    codegen::{
+        attributes_writeback, gen_core_log, gen_rtio_log, rpc_codegen_callback, ArtiqCodeGenerator,
+    },
+    symbol_resolver::{DeferredEvaluationStore, InnerResolver, PythonHelper, Resolver},
 };
-use symbol_resolver::{DeferredEvaluationStore, InnerResolver, PythonHelper, Resolver};
-use timeline::TimeFns;
+use tempfile::{self, TempDir};
 
 mod codegen;
 mod symbol_resolver;
 mod timeline;
 
+use timeline::TimeFns;
+
 #[derive(PartialEq, Clone, Copy)]
 enum Isa {
     Host,
@@ -142,32 +148,14 @@ impl Nac3 {
         module: &PyObject,
         registered_class_ids: &HashSet<u64>,
     ) -> PyResult<()> {
-        let (module_name, source_file, source) =
-            Python::with_gil(|py| -> PyResult<(String, String, String)> {
-                let module: &PyAny = module.extract(py)?;
-                let source_file = module.getattr("__file__");
-                let (source_file, source) = if let Ok(source_file) = source_file {
-                    let source_file = source_file.extract()?;
-                    (
-                        source_file,
-                        fs::read_to_string(source_file).map_err(|e| {
-                            exceptions::PyIOError::new_err(format!(
-                                "failed to read input file: {e}"
-                            ))
-                        })?,
-                    )
-                } else {
-                    // kernels submitted by content have no file
-                    // but still can provide source by StringLoader
-                    let get_src_fn = module
-                        .getattr("__loader__")?
-                        .extract::<PyObject>()?
-                        .getattr(py, "get_source")?;
-                    ("<expcontent>", get_src_fn.call1(py, (PyNone::get(py),))?.extract(py)?)
-                };
-                Ok((module.getattr("__name__")?.extract()?, source_file.to_string(), source))
-            })?;
+        let (module_name, source_file) = Python::with_gil(|py| -> PyResult<(String, String)> {
+            let module: &PyAny = module.extract(py)?;
+            Ok((module.getattr("__name__")?.extract()?, module.getattr("__file__")?.extract()?))
+        })?;
 
+        let source = fs::read_to_string(&source_file).map_err(|e| {
+            exceptions::PyIOError::new_err(format!("failed to read input file: {e}"))
+        })?;
         let parser_result = parse_program(&source, source_file.into())
             .map_err(|e| exceptions::PySyntaxError::new_err(format!("parse error: {e}")))?;
 
@@ -207,8 +195,10 @@ impl Nac3 {
                     body.retain(|stmt| {
                         if let StmtKind::FunctionDef { ref decorator_list, .. } = stmt.node {
                             decorator_list.iter().any(|decorator| {
-                                if let Some(id) = decorator_id_string(decorator) {
-                                    id == "kernel" || id == "portable" || id == "rpc"
+                                if let ExprKind::Name { id, .. } = decorator.node {
+                                    id.to_string() == "kernel"
+                                        || id.to_string() == "portable"
+                                        || id.to_string() == "rpc"
                                 } else {
                                     false
                                 }
@@ -221,8 +211,9 @@ impl Nac3 {
                 }
                 StmtKind::FunctionDef { ref decorator_list, .. } => {
                     decorator_list.iter().any(|decorator| {
-                        if let Some(id) = decorator_id_string(decorator) {
-                            id == "extern" || id == "kernel" || id == "portable" || id == "rpc"
+                        if let ExprKind::Name { id, .. } = decorator.node {
+                            let id = id.to_string();
+                            id == "extern" || id == "portable" || id == "kernel" || id == "rpc"
                         } else {
                             false
                         }
@@ -458,6 +449,7 @@ impl Nac3 {
                         pyid_to_type: pyid_to_type.clone(),
                         primitive_ids: self.primitive_ids.clone(),
                         global_value_ids: global_value_ids.clone(),
+                        class_names: Mutex::default(),
                         name_to_pyid: name_to_pyid.clone(),
                         module: module.clone(),
                         id_to_pyval: RwLock::default(),
@@ -488,25 +480,9 @@ impl Nac3 {
 
             match &stmt.node {
                 StmtKind::FunctionDef { decorator_list, .. } => {
-                    if decorator_list
-                        .iter()
-                        .any(|decorator| decorator_id_string(decorator) == Some("rpc".to_string()))
-                    {
-                        store_fun
-                            .call1(
-                                py,
-                                (
-                                    def_id.0.into_py(py),
-                                    module.getattr(py, name.to_string().as_str()).unwrap(),
-                                ),
-                            )
-                            .unwrap();
-                        let is_async = decorator_list.iter().any(|decorator| {
-                            decorator_get_flags(decorator)
-                                .iter()
-                                .any(|constant| *constant == Constant::Str("async".into()))
-                        });
-                        rpc_ids.push((None, def_id, is_async));
+                    if decorator_list.iter().any(|decorator| matches!(decorator.node, ExprKind::Name { id, .. } if id == "rpc".into())) {
+                        store_fun.call1(py, (def_id.0.into_py(py), module.getattr(py, name.to_string().as_str()).unwrap())).unwrap();
+                        rpc_ids.push((None, def_id));
                     }
                 }
                 StmtKind::ClassDef { name, body, .. } => {
@@ -514,26 +490,19 @@ impl Nac3 {
                     let class_obj = module.getattr(py, class_name.as_str()).unwrap();
                     for stmt in body {
                         if let StmtKind::FunctionDef { name, decorator_list, .. } = &stmt.node {
-                            if decorator_list.iter().any(|decorator| {
-                                decorator_id_string(decorator) == Some("rpc".to_string())
-                            }) {
-                                let is_async = decorator_list.iter().any(|decorator| {
-                                    decorator_get_flags(decorator)
-                                        .iter()
-                                        .any(|constant| *constant == Constant::Str("async".into()))
-                                });
+                            if decorator_list.iter().any(|decorator| matches!(decorator.node, ExprKind::Name { id, .. } if id == "rpc".into())) {
                                 if name == &"__init__".into() {
                                     return Err(CompileError::new_err(format!(
                                         "compilation failed\n----------\nThe constructor of class {} should not be decorated with rpc decorator (at {})",
                                         class_name, stmt.location
                                     )));
                                 }
-                                rpc_ids.push((Some((class_obj.clone(), *name)), def_id, is_async));
+                                rpc_ids.push((Some((class_obj.clone(), *name)), def_id));
                             }
                         }
                     }
                 }
-                _ => (),
+                _ => ()
             }
 
             let id = *name_to_pyid.get(&name).unwrap();
@@ -572,12 +541,13 @@ impl Nac3 {
             pyid_to_type: pyid_to_type.clone(),
             primitive_ids: self.primitive_ids.clone(),
             global_value_ids: global_value_ids.clone(),
+            class_names: Mutex::default(),
             id_to_pyval: RwLock::default(),
             id_to_primitive: RwLock::default(),
             field_to_val: RwLock::default(),
             name_to_pyid,
             module: module.to_object(py),
-            helper: helper.clone(),
+            helper,
             string_store: self.string_store.clone(),
             exception_ids: self.exception_ids.clone(),
             deferred_eval_store: self.deferred_eval_store.clone(),
@@ -589,8 +559,9 @@ impl Nac3 {
             .unwrap();
 
         // Process IRRT
-        let context = Context::create();
-        let irrt = load_irrt(&context, resolver.as_ref());
+        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() };
@@ -629,12 +600,13 @@ impl Nac3 {
         let top_level = Arc::new(composer.make_top_level_context());
 
         {
+            let rpc_codegen = rpc_codegen_callback();
             let defs = top_level.definitions.read();
-            for (class_data, id, is_async) in &rpc_ids {
+            for (class_data, id) in &rpc_ids {
                 let mut def = defs[id.0].write();
                 match &mut *def {
                     TopLevelDef::Function { codegen_callback, .. } => {
-                        *codegen_callback = Some(rpc_codegen_callback(*is_async));
+                        *codegen_callback = Some(rpc_codegen.clone());
                     }
                     TopLevelDef::Class { methods, .. } => {
                         let (class_def, method_name) = class_data.as_ref().unwrap();
@@ -645,7 +617,7 @@ impl Nac3 {
                             if let TopLevelDef::Function { codegen_callback, .. } =
                                 &mut *defs[id.0].write()
                             {
-                                *codegen_callback = Some(rpc_codegen_callback(*is_async));
+                                *codegen_callback = Some(rpc_codegen.clone());
                                 store_fun
                                     .call1(
                                         py,
@@ -660,11 +632,6 @@ impl Nac3 {
                             }
                         }
                     }
-                    TopLevelDef::Variable { .. } => {
-                        return Err(CompileError::new_err(String::from(
-                            "Unsupported @rpc annotation on global variable",
-                        )))
-                    }
                 }
             }
         }
@@ -685,12 +652,33 @@ impl Nac3 {
         let task = CodeGenTask {
             subst: Vec::default(),
             symbol_name: "__modinit__".to_string(),
+            body: instance.body,
+            signature,
+            resolver: resolver.clone(),
+            store,
+            unifier_index: instance.unifier_id,
+            calls: instance.calls,
+            id: 0,
+        };
+
+        let mut store = ConcreteTypeStore::new();
+        let mut cache = HashMap::new();
+        let signature = store.from_signature(
+            &mut composer.unifier,
+            &self.primitive,
+            &fun_signature,
+            &mut cache,
+        );
+        let signature = store.add_cty(signature);
+        let attributes_writeback_task = CodeGenTask {
+            subst: Vec::default(),
+            symbol_name: "attributes_writeback".to_string(),
             body: Arc::new(Vec::default()),
             signature,
             resolver,
             store,
             unifier_index: instance.unifier_id,
-            calls: instance.calls,
+            calls: Arc::new(HashMap::default()),
             id: 0,
         };
 
@@ -703,7 +691,7 @@ impl Nac3 {
             let buffer = buffer.as_slice().into();
             membuffer.lock().push(buffer);
         })));
-        let size_t = context
+        let size_t = Context::create()
             .ptr_sized_int_type(&self.get_llvm_target_machine().get_target_data(), None)
             .get_bit_width();
         let num_threads = if is_multithreaded() { 4 } else { 1 };
@@ -714,27 +702,19 @@ impl Nac3 {
             .collect();
 
         let membuffer = membuffers.clone();
-        let mut has_return = false;
         py.allow_threads(|| {
             let (registry, handles) =
                 WorkerRegistry::create_workers(threads, top_level.clone(), &self.llvm_options, &f);
+            registry.add_task(task);
+            registry.wait_tasks_complete(handles);
 
-            let mut generator = ArtiqCodeGenerator::new("main".to_string(), size_t, self.time_fns);
-            let context = Context::create();
-            let module = context.create_module("main");
+            let mut generator =
+                ArtiqCodeGenerator::new("attributes_writeback".to_string(), size_t, self.time_fns);
+            let context = inkwell::context::Context::create();
+            let module = context.create_module("attributes_writeback");
             let 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",
-                FlagBehavior::Warning,
-                context.i32_type().const_int(3, false),
-            );
-            module.add_basic_value_flag(
-                "Dwarf Version",
-                FlagBehavior::Warning,
-                context.i32_type().const_int(4, false),
-            );
             let builder = context.create_builder();
             let (_, module, _) = gen_func_impl(
                 &context,
@@ -742,27 +722,9 @@ impl Nac3 {
                 &registry,
                 builder,
                 module,
-                task,
+                attributes_writeback_task,
                 |generator, ctx| {
-                    assert_eq!(instance.body.len(), 1, "toplevel module should have 1 statement");
-                    let StmtKind::Expr { value: ref expr, .. } = instance.body[0].node else {
-                        unreachable!("toplevel statement must be an expression")
-                    };
-                    let ExprKind::Call { .. } = expr.node else {
-                        unreachable!("toplevel expression must be a function call")
-                    };
-
-                    let return_obj =
-                        generator.gen_expr(ctx, expr)?.map(|value| (expr.custom.unwrap(), value));
-                    has_return = return_obj.is_some();
-                    registry.wait_tasks_complete(handles);
-                    attributes_writeback(
-                        ctx,
-                        generator,
-                        inner_resolver.as_ref(),
-                        &host_attributes,
-                        return_obj,
-                    )
+                    attributes_writeback(ctx, generator, inner_resolver.as_ref(), &host_attributes)
                 },
             )
             .unwrap();
@@ -771,23 +733,35 @@ impl Nac3 {
             membuffer.lock().push(buffer);
         });
 
-        embedding_map.setattr("expects_return", has_return).unwrap();
-
         // Link all modules into `main`.
         let buffers = membuffers.lock();
         let main = context
-            .create_module_from_ir(MemoryBuffer::create_from_memory_range(
-                buffers.last().unwrap(),
-                "main",
-            ))
+            .create_module_from_ir(MemoryBuffer::create_from_memory_range(&buffers[0], "main"))
             .unwrap();
-        for buffer in buffers.iter().rev().skip(1) {
+        for buffer in buffers.iter().skip(1) {
             let other = context
                 .create_module_from_ir(MemoryBuffer::create_from_memory_range(buffer, "main"))
                 .unwrap();
 
             main.link_in_module(other).map_err(|err| CompileError::new_err(err.to_string()))?;
         }
+        let builder = context.create_builder();
+        let modinit_return = main
+            .get_function("__modinit__")
+            .unwrap()
+            .get_last_basic_block()
+            .unwrap()
+            .get_terminator()
+            .unwrap();
+        builder.position_before(&modinit_return);
+        builder
+            .build_call(
+                main.get_function("attributes_writeback").unwrap(),
+                &[],
+                "attributes_writeback",
+            )
+            .unwrap();
+
         main.link_in_module(irrt).map_err(|err| CompileError::new_err(err.to_string()))?;
 
         let mut function_iter = main.get_first_function();
@@ -822,20 +796,6 @@ impl Nac3 {
             panic!("Failed to run optimization for module `main`: {}", err.to_string());
         }
 
-        Python::with_gil(|py| {
-            let string_store = self.string_store.read();
-            let mut string_store_vec = string_store.iter().collect::<Vec<_>>();
-            string_store_vec.sort_by(|(_s1, key1), (_s2, key2)| key1.cmp(key2));
-            for (s, key) in string_store_vec {
-                let embed_key: i32 = helper.store_str.call1(py, (s,)).unwrap().extract(py).unwrap();
-                assert_eq!(
-                    embed_key, *key,
-                    "string {s} is out of sync between embedding map (key={embed_key}) and \
-                    the internal string store (key={key})"
-                );
-            }
-        });
-
         link_fn(&main)
     }
 
@@ -888,41 +848,6 @@ impl Nac3 {
     }
 }
 
-/// Retrieves the Name.id from a decorator, supports decorators with arguments.
-fn decorator_id_string(decorator: &Located<ExprKind>) -> Option<String> {
-    if let ExprKind::Name { id, .. } = decorator.node {
-        // Bare decorator
-        return Some(id.to_string());
-    } else if let ExprKind::Call { func, .. } = &decorator.node {
-        // Decorators that are calls (e.g. "@rpc()") have Call for the node,
-        // need to extract the id from within.
-        if let ExprKind::Name { id, .. } = func.node {
-            return Some(id.to_string());
-        }
-    }
-    None
-}
-
-/// Retrieves flags from a decorator, if any.
-fn decorator_get_flags(decorator: &Located<ExprKind>) -> Vec<Constant> {
-    let mut flags = vec![];
-    if let ExprKind::Call { keywords, .. } = &decorator.node {
-        for keyword in keywords {
-            if keyword.node.arg != Some("flags".into()) {
-                continue;
-            }
-            if let ExprKind::Set { elts } = &keyword.node.value.node {
-                for elt in elts {
-                    if let ExprKind::Constant { value, .. } = &elt.node {
-                        flags.push(value.clone());
-                    }
-                }
-            }
-        }
-    }
-    flags
-}
-
 fn link_with_lld(elf_filename: String, obj_filename: String) -> PyResult<()> {
     let linker_args = vec![
         "-shared".to_string(),
@@ -1085,48 +1010,6 @@ impl Nac3 {
         let working_directory = tempfile::Builder::new().prefix("nac3-").tempdir().unwrap();
         fs::write(working_directory.path().join("kernel.ld"), include_bytes!("kernel.ld")).unwrap();
 
-        let mut string_store: HashMap<String, i32> = HashMap::default();
-
-        // Keep this list of exceptions in sync with `EXCEPTION_ID_LOOKUP` in `artiq::firmware::ksupport::eh_artiq`
-        // The exceptions declared here must be defined in `artiq.coredevice.exceptions`
-        // Verify synchronization by running the test cases in `artiq.test.coredevice.test_exceptions`
-        let runtime_exception_names = [
-            "RTIOUnderflow",
-            "RTIOOverflow",
-            "RTIODestinationUnreachable",
-            "DMAError",
-            "I2CError",
-            "CacheError",
-            "SPIError",
-            "SubkernelError",
-            "0:AssertionError",
-            "0:AttributeError",
-            "0:IndexError",
-            "0:IOError",
-            "0:KeyError",
-            "0:NotImplementedError",
-            "0:OverflowError",
-            "0:RuntimeError",
-            "0:TimeoutError",
-            "0:TypeError",
-            "0:ValueError",
-            "0:ZeroDivisionError",
-            "0:LinAlgError",
-            "UnwrapNoneError",
-        ];
-
-        // Preallocate runtime exception names
-        for (i, name) in runtime_exception_names.iter().enumerate() {
-            let exn_name = if name.find(':').is_none() {
-                format!("0:artiq.coredevice.exceptions.{name}")
-            } else {
-                (*name).to_string()
-            };
-
-            let id = i32::try_from(i).unwrap();
-            string_store.insert(exn_name, id);
-        }
-
         Ok(Nac3 {
             isa,
             time_fns,
@@ -1136,7 +1019,7 @@ impl Nac3 {
             top_levels: Vec::default(),
             pyid_to_def: Arc::default(),
             working_directory,
-            string_store: Arc::new(string_store.into()),
+            string_store: Arc::default(),
             exception_ids: Arc::default(),
             deferred_eval_store: DeferredEvaluationStore::new(),
             llvm_options: CodeGenLLVMOptions {
@@ -1146,12 +1029,7 @@ impl Nac3 {
         })
     }
 
-    fn analyze(
-        &mut self,
-        functions: &PySet,
-        classes: &PySet,
-        content_modules: &PySet,
-    ) -> PyResult<()> {
+    fn analyze(&mut self, functions: &PySet, classes: &PySet) -> PyResult<()> {
         let (modules, class_ids) =
             Python::with_gil(|py| -> PyResult<(HashMap<u64, PyObject>, HashSet<u64>)> {
                 let mut modules: HashMap<u64, PyObject> = HashMap::new();
@@ -1161,21 +1039,13 @@ impl Nac3 {
                 let getmodule_fn = PyModule::import(py, "inspect")?.getattr("getmodule")?;
 
                 for function in functions {
-                    let module: PyObject = getmodule_fn.call1((function,))?.extract()?;
-                    if !module.is_none(py) {
-                        modules.insert(id_fn.call1((&module,))?.extract()?, module);
-                    }
+                    let module = getmodule_fn.call1((function,))?.extract()?;
+                    modules.insert(id_fn.call1((&module,))?.extract()?, module);
                 }
                 for class in classes {
-                    let module: PyObject = getmodule_fn.call1((class,))?.extract()?;
-                    if !module.is_none(py) {
-                        modules.insert(id_fn.call1((&module,))?.extract()?, module);
-                    }
-                    class_ids.insert(id_fn.call1((class,))?.extract()?);
-                }
-                for module in content_modules {
-                    let module: PyObject = module.extract()?;
+                    let module = getmodule_fn.call1((class,))?.extract()?;
                     modules.insert(id_fn.call1((&module,))?.extract()?, module);
+                    class_ids.insert(id_fn.call1((class,))?.extract()?);
                 }
                 Ok((modules, class_ids))
             })?;

nac3artiq/src/symbol_resolver.rs

@@ -1,32 +1,17 @@
-use std::{
-    collections::{HashMap, HashSet},
-    sync::{
-        atomic::{AtomicBool, Ordering::Relaxed},
-        Arc,
-    },
+use crate::PrimitivePythonId;
+use inkwell::{
+    module::Linkage,
+    types::BasicType,
+    values::{BasicValue, BasicValueEnum},
+    AddressSpace,
 };
-
 use itertools::Itertools;
-use parking_lot::RwLock;
-use pyo3::{
-    types::{PyDict, PyTuple},
-    PyAny, PyErr, PyObject, PyResult, Python,
-};
-
-use super::PrimitivePythonId;
 use nac3core::{
     codegen::{
-        types::{ndarray::NDArrayType, ProxyType},
-        values::ndarray::make_contiguous_strides,
+        model::*,
+        object::ndarray::{make_contiguous_strides, NDArray},
         CodeGenContext, CodeGenerator,
     },
-    inkwell::{
-        module::Linkage,
-        types::{BasicType, BasicTypeEnum},
-        values::BasicValueEnum,
-        AddressSpace,
-    },
-    nac3parser::ast::{self, StrRef},
     symbol_resolver::{StaticValue, SymbolResolver, SymbolValue, ValueEnum},
     toplevel::{
         helper::PrimDef,
@@ -38,6 +23,19 @@ use nac3core::{
         typedef::{into_var_map, iter_type_vars, Type, TypeEnum, TypeVar, Unifier, VarMap},
     },
 };
+use nac3parser::ast::{self, StrRef};
+use parking_lot::{Mutex, RwLock};
+use pyo3::{
+    types::{PyDict, PyTuple},
+    PyAny, PyErr, PyObject, PyResult, Python,
+};
+use std::{
+    collections::{HashMap, HashSet},
+    sync::{
+        atomic::{AtomicBool, Ordering::Relaxed},
+        Arc,
+    },
+};
 
 pub enum PrimitiveValue {
     I32(i32),
@@ -82,6 +80,7 @@ pub struct InnerResolver {
     pub id_to_primitive: RwLock<HashMap<u64, PrimitiveValue>>,
     pub field_to_val: RwLock<HashMap<ResolverField, Option<PyFieldHandle>>>,
     pub global_value_ids: Arc<RwLock<HashMap<u64, PyObject>>>,
+    pub class_names: Mutex<HashMap<StrRef, Type>>,
     pub pyid_to_def: Arc<RwLock<HashMap<u64, DefinitionId>>>,
     pub pyid_to_type: Arc<RwLock<HashMap<u64, Type>>>,
     pub primitive_ids: PrimitivePythonId,
@@ -1085,19 +1084,15 @@ impl InnerResolver {
             } else {
                 unreachable!("must be ndarray")
             };
-            let (ndarray_dtype, _) = unpack_ndarray_var_tys(&mut ctx.unifier, ndarray_ty);
-
-            let llvm_i8 = ctx.ctx.i8_type();
-            let llvm_pi8 = llvm_i8.ptr_type(AddressSpace::default());
-            let llvm_usize = generator.get_size_type(ctx.ctx);
-            let llvm_ndarray = NDArrayType::from_unifier_type(generator, ctx, ndarray_ty);
-            let dtype = llvm_ndarray.element_type();
+            let (ndarray_dtype, ndarray_ndims) =
+                unpack_ndarray_var_tys(&mut ctx.unifier, ndarray_ty);
 
+            let dtype = Any(ctx.get_llvm_type(generator, ndarray_dtype));
             {
                 if self.global_value_ids.read().contains_key(&id) {
                     let global = ctx.module.get_global(&id_str).unwrap_or_else(|| {
                         ctx.module.add_global(
-                            llvm_ndarray.as_base_type().get_element_type().into_struct_type(),
+                            Struct(NDArray).get_type(generator, ctx.ctx),
                             Some(AddressSpace::default()),
                             &id_str,
                         )
@@ -1107,14 +1102,26 @@ impl InnerResolver {
                 self.global_value_ids.write().insert(id, obj.into());
             }
 
-            let ndims = llvm_ndarray.ndims().unwrap();
+            let TypeEnum::TLiteral { values, .. } = &*ctx.unifier.get_ty_immutable(ndarray_ndims)
+            else {
+                unreachable!("Expected Literal for ndarray_ndims")
+            };
+
+            let ndarray_ndims = if values.len() == 1 {
+                values[0].clone()
+            } else {
+                todo!("Unpacking literal of more than one element unimplemented")
+            };
+            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(), ndims as usize);
 
             // The Rust type inferencer cannot figure this out
-            let shape_values = shape_tuple
+            let shape_values: Result<Vec<Instance<'ctx, Int<SizeT>>>, PyErr> = shape_tuple
                 .iter()
                 .enumerate()
                 .map(|(i, elem)| {
@@ -1124,32 +1131,33 @@ impl InnerResolver {
                             super::CompileError::new_err(format!("Error getting element {i}: {e}"))
                         })?
                         .unwrap();
-                    let value = value.into_int_value();
+                    let value = Int(SizeT).check_value(generator, ctx.ctx, value).unwrap();
                     Ok(value)
                 })
-                .collect::<Result<Vec<_>, PyErr>>()?;
+                .collect();
+            let shape_values = shape_values?;
 
             // Also use this opportunity to get the constant values of `shape_values` for calculating strides.
             let shape_u64s = shape_values
                 .iter()
                 .map(|dim| {
-                    assert!(dim.is_const());
-                    dim.get_zero_extended_constant().unwrap()
+                    assert!(dim.value.is_const());
+                    dim.value.get_zero_extended_constant().unwrap()
                 })
                 .collect_vec();
-            let shape_values = llvm_usize.const_array(&shape_values);
+            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(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: Vec<_> = (0..sz)
+            let data_values: Vec<Instance<'ctx, Any>> = (0..sz)
                 .map(|i| {
                     obj.getattr("flat")?.get_item(i).and_then(|elem| {
                         let value = self
@@ -1161,97 +1169,79 @@ impl InnerResolver {
                             })?
                             .unwrap();
 
-                        assert_eq!(value.get_type(), dtype);
+                        let value = dtype.check_value(generator, ctx.ctx, value).unwrap();
                         Ok(value)
                     })
                 })
                 .try_collect()?;
-            let data = data.into_iter();
-            let data = match dtype {
-                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!(),
-            };
+            let data = dtype.const_array(generator, ctx.ctx, &data_values);
 
             // create a global for ndarray.data and initialize it using the elements
             //
             // NOTE: NDArray's `data` is `u8*`. Here, `data_global` is an array of `dtype`.
             // We will have to cast it to an `u8*` later.
             let data_global = ctx.module.add_global(
-                dtype.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.size_of().unwrap();
+            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| llvm_usize.const_int(stride, false)).collect_vec();
-            let strides = llvm_usize.const_array(&strides);
+            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(
-                llvm_i8.array_type(ndims as u32),
+                Array { len: AnyLen(ndims as u32), item: Int(Byte) }.get_type(generator, ctx.ctx),
                 Some(AddressSpace::default()),
-                &format!("${id_str}.strides"),
+                &(id_str.clone() + ".strides"),
             );
-            strides_global.set_initializer(&strides);
+            strides_global.set_initializer(&strides.value);
 
             // create a global for the ndarray object and initialize it
+            // We are also doing [`Model::check_value`] instead of [`Model::believe_value`] to catch bugs.
 
             // NOTE: data_global is an array of dtype, we want a `u8*`.
-            let ndarray_data = data_global.as_pointer_value();
-            let ndarray_data = ctx.builder.build_pointer_cast(ndarray_data, llvm_pi8, "").unwrap();
+            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 = llvm_usize.const_int(itemsize, false);
+            let ndarray_itemsize = Int(SizeT).const_int(generator, ctx.ctx, itemsize);
 
-            let ndarray_ndims = llvm_usize.const_int(ndims, false);
+            let ndarray_ndims = Int(SizeT).const_int(generator, ctx.ctx, ndims);
 
-            let ndarray_shape = shape_global.as_pointer_value();
+            let ndarray_shape =
+                Ptr(Int(SizeT)).check_value(generator, ctx.ctx, shape_global).unwrap();
 
-            let ndarray_strides = strides_global.as_pointer_value();
+            let ndarray_strides =
+                Ptr(Int(SizeT)).check_value(generator, ctx.ctx, strides_global).unwrap();
 
-            let ndarray = llvm_ndarray
-                .as_base_type()
-                .get_element_type()
-                .into_struct_type()
-                .const_named_struct(&[
-                    ndarray_itemsize.into(),
-                    ndarray_ndims.into(),
-                    ndarray_shape.into(),
-                    ndarray_strides.into(),
-                    ndarray_data.into(),
-                ]);
+            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(
-                llvm_ndarray.as_base_type().get_element_type().into_struct_type(),
+                Struct(NDArray).get_type(generator, ctx.ctx),
                 Some(AddressSpace::default()),
                 &id_str,
             );
-            ndarray_global.set_initializer(&ndarray);
+            ndarray_global.set_initializer(&ndarray.value);
 
             Ok(Some(ndarray_global.as_pointer_value().into()))
         } else if ty_id == self.primitive_ids.tuple {
@@ -1514,7 +1504,6 @@ impl SymbolResolver for Resolver {
         &self,
         id: StrRef,
         _: &mut CodeGenContext<'ctx, '_>,
-        _: &mut dyn CodeGenerator,
     ) -> Option<ValueEnum<'ctx>> {
         let sym_value = {
             let id_to_val = self.0.id_to_pyval.read();
@@ -1576,7 +1565,10 @@ impl SymbolResolver for Resolver {
         if let Some(id) = string_store.get(s) {
             *id
         } else {
-            let id = i32::try_from(string_store.len()).unwrap();
+            let id = Python::with_gil(|py| -> PyResult<i32> {
+                self.0.helper.store_str.call1(py, (s,))?.extract(py)
+            })
+            .unwrap();
             string_store.insert(s.into(), id);
             id
         }

nac3artiq/src/timeline.rs

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

nac3ast/Cargo.toml

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

nac3ast/src/ast_gen.rs

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

nac3ast/src/lib.rs

@@ -1,4 +1,10 @@
-#![deny(future_incompatible, let_underscore, nonstandard_style, clippy::all)]
+#![deny(
+    future_incompatible,
+    let_underscore,
+    nonstandard_style,
+    rust_2024_compatibility,
+    clippy::all
+)]
 #![warn(clippy::pedantic)]
 #![allow(
     clippy::missing_errors_doc,
@@ -8,6 +14,9 @@
     clippy::wildcard_imports
 )]
 
+#[macro_use]
+extern crate lazy_static;
+
 mod ast_gen;
 mod constant;
 #[cfg(feature = "fold")]

nac3core/Cargo.toml

@@ -5,25 +5,22 @@ authors = ["M-Labs"]
 edition = "2021"
 
 [features]
-default = ["derive"]
-derive = ["dep:nac3core_derive"]
 no-escape-analysis = []
 
 [dependencies]
 itertools = "0.13"
 crossbeam = "0.8"
-indexmap = "2.6"
+indexmap = "2.2"
 parking_lot = "0.12"
-rayon = "1.10"
-nac3core_derive = { path = "nac3core_derive", optional = true }
+rayon = "1.8"
 nac3parser = { path = "../nac3parser" }
 strum = "0.26"
 strum_macros = "0.26"
 
 [dependencies.inkwell]
-version = "0.5"
+version = "0.4"
 default-features = false
-features = ["llvm14-0-prefer-dynamic", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
+features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
 
 [dev-dependencies]
 test-case = "1.2.0"

nac3core/build.rs

@@ -1,3 +1,4 @@
+use regex::Regex;
 use std::{
     env,
     fs::File,
@@ -6,8 +7,6 @@ use std::{
     process::{Command, Stdio},
 };
 
-use regex::Regex;
-
 fn main() {
     let out_dir = env::var("OUT_DIR").unwrap();
     let out_dir = Path::new(&out_dir);
@@ -23,7 +22,6 @@ fn main() {
         "--target=wasm32",
         "-x",
         "c++",
-        "-std=c++20",
         "-fno-discard-value-names",
         "-fno-exceptions",
         "-fno-rtti",
@@ -56,8 +54,9 @@ fn main() {
     let output = Command::new("clang-irrt")
         .args(flags)
         .output()
-        .inspect(|o| {
+        .map(|o| {
             assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
+            o
         })
         .unwrap();
 

nac3core/irrt/irrt.cpp

@@ -1,10 +1,16 @@
-#include "irrt/exception.hpp"
-#include "irrt/list.hpp"
-#include "irrt/math.hpp"
-#include "irrt/ndarray.hpp"
-#include "irrt/range.hpp"
-#include "irrt/slice.hpp"
-#include "irrt/ndarray/basic.hpp"
-#include "irrt/ndarray/def.hpp"
-#include "irrt/ndarray/iter.hpp"
-#include "irrt/ndarray/indexing.hpp"
+#include <irrt/exception.hpp>
+#include <irrt/int_types.hpp>
+#include <irrt/list.hpp>
+#include <irrt/math_util.hpp>
+#include <irrt/ndarray/array.hpp>
+#include <irrt/ndarray/basic.hpp>
+#include <irrt/ndarray/broadcast.hpp>
+#include <irrt/ndarray/def.hpp>
+#include <irrt/ndarray/indexing.hpp>
+#include <irrt/ndarray/iter.hpp>
+#include <irrt/ndarray/matmul.hpp>
+#include <irrt/ndarray/reshape.hpp>
+#include <irrt/ndarray/transpose.hpp>
+#include <irrt/original.hpp>
+#include <irrt/range.hpp>
+#include <irrt/slice.hpp>

nac3core/irrt/irrt/cslice.hpp

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

@@ -7,19 +7,17 @@
 #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 raise_debug_assert(SizeT, msg, param1, param2, param3)                                                         \
+    raise_exception(SizeT, EXN_ASSERTION_ERROR, "IRRT debug assert failed: " msg, param1, param2, param3);
 
-#define debug_assert_eq(SizeT, lhs, rhs)                                           \
-    if constexpr (IRRT_DEBUG_ASSERT_BOOL) {                                        \
-        if ((lhs) != (rhs)) {                                                      \
-            raise_debug_assert(SizeT, "LHS = {0}. RHS = {1}", lhs, rhs, NO_PARAM); \
-        }                                                                          \
+#define debug_assert_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 constexpr (IRRT_DEBUG_ASSERT_BOOL) {                                        \
-        if (!(expr)) {                                                             \
-            raise_debug_assert(SizeT, "Got false.", NO_PARAM, NO_PARAM, 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

@@ -1,22 +1,24 @@
 #pragma once
 
-#include "irrt/cslice.hpp"
-#include "irrt/int_types.hpp"
+#include <irrt/cslice.hpp>
+#include <irrt/cstr_util.hpp>
+#include <irrt/int_types.hpp>
 
 /**
  * @brief The int type of ARTIQ exception IDs.
  */
-using ExceptionId = int32_t;
+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;
+extern "C"
+{
+    ExceptionId EXN_INDEX_ERROR;
+    ExceptionId EXN_VALUE_ERROR;
+    ExceptionId EXN_ASSERTION_ERROR;
+    ExceptionId EXN_TYPE_ERROR;
 }
 
 /**
@@ -25,14 +27,15 @@ ExceptionId EXN_TYPE_ERROR;
  * 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);
+extern "C" void __nac3_raise(void *err);
 
-namespace {
+namespace
+{
 /**
  * @brief NAC3's Exception struct
  */
-template<typename SizeT>
-struct Exception {
+template <typename SizeT> struct Exception
+{
     ExceptionId id;
     CSlice<SizeT> filename;
     int32_t line;
@@ -42,35 +45,26 @@ struct Exception {
     int64_t params[3];
 };
 
-constexpr int64_t NO_PARAM = 0;
+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) {
+template <typename SizeT>
+void _raise_exception_helper(ExceptionId id, const char *filename, int32_t line, const char *function, const char *msg,
+                             int64_t param0, int64_t param1, int64_t param2)
+{
     Exception<SizeT> e = {
         .id = id,
-        .filename = {.base = reinterpret_cast<void*>(const_cast<char*>(filename)),
-                     .len = static_cast<SizeT>(__builtin_strlen(filename))},
+        .filename = {.base = (uint8_t *)filename, .len = (int32_t)cstr::length(filename)},
         .line = line,
         .column = 0,
-        .function = {.base = reinterpret_cast<void*>(const_cast<char*>(function)),
-                     .len = static_cast<SizeT>(__builtin_strlen(function))},
-        .msg = {.base = reinterpret_cast<void*>(const_cast<char*>(msg)),
-                .len = static_cast<SizeT>(__builtin_strlen(msg))},
+        .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(reinterpret_cast<void*>(&e));
+    __nac3_raise((void *)&e);
     __builtin_unreachable();
 }
-}  // namespace
 
 /**
  * @brief Raise an exception with location details (location in the IRRT source files).
@@ -81,5 +75,6 @@ void _raise_exception_helper(ExceptionId id,
  * `param0` to `param2` are optional format arguments of `msg`. They should be set to
  * `NO_PARAM` to indicate they are unused.
  */
-#define raise_exception(SizeT, id, msg, param0, param1, param2) \
+#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

@@ -1,27 +1,8 @@
 #pragma once
 
-#if __STDC_VERSION__ >= 202000
 using int8_t = _BitInt(8);
 using uint8_t = unsigned _BitInt(8);
 using int32_t = _BitInt(32);
 using uint32_t = unsigned _BitInt(32);
 using int64_t = _BitInt(64);
 using uint64_t = unsigned _BitInt(64);
-#else
-
-#pragma clang diagnostic push
-#pragma clang diagnostic ignored "-Wdeprecated-type"
-using int8_t = _ExtInt(8);
-using uint8_t = unsigned _ExtInt(8);
-using int32_t = _ExtInt(32);
-using uint32_t = unsigned _ExtInt(32);
-using int64_t = _ExtInt(64);
-using uint64_t = unsigned _ExtInt(64);
-#pragma clang diagnostic pop
-
-#endif
-
-// NDArray indices are always `uint32_t`.
-using NDIndexInt = uint32_t;
-// The type of an index or a value describing the length of a range/slice is always `int32_t`.
-using SliceIndex = int32_t;

nac3core/irrt/irrt/list.hpp

@@ -1,81 +1,19 @@
 #pragma once
 
-#include "irrt/int_types.hpp"
-#include "irrt/math_util.hpp"
+#include <irrt/int_types.hpp>
+#include <irrt/slice.hpp>
 
-extern "C" {
-// Handle list assignment and dropping part of the list when
-// both dest_step and src_step are +1.
-// - All the index must *not* be out-of-bound or negative,
-// - The end index is *inclusive*,
-// - The length of src and dest slice size should already
-// be checked: if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest)
-SliceIndex __nac3_list_slice_assign_var_size(SliceIndex dest_start,
-                                             SliceIndex dest_end,
-                                             SliceIndex dest_step,
-                                             void* dest_arr,
-                                             SliceIndex dest_arr_len,
-                                             SliceIndex src_start,
-                                             SliceIndex src_end,
-                                             SliceIndex src_step,
-                                             void* 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(static_cast<uint8_t*>(dest_arr) + dest_start * size,
-                              static_cast<uint8_t*>(src_arr) + src_start * size, src_len * size);
-        }
-        if (dest_len > 0) {
-            /* dropping */
-            __builtin_memmove(static_cast<uint8_t*>(dest_arr) + (dest_start + src_len) * size,
-                              static_cast<uint8_t*>(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) {
-        void* tmp = __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(static_cast<uint8_t*>(dest_arr) + dest_ind, static_cast<uint8_t*>(src_arr) + src_ind, 1);
-        } else if (size == 4) {
-            __builtin_memcpy(static_cast<uint8_t*>(dest_arr) + dest_ind * 4,
-                             static_cast<uint8_t*>(src_arr) + src_ind * 4, 4);
-        } else if (size == 8) {
-            __builtin_memcpy(static_cast<uint8_t*>(dest_arr) + dest_ind * 8,
-                             static_cast<uint8_t*>(src_arr) + src_ind * 8, 8);
-        } else {
-            /* memcpy for var size, cannot overlap after previous alloca */
-            __builtin_memcpy(static_cast<uint8_t*>(dest_arr) + dest_ind * size,
-                             static_cast<uint8_t*>(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(static_cast<uint8_t*>(dest_arr) + dest_ind * size,
-                          static_cast<uint8_t*>(dest_arr) + (dest_end + 1) * size,
-                          (dest_arr_len - dest_end - 1) * size);
-        return dest_arr_len - (dest_end - dest_ind) - 1;
-    }
-    return dest_arr_len;
-}
-}  // extern "C"
+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

nac3core/irrt/irrt/math.hpp

@@ -1,93 +0,0 @@
-#pragma once
-
-namespace {
-// adapted from GNU Scientific Library: https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
-// need to make sure `exp >= 0` before calling this function
-template<typename T>
-T __nac3_int_exp_impl(T base, T exp) {
-    T res = 1;
-    /* repeated squaring method */
-    do {
-        if (exp & 1) {
-            res *= base; /* for n odd */
-        }
-        exp >>= 1;
-        base *= base;
-    } while (exp);
-    return res;
-}
-}  // namespace
-
-#define DEF_nac3_int_exp_(T)                   \
-    T __nac3_int_exp_##T(T base, T exp) {      \
-        return __nac3_int_exp_impl(base, exp); \
-    }
-
-extern "C" {
-
-// Putting semicolons here to make clang-format not reformat this into
-// a stair shape.
-DEF_nac3_int_exp_(int32_t);
-DEF_nac3_int_exp_(int64_t);
-DEF_nac3_int_exp_(uint32_t);
-DEF_nac3_int_exp_(uint64_t);
-
-int32_t __nac3_isinf(double x) {
-    return __builtin_isinf(x);
-}
-
-int32_t __nac3_isnan(double x) {
-    return __builtin_isnan(x);
-}
-
-double tgamma(double arg);
-
-double __nac3_gamma(double z) {
-    // Handling for denormals
-    //     | x                 | Python gamma(x) | C tgamma(x) |
-    // --- | ----------------- | --------------- | ----------- |
-    // (1) | nan               | nan             | nan         |
-    // (2) | -inf              | -inf            | inf         |
-    // (3) | inf               | inf             | inf         |
-    // (4) | 0.0               | inf             | inf         |
-    // (5) | {-1.0, -2.0, ...} | inf             | nan         |
-
-    // (1)-(3)
-    if (__builtin_isinf(z) || __builtin_isnan(z)) {
-        return z;
-    }
-
-    double v = tgamma(z);
-
-    // (4)-(5)
-    return __builtin_isinf(v) || __builtin_isnan(v) ? __builtin_inf() : v;
-}
-
-double lgamma(double arg);
-
-double __nac3_gammaln(double x) {
-    // libm's handling of value overflows differs from scipy:
-    // - scipy: gammaln(-inf) -> -inf
-    // - libm : lgamma(-inf) -> inf
-
-    if (__builtin_isinf(x)) {
-        return x;
-    }
-
-    return lgamma(x);
-}
-
-double j0(double x);
-
-double __nac3_j0(double x) {
-    // libm's handling of value overflows differs from scipy:
-    // - scipy: j0(inf) -> nan
-    // - libm : j0(inf) -> 0.0
-
-    if (__builtin_isinf(x)) {
-        return __builtin_nan("");
-    }
-
-    return j0(x);
-}
-}  // namespace

nac3core/irrt/irrt/math_util.hpp

@@ -1,13 +1,14 @@
 #pragma once
 
-namespace {
-template<typename T>
-const T& max(const T& a, const T& b) {
+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) {
+template <typename T> const T &min(const T &a, const T &b)
+{
     return a > b ? b : a;
 }
-}  // namespace
+} // namespace

nac3core/irrt/irrt/ndarray.hpp

@@ -1,151 +0,0 @@
-#pragma once
-
-#include "irrt/int_types.hpp"
-
-// TODO: To be deleted since NDArray with strides is done.
-
-namespace {
-template<typename SizeT>
-SizeT __nac3_ndarray_calc_size_impl(const SizeT* list_data, SizeT list_len, SizeT begin_idx, SizeT end_idx) {
-    __builtin_assume(end_idx <= list_len);
-
-    SizeT num_elems = 1;
-    for (SizeT i = begin_idx; i < end_idx; ++i) {
-        SizeT val = list_data[i];
-        __builtin_assume(val > 0);
-        num_elems *= val;
-    }
-    return num_elems;
-}
-
-template<typename SizeT>
-void __nac3_ndarray_calc_nd_indices_impl(SizeT index, const SizeT* dims, SizeT num_dims, NDIndexInt* idxs) {
-    SizeT stride = 1;
-    for (SizeT dim = 0; dim < num_dims; dim++) {
-        SizeT i = num_dims - dim - 1;
-        __builtin_assume(dims[i] > 0);
-        idxs[i] = (index / stride) % dims[i];
-        stride *= dims[i];
-    }
-}
-
-template<typename SizeT>
-SizeT __nac3_ndarray_flatten_index_impl(const SizeT* dims,
-                                        SizeT num_dims,
-                                        const NDIndexInt* indices,
-                                        SizeT num_indices) {
-    SizeT idx = 0;
-    SizeT stride = 1;
-    for (SizeT i = 0; i < num_dims; ++i) {
-        SizeT ri = num_dims - i - 1;
-        if (ri < num_indices) {
-            idx += stride * indices[ri];
-        }
-
-        __builtin_assume(dims[i] > 0);
-        stride *= dims[ri];
-    }
-    return idx;
-}
-
-template<typename SizeT>
-void __nac3_ndarray_calc_broadcast_impl(const SizeT* lhs_dims,
-                                        SizeT lhs_ndims,
-                                        const SizeT* rhs_dims,
-                                        SizeT rhs_ndims,
-                                        SizeT* out_dims) {
-    SizeT max_ndims = lhs_ndims > rhs_ndims ? lhs_ndims : rhs_ndims;
-
-    for (SizeT i = 0; i < max_ndims; ++i) {
-        const SizeT* lhs_dim_sz = i < lhs_ndims ? &lhs_dims[lhs_ndims - i - 1] : nullptr;
-        const SizeT* rhs_dim_sz = i < rhs_ndims ? &rhs_dims[rhs_ndims - i - 1] : nullptr;
-        SizeT* out_dim = &out_dims[max_ndims - i - 1];
-
-        if (lhs_dim_sz == nullptr) {
-            *out_dim = *rhs_dim_sz;
-        } else if (rhs_dim_sz == nullptr) {
-            *out_dim = *lhs_dim_sz;
-        } else if (*lhs_dim_sz == 1) {
-            *out_dim = *rhs_dim_sz;
-        } else if (*rhs_dim_sz == 1) {
-            *out_dim = *lhs_dim_sz;
-        } else if (*lhs_dim_sz == *rhs_dim_sz) {
-            *out_dim = *lhs_dim_sz;
-        } else {
-            __builtin_unreachable();
-        }
-    }
-}
-
-template<typename SizeT>
-void __nac3_ndarray_calc_broadcast_idx_impl(const SizeT* src_dims,
-                                            SizeT src_ndims,
-                                            const NDIndexInt* in_idx,
-                                            NDIndexInt* out_idx) {
-    for (SizeT i = 0; i < src_ndims; ++i) {
-        SizeT src_i = src_ndims - i - 1;
-        out_idx[src_i] = src_dims[src_i] == 1 ? 0 : in_idx[src_i];
-    }
-}
-}  // namespace
-
-extern "C" {
-uint32_t __nac3_ndarray_calc_size(const uint32_t* list_data, uint32_t list_len, uint32_t begin_idx, uint32_t end_idx) {
-    return __nac3_ndarray_calc_size_impl(list_data, list_len, begin_idx, end_idx);
-}
-
-uint64_t
-__nac3_ndarray_calc_size64(const uint64_t* list_data, uint64_t list_len, uint64_t begin_idx, uint64_t end_idx) {
-    return __nac3_ndarray_calc_size_impl(list_data, list_len, begin_idx, end_idx);
-}
-
-void __nac3_ndarray_calc_nd_indices(uint32_t index, const uint32_t* dims, uint32_t num_dims, NDIndexInt* idxs) {
-    __nac3_ndarray_calc_nd_indices_impl(index, dims, num_dims, idxs);
-}
-
-void __nac3_ndarray_calc_nd_indices64(uint64_t index, const uint64_t* dims, uint64_t num_dims, NDIndexInt* idxs) {
-    __nac3_ndarray_calc_nd_indices_impl(index, dims, num_dims, idxs);
-}
-
-uint32_t
-__nac3_ndarray_flatten_index(const uint32_t* dims, uint32_t num_dims, const NDIndexInt* indices, uint32_t num_indices) {
-    return __nac3_ndarray_flatten_index_impl(dims, num_dims, indices, num_indices);
-}
-
-uint64_t __nac3_ndarray_flatten_index64(const uint64_t* dims,
-                                        uint64_t num_dims,
-                                        const NDIndexInt* indices,
-                                        uint64_t num_indices) {
-    return __nac3_ndarray_flatten_index_impl(dims, num_dims, indices, num_indices);
-}
-
-void __nac3_ndarray_calc_broadcast(const uint32_t* lhs_dims,
-                                   uint32_t lhs_ndims,
-                                   const uint32_t* rhs_dims,
-                                   uint32_t rhs_ndims,
-                                   uint32_t* out_dims) {
-    return __nac3_ndarray_calc_broadcast_impl(lhs_dims, lhs_ndims, rhs_dims, rhs_ndims, out_dims);
-}
-
-void __nac3_ndarray_calc_broadcast64(const uint64_t* lhs_dims,
-                                     uint64_t lhs_ndims,
-                                     const uint64_t* rhs_dims,
-                                     uint64_t rhs_ndims,
-                                     uint64_t* out_dims) {
-    return __nac3_ndarray_calc_broadcast_impl(lhs_dims, lhs_ndims, rhs_dims, rhs_ndims, out_dims);
-}
-
-void __nac3_ndarray_calc_broadcast_idx(const uint32_t* src_dims,
-                                       uint32_t src_ndims,
-                                       const NDIndexInt* in_idx,
-                                       NDIndexInt* out_idx) {
-    __nac3_ndarray_calc_broadcast_idx_impl(src_dims, src_ndims, in_idx, out_idx);
-}
-
-void __nac3_ndarray_calc_broadcast_idx64(const uint64_t* src_dims,
-                                         uint64_t src_ndims,
-                                         const NDIndexInt* in_idx,
-                                         NDIndexInt* out_idx) {
-    __nac3_ndarray_calc_broadcast_idx_impl(src_dims, src_ndims, in_idx, out_idx);
-}
-}  // namespace

nac3core/irrt/irrt/ndarray/array.hpp

@@ -0,0 +1,157 @@
+#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
+{
+/**
+ * @brief In the context of `np.array(<list>)`, deduce the ndarray's shape produced by `<list>` and raise
+ * an exception if there is anything wrong with `<shape>` (e.g., inconsistent dimensions `np.array([[1.0, 2.0], [3.0]])`)
+ *
+ * If this function finds no issues with `<list>`, the deduced shape is written to `shape`. The caller has the responsibility to
+ * allocate `[SizeT; ndims]` for `shape`. The caller must also initialize `shape` with `-1`s because of implementation details.
+ */
+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);
+        }
+    }
+}
+
+/**
+ * @brief See `set_and_validate_list_shape_helper`.
+ */
+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);
+}
+
+/**
+ * @brief In the context of `np.array(<list>)`, copied the contents stored in `list` to `ndarray`.
+ *
+ * `list` is assumed to be "legal". (i.e., no inconsistent dimensions)
+ *
+ * # Notes on `ndarray`
+ * The caller is responsible for allocating space for `ndarray`.
+ * Here is what this function expects from `ndarray` when called:
+ *   - `ndarray->data` has to be allocated, contiguous, and may contain uninitialized values.
+ *   - `ndarray->itemsize` has to be initialized.
+ *   - `ndarray->ndims` has to be initialized.
+ *   - `ndarray->shape` has to be initialized.
+ *   - `ndarray->strides` is ignored, but note that `ndarray->data` is contiguous.
+ * When this function call ends:
+ *   - `ndarray->data` is written with contents from `<list>`.
+ */
+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[scalar]`
+        // `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);
+        }
+    }
+}
+
+/**
+ * @brief See `write_list_to_array_helper`.
+ */
+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

@@ -1,23 +1,28 @@
 #pragma once
 
-#include "irrt/debug.hpp"
-#include "irrt/exception.hpp"
-#include "irrt/int_types.hpp"
-#include "irrt/ndarray/def.hpp"
+#include <irrt/debug.hpp>
+#include <irrt/exception.hpp>
+#include <irrt/int_types.hpp>
+#include <irrt/ndarray/def.hpp>
 
-namespace {
-namespace ndarray {
-namespace basic {
+namespace
+{
+namespace ndarray
+{
+namespace basic
+{
 /**
  * @brief Assert 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) {
+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}",
@@ -29,19 +34,21 @@ void assert_shape_no_negative(SizeT ndims, const SizeT* shape) {
 /**
  * @brief Assert that two shapes are the same in the context of writing output to an ndarray.
  */
-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) {
+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]) {
+    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 "
@@ -57,8 +64,8 @@ void assert_output_shape_same(SizeT ndarray_ndims,
  * @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) {
+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];
@@ -73,9 +80,10 @@ SizeT calc_size_from_shape(SizeT ndims, const SizeT* shape) {
  * @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++) {
+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];
 
@@ -89,8 +97,8 @@ void set_indices_by_nth(SizeT ndims, const SizeT* shape, SizeT* indices, SizeT n
  *
  * This function corresponds to `<an_ndarray>.size`
  */
-template<typename SizeT>
-SizeT size(const NDArray<SizeT>* ndarray) {
+template <typename SizeT> SizeT size(const NDArray<SizeT> *ndarray)
+{
     return calc_size_from_shape(ndarray->ndims, ndarray->shape);
 }
 
@@ -99,8 +107,8 @@ SizeT size(const NDArray<SizeT>* ndarray) {
  *
  * This function corresponds to `<an_ndarray>.nbytes`.
  */
-template<typename SizeT>
-SizeT nbytes(const NDArray<SizeT>* ndarray) {
+template <typename SizeT> SizeT nbytes(const NDArray<SizeT> *ndarray)
+{
     return size(ndarray) * ndarray->itemsize;
 }
 
@@ -111,35 +119,32 @@ SizeT nbytes(const NDArray<SizeT>* ndarray) {
  *
  * @param dst_length The length.
  */
-template<typename SizeT>
-SizeT len(const NDArray<SizeT>* ndarray) {
-    if (ndarray->ndims != 0) {
+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];
     }
-
-    // numpy prohibits `__len__` on unsized objects
-    raise_exception(SizeT, EXN_TYPE_ERROR, "len() of unsized object", NO_PARAM, NO_PARAM, NO_PARAM);
-    __builtin_unreachable();
 }
 
 /**
  * @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
+ * 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) {
+template <typename SizeT> bool is_c_contiguous(const NDArray<SizeT> *ndarray)
+{
     // 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
+    // - 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:
+    // 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:
@@ -155,17 +160,21 @@ bool is_c_contiguous(const NDArray<SizeT>* ndarray) {
     // 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) {
+    if (ndarray->ndims == 0)
+    {
         return true;
     }
 
-    if (ndarray->strides[ndarray->ndims - 1] != ndarray->itemsize) {
+    if (ndarray->strides[ndarray->ndims - 1] != ndarray->itemsize)
+    {
         return false;
     }
 
-    for (SizeT i = 1; i < ndarray->ndims; i++) {
+    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]) {
+        if (ndarray->strides[axis_i] != ndarray->shape[axis_i + 1] * ndarray->strides[axis_i + 1])
+        {
             return false;
         }
     }
@@ -178,11 +187,11 @@ bool is_c_contiguous(const NDArray<SizeT>* ndarray) {
  *
  * This function does no bound check.
  */
-template<typename SizeT>
-void* get_pelement_by_indices(const NDArray<SizeT>* ndarray, const SizeT* indices) {
-    void* element = ndarray->data;
+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 = static_cast<uint8_t*>(element) + indices[dim_i] * ndarray->strides[dim_i];
+        element += indices[dim_i] * ndarray->strides[dim_i];
     return element;
 }
 
@@ -191,13 +200,14 @@ void* get_pelement_by_indices(const NDArray<SizeT>* ndarray, const SizeT* indice
  *
  * This function does no bound check.
  */
-template<typename SizeT>
-void* get_nth_pelement(const NDArray<SizeT>* ndarray, SizeT nth) {
-    void* element = ndarray->data;
-    for (SizeT i = 0; i < ndarray->ndims; i++) {
+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 = static_cast<uint8_t*>(element) + ndarray->strides[axis] * (nth % dim);
+        element += ndarray->strides[axis] * (nth % dim);
         nth /= dim;
     }
     return element;
@@ -208,10 +218,11 @@ void* get_nth_pelement(const NDArray<SizeT>* ndarray, SizeT nth) {
  *
  * You might want to read https://ajcr.net/stride-guide-part-1/.
  */
-template<typename SizeT>
-void set_strides_by_shape(NDArray<SizeT>* ndarray) {
+template <typename SizeT> void set_strides_by_shape(NDArray<SizeT> *ndarray)
+{
     SizeT stride_product = 1;
-    for (SizeT i = 0; i < ndarray->ndims; i++) {
+    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];
@@ -224,8 +235,8 @@ void set_strides_by_shape(NDArray<SizeT>* 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, void* pelement, const void* pvalue) {
+template <typename SizeT> void set_pelement_value(NDArray<SizeT> *ndarray, uint8_t *pelement, const uint8_t *pvalue)
+{
     __builtin_memcpy(pelement, pvalue, ndarray->itemsize);
 }
 
@@ -234,109 +245,127 @@ void set_pelement_value(NDArray<SizeT>* ndarray, void* pelement, const void* pva
  *
  * 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) {
+template <typename SizeT> void copy_data(const NDArray<SizeT> *src_ndarray, NDArray<SizeT> *dst_ndarray)
+{
     // TODO: Make this faster with memcpy when we see a contiguous segment.
     // TODO: Handle overlapping.
 
     debug_assert_eq(SizeT, src_ndarray->itemsize, dst_ndarray->itemsize);
 
-    for (SizeT i = 0; i < size(src_ndarray); i++) {
+    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
+} // namespace basic
+} // namespace ndarray
+} // namespace
 
-extern "C" {
-using namespace ndarray::basic;
+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_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_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_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);
-}
+    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);
-}
+    uint32_t __nac3_ndarray_size(NDArray<int32_t> *ndarray)
+    {
+        return size(ndarray);
+    }
 
-uint64_t __nac3_ndarray_size64(NDArray<int64_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);
-}
+    uint32_t __nac3_ndarray_nbytes(NDArray<int32_t> *ndarray)
+    {
+        return nbytes(ndarray);
+    }
 
-uint64_t __nac3_ndarray_nbytes64(NDArray<int64_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);
-}
+    int32_t __nac3_ndarray_len(NDArray<int32_t> *ndarray)
+    {
+        return len(ndarray);
+    }
 
-int64_t __nac3_ndarray_len64(NDArray<int64_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_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);
-}
+    bool __nac3_ndarray_is_c_contiguous64(NDArray<int64_t> *ndarray)
+    {
+        return is_c_contiguous(ndarray);
+    }
 
-void* __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_pelement(const NDArray<int32_t> *ndarray, int32_t nth)
+    {
+        return get_nth_pelement(ndarray, nth);
+    }
 
-void* __nac3_ndarray_get_nth_pelement64(const NDArray<int64_t>* ndarray, int64_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);
+    }
 
-void* __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_indices(const NDArray<int32_t> *ndarray, int32_t *indices)
+    {
+        return get_pelement_by_indices(ndarray, indices);
+    }
 
-void* __nac3_ndarray_get_pelement_by_indices64(const NDArray<int64_t>* ndarray, int64_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_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_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_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);
-}
+    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

@@ -1,22 +1,21 @@
 #pragma once
 
-#include "irrt/int_types.hpp"
+#include <irrt/int_types.hpp>
 
-namespace {
+namespace
+{
 /**
  * @brief The NDArray object
  *
- * Official numpy implementation:
- * https://github.com/numpy/numpy/blob/735a477f0bc2b5b84d0e72d92f224bde78d4e069/doc/source/reference/c-api/types-and-structures.rst#pyarrayinterface
- *
- * Note that this implementation is based on `PyArrayInterface` rather of `PyArrayObject`. The
- * difference between `PyArrayInterface` and `PyArrayObject` (relevant to our implementation) is
- * that `PyArrayInterface` *has* `itemsize` and uses `void*` for its `data`, whereas `PyArrayObject`
- * does not require `itemsize` (probably using `strides[-1]` instead) and uses `char*` for its
- * `data`. There are also minor differences in the struct layout.
+ * Official numpy implementation: https://github.com/numpy/numpy/blob/735a477f0bc2b5b84d0e72d92f224bde78d4e069/doc/source/reference/c-api/types-and-structures.rst
  */
-template<typename SizeT>
-struct NDArray {
+template <typename SizeT> struct NDArray
+{
+    /**
+     * @brief The underlying data this `ndarray` is pointing to.
+     */
+    uint8_t *data;
+
     /**
      * @brief The number of bytes of a single element in `data`.
      */
@@ -32,7 +31,7 @@ struct NDArray {
      *
      * Note that it may contain 0.
      */
-    SizeT* shape;
+    SizeT *shape;
 
     /**
      * @brief Array strides, with length equal to `ndims`
@@ -41,11 +40,6 @@ struct NDArray {
      *
      * Note that `strides` can have negative values or contain 0.
      */
-    SizeT* strides;
-
-    /**
-     * @brief The underlying data this `ndarray` is pointing to.
-     */
-    void* data;
+    SizeT *strides;
 };
-}  // namespace
+} // namespace

nac3core/irrt/irrt/ndarray/indexing.hpp

@@ -1,13 +1,14 @@
 #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"
+#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 {
+namespace
+{
 typedef uint8_t NDIndexType;
 
 /**
@@ -47,7 +48,8 @@ const NDIndexType ND_INDEX_TYPE_ELLIPSIS = 3;
  *            ^^^^  ^  ^^^  ^^^^^^^^^^ each of these is represented by an NDIndex.
  * ```
  */
-struct NDIndex {
+struct NDIndex
+{
     /**
      * @brief Enum tag to specify the type of index.
      *
@@ -60,13 +62,16 @@ struct NDIndex {
      *
      * Please see the comment of each enum constant.
      */
-    uint8_t* data;
+    uint8_t *data;
 };
-}  // namespace
+} // namespace
 
-namespace {
-namespace ndarray {
-namespace indexing {
+namespace
+{
+namespace ndarray
+{
+namespace indexing
+{
 /**
  * @brief Perform ndarray "basic indexing" (https://numpy.org/doc/stable/user/basics.indexing.html#basic-indexing)
  *
@@ -94,8 +99,9 @@ namespace indexing {
  * @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) {
+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`.
@@ -105,28 +111,40 @@ void index(SizeT num_indices, const NDIndex* indices, const NDArray<SizeT>* src_
     // 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) {
+    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) {
+        }
+        else if (indices[i].type == ND_INDEX_TYPE_SLICE)
+        {
             num_indexed++;
-        } else if (indices[i].type == ND_INDEX_TYPE_NEWAXIS) {
+        }
+        else if (indices[i].type == ND_INDEX_TYPE_NEWAXIS)
+        {
             expected_dst_ndims++;
-        } else if (indices[i].type == ND_INDEX_TYPE_ELLIPSIS) {
+        }
+        else if (indices[i].type == ND_INDEX_TYPE_ELLIPSIS)
+        {
             num_ellipsis++;
-            if (num_ellipsis > 1) {
+            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 {
+        }
+        else
+        {
             __builtin_unreachable();
         }
     }
 
     debug_assert_eq(SizeT, expected_dst_ndims, dst_ndarray->ndims);
 
-    if (src_ndarray->ndims - num_indexed < 0) {
+    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",
@@ -136,60 +154,72 @@ void index(SizeT num_indices, const NDIndex* indices, const NDArray<SizeT>* src_
     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
+    // 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);
+    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) {
+            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 = static_cast<uint8_t*>(dst_ndarray->data) + k * src_ndarray->strides[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;
+        }
+        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 = static_cast<uint8_t*>(dst_ndarray->data) + (SizeT)range.start * src_ndarray->strides[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) {
+        }
+        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) {
+        }
+        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++) {
+            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 {
+        }
+        else
+        {
             __builtin_unreachable();
         }
     }
 
-    for (; dst_axis < dst_ndarray->ndims; dst_axis++, src_axis++) {
+    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];
     }
@@ -197,24 +227,23 @@ void index(SizeT num_indices, const NDIndex* indices, const NDArray<SizeT>* src_
     debug_assert_eq(SizeT, src_ndarray->ndims, src_axis);
     debug_assert_eq(SizeT, dst_ndarray->ndims, dst_axis);
 }
-}  // namespace indexing
-}  // namespace ndarray
-}  // namespace
+} // namespace indexing
+} // namespace ndarray
+} // namespace
 
-extern "C" {
-using namespace ndarray::indexing;
+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_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);
-}
+    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

@@ -1,55 +1,35 @@
 #pragma once
 
-#include "irrt/int_types.hpp"
-#include "irrt/ndarray/def.hpp"
+#include <irrt/int_types.hpp>
+#include <irrt/ndarray/def.hpp>
 
-namespace {
+namespace
+{
 /**
  * @brief Helper struct to enumerate through an ndarray *efficiently*.
  *
- * Example usage (in pseudo-code):
- * ```
- * // Suppose my_ndarray has been initialized, with shape [2, 3] and dtype `double`
- * NDIter nditer;
- * nditer.initialize(my_ndarray);
- * while (nditer.has_element()) {
- *    // This body is run 6 (= my_ndarray.size) times.
- *
- *    // [0, 0] -> [0, 1] -> [0, 2] -> [1, 0] -> [1, 1] -> [1, 2] -> end
- *    print(nditer.indices);
- *
- *    // 0 -> 1 -> 2 -> 3 -> 4 -> 5
- *    print(nditer.nth);
- *
- *    // <1st element> -> <2nd element> -> ... -> <6th element> -> end
- *    print(*((double *) nditer.element))
- *
- *    nditer.next(); // Go to next element.
- * }
- * ```
- *
  * Interesting cases:
- * - If `my_ndarray.ndims` == 0, there is one iteration.
- * - If `my_ndarray.shape` contains zeroes, there are no iterations.
+ * - If ndims == 0, there is one iteration.
+ * - If shape contains zeroes, there are no iterations.
  */
-template<typename SizeT>
-struct NDIter {
+template <typename SizeT> struct NDIter
+{
     // Information about the ndarray being iterated over.
     SizeT ndims;
-    SizeT* shape;
-    SizeT* strides;
+    SizeT *shape;
+    SizeT *strides;
 
     /**
      * @brief The current indices.
      *
      * Must be allocated by the caller.
      */
-    SizeT* indices;
+    SizeT *indices;
 
     /**
      * @brief The nth (0-based) index of the current indices.
      *
-     * Initially this is 0.
+     * Initially this is all 0s.
      */
     SizeT nth;
 
@@ -58,7 +38,7 @@ struct NDIter {
      *
      * Initially this points to first element of the ndarray.
      */
-    void* element;
+    uint8_t *element;
 
     /**
      * @brief Cache for the product of shape.
@@ -67,7 +47,11 @@ struct NDIter {
      */
     SizeT size;
 
-    void initialize(SizeT ndims, SizeT* shape, SizeT* strides, void* element, SizeT* indices) {
+    // TODO:: Not implemented: 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.
+
+    void initialize(SizeT ndims, SizeT *shape, SizeT *strides, uint8_t *element, SizeT *indices)
+    {
         this->ndims = ndims;
         this->shape = shape;
         this->strides = strides;
@@ -77,70 +61,79 @@ struct NDIter {
 
         // Compute size
         this->size = 1;
-        for (SizeT i = 0; i < ndims; i++) {
+        for (SizeT i = 0; i < ndims; i++)
+        {
             this->size *= shape[i];
         }
 
-        // `indices` starts on all 0s.
         for (SizeT axis = 0; axis < ndims; axis++)
             indices[axis] = 0;
         nth = 0;
     }
 
-    void initialize_by_ndarray(NDArray<SizeT>* ndarray, SizeT* indices) {
-        // NOTE: ndarray->data is pointing to the first element, and `NDIter`'s `element` should also point to the first
-        // element as well.
+    void initialize_by_ndarray(NDArray<SizeT> *ndarray, SizeT *indices)
+    {
         this->initialize(ndarray->ndims, ndarray->shape, ndarray->strides, ndarray->data, indices);
     }
 
-    // Is the current iteration valid?
-    // If true, then `element`, `indices` and `nth` contain details about the current element.
-    bool has_element() { return nth < size; }
+    bool has_next()
+    {
+        return nth < size;
+    }
 
-    // Go to the next element.
-    void next() {
-        for (SizeT i = 0; i < ndims; i++) {
+    void next()
+    {
+        for (SizeT i = 0; i < ndims; i++)
+        {
             SizeT axis = ndims - i - 1;
             indices[axis]++;
-            if (indices[axis] >= shape[axis]) {
+            if (indices[axis] >= shape[axis])
+            {
                 indices[axis] = 0;
 
-                // 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.
-                element = static_cast<void*>(reinterpret_cast<uint8_t*>(element) - strides[axis] * (shape[axis] - 1));
-            } else {
-                element = static_cast<void*>(reinterpret_cast<uint8_t*>(element) + strides[axis]);
+                // TODO: Can be optimized with backstrides.
+                element -= strides[axis] * (shape[axis] - 1);
+            }
+            else
+            {
+                element += strides[axis];
                 break;
             }
         }
         nth++;
     }
 };
-}  // namespace
+} // namespace
 
-extern "C" {
-void __nac3_nditer_initialize(NDIter<int32_t>* iter, NDArray<int32_t>* ndarray, int32_t* indices) {
-    iter->initialize_by_ndarray(ndarray, indices);
-}
+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);
-}
+    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_element(NDIter<int32_t>* iter) {
-    return iter->has_element();
-}
+    bool __nac3_nditer_has_next(NDIter<int32_t> *iter)
+    {
+        return iter->has_next();
+    }
 
-bool __nac3_nditer_has_element64(NDIter<int64_t>* iter) {
-    return iter->has_element();
-}
+    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_next(NDIter<int32_t> *iter)
+    {
+        iter->next();
+    }
 
-void __nac3_nditer_next64(NDIter<int64_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

@@ -1,12 +1,14 @@
 #pragma once
 
-#include "irrt/debug.hpp"
-#include "irrt/int_types.hpp"
+#include <irrt/debug.hpp>
+#include <irrt/int_types.hpp>
 
-namespace {
-namespace range {
-template<typename T>
-T len(T start, T stop, T step) {
+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)
@@ -16,13 +18,13 @@ T len(T start, T stop, T step) {
     else
         return 0;
 }
-}  // namespace range
+} // namespace range
 
 /**
  * @brief A Python range.
  */
-template<typename T>
-struct Range {
+template <typename T> struct Range
+{
     T start;
     T stop;
     T step;
@@ -30,18 +32,10 @@ struct Range {
     /**
      * @brief Calculate the `len()` of this range.
      */
-    template<typename SizeT>
-    T len() {
+    template <typename SizeT> T len()
+    {
         debug_assert(SizeT, step != 0);
         return range::len(start, stop, step);
     }
 };
-}  // namespace
-
-extern "C" {
-using namespace range;
-
-SliceIndex __nac3_range_slice_len(const SliceIndex start, const SliceIndex end, const SliceIndex step) {
-    return len(start, end, step);
-}
-}
+} // namespace

nac3core/irrt/irrt/slice.hpp

@@ -1,24 +1,29 @@
 #pragma once
 
-#include "irrt/debug.hpp"
-#include "irrt/exception.hpp"
-#include "irrt/int_types.hpp"
-#include "irrt/math_util.hpp"
-#include "irrt/range.hpp"
+#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 {
+namespace
+{
+namespace slice
+{
 /**
  * @brief Resolve a possibly negative index in a list of a known length.
  *
  * Returns -1 if the resolved index is out of the list's bounds.
  */
-template<typename T>
-T resolve_index_in_length(T length, T index) {
+template <typename T> T resolve_index_in_length(T length, T index)
+{
     T resolved = index < 0 ? length + index : index;
-    if (0 <= resolved && resolved < length) {
+    if (0 <= resolved && resolved < length)
+    {
         return resolved;
-    } else {
+    }
+    else
+    {
         return -1;
     }
 }
@@ -28,49 +33,51 @@ T resolve_index_in_length(T length, T index) {
  *
  * This is equivalent to `range(*slice(start, stop, step).indices(length))` in Python.
  */
-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) {
+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) {
+    if (step_is_negative)
+    {
         lower = -1;
         upper = length - 1;
-    } else {
+    }
+    else
+    {
         lower = 0;
         upper = length;
     }
 
-    if (start_defined) {
+    if (start_defined)
+    {
         *range_start = start < 0 ? max(lower, start + length) : min(upper, start);
-    } else {
+    }
+    else
+    {
         *range_start = step_is_negative ? upper : lower;
     }
 
-    if (stop_defined) {
+    if (stop_defined)
+    {
         *range_stop = stop < 0 ? max(lower, stop + length) : min(upper, stop);
-    } else {
+    }
+    else
+    {
         *range_stop = step_is_negative ? lower : upper;
     }
 }
-}  // namespace slice
+} // namespace slice
 
 /**
  * @brief A Python-like slice with **unresolved** indices.
  */
-template<typename T>
-struct Slice {
+template <typename T> struct Slice
+{
     bool start_defined;
     T start;
 
@@ -80,25 +87,32 @@ struct Slice {
     bool step_defined;
     T step;
 
-    Slice() { this->reset(); }
+    Slice()
+    {
+        this->reset();
+    }
 
-    void reset() {
+    void reset()
+    {
         this->start_defined = false;
         this->stop_defined = false;
         this->step_defined = false;
     }
 
-    void set_start(T start) {
+    void set_start(T start)
+    {
         this->start_defined = true;
         this->start = start;
     }
 
-    void set_stop(T stop) {
+    void set_stop(T stop)
+    {
         this->stop_defined = true;
         this->stop = stop;
     }
 
-    void set_step(T step) {
+    void set_step(T step)
+    {
         this->step_defined = true;
         this->step = step;
     }
@@ -108,8 +122,8 @@ struct Slice {
      *
      * In Python, this would be `range(*slice(start, stop, step).indices(length))`.
      */
-    template<typename SizeT>
-    Range<T> indices(T 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);
@@ -123,34 +137,22 @@ struct Slice {
     /**
      * @brief Like `.indices()` but with assertions.
      */
-    template<typename SizeT>
-    Range<T> indices_checked(T length) {
+    template <typename SizeT> Range<T> indices_checked(T length)
+    {
         // TODO: Switch to `SizeT length`
 
-        if (length < 0) {
+        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) {
+        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" {
-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;
-}
-}
+} // namespace

nac3core/nac3core_derive/Cargo.toml

@@ -1,21 +0,0 @@
-[package]
-name = "nac3core_derive"
-version = "0.1.0"
-edition = "2021"
-
-[lib]
-proc-macro = true
-
-[[test]]
-name = "structfields_tests"
-path = "tests/structfields_test.rs"
-
-[dev-dependencies]
-nac3core = { path = ".." }
-trybuild = { version = "1.0", features = ["diff"] }
-
-[dependencies]
-proc-macro2 = "1.0"
-proc-macro-error = "1.0"
-syn = "2.0"
-quote = "1.0"

nac3core/nac3core_derive/src/lib.rs

@@ -1,320 +0,0 @@
-use proc_macro::TokenStream;
-use proc_macro_error::{abort, proc_macro_error};
-use quote::quote;
-use syn::{
-    parse_macro_input, spanned::Spanned, Data, DataStruct, Expr, ExprField, ExprMethodCall,
-    ExprPath, GenericArgument, Ident, LitStr, Path, PathArguments, Type, TypePath,
-};
-
-/// Extracts all generic arguments of a [`Type`] into a [`Vec`].
-///
-/// Returns [`Some`] of a possibly-empty [`Vec`] if the path of `ty` matches with
-/// `expected_ty_name`, otherwise returns [`None`].
-fn extract_generic_args(expected_ty_name: &'static str, ty: &Type) -> Option<Vec<GenericArgument>> {
-    let Type::Path(TypePath { qself: None, path, .. }) = ty else {
-        return None;
-    };
-
-    let segments = &path.segments;
-    if segments.len() != 1 {
-        return None;
-    };
-
-    let segment = segments.iter().next().unwrap();
-    if segment.ident != expected_ty_name {
-        return None;
-    }
-
-    let PathArguments::AngleBracketed(path_args) = &segment.arguments else {
-        return Some(Vec::new());
-    };
-    let args = &path_args.args;
-
-    Some(args.iter().cloned().collect::<Vec<_>>())
-}
-
-/// Maps a `path` matching one of the `target_idents` into the `replacement` [`Ident`].
-fn map_path_to_ident(path: &Path, target_idents: &[&str], replacement: &str) -> Option<Ident> {
-    path.require_ident()
-        .ok()
-        .filter(|ident| target_idents.iter().any(|target| ident == target))
-        .map(|ident| Ident::new(replacement, ident.span()))
-}
-
-/// Extracts the left-hand side of a dot-expression.
-fn extract_dot_operand(expr: &Expr) -> Option<&Expr> {
-    match expr {
-        Expr::MethodCall(ExprMethodCall { receiver: operand, .. })
-        | Expr::Field(ExprField { base: operand, .. }) => Some(operand),
-        _ => None,
-    }
-}
-
-/// Replaces the top-level receiver of a dot-expression with an [`Ident`], returning `Some(&mut expr)` if the
-/// replacement is performed.
-///
-/// The top-level receiver is the left-most receiver expression, e.g. the top-level receiver of `a.b.c.foo()` is `a`.
-fn replace_top_level_receiver(expr: &mut Expr, ident: Ident) -> Option<&mut Expr> {
-    if let Expr::MethodCall(ExprMethodCall { receiver: operand, .. })
-    | Expr::Field(ExprField { base: operand, .. }) = expr
-    {
-        return if extract_dot_operand(operand).is_some() {
-            if replace_top_level_receiver(operand, ident).is_some() {
-                Some(expr)
-            } else {
-                None
-            }
-        } else {
-            *operand = Box::new(Expr::Path(ExprPath {
-                attrs: Vec::default(),
-                qself: None,
-                path: ident.into(),
-            }));
-
-            Some(expr)
-        };
-    }
-
-    None
-}
-
-/// Iterates all operands to the left-hand side of the `.` of an [expression][`Expr`], i.e. the container operand of all
-/// [`Expr::Field`] and the receiver operand of all [`Expr::MethodCall`].
-///
-/// The iterator will return the operand expressions in reverse order of appearance. For example, `a.b.c.func()` will
-/// return `vec![c, b, a]`.
-fn iter_dot_operands(expr: &Expr) -> impl Iterator<Item = &Expr> {
-    let mut o = extract_dot_operand(expr);
-
-    std::iter::from_fn(move || {
-        let this = o;
-        o = o.as_ref().and_then(|o| extract_dot_operand(o));
-
-        this
-    })
-}
-
-/// Normalizes a value expression for use when creating an instance of this structure, returning a
-/// [`proc_macro2::TokenStream`] of tokens representing the normalized expression.
-fn normalize_value_expr(expr: &Expr) -> proc_macro2::TokenStream {
-    match &expr {
-        Expr::Path(ExprPath { qself: None, path, .. }) => {
-            if let Some(ident) = map_path_to_ident(path, &["usize", "size_t"], "llvm_usize") {
-                quote! { #ident }
-            } else {
-                abort!(
-                    path,
-                    format!(
-                        "Expected one of `size_t`, `usize`, or an implicit call expression in #[value_type(...)], found {}", 
-                        quote!(#expr).to_string(),
-                    )
-                )
-            }
-        }
-
-        Expr::Call(_) => {
-            quote! { ctx.#expr }
-        }
-
-        Expr::MethodCall(_) => {
-            let base_receiver = iter_dot_operands(expr).last();
-
-            match base_receiver {
-                // `usize.{...}`, `size_t.{...}` -> Rewrite the identifiers to `llvm_usize`
-                Some(Expr::Path(ExprPath { qself: None, path, .. }))
-                    if map_path_to_ident(path, &["usize", "size_t"], "llvm_usize").is_some() =>
-                {
-                    let ident =
-                        map_path_to_ident(path, &["usize", "size_t"], "llvm_usize").unwrap();
-
-                    let mut expr = expr.clone();
-                    let expr = replace_top_level_receiver(&mut expr, ident).unwrap();
-
-                    quote!(#expr)
-                }
-
-                // `ctx.{...}`, `context.{...}` -> Rewrite the identifiers to `ctx`
-                Some(Expr::Path(ExprPath { qself: None, path, .. }))
-                    if map_path_to_ident(path, &["ctx", "context"], "ctx").is_some() =>
-                {
-                    let ident = map_path_to_ident(path, &["ctx", "context"], "ctx").unwrap();
-
-                    let mut expr = expr.clone();
-                    let expr = replace_top_level_receiver(&mut expr, ident).unwrap();
-
-                    quote!(#expr)
-                }
-
-                // No reserved identifier prefix -> Prepend `ctx.` to the entire expression
-                _ => quote! { ctx.#expr },
-            }
-        }
-
-        _ => {
-            abort!(
-                expr,
-                format!(
-                    "Expected one of `size_t`, `usize`, or an implicit call expression in #[value_type(...)], found {}", 
-                    quote!(#expr).to_string(),
-                )
-            )
-        }
-    }
-}
-
-/// Derives an implementation of `codegen::types::structure::StructFields`.
-///
-/// The benefit of using `#[derive(StructFields)]` is that all index- or order-dependent logic required by
-/// `impl StructFields` is automatically generated by this implementation, including the field index as required by
-/// `StructField::new` and the fields as returned by `StructFields::to_vec`.
-///
-/// # Prerequisites
-///
-/// In order to derive from [`StructFields`], you must implement (or derive) [`Eq`] and [`Copy`] as required by
-/// `StructFields`.
-///
-/// Moreover, `#[derive(StructFields)]` can only be used for `struct`s with named fields, and may only contain fields
-/// with either `StructField` or [`PhantomData`] types.
-///
-/// # Attributes for [`StructFields`]
-///
-/// Each `StructField` field must be declared with the `#[value_type(...)]` attribute. The argument of `value_type`
-/// accepts one of the following:
-///
-/// - An expression returning an instance of `inkwell::types::BasicType` (with or without the receiver `ctx`/`context`).
-/// For example, `context.i8_type()`, `ctx.i8_type()`, and `i8_type()` all refer to `i8`.
-/// - The reserved identifiers `usize` and `size_t` referring to an `inkwell::types::IntType` of the platform-dependent
-/// integer size. `usize` and `size_t` can also be used as the receiver to other method calls, e.g.
-/// `usize.array_type(3)`.
-///
-/// # Example
-///
-/// The following is an example of an LLVM slice implemented using `#[derive(StructFields)]`.
-///
-/// ```rust,ignore
-/// use nac3core::{
-///     codegen::types::structure::StructField,
-///     inkwell::{
-///         values::{IntValue, PointerValue},
-///         AddressSpace,
-///     },
-/// };
-/// use nac3core_derive::StructFields;
-///
-/// // All classes that implement StructFields must also implement Eq and Copy
-/// #[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-/// pub struct SliceValue<'ctx> {
-///     // Declares ptr have a value type of i8*
-///     //
-///     // Can also be written as `ctx.i8_type().ptr_type(...)` or `context.i8_type().ptr_type(...)`
-///     #[value_type(i8_type().ptr_type(AddressSpace::default()))]
-///     ptr: StructField<'ctx, PointerValue<'ctx>>,
-///
-///     // Declares len have a value type of usize, depending on the target compilation platform
-///     #[value_type(usize)]
-///     len: StructField<'ctx, IntValue<'ctx>>,
-/// }
-/// ```
-#[proc_macro_derive(StructFields, attributes(value_type))]
-#[proc_macro_error]
-pub fn derive(input: TokenStream) -> TokenStream {
-    let input = parse_macro_input!(input as syn::DeriveInput);
-    let ident = &input.ident;
-
-    let Data::Struct(DataStruct { fields, .. }) = &input.data else {
-        abort!(input, "Only structs with named fields are supported");
-    };
-    if let Err(err_span) =
-        fields
-            .iter()
-            .try_for_each(|field| if field.ident.is_some() { Ok(()) } else { Err(field.span()) })
-    {
-        abort!(err_span, "Only structs with named fields are supported");
-    };
-
-    // Check if struct<'ctx>
-    if input.generics.params.len() != 1 {
-        abort!(input.generics, "Expected exactly 1 generic parameter")
-    }
-
-    let phantom_info = fields
-        .iter()
-        .filter(|field| extract_generic_args("PhantomData", &field.ty).is_some())
-        .map(|field| field.ident.as_ref().unwrap())
-        .cloned()
-        .collect::<Vec<_>>();
-
-    let field_info = fields
-        .iter()
-        .filter(|field| extract_generic_args("PhantomData", &field.ty).is_none())
-        .map(|field| {
-            let ident = field.ident.as_ref().unwrap();
-            let ty = &field.ty;
-
-            let Some(_) = extract_generic_args("StructField", ty) else {
-                abort!(field, "Only StructField and PhantomData are allowed")
-            };
-
-            let attrs = &field.attrs;
-            let Some(value_type_attr) =
-                attrs.iter().find(|attr| attr.path().is_ident("value_type"))
-            else {
-                abort!(field, "Expected #[value_type(...)] attribute for field");
-            };
-
-            let Ok(value_type_expr) = value_type_attr.parse_args::<Expr>() else {
-                abort!(value_type_attr, "Expected expression in #[value_type(...)]");
-            };
-
-            let value_expr_toks = normalize_value_expr(&value_type_expr);
-
-            (ident.clone(), value_expr_toks)
-        })
-        .collect::<Vec<_>>();
-
-    // `<*>::new` impl of `StructField` and `PhantomData` for `StructFields::new`
-    let phantoms_create = phantom_info
-        .iter()
-        .map(|id| quote! { #id: ::std::marker::PhantomData })
-        .collect::<Vec<_>>();
-    let fields_create = field_info
-        .iter()
-        .map(|(id, ty)| {
-            let id_lit = LitStr::new(&id.to_string(), id.span());
-            quote! {
-                #id: ::nac3core::codegen::types::structure::StructField::create(
-                    &mut counter,
-                    #id_lit,
-                    #ty,
-                )
-            }
-        })
-        .collect::<Vec<_>>();
-
-    // `.into()` impl of `StructField` for `StructFields::to_vec`
-    let fields_into =
-        field_info.iter().map(|(id, _)| quote! { self.#id.into() }).collect::<Vec<_>>();
-
-    let impl_block = quote! {
-        impl<'ctx> ::nac3core::codegen::types::structure::StructFields<'ctx> for #ident<'ctx> {
-            fn new(ctx: impl ::nac3core::inkwell::context::AsContextRef<'ctx>, llvm_usize: ::nac3core::inkwell::types::IntType<'ctx>) -> Self {
-                let ctx = unsafe { ::nac3core::inkwell::context::ContextRef::new(ctx.as_ctx_ref()) };
-
-                let mut counter = ::nac3core::codegen::types::structure::FieldIndexCounter::default();
-
-                #ident {
-                    #(#fields_create),*
-                    #(#phantoms_create),*
-                }
-            }
-
-            fn to_vec(&self) -> ::std::vec::Vec<(&'static str, ::nac3core::inkwell::types::BasicTypeEnum<'ctx>)> {
-                vec![
-                    #(#fields_into),*
-                ]
-            }
-        }
-    };
-
-    impl_block.into()
-}

nac3core/nac3core_derive/tests/structfields_empty.rs

@@ -1,9 +0,0 @@
-use nac3core_derive::StructFields;
-use std::marker::PhantomData;
-
-#[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-pub struct EmptyValue<'ctx> {
-    _phantom: PhantomData<&'ctx ()>,
-}
-
-fn main() {}

nac3core/nac3core_derive/tests/structfields_ndarray.rs

@@ -1,20 +0,0 @@
-use nac3core::{
-    codegen::types::structure::StructField,
-    inkwell::{
-        values::{IntValue, PointerValue},
-        AddressSpace,
-    },
-};
-use nac3core_derive::StructFields;
-
-#[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-pub struct NDArrayValue<'ctx> {
-    #[value_type(usize)]
-    ndims: StructField<'ctx, IntValue<'ctx>>,
-    #[value_type(usize.ptr_type(AddressSpace::default()))]
-    shape: StructField<'ctx, PointerValue<'ctx>>,
-    #[value_type(i8_type().ptr_type(AddressSpace::default()))]
-    data: StructField<'ctx, PointerValue<'ctx>>,
-}
-
-fn main() {}

nac3core/nac3core_derive/tests/structfields_slice.rs

@@ -1,18 +0,0 @@
-use nac3core::{
-    codegen::types::structure::StructField,
-    inkwell::{
-        values::{IntValue, PointerValue},
-        AddressSpace,
-    },
-};
-use nac3core_derive::StructFields;
-
-#[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-pub struct SliceValue<'ctx> {
-    #[value_type(i8_type().ptr_type(AddressSpace::default()))]
-    ptr: StructField<'ctx, PointerValue<'ctx>>,
-    #[value_type(usize)]
-    len: StructField<'ctx, IntValue<'ctx>>,
-}
-
-fn main() {}

nac3core/nac3core_derive/tests/structfields_slice_context.rs

@@ -1,18 +0,0 @@
-use nac3core::{
-    codegen::types::structure::StructField,
-    inkwell::{
-        values::{IntValue, PointerValue},
-        AddressSpace,
-    },
-};
-use nac3core_derive::StructFields;
-
-#[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-pub struct SliceValue<'ctx> {
-    #[value_type(context.i8_type().ptr_type(AddressSpace::default()))]
-    ptr: StructField<'ctx, PointerValue<'ctx>>,
-    #[value_type(usize)]
-    len: StructField<'ctx, IntValue<'ctx>>,
-}
-
-fn main() {}

nac3core/nac3core_derive/tests/structfields_slice_ctx.rs

@@ -1,18 +0,0 @@
-use nac3core::{
-    codegen::types::structure::StructField,
-    inkwell::{
-        values::{IntValue, PointerValue},
-        AddressSpace,
-    },
-};
-use nac3core_derive::StructFields;
-
-#[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-pub struct SliceValue<'ctx> {
-    #[value_type(ctx.i8_type().ptr_type(AddressSpace::default()))]
-    ptr: StructField<'ctx, PointerValue<'ctx>>,
-    #[value_type(usize)]
-    len: StructField<'ctx, IntValue<'ctx>>,
-}
-
-fn main() {}

nac3core/nac3core_derive/tests/structfields_slice_sizet.rs

@@ -1,18 +0,0 @@
-use nac3core::{
-    codegen::types::structure::StructField,
-    inkwell::{
-        values::{IntValue, PointerValue},
-        AddressSpace,
-    },
-};
-use nac3core_derive::StructFields;
-
-#[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-pub struct SliceValue<'ctx> {
-    #[value_type(i8_type().ptr_type(AddressSpace::default()))]
-    ptr: StructField<'ctx, PointerValue<'ctx>>,
-    #[value_type(size_t)]
-    len: StructField<'ctx, IntValue<'ctx>>,
-}
-
-fn main() {}

nac3core/nac3core_derive/tests/structfields_test.rs

@@ -1,10 +0,0 @@
-#[test]
-fn test_parse_empty() {
-    let t = trybuild::TestCases::new();
-    t.pass("tests/structfields_empty.rs");
-    t.pass("tests/structfields_slice.rs");
-    t.pass("tests/structfields_slice_ctx.rs");
-    t.pass("tests/structfields_slice_context.rs");
-    t.pass("tests/structfields_slice_sizet.rs");
-    t.pass("tests/structfields_ndarray.rs");
-}

nac3core/src/codegen/concrete_type.rs

@@ -1,9 +1,3 @@
-use std::collections::HashMap;
-
-use indexmap::IndexMap;
-
-use nac3parser::ast::StrRef;
-
 use crate::{
     symbol_resolver::SymbolValue,
     toplevel::DefinitionId,
@@ -15,6 +9,10 @@ use crate::{
     },
 };
 
+use indexmap::IndexMap;
+use nac3parser::ast::StrRef;
+use std::collections::HashMap;
+
 pub struct ConcreteTypeStore {
     store: Vec<ConcreteTypeEnum>,
 }

nac3core/src/codegen/extern_fns.rs

@@ -1,10 +1,8 @@
-use inkwell::{
-    attributes::{Attribute, AttributeLoc},
-    values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue},
-};
+use inkwell::attributes::{Attribute, AttributeLoc};
+use inkwell::values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue};
 use itertools::Either;
 
-use super::CodeGenContext;
+use crate::codegen::CodeGenContext;
 
 /// Macro to generate extern function
 /// Both function return type and function parameter type are `FloatValue`

nac3core/src/codegen/generator.rs

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

nac3core/src/codegen/irrt/list.rs

@@ -1,162 +0,0 @@
-use inkwell::{
-    types::BasicTypeEnum,
-    values::{BasicValueEnum, CallSiteValue, IntValue},
-    AddressSpace, IntPredicate,
-};
-use itertools::Either;
-
-use super::calculate_len_for_slice_range;
-use crate::codegen::{
-    macros::codegen_unreachable,
-    values::{ArrayLikeValue, ListValue},
-    CodeGenContext, CodeGenerator,
-};
-
-/// This function handles 'end' **inclusively**.
-/// Order of tuples `assign_idx` and `value_idx` is ('start', 'end', 'step').
-/// Negative index should be handled before entering this function
-pub fn list_slice_assignment<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &mut G,
-    ctx: &mut CodeGenContext<'ctx, '_>,
-    ty: BasicTypeEnum<'ctx>,
-    dest_arr: ListValue<'ctx>,
-    dest_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
-    src_arr: ListValue<'ctx>,
-    src_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
-) {
-    let size_ty = generator.get_size_type(ctx.ctx);
-    let int8_ptr = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
-    let int32 = ctx.ctx.i32_type();
-    let (fun_symbol, elem_ptr_type) = ("__nac3_list_slice_assign_var_size", int8_ptr);
-    let slice_assign_fun = {
-        let ty_vec = vec![
-            int32.into(),         // dest start idx
-            int32.into(),         // dest end idx
-            int32.into(),         // dest step
-            elem_ptr_type.into(), // dest arr ptr
-            int32.into(),         // dest arr len
-            int32.into(),         // src start idx
-            int32.into(),         // src end idx
-            int32.into(),         // src step
-            elem_ptr_type.into(), // src arr ptr
-            int32.into(),         // src arr len
-            int32.into(),         // size
-        ];
-        ctx.module.get_function(fun_symbol).unwrap_or_else(|| {
-            let fn_t = int32.fn_type(ty_vec.as_slice(), false);
-            ctx.module.add_function(fun_symbol, fn_t, None)
-        })
-    };
-
-    let zero = int32.const_zero();
-    let one = int32.const_int(1, false);
-    let dest_arr_ptr = dest_arr.data().base_ptr(ctx, generator);
-    let dest_arr_ptr =
-        ctx.builder.build_pointer_cast(dest_arr_ptr, elem_ptr_type, "dest_arr_ptr_cast").unwrap();
-    let dest_len = dest_arr.load_size(ctx, Some("dest.len"));
-    let dest_len = ctx.builder.build_int_truncate_or_bit_cast(dest_len, int32, "srclen32").unwrap();
-    let src_arr_ptr = src_arr.data().base_ptr(ctx, generator);
-    let src_arr_ptr =
-        ctx.builder.build_pointer_cast(src_arr_ptr, elem_ptr_type, "src_arr_ptr_cast").unwrap();
-    let src_len = src_arr.load_size(ctx, Some("src.len"));
-    let src_len = ctx.builder.build_int_truncate_or_bit_cast(src_len, int32, "srclen32").unwrap();
-
-    // index in bound and positive should be done
-    // assert if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest), and
-    // throw exception if not satisfied
-    let src_end = ctx
-        .builder
-        .build_select(
-            ctx.builder.build_int_compare(IntPredicate::SLT, src_idx.2, zero, "is_neg").unwrap(),
-            ctx.builder.build_int_sub(src_idx.1, one, "e_min_one").unwrap(),
-            ctx.builder.build_int_add(src_idx.1, one, "e_add_one").unwrap(),
-            "final_e",
-        )
-        .map(BasicValueEnum::into_int_value)
-        .unwrap();
-    let dest_end = ctx
-        .builder
-        .build_select(
-            ctx.builder.build_int_compare(IntPredicate::SLT, dest_idx.2, zero, "is_neg").unwrap(),
-            ctx.builder.build_int_sub(dest_idx.1, one, "e_min_one").unwrap(),
-            ctx.builder.build_int_add(dest_idx.1, one, "e_add_one").unwrap(),
-            "final_e",
-        )
-        .map(BasicValueEnum::into_int_value)
-        .unwrap();
-    let src_slice_len =
-        calculate_len_for_slice_range(generator, ctx, src_idx.0, src_end, src_idx.2);
-    let dest_slice_len =
-        calculate_len_for_slice_range(generator, ctx, dest_idx.0, dest_end, dest_idx.2);
-    let src_eq_dest = ctx
-        .builder
-        .build_int_compare(IntPredicate::EQ, src_slice_len, dest_slice_len, "slice_src_eq_dest")
-        .unwrap();
-    let src_slt_dest = ctx
-        .builder
-        .build_int_compare(IntPredicate::SLT, src_slice_len, dest_slice_len, "slice_src_slt_dest")
-        .unwrap();
-    let dest_step_eq_one = ctx
-        .builder
-        .build_int_compare(
-            IntPredicate::EQ,
-            dest_idx.2,
-            dest_idx.2.get_type().const_int(1, false),
-            "slice_dest_step_eq_one",
-        )
-        .unwrap();
-    let cond_1 = ctx.builder.build_and(dest_step_eq_one, src_slt_dest, "slice_cond_1").unwrap();
-    let cond = ctx.builder.build_or(src_eq_dest, cond_1, "slice_cond").unwrap();
-    ctx.make_assert(
-        generator,
-        cond,
-        "0:ValueError",
-        "attempt to assign sequence of size {0} to slice of size {1} with step size {2}",
-        [Some(src_slice_len), Some(dest_slice_len), Some(dest_idx.2)],
-        ctx.current_loc,
-    );
-
-    let new_len = {
-        let args = vec![
-            dest_idx.0.into(),   // dest start idx
-            dest_idx.1.into(),   // dest end idx
-            dest_idx.2.into(),   // dest step
-            dest_arr_ptr.into(), // dest arr ptr
-            dest_len.into(),     // dest arr len
-            src_idx.0.into(),    // src start idx
-            src_idx.1.into(),    // src end idx
-            src_idx.2.into(),    // src step
-            src_arr_ptr.into(),  // src arr ptr
-            src_len.into(),      // src arr len
-            {
-                let s = match ty {
-                    BasicTypeEnum::FloatType(t) => t.size_of(),
-                    BasicTypeEnum::IntType(t) => t.size_of(),
-                    BasicTypeEnum::PointerType(t) => t.size_of(),
-                    BasicTypeEnum::StructType(t) => t.size_of().unwrap(),
-                    _ => codegen_unreachable!(ctx),
-                };
-                ctx.builder.build_int_truncate_or_bit_cast(s, int32, "size").unwrap()
-            }
-            .into(),
-        ];
-        ctx.builder
-            .build_call(slice_assign_fun, args.as_slice(), "slice_assign")
-            .map(CallSiteValue::try_as_basic_value)
-            .map(|v| v.map_left(BasicValueEnum::into_int_value))
-            .map(Either::unwrap_left)
-            .unwrap()
-    };
-    // update length
-    let need_update =
-        ctx.builder.build_int_compare(IntPredicate::NE, new_len, dest_len, "need_update").unwrap();
-    let current = ctx.builder.get_insert_block().unwrap().get_parent().unwrap();
-    let update_bb = ctx.ctx.append_basic_block(current, "update");
-    let cont_bb = ctx.ctx.append_basic_block(current, "cont");
-    ctx.builder.build_conditional_branch(need_update, update_bb, cont_bb).unwrap();
-    ctx.builder.position_at_end(update_bb);
-    let new_len = ctx.builder.build_int_z_extend_or_bit_cast(new_len, size_ty, "new_len").unwrap();
-    dest_arr.store_size(ctx, generator, new_len);
-    ctx.builder.build_unconditional_branch(cont_bb).unwrap();
-    ctx.builder.position_at_end(cont_bb);
-}

nac3core/src/codegen/irrt/math.rs

@@ -1,152 +0,0 @@
-use inkwell::{
-    values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue},
-    IntPredicate,
-};
-use itertools::Either;
-
-use crate::codegen::{
-    macros::codegen_unreachable,
-    {CodeGenContext, CodeGenerator},
-};
-
-// repeated squaring method adapted from GNU Scientific Library:
-// https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
-pub fn integer_power<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &mut G,
-    ctx: &mut CodeGenContext<'ctx, '_>,
-    base: IntValue<'ctx>,
-    exp: IntValue<'ctx>,
-    signed: bool,
-) -> IntValue<'ctx> {
-    let symbol = match (base.get_type().get_bit_width(), exp.get_type().get_bit_width(), signed) {
-        (32, 32, true) => "__nac3_int_exp_int32_t",
-        (64, 64, true) => "__nac3_int_exp_int64_t",
-        (32, 32, false) => "__nac3_int_exp_uint32_t",
-        (64, 64, false) => "__nac3_int_exp_uint64_t",
-        _ => codegen_unreachable!(ctx),
-    };
-    let base_type = base.get_type();
-    let pow_fun = ctx.module.get_function(symbol).unwrap_or_else(|| {
-        let fn_type = base_type.fn_type(&[base_type.into(), base_type.into()], false);
-        ctx.module.add_function(symbol, fn_type, None)
-    });
-    // throw exception when exp < 0
-    let ge_zero = ctx
-        .builder
-        .build_int_compare(
-            IntPredicate::SGE,
-            exp,
-            exp.get_type().const_zero(),
-            "assert_int_pow_ge_0",
-        )
-        .unwrap();
-    ctx.make_assert(
-        generator,
-        ge_zero,
-        "0:ValueError",
-        "integer power must be positive or zero",
-        [None, None, None],
-        ctx.current_loc,
-    );
-    ctx.builder
-        .build_call(pow_fun, &[base.into(), exp.into()], "call_int_pow")
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_int_value))
-        .map(Either::unwrap_left)
-        .unwrap()
-}
-
-/// Generates a call to `isinf` in IR. Returns an `i1` representing the result.
-pub fn call_isinf<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &mut G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    v: FloatValue<'ctx>,
-) -> IntValue<'ctx> {
-    let intrinsic_fn = ctx.module.get_function("__nac3_isinf").unwrap_or_else(|| {
-        let fn_type = ctx.ctx.i32_type().fn_type(&[ctx.ctx.f64_type().into()], false);
-        ctx.module.add_function("__nac3_isinf", fn_type, None)
-    });
-
-    let ret = ctx
-        .builder
-        .build_call(intrinsic_fn, &[v.into()], "isinf")
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_int_value))
-        .map(Either::unwrap_left)
-        .unwrap();
-
-    generator.bool_to_i1(ctx, ret)
-}
-
-/// Generates a call to `isnan` in IR. Returns an `i1` representing the result.
-pub fn call_isnan<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &mut G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    v: FloatValue<'ctx>,
-) -> IntValue<'ctx> {
-    let intrinsic_fn = ctx.module.get_function("__nac3_isnan").unwrap_or_else(|| {
-        let fn_type = ctx.ctx.i32_type().fn_type(&[ctx.ctx.f64_type().into()], false);
-        ctx.module.add_function("__nac3_isnan", fn_type, None)
-    });
-
-    let ret = ctx
-        .builder
-        .build_call(intrinsic_fn, &[v.into()], "isnan")
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_int_value))
-        .map(Either::unwrap_left)
-        .unwrap();
-
-    generator.bool_to_i1(ctx, ret)
-}
-
-/// Generates a call to `gamma` in IR. Returns an `f64` representing the result.
-pub fn call_gamma<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
-    let llvm_f64 = ctx.ctx.f64_type();
-
-    let intrinsic_fn = ctx.module.get_function("__nac3_gamma").unwrap_or_else(|| {
-        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
-        ctx.module.add_function("__nac3_gamma", fn_type, None)
-    });
-
-    ctx.builder
-        .build_call(intrinsic_fn, &[v.into()], "gamma")
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_float_value))
-        .map(Either::unwrap_left)
-        .unwrap()
-}
-
-/// Generates a call to `gammaln` in IR. Returns an `f64` representing the result.
-pub fn call_gammaln<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
-    let llvm_f64 = ctx.ctx.f64_type();
-
-    let intrinsic_fn = ctx.module.get_function("__nac3_gammaln").unwrap_or_else(|| {
-        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
-        ctx.module.add_function("__nac3_gammaln", fn_type, None)
-    });
-
-    ctx.builder
-        .build_call(intrinsic_fn, &[v.into()], "gammaln")
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_float_value))
-        .map(Either::unwrap_left)
-        .unwrap()
-}
-
-/// Generates a call to `j0` in IR. Returns an `f64` representing the result.
-pub fn call_j0<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
-    let llvm_f64 = ctx.ctx.f64_type();
-
-    let intrinsic_fn = ctx.module.get_function("__nac3_j0").unwrap_or_else(|| {
-        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
-        ctx.module.add_function("__nac3_j0", fn_type, None)
-    });
-
-    ctx.builder
-        .build_call(intrinsic_fn, &[v.into()], "j0")
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_float_value))
-        .map(Either::unwrap_left)
-        .unwrap()
-}

nac3core/src/codegen/irrt/mod.rs

@@ -1,29 +1,29 @@
+use crate::{symbol_resolver::SymbolResolver, typecheck::typedef::Type};
+
+use super::{
+    classes::{ArrayLikeValue, ListValue},
+    model::*,
+    object::{
+        list::List,
+        ndarray::{broadcast::ShapeEntry, indexing::NDIndex, nditer::NDIter, NDArray},
+    },
+    CodeGenContext, CodeGenerator,
+};
+use function::CallFunction;
 use inkwell::{
     attributes::{Attribute, AttributeLoc},
     context::Context,
     memory_buffer::MemoryBuffer,
     module::Module,
-    values::{BasicValue, BasicValueEnum, IntValue},
-    IntPredicate,
+    types::BasicTypeEnum,
+    values::{BasicValue, BasicValueEnum, CallSiteValue, FloatValue, IntValue},
+    AddressSpace, IntPredicate,
 };
-
+use itertools::Either;
 use nac3parser::ast::Expr;
 
-use super::{CodeGenContext, CodeGenerator};
-use crate::{symbol_resolver::SymbolResolver, typecheck::typedef::Type};
-pub use list::*;
-pub use math::*;
-pub use range::*;
-pub use slice::*;
-
-mod list;
-mod math;
-pub mod ndarray;
-mod range;
-mod slice;
-
 #[must_use]
-pub fn load_irrt<'ctx>(ctx: &'ctx Context, symbol_resolver: &dyn SymbolResolver) -> Module<'ctx> {
+pub fn load_irrt(ctx: &Context) -> Module {
     let bitcode_buf = MemoryBuffer::create_from_memory_range(
         include_bytes!(concat!(env!("OUT_DIR"), "/irrt.bc")),
         "irrt_bitcode_buffer",
@@ -39,47 +39,89 @@ pub fn load_irrt<'ctx>(ctx: &'ctx Context, symbol_resolver: &dyn SymbolResolver)
         let function = irrt_mod.get_function(symbol).unwrap();
         function.add_attribute(AttributeLoc::Function, ctx.create_enum_attribute(inline_attr, 0));
     }
-
-    // Initialize all global `EXN_*` exception IDs in IRRT with the [`SymbolResolver`].
-    let exn_id_type = ctx.i32_type();
-    let errors = &[
-        ("EXN_INDEX_ERROR", "0:IndexError"),
-        ("EXN_VALUE_ERROR", "0:ValueError"),
-        ("EXN_ASSERTION_ERROR", "0:AssertionError"),
-        ("EXN_TYPE_ERROR", "0:TypeError"),
-    ];
-    for (irrt_name, symbol_name) in errors {
-        let exn_id = symbol_resolver.get_string_id(symbol_name);
-        let exn_id = exn_id_type.const_int(exn_id as u64, false).as_basic_value_enum();
-
-        let global = irrt_mod.get_global(irrt_name).unwrap_or_else(|| {
-            panic!("Exception symbol name '{irrt_name}' should exist in the IRRT LLVM module")
-        });
-        global.set_initializer(&exn_id);
-    }
-
     irrt_mod
 }
 
-/// Returns the name of a function which contains variants for 32-bit and 64-bit `size_t`.
-///
-/// - When [`TypeContext::size_type`] is 32-bits, the function name is `fn_name}`.
-/// - When [`TypeContext::size_type`] is 64-bits, the function name is `{fn_name}64`.
-#[must_use]
-pub fn get_usize_dependent_function_name<G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'_, '_>,
-    name: &str,
-) -> String {
-    let mut name = name.to_owned();
-    match generator.get_size_type(ctx.ctx).get_bit_width() {
-        32 => {}
-        64 => name.push_str("64"),
-        bit_width => {
-            panic!("Unsupported int type bit width {bit_width}, must be either 32-bits or 64-bits")
-        }
-    }
-    name
+// repeated squaring method adapted from GNU Scientific Library:
+// https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
+pub fn integer_power<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    base: IntValue<'ctx>,
+    exp: IntValue<'ctx>,
+    signed: bool,
+) -> IntValue<'ctx> {
+    let symbol = match (base.get_type().get_bit_width(), exp.get_type().get_bit_width(), signed) {
+        (32, 32, true) => "__nac3_int_exp_int32_t",
+        (64, 64, true) => "__nac3_int_exp_int64_t",
+        (32, 32, false) => "__nac3_int_exp_uint32_t",
+        (64, 64, false) => "__nac3_int_exp_uint64_t",
+        _ => unreachable!(),
+    };
+    let base_type = base.get_type();
+    let pow_fun = ctx.module.get_function(symbol).unwrap_or_else(|| {
+        let fn_type = base_type.fn_type(&[base_type.into(), base_type.into()], false);
+        ctx.module.add_function(symbol, fn_type, None)
+    });
+    // throw exception when exp < 0
+    let ge_zero = ctx
+        .builder
+        .build_int_compare(
+            IntPredicate::SGE,
+            exp,
+            exp.get_type().const_zero(),
+            "assert_int_pow_ge_0",
+        )
+        .unwrap();
+    ctx.make_assert(
+        generator,
+        ge_zero,
+        "0:ValueError",
+        "integer power must be positive or zero",
+        [None, None, None],
+        ctx.current_loc,
+    );
+    ctx.builder
+        .build_call(pow_fun, &[base.into(), exp.into()], "call_int_pow")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+pub fn calculate_len_for_slice_range<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    start: IntValue<'ctx>,
+    end: IntValue<'ctx>,
+    step: IntValue<'ctx>,
+) -> IntValue<'ctx> {
+    const SYMBOL: &str = "__nac3_range_slice_len";
+    let len_func = ctx.module.get_function(SYMBOL).unwrap_or_else(|| {
+        let i32_t = ctx.ctx.i32_type();
+        let fn_t = i32_t.fn_type(&[i32_t.into(), i32_t.into(), i32_t.into()], false);
+        ctx.module.add_function(SYMBOL, fn_t, None)
+    });
+
+    // assert step != 0, throw exception if not
+    let not_zero = ctx
+        .builder
+        .build_int_compare(IntPredicate::NE, step, step.get_type().const_zero(), "range_step_ne")
+        .unwrap();
+    ctx.make_assert(
+        generator,
+        not_zero,
+        "0:ValueError",
+        "step must not be zero",
+        [None, None, None],
+        ctx.current_loc,
+    );
+    ctx.builder
+        .build_call(len_func, &[start.into(), end.into(), step.into()], "calc_len")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap()
 }
 
 /// NOTE: the output value of the end index of this function should be compared ***inclusively***,
@@ -247,3 +289,599 @@ pub fn handle_slice_indices<'ctx, G: CodeGenerator>(
         }
     }))
 }
+
+/// this function allows index out of range, since python
+/// allows index out of range in slice (`a = [1,2,3]; a[1:10] == [2,3]`).
+pub fn handle_slice_index_bound<'ctx, G: CodeGenerator>(
+    i: &Expr<Option<Type>>,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    generator: &mut G,
+    length: IntValue<'ctx>,
+) -> Result<Option<IntValue<'ctx>>, String> {
+    const SYMBOL: &str = "__nac3_slice_index_bound";
+    let func = ctx.module.get_function(SYMBOL).unwrap_or_else(|| {
+        let i32_t = ctx.ctx.i32_type();
+        let fn_t = i32_t.fn_type(&[i32_t.into(), i32_t.into()], false);
+        ctx.module.add_function(SYMBOL, fn_t, None)
+    });
+
+    let i = if let Some(v) = generator.gen_expr(ctx, i)? {
+        v.to_basic_value_enum(ctx, generator, i.custom.unwrap())?
+    } else {
+        return Ok(None);
+    };
+    Ok(Some(
+        ctx.builder
+            .build_call(func, &[i.into(), length.into()], "bounded_ind")
+            .map(CallSiteValue::try_as_basic_value)
+            .map(|v| v.map_left(BasicValueEnum::into_int_value))
+            .map(Either::unwrap_left)
+            .unwrap(),
+    ))
+}
+
+/// This function handles 'end' **inclusively**.
+/// Order of tuples `assign_idx` and `value_idx` is ('start', 'end', 'step').
+/// Negative index should be handled before entering this function
+pub fn list_slice_assignment<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ty: BasicTypeEnum<'ctx>,
+    dest_arr: ListValue<'ctx>,
+    dest_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
+    src_arr: ListValue<'ctx>,
+    src_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
+) {
+    let size_ty = generator.get_size_type(ctx.ctx);
+    let int8_ptr = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
+    let int32 = ctx.ctx.i32_type();
+    let (fun_symbol, elem_ptr_type) = ("__nac3_list_slice_assign_var_size", int8_ptr);
+    let slice_assign_fun = {
+        let ty_vec = vec![
+            int32.into(),         // dest start idx
+            int32.into(),         // dest end idx
+            int32.into(),         // dest step
+            elem_ptr_type.into(), // dest arr ptr
+            int32.into(),         // dest arr len
+            int32.into(),         // src start idx
+            int32.into(),         // src end idx
+            int32.into(),         // src step
+            elem_ptr_type.into(), // src arr ptr
+            int32.into(),         // src arr len
+            int32.into(),         // size
+        ];
+        ctx.module.get_function(fun_symbol).unwrap_or_else(|| {
+            let fn_t = int32.fn_type(ty_vec.as_slice(), false);
+            ctx.module.add_function(fun_symbol, fn_t, None)
+        })
+    };
+
+    let zero = int32.const_zero();
+    let one = int32.const_int(1, false);
+    let dest_arr_ptr = dest_arr.data().base_ptr(ctx, generator);
+    let dest_arr_ptr =
+        ctx.builder.build_pointer_cast(dest_arr_ptr, elem_ptr_type, "dest_arr_ptr_cast").unwrap();
+    let dest_len = dest_arr.load_size(ctx, Some("dest.len"));
+    let dest_len = ctx.builder.build_int_truncate_or_bit_cast(dest_len, int32, "srclen32").unwrap();
+    let src_arr_ptr = src_arr.data().base_ptr(ctx, generator);
+    let src_arr_ptr =
+        ctx.builder.build_pointer_cast(src_arr_ptr, elem_ptr_type, "src_arr_ptr_cast").unwrap();
+    let src_len = src_arr.load_size(ctx, Some("src.len"));
+    let src_len = ctx.builder.build_int_truncate_or_bit_cast(src_len, int32, "srclen32").unwrap();
+
+    // index in bound and positive should be done
+    // assert if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest), and
+    // throw exception if not satisfied
+    let src_end = ctx
+        .builder
+        .build_select(
+            ctx.builder.build_int_compare(IntPredicate::SLT, src_idx.2, zero, "is_neg").unwrap(),
+            ctx.builder.build_int_sub(src_idx.1, one, "e_min_one").unwrap(),
+            ctx.builder.build_int_add(src_idx.1, one, "e_add_one").unwrap(),
+            "final_e",
+        )
+        .map(BasicValueEnum::into_int_value)
+        .unwrap();
+    let dest_end = ctx
+        .builder
+        .build_select(
+            ctx.builder.build_int_compare(IntPredicate::SLT, dest_idx.2, zero, "is_neg").unwrap(),
+            ctx.builder.build_int_sub(dest_idx.1, one, "e_min_one").unwrap(),
+            ctx.builder.build_int_add(dest_idx.1, one, "e_add_one").unwrap(),
+            "final_e",
+        )
+        .map(BasicValueEnum::into_int_value)
+        .unwrap();
+    let src_slice_len =
+        calculate_len_for_slice_range(generator, ctx, src_idx.0, src_end, src_idx.2);
+    let dest_slice_len =
+        calculate_len_for_slice_range(generator, ctx, dest_idx.0, dest_end, dest_idx.2);
+    let src_eq_dest = ctx
+        .builder
+        .build_int_compare(IntPredicate::EQ, src_slice_len, dest_slice_len, "slice_src_eq_dest")
+        .unwrap();
+    let src_slt_dest = ctx
+        .builder
+        .build_int_compare(IntPredicate::SLT, src_slice_len, dest_slice_len, "slice_src_slt_dest")
+        .unwrap();
+    let dest_step_eq_one = ctx
+        .builder
+        .build_int_compare(
+            IntPredicate::EQ,
+            dest_idx.2,
+            dest_idx.2.get_type().const_int(1, false),
+            "slice_dest_step_eq_one",
+        )
+        .unwrap();
+    let cond_1 = ctx.builder.build_and(dest_step_eq_one, src_slt_dest, "slice_cond_1").unwrap();
+    let cond = ctx.builder.build_or(src_eq_dest, cond_1, "slice_cond").unwrap();
+    ctx.make_assert(
+        generator,
+        cond,
+        "0:ValueError",
+        "attempt to assign sequence of size {0} to slice of size {1} with step size {2}",
+        [Some(src_slice_len), Some(dest_slice_len), Some(dest_idx.2)],
+        ctx.current_loc,
+    );
+
+    let new_len = {
+        let args = vec![
+            dest_idx.0.into(),   // dest start idx
+            dest_idx.1.into(),   // dest end idx
+            dest_idx.2.into(),   // dest step
+            dest_arr_ptr.into(), // dest arr ptr
+            dest_len.into(),     // dest arr len
+            src_idx.0.into(),    // src start idx
+            src_idx.1.into(),    // src end idx
+            src_idx.2.into(),    // src step
+            src_arr_ptr.into(),  // src arr ptr
+            src_len.into(),      // src arr len
+            {
+                let s = match ty {
+                    BasicTypeEnum::FloatType(t) => t.size_of(),
+                    BasicTypeEnum::IntType(t) => t.size_of(),
+                    BasicTypeEnum::PointerType(t) => t.size_of(),
+                    BasicTypeEnum::StructType(t) => t.size_of().unwrap(),
+                    _ => unreachable!(),
+                };
+                ctx.builder.build_int_truncate_or_bit_cast(s, int32, "size").unwrap()
+            }
+            .into(),
+        ];
+        ctx.builder
+            .build_call(slice_assign_fun, args.as_slice(), "slice_assign")
+            .map(CallSiteValue::try_as_basic_value)
+            .map(|v| v.map_left(BasicValueEnum::into_int_value))
+            .map(Either::unwrap_left)
+            .unwrap()
+    };
+    // update length
+    let need_update =
+        ctx.builder.build_int_compare(IntPredicate::NE, new_len, dest_len, "need_update").unwrap();
+    let current = ctx.builder.get_insert_block().unwrap().get_parent().unwrap();
+    let update_bb = ctx.ctx.append_basic_block(current, "update");
+    let cont_bb = ctx.ctx.append_basic_block(current, "cont");
+    ctx.builder.build_conditional_branch(need_update, update_bb, cont_bb).unwrap();
+    ctx.builder.position_at_end(update_bb);
+    let new_len = ctx.builder.build_int_z_extend_or_bit_cast(new_len, size_ty, "new_len").unwrap();
+    dest_arr.store_size(ctx, generator, new_len);
+    ctx.builder.build_unconditional_branch(cont_bb).unwrap();
+    ctx.builder.position_at_end(cont_bb);
+}
+
+/// Generates a call to `isinf` in IR. Returns an `i1` representing the result.
+pub fn call_isinf<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &CodeGenContext<'ctx, '_>,
+    v: FloatValue<'ctx>,
+) -> IntValue<'ctx> {
+    let intrinsic_fn = ctx.module.get_function("__nac3_isinf").unwrap_or_else(|| {
+        let fn_type = ctx.ctx.i32_type().fn_type(&[ctx.ctx.f64_type().into()], false);
+        ctx.module.add_function("__nac3_isinf", fn_type, None)
+    });
+
+    let ret = ctx
+        .builder
+        .build_call(intrinsic_fn, &[v.into()], "isinf")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap();
+
+    generator.bool_to_i1(ctx, ret)
+}
+
+/// Generates a call to `isnan` in IR. Returns an `i1` representing the result.
+pub fn call_isnan<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &CodeGenContext<'ctx, '_>,
+    v: FloatValue<'ctx>,
+) -> IntValue<'ctx> {
+    let intrinsic_fn = ctx.module.get_function("__nac3_isnan").unwrap_or_else(|| {
+        let fn_type = ctx.ctx.i32_type().fn_type(&[ctx.ctx.f64_type().into()], false);
+        ctx.module.add_function("__nac3_isnan", fn_type, None)
+    });
+
+    let ret = ctx
+        .builder
+        .build_call(intrinsic_fn, &[v.into()], "isnan")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap();
+
+    generator.bool_to_i1(ctx, ret)
+}
+
+/// Generates a call to `gamma` in IR. Returns an `f64` representing the result.
+pub fn call_gamma<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
+    let llvm_f64 = ctx.ctx.f64_type();
+
+    let intrinsic_fn = ctx.module.get_function("__nac3_gamma").unwrap_or_else(|| {
+        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
+        ctx.module.add_function("__nac3_gamma", fn_type, None)
+    });
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[v.into()], "gamma")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_float_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Generates a call to `gammaln` in IR. Returns an `f64` representing the result.
+pub fn call_gammaln<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
+    let llvm_f64 = ctx.ctx.f64_type();
+
+    let intrinsic_fn = ctx.module.get_function("__nac3_gammaln").unwrap_or_else(|| {
+        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
+        ctx.module.add_function("__nac3_gammaln", fn_type, None)
+    });
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[v.into()], "gammaln")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_float_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Generates a call to `j0` in IR. Returns an `f64` representing the result.
+pub fn call_j0<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
+    let llvm_f64 = ctx.ctx.f64_type();
+
+    let intrinsic_fn = ctx.module.get_function("__nac3_j0").unwrap_or_else(|| {
+        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
+        ctx.module.add_function("__nac3_j0", fn_type, None)
+    });
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[v.into()], "j0")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_float_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Initialize all global `EXN_*` exception IDs in IRRT with the [`SymbolResolver`].
+pub fn setup_irrt_exceptions<'ctx>(
+    ctx: &'ctx Context,
+    module: &Module<'ctx>,
+    symbol_resolver: &dyn SymbolResolver,
+) {
+    let exn_id_type = ctx.i32_type();
+
+    let errors = &[
+        ("EXN_INDEX_ERROR", "0:IndexError"),
+        ("EXN_VALUE_ERROR", "0:ValueError"),
+        ("EXN_ASSERTION_ERROR", "0:AssertionError"),
+        ("EXN_TYPE_ERROR", "0:TypeError"),
+    ];
+
+    for (irrt_name, symbol_name) in errors {
+        let exn_id = symbol_resolver.get_string_id(symbol_name);
+        let exn_id = exn_id_type.const_int(exn_id as u64, false).as_basic_value_enum();
+
+        let global = 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);
+    }
+}
+
+// When [`TypeContext::size_type`] is 32-bits, the function name is "{fn_name}".
+// When [`TypeContext::size_type`] is 64-bits, the function name is "{fn_name}64".
+#[must_use]
+pub fn get_sizet_dependent_function_name<G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &CodeGenContext<'_, '_>,
+    name: &str,
+) -> String {
+    let mut name = name.to_owned();
+    match generator.get_size_type(ctx.ctx).get_bit_width() {
+        32 => {}
+        64 => name.push_str("64"),
+        bit_width => {
+            panic!("Unsupported int type bit width {bit_width}, must be either 32-bits or 64-bits")
+        }
+    }
+    name
+}
+
+pub fn call_nac3_ndarray_util_assert_shape_no_negative<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndims: Instance<'ctx, Int<SizeT>>,
+    shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(
+        generator,
+        ctx,
+        "__nac3_ndarray_util_assert_shape_no_negative",
+    );
+    CallFunction::begin(generator, ctx, &name).arg(ndims).arg(shape).returning_void();
+}
+
+pub fn call_nac3_ndarray_util_assert_output_shape_same<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray_ndims: Instance<'ctx, Int<SizeT>>,
+    ndarray_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    output_ndims: Instance<'ctx, Int<SizeT>>,
+    output_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(
+        generator,
+        ctx,
+        "__nac3_ndarray_util_assert_output_shape_same",
+    );
+    CallFunction::begin(generator, ctx, &name)
+        .arg(ndarray_ndims)
+        .arg(ndarray_shape)
+        .arg(output_ndims)
+        .arg(output_shape)
+        .returning_void();
+}
+
+pub fn call_nac3_ndarray_size<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) -> Instance<'ctx, Int<SizeT>> {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_size");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("size")
+}
+
+pub fn call_nac3_ndarray_nbytes<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) -> Instance<'ctx, Int<SizeT>> {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_nbytes");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("nbytes")
+}
+
+pub fn call_nac3_ndarray_len<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) -> Instance<'ctx, Int<SizeT>> {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_len");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("len")
+}
+
+pub fn call_nac3_ndarray_is_c_contiguous<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) -> Instance<'ctx, Int<Bool>> {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_is_c_contiguous");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_auto("is_c_contiguous")
+}
+
+pub fn call_nac3_ndarray_get_nth_pelement<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    index: Instance<'ctx, Int<SizeT>>,
+) -> Instance<'ctx, Ptr<Int<Byte>>> {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_get_nth_pelement");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).arg(index).returning_auto("pelement")
+}
+
+pub fn call_nac3_ndarray_get_pelement_by_indices<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    indices: Instance<'ctx, Ptr<Int<SizeT>>>,
+) -> Instance<'ctx, Ptr<Int<Byte>>> {
+    let name =
+        get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_get_pelement_by_indices");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).arg(indices).returning_auto("pelement")
+}
+
+pub fn call_nac3_ndarray_set_strides_by_shape<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name =
+        get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_set_strides_by_shape");
+    CallFunction::begin(generator, ctx, &name).arg(ndarray).returning_void();
+}
+
+pub fn call_nac3_ndarray_copy_data<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_copy_data");
+    CallFunction::begin(generator, ctx, &name).arg(src_ndarray).arg(dst_ndarray).returning_void();
+}
+
+pub fn call_nac3_nditer_initialize<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    indices: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_initialize");
+    CallFunction::begin(generator, ctx, &name).arg(iter).arg(ndarray).arg(indices).returning_void();
+}
+
+pub fn call_nac3_nditer_has_next<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
+) -> Instance<'ctx, Int<Bool>> {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_has_next");
+    CallFunction::begin(generator, ctx, &name).arg(iter).returning_auto("has_next")
+}
+
+pub fn call_nac3_nditer_next<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    iter: Instance<'ctx, Ptr<Struct<NDIter>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_nditer_next");
+    CallFunction::begin(generator, ctx, &name).arg(iter).returning_void();
+}
+
+pub fn call_nac3_ndarray_index<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    num_indices: Instance<'ctx, Int<SizeT>>,
+    indices: Instance<'ctx, Ptr<Struct<NDIndex>>>,
+    src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_index");
+    CallFunction::begin(generator, ctx, &name)
+        .arg(num_indices)
+        .arg(indices)
+        .arg(src_ndarray)
+        .arg(dst_ndarray)
+        .returning_void();
+}
+
+pub fn call_nac3_ndarray_array_set_and_validate_list_shape<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    list: Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>>,
+    ndims: Instance<'ctx, Int<SizeT>>,
+    shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(
+        generator,
+        ctx,
+        "__nac3_ndarray_array_set_and_validate_list_shape",
+    );
+    CallFunction::begin(generator, ctx, &name).arg(list).arg(ndims).arg(shape).returning_void();
+}
+
+pub fn call_nac3_ndarray_array_write_list_to_array<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    list: Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>>,
+    ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name = get_sizet_dependent_function_name(
+        generator,
+        ctx,
+        "__nac3_ndarray_array_write_list_to_array",
+    );
+    CallFunction::begin(generator, ctx, &name).arg(list).arg(ndarray).returning_void();
+}
+
+pub fn call_nac3_ndarray_reshape_resolve_and_check_new_shape<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    size: Instance<'ctx, Int<SizeT>>,
+    new_ndims: Instance<'ctx, Int<SizeT>>,
+    new_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(
+        generator,
+        ctx,
+        "__nac3_ndarray_reshape_resolve_and_check_new_shape",
+    );
+    CallFunction::begin(generator, ctx, &name)
+        .arg(size)
+        .arg(new_ndims)
+        .arg(new_shape)
+        .returning_void();
+}
+
+pub fn call_nac3_ndarray_broadcast_to<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_broadcast_to");
+    CallFunction::begin(generator, ctx, &name).arg(src_ndarray).arg(dst_ndarray).returning_void();
+}
+
+pub fn call_nac3_ndarray_broadcast_shapes<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    num_shape_entries: Instance<'ctx, Int<SizeT>>,
+    shape_entries: Instance<'ctx, Ptr<Struct<ShapeEntry>>>,
+    dst_ndims: Instance<'ctx, Int<SizeT>>,
+    dst_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_broadcast_shapes");
+    CallFunction::begin(generator, ctx, &name)
+        .arg(num_shape_entries)
+        .arg(shape_entries)
+        .arg(dst_ndims)
+        .arg(dst_shape)
+        .returning_void();
+}
+
+pub fn call_nac3_ndarray_transpose<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    src_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    dst_ndarray: Instance<'ctx, Ptr<Struct<NDArray>>>,
+    num_axes: Instance<'ctx, Int<SizeT>>,
+    axes: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name = get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_transpose");
+    CallFunction::begin(generator, ctx, &name)
+        .arg(src_ndarray)
+        .arg(dst_ndarray)
+        .arg(num_axes)
+        .arg(axes)
+        .returning_void();
+}
+
+#[allow(clippy::too_many_arguments)]
+pub fn call_nac3_ndarray_matmul_calculate_shapes<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    a_ndims: Instance<'ctx, Int<SizeT>>,
+    a_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    b_ndims: Instance<'ctx, Int<SizeT>>,
+    b_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    final_ndims: Instance<'ctx, Int<SizeT>>,
+    new_a_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    new_b_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    dst_shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+) {
+    let name =
+        get_sizet_dependent_function_name(generator, ctx, "__nac3_ndarray_matmul_calculate_shapes");
+    CallFunction::begin(generator, ctx, &name)
+        .arg(a_ndims)
+        .arg(a_shape)
+        .arg(b_ndims)
+        .arg(b_shape)
+        .arg(final_ndims)
+        .arg(new_a_shape)
+        .arg(new_b_shape)
+        .arg(dst_shape)
+        .returning_void();
+}

nac3core/src/codegen/irrt/ndarray/basic.rs

@@ -1,250 +0,0 @@
-use inkwell::{
-    values::{BasicValueEnum, IntValue, PointerValue},
-    AddressSpace,
-};
-
-use crate::codegen::{
-    expr::{create_and_call_function, infer_and_call_function},
-    irrt::get_usize_dependent_function_name,
-    types::ProxyType,
-    values::{ndarray::NDArrayValue, ProxyValue},
-    CodeGenContext, CodeGenerator,
-};
-
-pub fn call_nac3_ndarray_util_assert_shape_no_negative<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndims: IntValue<'ctx>,
-    shape: PointerValue<'ctx>,
-) {
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    let name = get_usize_dependent_function_name(
-        generator,
-        ctx,
-        "__nac3_ndarray_util_assert_shape_no_negative",
-    );
-
-    create_and_call_function(
-        ctx,
-        &name,
-        Some(llvm_usize.into()),
-        &[(llvm_usize.into(), ndims.into()), (llvm_pusize.into(), shape.into())],
-        None,
-        None,
-    );
-}
-
-pub fn call_nac3_ndarray_util_assert_output_shape_same<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray_ndims: IntValue<'ctx>,
-    ndarray_shape: PointerValue<'ctx>,
-    output_ndims: IntValue<'ctx>,
-    output_shape: IntValue<'ctx>,
-) {
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    let name = get_usize_dependent_function_name(
-        generator,
-        ctx,
-        "__nac3_ndarray_util_assert_output_shape_same",
-    );
-
-    create_and_call_function(
-        ctx,
-        &name,
-        Some(llvm_usize.into()),
-        &[
-            (llvm_usize.into(), ndarray_ndims.into()),
-            (llvm_pusize.into(), ndarray_shape.into()),
-            (llvm_usize.into(), output_ndims.into()),
-            (llvm_pusize.into(), output_shape.into()),
-        ],
-        None,
-        None,
-    );
-}
-
-pub fn call_nac3_ndarray_size<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-) -> IntValue<'ctx> {
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_ndarray = ndarray.get_type().as_base_type();
-
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_size");
-
-    create_and_call_function(
-        ctx,
-        &name,
-        Some(llvm_usize.into()),
-        &[(llvm_ndarray.into(), ndarray.as_base_value().into())],
-        Some("size"),
-        None,
-    )
-    .map(BasicValueEnum::into_int_value)
-    .unwrap()
-}
-
-pub fn call_nac3_ndarray_nbytes<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-) -> IntValue<'ctx> {
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_ndarray = ndarray.get_type().as_base_type();
-
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_nbytes");
-
-    create_and_call_function(
-        ctx,
-        &name,
-        Some(llvm_usize.into()),
-        &[(llvm_ndarray.into(), ndarray.as_base_value().into())],
-        Some("nbytes"),
-        None,
-    )
-    .map(BasicValueEnum::into_int_value)
-    .unwrap()
-}
-
-pub fn call_nac3_ndarray_len<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-) -> IntValue<'ctx> {
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_ndarray = ndarray.get_type().as_base_type();
-
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_len");
-
-    create_and_call_function(
-        ctx,
-        &name,
-        Some(llvm_usize.into()),
-        &[(llvm_ndarray.into(), ndarray.as_base_value().into())],
-        Some("len"),
-        None,
-    )
-    .map(BasicValueEnum::into_int_value)
-    .unwrap()
-}
-
-pub fn call_nac3_ndarray_is_c_contiguous<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-) -> IntValue<'ctx> {
-    let llvm_i1 = ctx.ctx.bool_type();
-    let llvm_ndarray = ndarray.get_type().as_base_type();
-
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_is_c_contiguous");
-
-    create_and_call_function(
-        ctx,
-        &name,
-        Some(llvm_i1.into()),
-        &[(llvm_ndarray.into(), ndarray.as_base_value().into())],
-        Some("is_c_contiguous"),
-        None,
-    )
-    .map(BasicValueEnum::into_int_value)
-    .unwrap()
-}
-
-pub fn call_nac3_ndarray_get_nth_pelement<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-    index: IntValue<'ctx>,
-) -> PointerValue<'ctx> {
-    let llvm_i8 = ctx.ctx.i8_type();
-    let llvm_pi8 = llvm_i8.ptr_type(AddressSpace::default());
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_ndarray = ndarray.get_type().as_base_type();
-
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_get_nth_pelement");
-
-    create_and_call_function(
-        ctx,
-        &name,
-        Some(llvm_pi8.into()),
-        &[(llvm_ndarray.into(), ndarray.as_base_value().into()), (llvm_usize.into(), index.into())],
-        Some("pelement"),
-        None,
-    )
-    .map(BasicValueEnum::into_pointer_value)
-    .unwrap()
-}
-
-pub fn call_nac3_ndarray_get_pelement_by_indices<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-    indices: PointerValue<'ctx>,
-) -> PointerValue<'ctx> {
-    let llvm_i8 = ctx.ctx.i8_type();
-    let llvm_pi8 = llvm_i8.ptr_type(AddressSpace::default());
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-    let llvm_ndarray = ndarray.get_type().as_base_type();
-
-    let name =
-        get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_get_pelement_by_indices");
-
-    create_and_call_function(
-        ctx,
-        &name,
-        Some(llvm_pi8.into()),
-        &[
-            (llvm_ndarray.into(), ndarray.as_base_value().into()),
-            (llvm_pusize.into(), indices.into()),
-        ],
-        Some("pelement"),
-        None,
-    )
-    .map(BasicValueEnum::into_pointer_value)
-    .unwrap()
-}
-
-pub fn call_nac3_ndarray_set_strides_by_shape<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-) {
-    let llvm_ndarray = ndarray.get_type().as_base_type();
-
-    let name =
-        get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_set_strides_by_shape");
-
-    create_and_call_function(
-        ctx,
-        &name,
-        None,
-        &[(llvm_ndarray.into(), ndarray.as_base_value().into())],
-        None,
-        None,
-    );
-}
-
-pub fn call_nac3_ndarray_copy_data<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    src_ndarray: NDArrayValue<'ctx>,
-    dst_ndarray: NDArrayValue<'ctx>,
-) {
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_copy_data");
-
-    infer_and_call_function(
-        ctx,
-        &name,
-        None,
-        &[src_ndarray.as_base_value().into(), dst_ndarray.as_base_value().into()],
-        None,
-        None,
-    );
-}

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

@@ -1,29 +0,0 @@
-use crate::codegen::{
-    expr::infer_and_call_function,
-    irrt::get_usize_dependent_function_name,
-    values::{ndarray::NDArrayValue, ArrayLikeValue, ArraySliceValue, ProxyValue},
-    CodeGenContext, CodeGenerator,
-};
-
-pub fn call_nac3_ndarray_index<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    indices: ArraySliceValue<'ctx>,
-    src_ndarray: NDArrayValue<'ctx>,
-    dst_ndarray: NDArrayValue<'ctx>,
-) {
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_index");
-    infer_and_call_function(
-        ctx,
-        &name,
-        None,
-        &[
-            indices.size(ctx, generator).into(),
-            indices.base_ptr(ctx, generator).into(),
-            src_ndarray.as_base_value().into(),
-            dst_ndarray.as_base_value().into(),
-        ],
-        None,
-        None,
-    );
-}

nac3core/src/codegen/irrt/ndarray/iter.rs

@@ -1,70 +0,0 @@
-use inkwell::{
-    values::{BasicValueEnum, IntValue},
-    AddressSpace,
-};
-
-use crate::codegen::{
-    expr::{create_and_call_function, infer_and_call_function},
-    irrt::get_usize_dependent_function_name,
-    types::ProxyType,
-    values::{
-        ndarray::{NDArrayValue, NDIterValue},
-        ArrayLikeValue, ArraySliceValue, ProxyValue,
-    },
-    CodeGenContext, CodeGenerator,
-};
-
-pub fn call_nac3_nditer_initialize<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    iter: NDIterValue<'ctx>,
-    ndarray: NDArrayValue<'ctx>,
-    indices: ArraySliceValue<'ctx>,
-) {
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_nditer_initialize");
-
-    create_and_call_function(
-        ctx,
-        &name,
-        None,
-        &[
-            (iter.get_type().as_base_type().into(), iter.as_base_value().into()),
-            (ndarray.get_type().as_base_type().into(), ndarray.as_base_value().into()),
-            (llvm_pusize.into(), indices.base_ptr(ctx, generator).into()),
-        ],
-        None,
-        None,
-    );
-}
-
-pub fn call_nac3_nditer_has_element<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    iter: NDIterValue<'ctx>,
-) -> IntValue<'ctx> {
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_nditer_has_element");
-
-    infer_and_call_function(
-        ctx,
-        &name,
-        Some(ctx.ctx.bool_type().into()),
-        &[iter.as_base_value().into()],
-        None,
-        None,
-    )
-    .map(BasicValueEnum::into_int_value)
-    .unwrap()
-}
-
-pub fn call_nac3_nditer_next<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    iter: NDIterValue<'ctx>,
-) {
-    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_nditer_next");
-
-    infer_and_call_function(ctx, &name, None, &[iter.as_base_value().into()], None, None);
-}

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

@@ -1,391 +0,0 @@
-use inkwell::{
-    types::IntType,
-    values::{BasicValueEnum, CallSiteValue, IntValue},
-    AddressSpace, IntPredicate,
-};
-use itertools::Either;
-
-use crate::codegen::{
-    llvm_intrinsics,
-    macros::codegen_unreachable,
-    stmt::gen_for_callback_incrementing,
-    values::{
-        ndarray::NDArrayValue, ArrayLikeIndexer, ArrayLikeValue, ArraySliceValue,
-        TypedArrayLikeAccessor, TypedArrayLikeAdapter, UntypedArrayLikeAccessor,
-    },
-    CodeGenContext, CodeGenerator,
-};
-pub use basic::*;
-pub use indexing::*;
-pub use iter::*;
-
-mod basic;
-mod indexing;
-mod iter;
-
-/// Generates a call to `__nac3_ndarray_calc_size`. Returns an [`IntValue`] representing the
-/// calculated total size.
-///
-/// * `dims` - An [`ArrayLikeIndexer`] containing the size of each dimension.
-/// * `range` - The dimension index to begin and end (exclusively) calculating the dimensions for,
-///   or [`None`] if starting from the first dimension and ending at the last dimension
-///   respectively.
-pub fn call_ndarray_calc_size<'ctx, G, Dims>(
-    generator: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    dims: &Dims,
-    (begin, end): (Option<IntValue<'ctx>>, Option<IntValue<'ctx>>),
-) -> IntValue<'ctx>
-where
-    G: CodeGenerator + ?Sized,
-    Dims: ArrayLikeIndexer<'ctx>,
-{
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    let ndarray_calc_size_fn_name = match llvm_usize.get_bit_width() {
-        32 => "__nac3_ndarray_calc_size",
-        64 => "__nac3_ndarray_calc_size64",
-        bw => codegen_unreachable!(ctx, "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`]
-/// containing `i32` indices of the flattened index.
-///
-/// * `index` - The index to compute the multidimensional index for.
-/// * `ndarray` - LLVM pointer to the `NDArray`. This value must be the LLVM representation of an
-///   `NDArray`.
-pub fn call_ndarray_calc_nd_indices<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &G,
-    ctx: &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() {
-        32 => "__nac3_ndarray_calc_nd_indices",
-        64 => "__nac3_ndarray_calc_nd_indices64",
-        bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
-    };
-    let ndarray_calc_nd_indices_fn =
-        ctx.module.get_function(ndarray_calc_nd_indices_fn_name).unwrap_or_else(|| {
-            let fn_type = llvm_void.fn_type(
-                &[llvm_usize.into(), llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into()],
-                false,
-            );
-
-            ctx.module.add_function(ndarray_calc_nd_indices_fn_name, fn_type, None)
-        });
-
-    let ndarray_num_dims = ndarray.load_ndims(ctx);
-    let ndarray_dims = ndarray.shape();
-
-    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>(
-    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 => codegen_unreachable!(ctx, "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.shape();
-
-    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: &G,
-    ctx: &CodeGenContext<'ctx, '_>,
-    ndarray: NDArrayValue<'ctx>,
-    indices: &Index,
-) -> IntValue<'ctx>
-where
-    G: CodeGenerator + ?Sized,
-    Index: ArrayLikeIndexer<'ctx>,
-{
-    call_ndarray_flatten_index_impl(generator, ctx, ndarray, indices)
-}
-
-/// Generates a call to `__nac3_ndarray_calc_broadcast`. Returns a tuple containing the number of
-/// dimension and size of each dimension of the resultant `ndarray`.
-pub fn call_ndarray_calc_broadcast<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &mut G,
-    ctx: &mut CodeGenContext<'ctx, '_>,
-    lhs: NDArrayValue<'ctx>,
-    rhs: NDArrayValue<'ctx>,
-) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
-        32 => "__nac3_ndarray_calc_broadcast",
-        64 => "__nac3_ndarray_calc_broadcast64",
-        bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
-    };
-    let ndarray_calc_broadcast_fn =
-        ctx.module.get_function(ndarray_calc_broadcast_fn_name).unwrap_or_else(|| {
-            let fn_type = llvm_usize.fn_type(
-                &[
-                    llvm_pusize.into(),
-                    llvm_usize.into(),
-                    llvm_pusize.into(),
-                    llvm_usize.into(),
-                    llvm_pusize.into(),
-                ],
-                false,
-            );
-
-            ctx.module.add_function(ndarray_calc_broadcast_fn_name, fn_type, None)
-        });
-
-    let lhs_ndims = lhs.load_ndims(ctx);
-    let rhs_ndims = rhs.load_ndims(ctx);
-    let min_ndims = llvm_intrinsics::call_int_umin(ctx, lhs_ndims, rhs_ndims, None);
-
-    gen_for_callback_incrementing(
-        generator,
-        ctx,
-        None,
-        llvm_usize.const_zero(),
-        (min_ndims, false),
-        |generator, ctx, _, idx| {
-            let idx = ctx.builder.build_int_sub(min_ndims, idx, "").unwrap();
-            let (lhs_dim_sz, rhs_dim_sz) = unsafe {
-                (
-                    lhs.shape().get_typed_unchecked(ctx, generator, &idx, None),
-                    rhs.shape().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.shape().base_ptr(ctx, generator);
-    let lhs_ndims = lhs.load_ndims(ctx);
-    let rhs_dims = rhs.shape().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`]
-/// 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>,
-    broadcast_idx: &BroadcastIdx,
-) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
-    let llvm_i32 = ctx.ctx.i32_type();
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
-    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
-
-    let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
-        32 => "__nac3_ndarray_calc_broadcast_idx",
-        64 => "__nac3_ndarray_calc_broadcast_idx64",
-        bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
-    };
-    let ndarray_calc_broadcast_fn =
-        ctx.module.get_function(ndarray_calc_broadcast_fn_name).unwrap_or_else(|| {
-            let fn_type = llvm_usize.fn_type(
-                &[llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into(), llvm_pi32.into()],
-                false,
-            );
-
-            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.shape().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()),
-    )
-}

nac3core/src/codegen/irrt/range.rs

@@ -1,42 +0,0 @@
-use inkwell::{
-    values::{BasicValueEnum, CallSiteValue, IntValue},
-    IntPredicate,
-};
-use itertools::Either;
-
-use crate::codegen::{CodeGenContext, CodeGenerator};
-
-pub fn calculate_len_for_slice_range<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &mut G,
-    ctx: &mut CodeGenContext<'ctx, '_>,
-    start: IntValue<'ctx>,
-    end: IntValue<'ctx>,
-    step: IntValue<'ctx>,
-) -> IntValue<'ctx> {
-    const SYMBOL: &str = "__nac3_range_slice_len";
-    let len_func = ctx.module.get_function(SYMBOL).unwrap_or_else(|| {
-        let i32_t = ctx.ctx.i32_type();
-        let fn_t = i32_t.fn_type(&[i32_t.into(), i32_t.into(), i32_t.into()], false);
-        ctx.module.add_function(SYMBOL, fn_t, None)
-    });
-
-    // assert step != 0, throw exception if not
-    let not_zero = ctx
-        .builder
-        .build_int_compare(IntPredicate::NE, step, step.get_type().const_zero(), "range_step_ne")
-        .unwrap();
-    ctx.make_assert(
-        generator,
-        not_zero,
-        "0:ValueError",
-        "step must not be zero",
-        [None, None, None],
-        ctx.current_loc,
-    );
-    ctx.builder
-        .build_call(len_func, &[start.into(), end.into(), step.into()], "calc_len")
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_int_value))
-        .map(Either::unwrap_left)
-        .unwrap()
-}

nac3core/src/codegen/irrt/slice.rs

@@ -1,39 +0,0 @@
-use inkwell::values::{BasicValueEnum, CallSiteValue, IntValue};
-use itertools::Either;
-
-use nac3parser::ast::Expr;
-
-use crate::{
-    codegen::{CodeGenContext, CodeGenerator},
-    typecheck::typedef::Type,
-};
-
-/// this function allows index out of range, since python
-/// allows index out of range in slice (`a = [1,2,3]; a[1:10] == [2,3]`).
-pub fn handle_slice_index_bound<'ctx, G: CodeGenerator>(
-    i: &Expr<Option<Type>>,
-    ctx: &mut CodeGenContext<'ctx, '_>,
-    generator: &mut G,
-    length: IntValue<'ctx>,
-) -> Result<Option<IntValue<'ctx>>, String> {
-    const SYMBOL: &str = "__nac3_slice_index_bound";
-    let func = ctx.module.get_function(SYMBOL).unwrap_or_else(|| {
-        let i32_t = ctx.ctx.i32_type();
-        let fn_t = i32_t.fn_type(&[i32_t.into(), i32_t.into()], false);
-        ctx.module.add_function(SYMBOL, fn_t, None)
-    });
-
-    let i = if let Some(v) = generator.gen_expr(ctx, i)? {
-        v.to_basic_value_enum(ctx, generator, i.custom.unwrap())?
-    } else {
-        return Ok(None);
-    };
-    Ok(Some(
-        ctx.builder
-            .build_call(func, &[i.into(), length.into()], "bounded_ind")
-            .map(CallSiteValue::try_as_basic_value)
-            .map(|v| v.map_left(BasicValueEnum::into_int_value))
-            .map(Either::unwrap_left)
-            .unwrap(),
-    ))
-}

nac3core/src/codegen/llvm_intrinsics.rs

@@ -1,14 +1,12 @@
-use inkwell::{
-    context::Context,
-    intrinsics::Intrinsic,
-    types::{AnyTypeEnum::IntType, FloatType},
-    values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue, PointerValue},
-    AddressSpace,
-};
+use crate::codegen::CodeGenContext;
+use inkwell::context::Context;
+use inkwell::intrinsics::Intrinsic;
+use inkwell::types::AnyTypeEnum::IntType;
+use inkwell::types::FloatType;
+use inkwell::values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue, PointerValue};
+use inkwell::AddressSpace;
 use itertools::Either;
 
-use super::CodeGenContext;
-
 /// Returns the string representation for the floating-point type `ft` when used in intrinsic
 /// functions.
 fn get_float_intrinsic_repr(ctx: &Context, ft: FloatType) -> &'static str {
@@ -185,7 +183,7 @@ pub fn call_memcpy_generic<'ctx>(
         dest
     } else {
         ctx.builder
-            .build_bit_cast(dest, llvm_p0i8, "")
+            .build_bitcast(dest, llvm_p0i8, "")
             .map(BasicValueEnum::into_pointer_value)
             .unwrap()
     };
@@ -193,7 +191,7 @@ pub fn call_memcpy_generic<'ctx>(
         src
     } else {
         ctx.builder
-            .build_bit_cast(src, llvm_p0i8, "")
+            .build_bitcast(src, llvm_p0i8, "")
             .map(BasicValueEnum::into_pointer_value)
             .unwrap()
     };
@@ -201,49 +199,6 @@ pub fn call_memcpy_generic<'ctx>(
     call_memcpy(ctx, dest, src, len, is_volatile);
 }
 
-/// Invokes the `llvm.memcpy` intrinsic.
-///
-/// Unlike [`call_memcpy`], this function accepts any type of pointer value. If `dest` or `src` is
-/// not a pointer to an integer, the pointer(s) will be cast to `i8*` before invoking `memcpy`.
-/// Moreover, `len` now refers to the number of elements to copy (rather than number of bytes to
-/// copy).
-pub fn call_memcpy_generic_array<'ctx>(
-    ctx: &CodeGenContext<'ctx, '_>,
-    dest: PointerValue<'ctx>,
-    src: PointerValue<'ctx>,
-    len: IntValue<'ctx>,
-    is_volatile: IntValue<'ctx>,
-) {
-    let llvm_i8 = ctx.ctx.i8_type();
-    let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
-    let llvm_sizeof_expr_t = llvm_i8.size_of().get_type();
-
-    let dest_elem_t = dest.get_type().get_element_type();
-    let src_elem_t = src.get_type().get_element_type();
-
-    let dest = if matches!(dest_elem_t, IntType(t) if t.get_bit_width() == 8) {
-        dest
-    } else {
-        ctx.builder
-            .build_bit_cast(dest, llvm_p0i8, "")
-            .map(BasicValueEnum::into_pointer_value)
-            .unwrap()
-    };
-    let src = if matches!(src_elem_t, IntType(t) if t.get_bit_width() == 8) {
-        src
-    } else {
-        ctx.builder
-            .build_bit_cast(src, llvm_p0i8, "")
-            .map(BasicValueEnum::into_pointer_value)
-            .unwrap()
-    };
-
-    let len = ctx.builder.build_int_z_extend_or_bit_cast(len, llvm_sizeof_expr_t, "").unwrap();
-    let len = ctx.builder.build_int_mul(len, src_elem_t.size_of().unwrap(), "").unwrap();
-
-    call_memcpy(ctx, dest, src, len, is_volatile);
-}
-
 /// Macro to find and generate build call for llvm intrinsic (body of llvm intrinsic function)
 ///
 /// Arguments:
@@ -386,25 +341,3 @@ pub fn call_float_powi<'ctx>(
         .map(Either::unwrap_left)
         .unwrap()
 }
-
-/// Invokes the [`llvm.ctpop`](https://llvm.org/docs/LangRef.html#llvm-ctpop-intrinsic) intrinsic.
-pub fn call_int_ctpop<'ctx>(
-    ctx: &CodeGenContext<'ctx, '_>,
-    src: IntValue<'ctx>,
-    name: Option<&str>,
-) -> IntValue<'ctx> {
-    const FN_NAME: &str = "llvm.ctpop";
-
-    let llvm_src_t = src.get_type();
-
-    let intrinsic_fn = Intrinsic::find(FN_NAME)
-        .and_then(|intrinsic| intrinsic.get_declaration(&ctx.module, &[llvm_src_t.into()]))
-        .unwrap();
-
-    ctx.builder
-        .build_call(intrinsic_fn, &[src.into()], name.unwrap_or_default())
-        .map(CallSiteValue::try_as_basic_value)
-        .map(|v| v.map_left(BasicValueEnum::into_int_value))
-        .map(Either::unwrap_left)
-        .unwrap()
-}

nac3core/src/codegen/mod.rs

@@ -1,12 +1,12 @@
-use std::{
-    collections::{HashMap, HashSet},
-    sync::{
-        atomic::{AtomicBool, Ordering},
-        Arc,
+use crate::{
+    codegen::classes::{ListType, ProxyType, RangeType},
+    symbol_resolver::{StaticValue, SymbolResolver},
+    toplevel::{helper::PrimDef, TopLevelContext, TopLevelDef},
+    typecheck::{
+        type_inferencer::{CodeLocation, PrimitiveStore},
+        typedef::{CallId, FuncArg, Type, TypeEnum, Unifier},
     },
-    thread,
 };
-
 use crossbeam::channel::{unbounded, Receiver, Sender};
 use inkwell::{
     attributes::{Attribute, AttributeLoc},
@@ -24,56 +24,35 @@ use inkwell::{
     AddressSpace, IntPredicate, OptimizationLevel,
 };
 use itertools::Itertools;
-use parking_lot::{Condvar, Mutex};
-
+use model::*;
 use nac3parser::ast::{Location, Stmt, StrRef};
-
-use crate::{
-    symbol_resolver::{StaticValue, SymbolResolver},
-    toplevel::{
-        helper::{extract_ndims, PrimDef},
-        numpy::unpack_ndarray_var_tys,
-        TopLevelContext, TopLevelDef,
-    },
-    typecheck::{
-        type_inferencer::{CodeLocation, PrimitiveStore},
-        typedef::{CallId, FuncArg, Type, TypeEnum, Unifier},
-    },
+use object::ndarray::NDArray;
+use parking_lot::{Condvar, Mutex};
+use std::collections::{HashMap, HashSet};
+use std::sync::{
+    atomic::{AtomicBool, Ordering},
+    Arc,
 };
-use concrete_type::{ConcreteType, ConcreteTypeEnum, ConcreteTypeStore};
-pub use generator::{CodeGenerator, DefaultCodeGenerator};
-use types::{ndarray::NDArrayType, ListType, ProxyType, RangeType};
+use std::thread;
 
 pub mod builtin_fns;
+pub mod classes;
 pub mod concrete_type;
 pub mod expr;
 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;
-pub mod types;
-pub mod values;
 
 #[cfg(test)]
 mod test;
 
-mod macros {
-    /// Codegen-variant of [`std::unreachable`] which accepts an instance of [`CodeGenContext`] as
-    /// its first argument to provide Python source information to indicate the codegen location
-    /// causing the assertion.
-    macro_rules! codegen_unreachable {
-        ($ctx:expr $(,)?) => {
-            std::unreachable!("unreachable code while processing {}", &$ctx.current_loc)
-        };
-        ($ctx:expr, $($arg:tt)*) => {
-            std::unreachable!("unreachable code while processing {}: {}", &$ctx.current_loc, std::format!("{}", std::format_args!($($arg)+)))
-        };
-    }
-
-    pub(crate) use codegen_unreachable;
-}
+use concrete_type::{ConcreteType, ConcreteTypeEnum, ConcreteTypeStore};
+pub use generator::{CodeGenerator, DefaultCodeGenerator};
 
 #[derive(Default)]
 pub struct StaticValueStore {
@@ -514,13 +493,7 @@ fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
                         }
 
                         TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
-                            let (dtype, ndims) = unpack_ndarray_var_tys(unifier, ty);
-                            let ndims = extract_ndims(unifier, ndims);
-                            let element_type = get_llvm_type(
-                                ctx, module, generator, unifier, top_level, type_cache, dtype,
-                            );
-
-                            NDArrayType::new(generator, ctx, element_type, Some(ndims)).as_base_type().into()
+                            Ptr(Struct(NDArray)).get_type(generator, ctx).as_basic_type_enum()
                         }
 
                         _ => unreachable!(
@@ -858,9 +831,10 @@ pub fn gen_func_impl<
     builder.position_at_end(init_bb);
     let body_bb = context.append_basic_block(fn_val, "body");
 
-    // Store non-vararg argument values into local variables
     let mut var_assignment = HashMap::new();
     let offset = u32::from(has_sret);
+
+    // 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(
@@ -1124,106 +1098,3 @@ fn gen_in_range_check<'ctx>(
 fn get_va_count_arg_name(arg_name: StrRef) -> StrRef {
     format!("__{}_va_count", &arg_name).into()
 }
-
-/// Returns the alignment of the type.
-///
-/// This is necessary as `get_alignment` is not implemented as part of [`BasicType`].
-pub fn get_type_alignment<'ctx>(ty: impl Into<BasicTypeEnum<'ctx>>) -> IntValue<'ctx> {
-    match ty.into() {
-        BasicTypeEnum::ArrayType(ty) => ty.get_alignment(),
-        BasicTypeEnum::FloatType(ty) => ty.get_alignment(),
-        BasicTypeEnum::IntType(ty) => ty.get_alignment(),
-        BasicTypeEnum::PointerType(ty) => ty.get_alignment(),
-        BasicTypeEnum::StructType(ty) => ty.get_alignment(),
-        BasicTypeEnum::VectorType(ty) => ty.get_alignment(),
-    }
-}
-
-/// Inserts an `alloca` instruction with allocation `size` given in bytes and the alignment of the
-/// given type.
-///
-/// The returned [`PointerValue`] will have a type of `i8*`, a size of at least `size`, and will be
-/// aligned with the alignment of `align_ty`.
-pub fn type_aligned_alloca<'ctx, G: CodeGenerator + ?Sized>(
-    generator: &mut G,
-    ctx: &mut CodeGenContext<'ctx, '_>,
-    align_ty: impl Into<BasicTypeEnum<'ctx>>,
-    size: IntValue<'ctx>,
-    name: Option<&str>,
-) -> PointerValue<'ctx> {
-    /// Round `val` up to its modulo `power_of_two`.
-    fn round_up<'ctx>(
-        ctx: &CodeGenContext<'ctx, '_>,
-        val: IntValue<'ctx>,
-        power_of_two: IntValue<'ctx>,
-    ) -> IntValue<'ctx> {
-        debug_assert_eq!(
-            val.get_type().get_bit_width(),
-            power_of_two.get_type().get_bit_width(),
-            "`val` ({}) and `power_of_two` ({}) must be the same type",
-            val.get_type(),
-            power_of_two.get_type(),
-        );
-
-        let llvm_val_t = val.get_type();
-
-        let max_rem =
-            ctx.builder.build_int_sub(power_of_two, llvm_val_t.const_int(1, false), "").unwrap();
-        ctx.builder
-            .build_and(
-                ctx.builder.build_int_add(val, max_rem, "").unwrap(),
-                ctx.builder.build_not(max_rem, "").unwrap(),
-                "",
-            )
-            .unwrap()
-    }
-
-    let llvm_i8 = ctx.ctx.i8_type();
-    let llvm_pi8 = llvm_i8.ptr_type(AddressSpace::default());
-    let llvm_usize = generator.get_size_type(ctx.ctx);
-    let align_ty = align_ty.into();
-
-    let size = ctx.builder.build_int_truncate_or_bit_cast(size, llvm_usize, "").unwrap();
-
-    debug_assert_eq!(
-        size.get_type().get_bit_width(),
-        llvm_usize.get_bit_width(),
-        "Expected size_t ({}) for parameter `size` of `aligned_alloca`, got {}",
-        llvm_usize,
-        size.get_type(),
-    );
-
-    let alignment = get_type_alignment(align_ty);
-    let alignment = ctx.builder.build_int_truncate_or_bit_cast(alignment, llvm_usize, "").unwrap();
-
-    if ctx.registry.llvm_options.opt_level == OptimizationLevel::None {
-        let alignment_bitcount = llvm_intrinsics::call_int_ctpop(ctx, alignment, None);
-
-        ctx.make_assert(
-            generator,
-            ctx.builder
-                .build_int_compare(
-                    IntPredicate::EQ,
-                    alignment_bitcount,
-                    alignment_bitcount.get_type().const_int(1, false),
-                    "",
-                )
-                .unwrap(),
-            "0:AssertionError",
-            "Expected power-of-two alignment for aligned_alloca, got {0}",
-            [Some(alignment), None, None],
-            ctx.current_loc,
-        );
-    }
-
-    let buffer_size = round_up(ctx, size, alignment);
-    let aligned_slices = ctx.builder.build_int_unsigned_div(buffer_size, alignment, "").unwrap();
-
-    // Just to be absolutely sure, alloca in [i8 x alignment] slices
-    let buffer = ctx.builder.build_array_alloca(align_ty, aligned_slices, "").unwrap();
-
-    ctx.builder
-        .build_bit_cast(buffer, llvm_pi8, name.unwrap_or_default())
-        .map(BasicValueEnum::into_pointer_value)
-        .unwrap()
-}

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 it is infeasible to use model abstractions.
+#[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 +1,143 @@
+use std::fmt;
+
+use inkwell::{
+    context::Context,
+    types::{ArrayType, BasicType, BasicTypeEnum},
+    values::{ArrayValue, IntValue},
+};
+
+use crate::codegen::{CodeGenContext, CodeGenerator};
+
+use super::*;
+
+/// Trait for Rust structs identifying length values 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`].
+///
+/// `Len` should be of a [`LenKind`] and `Item` should be a of [`Model`].
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Array<Len, Item> {
+    /// Length of this array.
+    pub len: Len,
+    /// [`Model`] of the array items.
+    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 +1,202 @@
+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 [`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 their types, it is possible
+/// to tell the [`BasicType`]s 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, for example, 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.
+    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.
+    ///
+    /// 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 the [`BasicValue`] 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.
+    ///
+    /// It is guaranteed the [`BasicType`] of `value` is consistent with that of `model`.
+    pub value: M::Value,
+}

nac3core/src/codegen/model/float.rs

@@ -0,0 +1,90 @@
+use std::fmt;
+
+use inkwell::{
+    context::Context,
+    types::{BasicType, 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: 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 +1,122 @@
+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.
+///
+/// ### Usage
+///
+/// The syntax is like this:
+/// ```ignore
+/// let result = CallFunction::begin("my_function_name")
+///   .attrs(...)
+///   .arg(arg1)
+///   .arg(arg2)
+///   .arg(arg3)
+///   .returning("my_function_result", Int32);
+/// ```
+///
+/// The function `my_function_name` is called when `.returning()` (or its variants) is called, returning
+/// the result as an `Instance<'ctx, Int<Int32>>`.
+///
+/// If `my_function_name` has not been declared in `ctx.module`, once `.returning()` is called, a function
+/// declaration of `my_function_name` is added to `ctx.module`, where the [`FunctionType`] is deduced from
+/// the argument types and returning type.
+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 +1,219 @@
+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`].
+///
+/// `Item` is the element type this pointer is pointing to, and should be of a [`Model`].
+///
+// 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)
+    }
+
+    /// Cast this pointer to `uint8_t*`
+    pub fn cast_to_pi8<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Ptr<Int<Byte>>> {
+        Ptr(Int(Byte)).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 +1,359 @@
+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 calculate 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 field types and construct the LLVM struct type with [`Context::struct_type`].
+struct TypeFieldTraversal<'ctx, 'a, G: CodeGenerator + ?Sized> {
+    generator: &'a G,
+    ctx: &'ctx Context,
+    /// The collected field types so far in exact order.
+    field_types: Vec<BasicTypeEnum<'ctx>>,
+}
+
+impl<'ctx, 'a, G: CodeGenerator + ?Sized> FieldTraversal<'ctx> for TypeFieldTraversal<'ctx, 'a, G> {
+    type Out<M> = (); // Checking types return nothing.
+
+    fn add<M: Model<'ctx>>(&mut self, _name: &'static str, model: M) -> Self::Out<M> {
+        let t = model.get_type(self.generator, self.ctx).as_basic_type_enum();
+        self.field_types.push(t);
+    }
+}
+
+/// A traversal to check the types of fields.
+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.
+    gep_index_counter: 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 gep_index = self.gep_index_counter;
+        self.gep_index_counter += 1;
+
+        if let Some(t) = self.scrutinee.get_field_type_at_index(gep_index) {
+            if let Err(err) = model.check_type(self.generator, self.ctx, t) {
+                self.errors
+                    .push(err.under_context(format!("field #{gep_index} '{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:
+/// ```ignore
+/// type F64NDArray = Struct<ContiguousNDArray<Float<Float64>>>; // Type alias for leaner documentation
+/// let model: F64NDArray = Struct(ContigousNDArray { item: Float(Float64) });
+/// 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)
+///
+/// ```ignore
+/// // 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 through all fields of this [`StructKind`].
+    ///
+    /// Only used internally in this module for implementing other components.
+    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.
+    ///
+    /// Only used internally in this module for implementing other components.
+    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)
+    }
+}
+
+/// A model for LLVM struct.
+///
+/// `S` should be of a [`StructKind`].
+#[derive(Debug, Clone, Copy, Default)]
+pub struct Struct<S>(pub S);
+
+impl<'ctx, S: StructKind<'ctx>> Struct<S> {
+    /// Create a constant struct value from its fields.
+    ///
+    /// This function also validates `fields` and panic when there is something wrong.
+    pub fn const_struct<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        fields: &[BasicValueEnum<'ctx>],
+    ) -> Instance<'ctx, Self> {
+        // NOTE: There *could* have been a functor `F<M> = Instance<'ctx, M>` for `S::Fields<F>`
+        // to create a more user-friendly interface, but Rust's type system is not sophisticated enough
+        // and if you try doing that Rust would force you put lifetimes everywhere.
+        let val = ctx.const_struct(fields, false);
+        self.check_value(generator, ctx, val).unwrap()
+    }
+}
+
+impl<'ctx, S: StructKind<'ctx>> Model<'ctx> for Struct<S> {
+    type Value = StructValue<'ctx>;
+    type Type = StructType<'ctx>;
+
+    fn get_type<G: CodeGenerator + ?Sized>(&self, generator: &G, ctx: &'ctx Context) -> Self::Type {
+        self.0.get_struct_type(generator, ctx)
+    }
+
+    fn check_type<T: BasicType<'ctx>, G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &'ctx Context,
+        ty: T,
+    ) -> Result<(), ModelError> {
+        let ty = ty.as_basic_type_enum();
+        let Ok(ty) = StructType::try_from(ty) else {
+            return Err(ModelError(format!("Expecting StructType, but got {ty:?}")));
+        };
+
+        // Check each field individually.
+        let mut traversal = CheckTypeFieldTraversal {
+            generator,
+            ctx,
+            gep_index_counter: 0,
+            errors: Vec::new(),
+            scrutinee: ty,
+        };
+        self.0.traverse_fields(&mut traversal);
+
+        // Check the number of fields.
+        let exp_num_fields = traversal.gep_index_counter;
+        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 +1,42 @@
+use crate::codegen::{
+    stmt::{gen_for_callback_incrementing, BreakContinueHooks},
+    CodeGenContext, CodeGenerator,
+};
+
+use super::*;
+
+/// Like [`gen_for_callback_incrementing`] with [`Model`] abstractions.
+///
+/// `stop` is not included.
+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;
+
+/// A NAC3 LLVM Python object of any type.
+#[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/list.rs

@@ -0,0 +1,87 @@
+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"),
+        }
+    }
+}
+
+impl<'ctx, Item: Model<'ctx>> Instance<'ctx, Ptr<Struct<List<Item>>>> {
+    /// Cast the items pointer to `uint8_t*`.
+    pub fn with_pi8_items<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> Instance<'ctx, Ptr<Struct<List<Int<Byte>>>>> {
+        self.pointer_cast(generator, ctx, Struct(List { item: Int(Byte) }))
+    }
+}
+
+/// 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)
+    }
+}

nac3core/src/codegen/object/mod.rs

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

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

@@ -0,0 +1,184 @@
+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},
+};
+
+/// Get the expected `dtype` and `ndims` of the ndarray returned by `np_array(list)`.
+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> {
+    /// Implementation of `np_array(<list>, copy=True)`
+    fn make_np_array_list_copy_true_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.instance.with_pi8_items(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
+    }
+
+    /// Implementation of `np_array(<list>, copy=None)`
+    fn make_np_array_list_copy_none_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.instance.get(generator, ctx, |f| f.items).cast_to_pi8(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_true_impl(generator, ctx, list)
+        }
+    }
+
+    /// Implementation of `np_array(<list>, copy=copy)`
+    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_true_impl(generator, ctx, list);
+                Ok(Some(ndarray.instance.value))
+            },
+            |generator, ctx| {
+                let ndarray =
+                    NDArrayObject::make_np_array_list_copy_none_impl(generator, ctx, list);
+                Ok(Some(ndarray.instance.value))
+            },
+        )
+        .unwrap()
+        .unwrap();
+
+        NDArrayObject::from_value_and_unpacked_types(generator, ctx, ndarray, dtype, ndims)
+    }
+
+    /// Implementation of `np_array(<ndarray>, copy=copy)`.
+    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, "").into()
+    } else {
+        panic!("unrecognized dtype: {}", ctx.unifier.stringify(dtype));
+    }
+}
+
+/// Get the one value in `np.ones()` of a `dtype`.
+fn ndarray_one_value<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    dtype: Type,
+) -> BasicValueEnum<'ctx> {
+    if [ctx.primitives.int32, ctx.primitives.uint32]
+        .iter()
+        .any(|ty| ctx.unifier.unioned(dtype, *ty))
+    {
+        let is_signed = ctx.unifier.unioned(dtype, ctx.primitives.int32);
+        ctx.ctx.i32_type().const_int(1, is_signed).into()
+    } else if [ctx.primitives.int64, ctx.primitives.uint64]
+        .iter()
+        .any(|ty| ctx.unifier.unioned(dtype, *ty))
+    {
+        let is_signed = ctx.unifier.unioned(dtype, ctx.primitives.int64);
+        ctx.ctx.i64_type().const_int(1, is_signed).into()
+    } else if ctx.unifier.unioned(dtype, ctx.primitives.float) {
+        ctx.ctx.f64_type().const_float(1.0).into()
+    } else if ctx.unifier.unioned(dtype, ctx.primitives.bool) {
+        ctx.ctx.bool_type().const_int(1, false).into()
+    } else if ctx.unifier.unioned(dtype, ctx.primitives.str) {
+        ctx.gen_string(generator, "1").into()
+    } else {
+        panic!("unrecognized dtype: {}", ctx.unifier.stringify(dtype));
+    }
+}
+
+impl<'ctx> NDArrayObject<'ctx> {
+    /// Create an ndarray like `np.empty`.
+    pub fn make_np_empty<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        dtype: Type,
+        ndims: u64,
+        shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    ) -> Self {
+        // Validate `shape`
+        let ndims_llvm = Int(SizeT).const_int(generator, ctx.ctx, ndims);
+        call_nac3_ndarray_util_assert_shape_no_negative(generator, ctx, ndims_llvm, shape);
+
+        let ndarray = NDArrayObject::alloca(generator, ctx, dtype, ndims);
+        ndarray.copy_shape_from_array(generator, ctx, shape);
+        ndarray.create_data(generator, ctx);
+
+        ndarray
+    }
+
+    /// Create an ndarray like `np.full`.
+    pub fn make_np_full<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        dtype: Type,
+        ndims: u64,
+        shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+        fill_value: BasicValueEnum<'ctx>,
+    ) -> Self {
+        let ndarray = NDArrayObject::make_np_empty(generator, ctx, dtype, ndims, shape);
+        ndarray.fill(generator, ctx, fill_value);
+        ndarray
+    }
+
+    /// Create an ndarray like `np.zero`.
+    pub fn make_np_zeros<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        dtype: Type,
+        ndims: u64,
+        shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    ) -> Self {
+        let fill_value = ndarray_zero_value(generator, ctx, dtype);
+        NDArrayObject::make_np_full(generator, ctx, dtype, ndims, shape, fill_value)
+    }
+
+    /// Create an ndarray like `np.ones`.
+    pub fn make_np_ones<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        dtype: Type,
+        ndims: u64,
+        shape: Instance<'ctx, Ptr<Int<SizeT>>>,
+    ) -> Self {
+        let fill_value = ndarray_one_value(generator, ctx, dtype);
+        NDArrayObject::make_np_full(generator, ctx, dtype, ndims, shape, fill_value)
+    }
+
+    /// Create an ndarray like `np.eye`.
+    pub fn make_np_eye<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        dtype: Type,
+        nrows: Instance<'ctx, Int<SizeT>>,
+        ncols: Instance<'ctx, Int<SizeT>>,
+        offset: Instance<'ctx, Int<SizeT>>,
+    ) -> Self {
+        let ndzero = ndarray_zero_value(generator, ctx, dtype);
+        let ndone = ndarray_one_value(generator, ctx, dtype);
+
+        let ndarray = NDArrayObject::alloca_dynamic_shape(generator, ctx, dtype, &[nrows, ncols]);
+
+        // Create data and make the matrix like look np.eye()
+        ndarray.create_data(generator, ctx);
+        ndarray
+            .foreach(generator, ctx, |generator, ctx, _hooks, nditer| {
+                // NOTE: rows and cols can never be zero here, since this ndarray's `np.size` would be zero
+                // and this loop would not execute.
+
+                // Load up `row_i` and `col_i` from indices.
+                let row_i = nditer.get_indices().get_index_const(generator, ctx, 0);
+                let col_i = nditer.get_indices().get_index_const(generator, ctx, 1);
+
+                let be_one = row_i.add(ctx, offset).compare(ctx, IntPredicate::EQ, col_i);
+                let value = ctx.builder.build_select(be_one.value, ndone, ndzero, "value").unwrap();
+
+                let p = nditer.get_pointer(generator, ctx);
+                ctx.builder.build_store(p, value).unwrap();
+
+                Ok(())
+            })
+            .unwrap();
+
+        ndarray
+    }
+
+    /// Create an ndarray like `np.identity`.
+    pub fn make_np_identity<G: CodeGenerator + ?Sized>(
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        dtype: Type,
+        size: Instance<'ctx, Int<SizeT>>,
+    ) -> Self {
+        // Convenient implementation
+        let offset = Int(SizeT).const_0(generator, ctx.ctx);
+        NDArrayObject::make_np_eye(generator, ctx, dtype, size, size, offset)
+    }
+}

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

@@ -0,0 +1,226 @@
+use crate::codegen::{
+    irrt::call_nac3_ndarray_index,
+    model::*,
+    object::utils::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 representing a [`NDIndex`].
+#[derive(Debug, Clone)]
+pub enum RustNDIndex<'ctx> {
+    SingleElement(Instance<'ctx, Int<Int32>>),
+    Slice(RustSlice<'ctx, Int32>),
+    NewAxis,
+    Ellipsis,
+}
+
+impl<'ctx> RustNDIndex<'ctx> {
+    /// Get the value to set `NDIndex::type` for this variant.
+    fn get_type_id(&self) -> u64 {
+        // Defined in IRRT, must be in sync
+        match self {
+            RustNDIndex::SingleElement(_) => 0,
+            RustNDIndex::Slice(_) => 1,
+            RustNDIndex::NewAxis => 2,
+            RustNDIndex::Ellipsis => 3,
+        }
+    }
+
+    /// Write the contents to an LLVM [`NDIndex`].
+    fn write_to_ndindex<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &CodeGenContext<'ctx, '_>,
+        dst_ndindex_ptr: Instance<'ctx, Ptr<Struct<NDIndex>>>,
+    ) {
+        // Set `dst_ndindex_ptr->type`
+        dst_ndindex_ptr.gep(ctx, |f| f.type_).store(
+            ctx,
+            Int(NDIndexType::default()).const_int(generator, ctx.ctx, self.get_type_id()),
+        );
+
+        // Set `dst_ndindex_ptr->data`
+        match self {
+            RustNDIndex::SingleElement(in_index) => {
+                let index_ptr = Int(Int32).alloca(generator, ctx);
+                index_ptr.store(ctx, *in_index);
+
+                dst_ndindex_ptr
+                    .gep(ctx, |f| f.data)
+                    .store(ctx, index_ptr.pointer_cast(generator, ctx, Int(Byte)));
+            }
+            RustNDIndex::Slice(in_rust_slice) => {
+                let user_slice_ptr = Struct(Slice(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 the array pointer.
+    pub fn make_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 expected `ndims` after indexing with `indices`.
+    #[must_use]
+    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::make_ndindices(generator, ctx, indices);
+        call_nac3_ndarray_index(
+            generator,
+            ctx,
+            num_indices,
+            indices,
+            self.instance,
+            dst_ndarray.instance,
+        );
+
+        dst_ndarray
+    }
+}
+
+pub mod util {
+    use itertools::Itertools;
+    use nac3parser::ast::{Expr, ExprKind};
+
+    use crate::{
+        codegen::{
+            expr::gen_slice, model::*, object::utils::slice::RustSlice, 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 (lower, upper, step) = gen_slice(generator, ctx, lower, upper, step)?;
+                RustNDIndex::Slice(RustSlice { int_kind: Int32, start: lower, stop: upper, step })
+            } else {
+                // Treat and handle everything else as a single element index.
+                let index = generator.gen_expr(ctx, index_expr)?.unwrap().to_basic_value_enum(
+                    ctx,
+                    generator,
+                    ctx.primitives.int32, // Must be int32, this checks for illegal values
+                )?;
+                let index = Int(Int32).check_value(generator, ctx.ctx, index).unwrap();
+
+                RustNDIndex::SingleElement(index)
+            };
+            rust_ndindices.push(ndindex);
+        }
+        Ok(rust_ndindices)
+    }
+}

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 +1,668 @@
+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 onto the stack.
+    ///
+    /// The returned ndarray's content will be:
+    /// - `data`: uninitialized.
+    /// - `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 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 `<ndarray>.strides`.
+    ///
+    /// 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 +1,177 @@
+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 with a convenient interface to interact with [`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 }
+    }
+
+    /// Is there a next element?
+    ///
+    /// If `ndarray` is unsized, this returns true only for the first iteration.
+    /// If `ndarray` is 0-sized, this always returns false.
+    #[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)
+    }
+
+    /// Go to the next element. If `has_next()` is false, then this has undefined behavior.
+    ///
+    /// If `ndarray` is unsized, this can only be called once.
+    /// If `ndarray` is 0-sized, this can never be called.
+    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 if this ndarray were a flat ndarray.
+    #[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` has access to [`BreakContinueHooks`] to short-circuit and [`NDIterHandle`] to
+    /// get properties of the current iteration (e.g., the current element, indices, etc.)
+    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/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 {
+            // Extend the dimensions with np.newaxis.
+            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/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/object/utils/mod.rs

@@ -0,0 +1 @@
+pub mod slice;

nac3core/src/codegen/object/utils/slice.rs

@@ -0,0 +1,125 @@
+use crate::codegen::{model::*, CodeGenContext, CodeGenerator};
+
+/// 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_),
+        }
+    }
+}
+
+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/stmt.rs

@@ -1,3 +1,27 @@
+use super::{
+    super::symbol_resolver::ValueEnum,
+    expr::destructure_range,
+    irrt::{handle_slice_indices, list_slice_assignment},
+    object::{
+        any::AnyObject,
+        ndarray::{
+            indexing::util::gen_ndarray_subscript_ndindices, NDArrayObject, ScalarOrNDArray,
+        },
+    },
+    CodeGenContext, CodeGenerator,
+};
+use crate::{
+    codegen::{
+        classes::{ArrayLikeIndexer, ArraySliceValue, ListValue, RangeValue},
+        expr::gen_binop_expr,
+        gen_in_range_check,
+    },
+    toplevel::{DefinitionId, TopLevelDef},
+    typecheck::{
+        magic_methods::Binop,
+        typedef::{iter_type_vars, FunSignature, Type, TypeEnum},
+    },
+};
 use inkwell::{
     attributes::{Attribute, AttributeLoc},
     basic_block::BasicBlock,
@@ -6,28 +30,10 @@ use inkwell::{
     IntPredicate,
 };
 use itertools::{izip, Itertools};
-
 use nac3parser::ast::{
     Constant, ExcepthandlerKind, Expr, ExprKind, Location, Stmt, StmtKind, StrRef,
 };
 
-use super::{
-    expr::{destructure_range, gen_binop_expr},
-    gen_in_range_check,
-    irrt::{handle_slice_indices, list_slice_assignment},
-    macros::codegen_unreachable,
-    values::{ArrayLikeIndexer, ArraySliceValue, ListValue, RangeValue},
-    CodeGenContext, CodeGenerator,
-};
-use crate::{
-    symbol_resolver::ValueEnum,
-    toplevel::{DefinitionId, TopLevelDef},
-    typecheck::{
-        magic_methods::Binop,
-        typedef::{iter_type_vars, FunSignature, Type, TypeEnum},
-    },
-};
-
 /// See [`CodeGenerator::gen_var_alloc`].
 pub fn gen_var<'ctx>(
     ctx: &mut CodeGenContext<'ctx, '_>,
@@ -121,7 +127,7 @@ pub fn gen_store_target<'ctx, G: CodeGenerator>(
                 return Ok(None);
             };
             let BasicValueEnum::PointerValue(ptr) = val else {
-                codegen_unreachable!(ctx);
+                unreachable!();
             };
             unsafe {
                 ctx.builder.build_in_bounds_gep(
@@ -135,7 +141,7 @@ pub fn gen_store_target<'ctx, G: CodeGenerator>(
             }
             .unwrap()
         }
-        _ => codegen_unreachable!(ctx),
+        _ => unreachable!(),
     }))
 }
 
@@ -176,14 +182,6 @@ pub fn gen_assign<'ctx, G: CodeGenerator>(
                 }
             }
             let val = value.to_basic_value_enum(ctx, generator, target.custom.unwrap())?;
-
-            // Perform i1 <-> i8 conversion as needed
-            let val = if ctx.unifier.unioned(target.custom.unwrap(), ctx.primitives.bool) {
-                generator.bool_to_i8(ctx, val.into_int_value()).into()
-            } else {
-                val
-            };
-
             ctx.builder.build_store(ptr, val).unwrap();
         }
     };
@@ -201,12 +199,12 @@ pub fn gen_assign_target_list<'ctx, G: CodeGenerator>(
     // Deconstruct the tuple `value`
     let BasicValueEnum::StructValue(tuple) = value.to_basic_value_enum(ctx, generator, value_ty)?
     else {
-        codegen_unreachable!(ctx)
+        unreachable!()
     };
 
     // NOTE: Currently, RHS's type is forced to be a Tuple by the type inferencer.
     let TypeEnum::TTuple { ty: tuple_tys, .. } = &*ctx.unifier.get_ty(value_ty) else {
-        codegen_unreachable!(ctx);
+        unreachable!();
     };
 
     assert_eq!(tuple.get_type().count_fields() as usize, tuple_tys.len());
@@ -266,7 +264,7 @@ pub fn gen_assign_target_list<'ctx, G: CodeGenerator>(
             // Now assign with that sub-tuple to the starred target.
             generator.gen_assign(ctx, target, ValueEnum::Dynamic(sub_tuple_val), sub_tuple_ty)?;
         } else {
-            codegen_unreachable!(ctx) // The typechecker ensures this
+            unreachable!() // The typechecker ensures this
         }
 
         // Handle assignment after the starred target
@@ -310,13 +308,11 @@ pub fn gen_setitem<'ctx, G: CodeGenerator>(
                 .unwrap()
                 .to_basic_value_enum(ctx, generator, target_ty)?
                 .into_pointer_value();
-            let target = ListValue::from_pointer_value(target, llvm_usize, None);
+            let target = ListValue::from_ptr_val(target, llvm_usize, None);
 
             if let ExprKind::Slice { .. } = &key.node {
                 // Handle assigning to a slice
-                let ExprKind::Slice { lower, upper, step } = &key.node else {
-                    codegen_unreachable!(ctx)
-                };
+                let ExprKind::Slice { lower, upper, step } = &key.node else { unreachable!() };
                 let Some((start, end, step)) = handle_slice_indices(
                     lower,
                     upper,
@@ -331,7 +327,7 @@ pub fn gen_setitem<'ctx, G: CodeGenerator>(
 
                 let value =
                     value.to_basic_value_enum(ctx, generator, value_ty)?.into_pointer_value();
-                let value = ListValue::from_pointer_value(value, llvm_usize, None);
+                let value = ListValue::from_ptr_val(value, llvm_usize, None);
 
                 let target_item_ty = ctx.get_llvm_type(generator, target_item_ty);
                 let Some(src_ind) = handle_slice_indices(
@@ -411,7 +407,47 @@ pub fn gen_setitem<'ctx, G: CodeGenerator>(
             if *obj_id == ctx.primitives.ndarray.obj_id(&ctx.unifier).unwrap() =>
         {
             // Handle NDArray item assignment
-            todo!("ndarray subscript assignment is not yet implemented");
+            // Process target
+            let target = generator
+                .gen_expr(ctx, target)?
+                .unwrap()
+                .to_basic_value_enum(ctx, generator, target_ty)?;
+            let target = AnyObject { value: target, ty: target_ty };
+
+            // Process key
+            let key = gen_ndarray_subscript_ndindices(generator, ctx, key)?;
+
+            // Process value
+            let value = value.to_basic_value_enum(ctx, generator, value_ty)?;
+            let value = AnyObject { value, ty: value_ty };
+
+            /*
+                Reference code:
+                ```python
+                target = target[key]
+                value = np.asarray(value)
+
+                shape = np.broadcast_shape((target, value))
+
+                target = np.broadcast_to(target, shape)
+                value = np.broadcast_to(value, shape)
+
+                ...and finally copy 1-1 from value to target.
+                ```
+            */
+
+            let target = NDArrayObject::from_object(generator, ctx, target);
+            let target = target.index(generator, ctx, &key);
+
+            let value =
+                ScalarOrNDArray::split_object(generator, ctx, value).to_ndarray(generator, ctx);
+
+            let broadcast_result = NDArrayObject::broadcast(generator, ctx, &[target, value]);
+
+            let target = broadcast_result.ndarrays[0];
+            let value = broadcast_result.ndarrays[1];
+
+            target.copy_data_from(generator, ctx, value);
         }
         _ => {
             panic!("encountered unknown target type: {}", ctx.unifier.stringify(target_ty));
@@ -426,9 +462,7 @@ pub fn gen_for<G: CodeGenerator>(
     ctx: &mut CodeGenContext<'_, '_>,
     stmt: &Stmt<Option<Type>>,
 ) -> Result<(), String> {
-    let StmtKind::For { iter, target, body, orelse, .. } = &stmt.node else {
-        codegen_unreachable!(ctx)
-    };
+    let StmtKind::For { iter, target, body, orelse, .. } = &stmt.node else { unreachable!() };
 
     // var_assignment static values may be changed in another branch
     // if so, remove the static value as it may not be correct in this branch
@@ -463,15 +497,14 @@ pub fn gen_for<G: CodeGenerator>(
         TypeEnum::TObj { obj_id, .. }
             if *obj_id == ctx.primitives.range.obj_id(&ctx.unifier).unwrap() =>
         {
-            let iter_val =
-                RangeValue::from_pointer_value(iter_val.into_pointer_value(), Some("range"));
+            let iter_val = RangeValue::from_ptr_val(iter_val.into_pointer_value(), Some("range"));
             // Internal variable for loop; Cannot be assigned
             let i = generator.gen_var_alloc(ctx, int32.into(), Some("for.i.addr"))?;
             // Variable declared in "target" expression of the loop; Can be reassigned *or* shadowed
             let Some(target_i) =
                 generator.gen_store_target(ctx, target, Some("for.target.addr"))?
             else {
-                codegen_unreachable!(ctx)
+                unreachable!()
             };
             let (start, stop, step) = destructure_range(ctx, iter_val);
 
@@ -914,7 +947,7 @@ pub fn gen_while<G: CodeGenerator>(
     ctx: &mut CodeGenContext<'_, '_>,
     stmt: &Stmt<Option<Type>>,
 ) -> Result<(), String> {
-    let StmtKind::While { test, body, orelse, .. } = &stmt.node else { codegen_unreachable!(ctx) };
+    let StmtKind::While { test, body, orelse, .. } = &stmt.node else { unreachable!() };
 
     // var_assignment static values may be changed in another branch
     // if so, remove the static value as it may not be correct in this branch
@@ -944,7 +977,7 @@ pub fn gen_while<G: CodeGenerator>(
 
         return Ok(());
     };
-    let BasicValueEnum::IntValue(test) = test else { codegen_unreachable!(ctx) };
+    let BasicValueEnum::IntValue(test) = test else { unreachable!() };
 
     ctx.builder
         .build_conditional_branch(generator.bool_to_i1(ctx, test), body_bb, orelse_bb)
@@ -1092,7 +1125,7 @@ pub fn gen_if<G: CodeGenerator>(
     ctx: &mut CodeGenContext<'_, '_>,
     stmt: &Stmt<Option<Type>>,
 ) -> Result<(), String> {
-    let StmtKind::If { test, body, orelse, .. } = &stmt.node else { codegen_unreachable!(ctx) };
+    let StmtKind::If { test, body, orelse, .. } = &stmt.node else { unreachable!() };
 
     // var_assignment static values may be changed in another branch
     // if so, remove the static value as it may not be correct in this branch
@@ -1215,11 +1248,11 @@ pub fn exn_constructor<'ctx>(
     let zelf_id = if let TypeEnum::TObj { obj_id, .. } = &*ctx.unifier.get_ty(zelf_ty) {
         obj_id.0
     } else {
-        codegen_unreachable!(ctx)
+        unreachable!()
     };
     let defs = ctx.top_level.definitions.read();
     let def = defs[zelf_id].read();
-    let TopLevelDef::Class { name: zelf_name, .. } = &*def else { codegen_unreachable!(ctx) };
+    let TopLevelDef::Class { name: zelf_name, .. } = &*def else { unreachable!() };
     let exception_name = format!("{}:{}", ctx.resolver.get_exception_id(zelf_id), zelf_name);
     unsafe {
         let id_ptr = ctx.builder.build_in_bounds_gep(zelf, &[zero, zero], "exn.id").unwrap();
@@ -1327,7 +1360,7 @@ pub fn gen_try<'ctx, 'a, G: CodeGenerator>(
     target: &Stmt<Option<Type>>,
 ) -> Result<(), String> {
     let StmtKind::Try { body, handlers, orelse, finalbody, .. } = &target.node else {
-        codegen_unreachable!(ctx)
+        unreachable!()
     };
 
     // if we need to generate anything related to exception, we must have personality defined
@@ -1404,7 +1437,7 @@ pub fn gen_try<'ctx, 'a, G: CodeGenerator>(
             if let TypeEnum::TObj { obj_id, .. } = &*ctx.unifier.get_ty(type_.custom.unwrap()) {
                 *obj_id
             } else {
-                codegen_unreachable!(ctx)
+                unreachable!()
             };
         let exception_name = format!("{}:{}", ctx.resolver.get_exception_id(obj_id.0), exn_name);
         let exn_id = ctx.resolver.get_string_id(&exception_name);
@@ -1676,23 +1709,6 @@ pub fn gen_return<G: CodeGenerator>(
     } else {
         None
     };
-
-    // Remap boolean return type into i1
-    let value = value.map(|ret_val| {
-        // The "return type" of a sret function is in the first parameter
-        let expected_ty = if ctx.need_sret {
-            func.get_type().get_param_types()[0]
-        } else {
-            func.get_type().get_return_type().unwrap()
-        };
-
-        if matches!(expected_ty, BasicTypeEnum::IntType(ty) if ty.get_bit_width() == 1) {
-            generator.bool_to_i1(ctx, ret_val.into_int_value()).into()
-        } else {
-            ret_val
-        }
-    });
-
     if let Some(return_target) = ctx.return_target {
         if let Some(value) = value {
             ctx.builder.build_store(ctx.return_buffer.unwrap(), value).unwrap();
@@ -1703,6 +1719,25 @@ pub fn gen_return<G: CodeGenerator>(
         ctx.builder.build_store(ctx.return_buffer.unwrap(), value.unwrap()).unwrap();
         ctx.builder.build_return(None).unwrap();
     } else {
+        // Remap boolean return type into i1
+        let value = value.map(|v| {
+            let expected_ty = func.get_type().get_return_type().unwrap();
+            let ret_val = v.as_basic_value_enum();
+
+            if expected_ty.is_int_type() && ret_val.is_int_value() {
+                let ret_type = expected_ty.into_int_type();
+                let ret_val = ret_val.into_int_value();
+
+                if ret_type.get_bit_width() == 1 && ret_val.get_type().get_bit_width() != 1 {
+                    generator.bool_to_i1(ctx, ret_val)
+                } else {
+                    ret_val
+                }
+                .into()
+            } else {
+                ret_val
+            }
+        });
         let value = value.as_ref().map(|v| v as &dyn BasicValue);
         ctx.builder.build_return(value).unwrap();
     }
@@ -1771,30 +1806,7 @@ pub fn gen_stmt<G: CodeGenerator>(
         StmtKind::Try { .. } => gen_try(generator, ctx, stmt)?,
         StmtKind::Raise { exc, .. } => {
             if let Some(exc) = exc {
-                let exn = if let ExprKind::Name { id, .. } = &exc.node {
-                    // Handle "raise Exception" short form
-                    let def_id = ctx.resolver.get_identifier_def(*id).map_err(|e| {
-                        format!("{} (at {})", e.iter().next().unwrap(), exc.location)
-                    })?;
-                    let def = ctx.top_level.definitions.read();
-                    let TopLevelDef::Class { constructor, .. } = *def[def_id.0].read() else {
-                        return Err(format!("Failed to resolve symbol {id} (at {})", exc.location));
-                    };
-
-                    let TypeEnum::TFunc(signature) =
-                        ctx.unifier.get_ty(constructor.unwrap()).as_ref().clone()
-                    else {
-                        return Err(format!("Failed to resolve symbol {id} (at {})", exc.location));
-                    };
-
-                    generator
-                        .gen_call(ctx, None, (&signature, def_id), Vec::default())?
-                        .map(Into::into)
-                } else {
-                    generator.gen_expr(ctx, exc)?
-                };
-
-                let exc = if let Some(v) = exn {
+                let exc = if let Some(v) = generator.gen_expr(ctx, exc)? {
                     v.to_basic_value_enum(ctx, generator, exc.custom.unwrap())?
                 } else {
                     return Ok(());
@@ -1829,37 +1841,6 @@ pub fn gen_stmt<G: CodeGenerator>(
                 stmt.location,
             );
         }
-        StmtKind::Global { names, .. } => {
-            let registered_globals = ctx
-                .top_level
-                .definitions
-                .read()
-                .iter()
-                .filter_map(|def| {
-                    if let TopLevelDef::Variable { simple_name, ty, .. } = &*def.read() {
-                        Some((*simple_name, *ty))
-                    } else {
-                        None
-                    }
-                })
-                .collect_vec();
-
-            for id in names {
-                let Some((_, ty)) = registered_globals.iter().find(|(name, _)| name == id) else {
-                    return Err(format!("{id} is not a global at {}", stmt.location));
-                };
-
-                let resolver = ctx.resolver.clone();
-                let ptr = resolver
-                    .get_symbol_value(*id, ctx, generator)
-                    .map(|val| val.to_basic_value_enum(ctx, generator, *ty))
-                    .transpose()?
-                    .map(BasicValueEnum::into_pointer_value)
-                    .unwrap();
-
-                ctx.var_assignment.insert(*id, (ptr, None, 0));
-            }
-        }
         _ => unimplemented!(),
     };
     Ok(())

nac3core/src/codegen/test.rs

@@ -1,37 +1,34 @@
-use std::{
-    collections::{HashMap, HashSet},
-    sync::Arc,
-};
-
-use indexmap::IndexMap;
-use indoc::indoc;
-use inkwell::{
-    targets::{InitializationConfig, Target},
-    OptimizationLevel,
-};
-use nac3parser::{
-    ast::{fold::Fold, FileName, StrRef},
-    parser::parse_program,
-};
-use parking_lot::RwLock;
-
-use super::{
-    concrete_type::ConcreteTypeStore,
-    types::{ndarray::NDArrayType, ListType, ProxyType, RangeType},
-    CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask, CodeGenerator,
-    DefaultCodeGenerator, WithCall, WorkerRegistry,
-};
 use crate::{
+    codegen::{
+        classes::{ListType, ProxyType, RangeType},
+        concrete_type::ConcreteTypeStore,
+        CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask,
+        CodeGenerator, DefaultCodeGenerator, WithCall, WorkerRegistry,
+    },
     symbol_resolver::{SymbolResolver, ValueEnum},
     toplevel::{
         composer::{ComposerConfig, TopLevelComposer},
         DefinitionId, FunInstance, TopLevelContext, TopLevelDef,
     },
     typecheck::{
-        type_inferencer::{FunctionData, IdentifierInfo, Inferencer, PrimitiveStore},
+        type_inferencer::{FunctionData, Inferencer, PrimitiveStore},
         typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
     },
 };
+use indexmap::IndexMap;
+use indoc::indoc;
+use inkwell::{
+    targets::{InitializationConfig, Target},
+    OptimizationLevel,
+};
+use nac3parser::ast::FileName;
+use nac3parser::{
+    ast::{fold::Fold, StrRef},
+    parser::parse_program,
+};
+use parking_lot::RwLock;
+use std::collections::{HashMap, HashSet};
+use std::sync::Arc;
 
 struct Resolver {
     id_to_type: HashMap<StrRef, Type>,
@@ -67,7 +64,6 @@ impl SymbolResolver for Resolver {
         &self,
         _: StrRef,
         _: &mut CodeGenContext<'ctx, '_>,
-        _: &mut dyn CodeGenerator,
     ) -> Option<ValueEnum<'ctx>> {
         unimplemented!()
     }
@@ -142,8 +138,7 @@ fn test_primitives() {
     };
     let mut virtual_checks = Vec::new();
     let mut calls = HashMap::new();
-    let mut identifiers: HashMap<_, _> =
-        ["a".into(), "b".into()].map(|id| (id, IdentifierInfo::default())).into();
+    let mut identifiers: HashSet<_> = ["a".into(), "b".into()].into();
     let mut inferencer = Inferencer {
         top_level: &top_level,
         function_data: &mut function_data,
@@ -322,8 +317,7 @@ fn test_simple_call() {
     };
     let mut virtual_checks = Vec::new();
     let mut calls = HashMap::new();
-    let mut identifiers: HashMap<_, _> =
-        ["a".into(), "foo".into()].map(|id| (id, IdentifierInfo::default())).into();
+    let mut identifiers: HashSet<_> = ["a".into(), "foo".into()].into();
     let mut inferencer = Inferencer {
         top_level: &top_level,
         function_data: &mut function_data,
@@ -452,7 +446,7 @@ fn test_classes_list_type_new() {
     let llvm_usize = generator.get_size_type(&ctx);
 
     let llvm_list = ListType::new(&generator, &ctx, llvm_i32.into());
-    assert!(ListType::is_representable(llvm_list.as_base_type(), llvm_usize).is_ok());
+    assert!(ListType::is_type(llvm_list.as_base_type(), llvm_usize).is_ok());
 }
 
 #[test]
@@ -460,17 +454,5 @@ fn test_classes_range_type_new() {
     let ctx = inkwell::context::Context::create();
 
     let llvm_range = RangeType::new(&ctx);
-    assert!(RangeType::is_representable(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(), None);
-    assert!(NDArrayType::is_representable(llvm_ndarray.as_base_type(), llvm_usize).is_ok());
+    assert!(RangeType::is_type(llvm_range.as_base_type()).is_ok());
 }

nac3core/src/codegen/types/list.rs

@@ -1,206 +0,0 @@
-use inkwell::{
-    context::Context,
-    types::{AnyTypeEnum, BasicType, BasicTypeEnum, IntType, PointerType},
-    values::IntValue,
-    AddressSpace,
-};
-
-use super::ProxyType;
-use crate::codegen::{
-    values::{ArraySliceValue, ListValue, ProxyValue},
-    CodeGenContext, CodeGenerator,
-};
-
-/// Proxy type for a `list` type in LLVM.
-#[derive(Debug, PartialEq, Eq, Clone, Copy)]
-pub struct ListType<'ctx> {
-    ty: PointerType<'ctx>,
-    llvm_usize: IntType<'ctx>,
-}
-
-impl<'ctx> ListType<'ctx> {
-    /// Checks whether `llvm_ty` represents a `list` type, returning [Err] if it does not.
-    pub fn is_representable(
-        llvm_ty: PointerType<'ctx>,
-        llvm_usize: IntType<'ctx>,
-    ) -> Result<(), String> {
-        let llvm_list_ty = llvm_ty.get_element_type();
-        let AnyTypeEnum::StructType(llvm_list_ty) = llvm_list_ty else {
-            return Err(format!("Expected struct type for `list` type, got {llvm_list_ty}"));
-        };
-        if llvm_list_ty.count_fields() != 2 {
-            return Err(format!(
-                "Expected 2 fields in `list`, got {}",
-                llvm_list_ty.count_fields()
-            ));
-        }
-
-        let list_size_ty = llvm_list_ty.get_field_type_at_index(0).unwrap();
-        let Ok(_) = PointerType::try_from(list_size_ty) else {
-            return Err(format!("Expected pointer type for `list.0`, got {list_size_ty}"));
-        };
-
-        let list_data_ty = llvm_list_ty.get_field_type_at_index(1).unwrap();
-        let Ok(list_data_ty) = IntType::try_from(list_data_ty) else {
-            return Err(format!("Expected int type for `list.1`, got {list_data_ty}"));
-        };
-        if list_data_ty.get_bit_width() != llvm_usize.get_bit_width() {
-            return Err(format!(
-                "Expected {}-bit int type for `list.1`, got {}-bit int",
-                llvm_usize.get_bit_width(),
-                list_data_ty.get_bit_width()
-            ));
-        }
-
-        Ok(())
-    }
-
-    /// Creates an LLVM type corresponding to the expected structure of a `List`.
-    #[must_use]
-    fn llvm_type(
-        ctx: &'ctx Context,
-        element_type: BasicTypeEnum<'ctx>,
-        llvm_usize: IntType<'ctx>,
-    ) -> PointerType<'ctx> {
-        // struct List { data: T*, size: size_t }
-        let field_tys = [element_type.ptr_type(AddressSpace::default()).into(), llvm_usize.into()];
-
-        ctx.struct_type(&field_tys, false).ptr_type(AddressSpace::default())
-    }
-
-    /// Creates an instance of [`ListType`].
-    #[must_use]
-    pub fn new<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        element_type: BasicTypeEnum<'ctx>,
-    ) -> Self {
-        let llvm_usize = generator.get_size_type(ctx);
-        let llvm_list = Self::llvm_type(ctx, element_type, llvm_usize);
-
-        ListType::from_type(llvm_list, llvm_usize)
-    }
-
-    /// Creates an [`ListType`] from a [`PointerType`].
-    #[must_use]
-    pub fn from_type(ptr_ty: PointerType<'ctx>, llvm_usize: IntType<'ctx>) -> Self {
-        debug_assert!(Self::is_representable(ptr_ty, llvm_usize).is_ok());
-
-        ListType { ty: ptr_ty, llvm_usize }
-    }
-
-    /// Returns the type of the `size` field of this `list` type.
-    #[must_use]
-    pub fn size_type(&self) -> IntType<'ctx> {
-        self.as_base_type()
-            .get_element_type()
-            .into_struct_type()
-            .get_field_type_at_index(1)
-            .map(BasicTypeEnum::into_int_type)
-            .unwrap()
-    }
-
-    /// Returns the element type of this `list` type.
-    #[must_use]
-    pub fn element_type(&self) -> AnyTypeEnum<'ctx> {
-        self.as_base_type()
-            .get_element_type()
-            .into_struct_type()
-            .get_field_type_at_index(0)
-            .map(BasicTypeEnum::into_pointer_type)
-            .map(PointerType::get_element_type)
-            .unwrap()
-    }
-
-    /// Allocates an instance of [`ListValue`] as if by calling `alloca` on the base type.
-    #[must_use]
-    pub fn alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        name: Option<&'ctx str>,
-    ) -> <Self as ProxyType<'ctx>>::Value {
-        <Self as ProxyType<'ctx>>::Value::from_pointer_value(
-            self.raw_alloca(generator, ctx, name),
-            self.llvm_usize,
-            name,
-        )
-    }
-
-    /// Converts an existing value into a [`ListValue`].
-    #[must_use]
-    pub fn map_value(
-        &self,
-        value: <<Self as ProxyType<'ctx>>::Value as ProxyValue<'ctx>>::Base,
-        name: Option<&'ctx str>,
-    ) -> <Self as ProxyType<'ctx>>::Value {
-        <Self as ProxyType<'ctx>>::Value::from_pointer_value(value, self.llvm_usize, name)
-    }
-}
-
-impl<'ctx> ProxyType<'ctx> for ListType<'ctx> {
-    type Base = PointerType<'ctx>;
-    type Value = ListValue<'ctx>;
-
-    fn is_type<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        llvm_ty: impl BasicType<'ctx>,
-    ) -> Result<(), String> {
-        if let BasicTypeEnum::PointerType(ty) = llvm_ty.as_basic_type_enum() {
-            <Self as ProxyType<'ctx>>::is_representable(generator, ctx, ty)
-        } else {
-            Err(format!("Expected pointer type, got {llvm_ty:?}"))
-        }
-    }
-
-    fn is_representable<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        llvm_ty: Self::Base,
-    ) -> Result<(), String> {
-        Self::is_representable(llvm_ty, generator.get_size_type(ctx))
-    }
-
-    fn raw_alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        name: Option<&'ctx str>,
-    ) -> <Self::Value as ProxyValue<'ctx>>::Base {
-        generator
-            .gen_var_alloc(
-                ctx,
-                self.as_base_type().get_element_type().into_struct_type().into(),
-                name,
-            )
-            .unwrap()
-    }
-
-    fn array_alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        size: IntValue<'ctx>,
-        name: Option<&'ctx str>,
-    ) -> ArraySliceValue<'ctx> {
-        generator
-            .gen_array_var_alloc(
-                ctx,
-                self.as_base_type().get_element_type().into_struct_type().into(),
-                size,
-                name,
-            )
-            .unwrap()
-    }
-
-    fn as_base_type(&self) -> Self::Base {
-        self.ty
-    }
-}
-
-impl<'ctx> From<ListType<'ctx>> for PointerType<'ctx> {
-    fn from(value: ListType<'ctx>) -> Self {
-        value.as_base_type()
-    }
-}

nac3core/src/codegen/types/mod.rs

@@ -1,76 +0,0 @@
-//! This module contains abstraction over all intrinsic composite types of NAC3.
-//!
-//! # `raw_alloca` vs `alloca` vs `construct`
-//!
-//! There are three ways of creating a new object instance using the abstractions provided by this
-//! module.
-//!
-//! - `raw_alloca`: Allocates the object on the stack, returning an instance of
-//!   [`impl BasicValue`][inkwell::values::BasicValue]. This is similar to a `malloc` expression in
-//!   C++ but the object is allocated on the stack.
-//! - `alloca`: Similar to `raw_alloca`, but also wraps the allocated object with
-//!   [`<Self as ProxyType<'ctx>>::Value`][ProxyValue], and returns the wrapped object. The returned
-//!   object will not initialize any value or fields. This is similar to a type-safe `malloc`
-//!   expression in C++ but the object is allocated on the stack.
-//! - `construct`: Similar to `alloca`, but performs some initialization on the value or fields of
-//!   the returned object. This is similar to a `new` expression in C++ but the object is allocated
-//!   on the stack.
-
-use inkwell::{context::Context, types::BasicType, values::IntValue};
-
-use super::{
-    values::{ArraySliceValue, ProxyValue},
-    {CodeGenContext, CodeGenerator},
-};
-pub use list::*;
-pub use range::*;
-
-mod list;
-pub mod ndarray;
-mod range;
-pub mod structure;
-pub mod utils;
-
-/// A LLVM type that is used to represent a corresponding type in NAC3.
-pub trait ProxyType<'ctx>: Into<Self::Base> {
-    /// The LLVM type of which values of this type possess. This is usually a
-    /// [LLVM pointer type][PointerType] for any non-primitive types.
-    type Base: BasicType<'ctx>;
-
-    /// The type of values represented by this type.
-    type Value: ProxyValue<'ctx, Type = Self>;
-
-    fn is_type<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        llvm_ty: impl BasicType<'ctx>,
-    ) -> Result<(), String>;
-
-    /// Checks whether `llvm_ty` can be represented by this [`ProxyType`].
-    fn is_representable<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        llvm_ty: Self::Base,
-    ) -> Result<(), String>;
-
-    /// Creates a new value of this type, returning the LLVM instance of this value.
-    fn raw_alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        name: Option<&'ctx str>,
-    ) -> <Self::Value as ProxyValue<'ctx>>::Base;
-
-    /// Creates a new array value of this type, returning an [`ArraySliceValue`] encapsulating the
-    /// resulting array.
-    fn array_alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        size: IntValue<'ctx>,
-        name: Option<&'ctx str>,
-    ) -> ArraySliceValue<'ctx>;
-
-    /// Returns the [base type][Self::Base] of this proxy.
-    fn as_base_type(&self) -> Self::Base;
-}

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

@@ -1,257 +0,0 @@
-use inkwell::{
-    context::Context,
-    types::{AnyTypeEnum, BasicType, BasicTypeEnum, IntType, PointerType},
-    values::{IntValue, PointerValue},
-    AddressSpace,
-};
-use itertools::Itertools;
-
-use nac3core_derive::StructFields;
-
-use crate::{
-    codegen::{
-        types::{
-            structure::{
-                check_struct_type_matches_fields, FieldIndexCounter, StructField, StructFields,
-            },
-            ProxyType,
-        },
-        values::{ndarray::ContiguousNDArrayValue, ArraySliceValue, ProxyValue},
-        CodeGenContext, CodeGenerator,
-    },
-    toplevel::numpy::unpack_ndarray_var_tys,
-    typecheck::typedef::Type,
-};
-
-#[derive(Debug, PartialEq, Eq, Clone, Copy)]
-pub struct ContiguousNDArrayType<'ctx> {
-    ty: PointerType<'ctx>,
-    item: BasicTypeEnum<'ctx>,
-    llvm_usize: IntType<'ctx>,
-}
-
-#[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-pub struct ContiguousNDArrayFields<'ctx> {
-    #[value_type(usize)]
-    pub ndims: StructField<'ctx, IntValue<'ctx>>,
-    #[value_type(usize.ptr_type(AddressSpace::default()))]
-    pub shape: StructField<'ctx, PointerValue<'ctx>>,
-    #[value_type(i8_type().ptr_type(AddressSpace::default()))]
-    pub data: StructField<'ctx, PointerValue<'ctx>>,
-}
-
-impl<'ctx> ContiguousNDArrayFields<'ctx> {
-    #[must_use]
-    pub fn new_typed(item: BasicTypeEnum<'ctx>, llvm_usize: IntType<'ctx>) -> Self {
-        let mut counter = FieldIndexCounter::default();
-
-        ContiguousNDArrayFields {
-            ndims: StructField::create(&mut counter, "ndims", llvm_usize),
-            shape: StructField::create(
-                &mut counter,
-                "shape",
-                llvm_usize.ptr_type(AddressSpace::default()),
-            ),
-            data: StructField::create(&mut counter, "data", item.ptr_type(AddressSpace::default())),
-        }
-    }
-}
-
-impl<'ctx> ContiguousNDArrayType<'ctx> {
-    /// Checks whether `llvm_ty` represents a `ndarray` type, returning [Err] if it does not.
-    pub fn is_representable(
-        llvm_ty: PointerType<'ctx>,
-        llvm_usize: IntType<'ctx>,
-    ) -> Result<(), String> {
-        let ctx = llvm_ty.get_context();
-
-        let llvm_ty = llvm_ty.get_element_type();
-        let AnyTypeEnum::StructType(llvm_ty) = llvm_ty else {
-            return Err(format!(
-                "Expected struct type for `ContiguousNDArray` type, got {llvm_ty}"
-            ));
-        };
-
-        let fields = ContiguousNDArrayFields::new(ctx, llvm_usize);
-
-        check_struct_type_matches_fields(
-            fields,
-            llvm_ty,
-            "ContiguousNDArray",
-            &[(fields.data.name(), &|ty| {
-                if ty.is_pointer_type() {
-                    Ok(())
-                } else {
-                    Err(format!("Expected T* for `ContiguousNDArray.data`, got {ty}"))
-                }
-            })],
-        )
-    }
-
-    /// Returns an instance of [`StructFields`] containing all field accessors for this type.
-    #[must_use]
-    fn fields(
-        item: BasicTypeEnum<'ctx>,
-        llvm_usize: IntType<'ctx>,
-    ) -> ContiguousNDArrayFields<'ctx> {
-        ContiguousNDArrayFields::new_typed(item, llvm_usize)
-    }
-
-    /// See [`NDArrayType::fields`].
-    // TODO: Move this into e.g. StructProxyType
-    #[must_use]
-    pub fn get_fields(&self) -> ContiguousNDArrayFields<'ctx> {
-        Self::fields(self.item, self.llvm_usize)
-    }
-
-    /// Creates an LLVM type corresponding to the expected structure of an `NDArray`.
-    #[must_use]
-    fn llvm_type(
-        ctx: &'ctx Context,
-        item: BasicTypeEnum<'ctx>,
-        llvm_usize: IntType<'ctx>,
-    ) -> PointerType<'ctx> {
-        let field_tys =
-            Self::fields(item, llvm_usize).into_iter().map(|field| field.1).collect_vec();
-
-        ctx.struct_type(&field_tys, false).ptr_type(AddressSpace::default())
-    }
-
-    /// Creates an instance of [`ContiguousNDArrayType`].
-    #[must_use]
-    pub fn new<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        item: BasicTypeEnum<'ctx>,
-    ) -> Self {
-        let llvm_usize = generator.get_size_type(ctx);
-        let llvm_cndarray = Self::llvm_type(ctx, item, llvm_usize);
-
-        Self { ty: llvm_cndarray, item, llvm_usize }
-    }
-
-    /// Creates an [`ContiguousNDArrayType`] from a [unifier type][Type].
-    #[must_use]
-    pub fn from_unifier_type<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        ty: Type,
-    ) -> Self {
-        let (dtype, _) = unpack_ndarray_var_tys(&mut ctx.unifier, ty);
-
-        let llvm_dtype = ctx.get_llvm_type(generator, dtype);
-        let llvm_usize = generator.get_size_type(ctx.ctx);
-
-        Self { ty: Self::llvm_type(ctx.ctx, llvm_dtype, llvm_usize), item: llvm_dtype, llvm_usize }
-    }
-
-    /// Creates an [`ContiguousNDArrayType`] from a [`PointerType`] representing an `NDArray`.
-    #[must_use]
-    pub fn from_type(
-        ptr_ty: PointerType<'ctx>,
-        item: BasicTypeEnum<'ctx>,
-        llvm_usize: IntType<'ctx>,
-    ) -> Self {
-        debug_assert!(Self::is_representable(ptr_ty, llvm_usize).is_ok());
-
-        Self { ty: ptr_ty, item, llvm_usize }
-    }
-
-    /// Allocates an instance of [`ContiguousNDArrayValue`] as if by calling `alloca` on the base type.
-    #[must_use]
-    pub fn alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        name: Option<&'ctx str>,
-    ) -> <Self as ProxyType<'ctx>>::Value {
-        <Self as ProxyType<'ctx>>::Value::from_pointer_value(
-            self.raw_alloca(generator, ctx, name),
-            self.item,
-            self.llvm_usize,
-            name,
-        )
-    }
-
-    /// Converts an existing value into a [`ContiguousNDArrayValue`].
-    #[must_use]
-    pub fn map_value(
-        &self,
-        value: <<Self as ProxyType<'ctx>>::Value as ProxyValue<'ctx>>::Base,
-        name: Option<&'ctx str>,
-    ) -> <Self as ProxyType<'ctx>>::Value {
-        <Self as ProxyType<'ctx>>::Value::from_pointer_value(
-            value,
-            self.item,
-            self.llvm_usize,
-            name,
-        )
-    }
-}
-
-impl<'ctx> ProxyType<'ctx> for ContiguousNDArrayType<'ctx> {
-    type Base = PointerType<'ctx>;
-    type Value = ContiguousNDArrayValue<'ctx>;
-
-    fn is_type<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        llvm_ty: impl BasicType<'ctx>,
-    ) -> Result<(), String> {
-        if let BasicTypeEnum::PointerType(ty) = llvm_ty.as_basic_type_enum() {
-            <Self as ProxyType<'ctx>>::is_representable(generator, ctx, ty)
-        } else {
-            Err(format!("Expected pointer type, got {llvm_ty:?}"))
-        }
-    }
-
-    fn is_representable<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        llvm_ty: Self::Base,
-    ) -> Result<(), String> {
-        Self::is_representable(llvm_ty, generator.get_size_type(ctx))
-    }
-
-    fn raw_alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        name: Option<&'ctx str>,
-    ) -> <Self::Value as ProxyValue<'ctx>>::Base {
-        generator
-            .gen_var_alloc(
-                ctx,
-                self.as_base_type().get_element_type().into_struct_type().into(),
-                name,
-            )
-            .unwrap()
-    }
-
-    fn array_alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        size: IntValue<'ctx>,
-        name: Option<&'ctx str>,
-    ) -> ArraySliceValue<'ctx> {
-        generator
-            .gen_array_var_alloc(
-                ctx,
-                self.as_base_type().get_element_type().into_struct_type().into(),
-                size,
-                name,
-            )
-            .unwrap()
-    }
-
-    fn as_base_type(&self) -> Self::Base {
-        self.ty
-    }
-}
-
-impl<'ctx> From<ContiguousNDArrayType<'ctx>> for PointerType<'ctx> {
-    fn from(value: ContiguousNDArrayType<'ctx>) -> Self {
-        value.as_base_type()
-    }
-}

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

@@ -1,215 +0,0 @@
-use inkwell::{
-    context::{AsContextRef, Context},
-    types::{AnyTypeEnum, BasicType, BasicTypeEnum, IntType, PointerType},
-    values::{IntValue, PointerValue},
-    AddressSpace,
-};
-use itertools::Itertools;
-
-use nac3core_derive::StructFields;
-
-use crate::codegen::{
-    types::{
-        structure::{check_struct_type_matches_fields, StructField, StructFields},
-        ProxyType,
-    },
-    values::{
-        ndarray::{NDIndexValue, RustNDIndex},
-        ArrayLikeIndexer, ArraySliceValue, ProxyValue,
-    },
-    CodeGenContext, CodeGenerator,
-};
-
-#[derive(Debug, PartialEq, Eq, Clone, Copy)]
-pub struct NDIndexType<'ctx> {
-    ty: PointerType<'ctx>,
-    llvm_usize: IntType<'ctx>,
-}
-
-#[derive(PartialEq, Eq, Clone, Copy, StructFields)]
-pub struct NDIndexStructFields<'ctx> {
-    #[value_type(i8_type())]
-    pub type_: StructField<'ctx, IntValue<'ctx>>,
-    #[value_type(i8_type().ptr_type(AddressSpace::default()))]
-    pub data: StructField<'ctx, PointerValue<'ctx>>,
-}
-
-impl<'ctx> NDIndexType<'ctx> {
-    /// Checks whether `llvm_ty` represents a `ndindex` type, returning [Err] if it does not.
-    pub fn is_representable(
-        llvm_ty: PointerType<'ctx>,
-        llvm_usize: IntType<'ctx>,
-    ) -> Result<(), String> {
-        let ctx = llvm_ty.get_context();
-
-        let llvm_ty = llvm_ty.get_element_type();
-        let AnyTypeEnum::StructType(llvm_ty) = llvm_ty else {
-            return Err(format!(
-                "Expected struct type for `ContiguousNDArray` type, got {llvm_ty}"
-            ));
-        };
-
-        let fields = NDIndexStructFields::new(ctx, llvm_usize);
-
-        check_struct_type_matches_fields(fields, llvm_ty, "NDIndex", &[])
-    }
-
-    #[must_use]
-    fn fields(
-        ctx: impl AsContextRef<'ctx>,
-        llvm_usize: IntType<'ctx>,
-    ) -> NDIndexStructFields<'ctx> {
-        NDIndexStructFields::new(ctx, llvm_usize)
-    }
-
-    #[must_use]
-    pub fn get_fields(&self) -> NDIndexStructFields<'ctx> {
-        Self::fields(self.ty.get_context(), self.llvm_usize)
-    }
-
-    #[must_use]
-    fn llvm_type(ctx: &'ctx Context, llvm_usize: IntType<'ctx>) -> PointerType<'ctx> {
-        let field_tys =
-            Self::fields(ctx, llvm_usize).into_iter().map(|field| field.1).collect_vec();
-
-        ctx.struct_type(&field_tys, false).ptr_type(AddressSpace::default())
-    }
-
-    #[must_use]
-    pub fn new<G: CodeGenerator + ?Sized>(generator: &G, ctx: &'ctx Context) -> Self {
-        let llvm_usize = generator.get_size_type(ctx);
-        let llvm_ndindex = Self::llvm_type(ctx, llvm_usize);
-
-        Self { ty: llvm_ndindex, llvm_usize }
-    }
-
-    #[must_use]
-    pub fn from_type(ptr_ty: PointerType<'ctx>, llvm_usize: IntType<'ctx>) -> Self {
-        debug_assert!(Self::is_representable(ptr_ty, llvm_usize).is_ok());
-
-        Self { ty: ptr_ty, llvm_usize }
-    }
-
-    #[must_use]
-    pub fn alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        name: Option<&'ctx str>,
-    ) -> <Self as ProxyType<'ctx>>::Value {
-        <Self as ProxyType<'ctx>>::Value::from_pointer_value(
-            self.raw_alloca(generator, ctx, name),
-            self.llvm_usize,
-            name,
-        )
-    }
-
-    /// Serialize a list of [`RustNDIndex`] as a newly allocated LLVM array of [`NDIndexValue`].
-    #[must_use]
-    pub fn construct_ndindices<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        in_ndindices: &[RustNDIndex<'ctx>],
-    ) -> ArraySliceValue<'ctx> {
-        // Allocate the LLVM ndindices.
-        let num_ndindices = self.llvm_usize.const_int(in_ndindices.len() as u64, false);
-        let ndindices = self.array_alloca(generator, ctx, num_ndindices, None);
-
-        // Initialize all of them.
-        for (i, in_ndindex) in in_ndindices.iter().enumerate() {
-            let pndindex = unsafe {
-                ndindices.ptr_offset_unchecked(
-                    ctx,
-                    generator,
-                    &ctx.ctx.i64_type().const_int(u64::try_from(i).unwrap(), false),
-                    None,
-                )
-            };
-
-            in_ndindex.write_to_ndindex(
-                generator,
-                ctx,
-                NDIndexValue::from_pointer_value(pndindex, self.llvm_usize, None),
-            );
-        }
-
-        ndindices
-    }
-
-    #[must_use]
-    pub fn map_value(
-        &self,
-        value: <<Self as ProxyType<'ctx>>::Value as ProxyValue<'ctx>>::Base,
-        name: Option<&'ctx str>,
-    ) -> <Self as ProxyType<'ctx>>::Value {
-        <Self as ProxyType<'ctx>>::Value::from_pointer_value(value, self.llvm_usize, name)
-    }
-}
-
-impl<'ctx> ProxyType<'ctx> for NDIndexType<'ctx> {
-    type Base = PointerType<'ctx>;
-    type Value = NDIndexValue<'ctx>;
-
-    fn is_type<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        llvm_ty: impl BasicType<'ctx>,
-    ) -> Result<(), String> {
-        if let BasicTypeEnum::PointerType(ty) = llvm_ty.as_basic_type_enum() {
-            <Self as ProxyType<'ctx>>::is_representable(generator, ctx, ty)
-        } else {
-            Err(format!("Expected pointer type, got {llvm_ty:?}"))
-        }
-    }
-
-    fn is_representable<G: CodeGenerator + ?Sized>(
-        generator: &G,
-        ctx: &'ctx Context,
-        llvm_ty: Self::Base,
-    ) -> Result<(), String> {
-        Self::is_representable(llvm_ty, generator.get_size_type(ctx))
-    }
-
-    fn raw_alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        name: Option<&'ctx str>,
-    ) -> <Self::Value as ProxyValue<'ctx>>::Base {
-        generator
-            .gen_var_alloc(
-                ctx,
-                self.as_base_type().get_element_type().into_struct_type().into(),
-                name,
-            )
-            .unwrap()
-    }
-
-    fn array_alloca<G: CodeGenerator + ?Sized>(
-        &self,
-        generator: &mut G,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        size: IntValue<'ctx>,
-        name: Option<&'ctx str>,
-    ) -> ArraySliceValue<'ctx> {
-        generator
-            .gen_array_var_alloc(
-                ctx,
-                self.as_base_type().get_element_type().into_struct_type().into(),
-                size,
-                name,
-            )
-            .unwrap()
-    }
-
-    fn as_base_type(&self) -> Self::Base {
-        self.ty
-    }
-}
-
-impl<'ctx> From<NDIndexType<'ctx>> for PointerType<'ctx> {
-    fn from(value: NDIndexType<'ctx>) -> Self {
-        value.as_base_type()
-    }
-}