WIP

flake.nix

@@ -107,18 +107,18 @@
             (pkgs.fetchFromGitHub {
               owner = "m-labs";
               repo = "sipyco";
-              rev = "094a6cd63ffa980ef63698920170e50dc9ba77fd";
+              rev = "939f84f9b5eef7efbf7423c735d1834783b6140e";
-              sha256 = "sha256-PPnAyDedUQ7Og/Cby9x5OT9wMkNGTP8GS53V6N/dk4w=";
+              sha256 = "sha256-15Nun4EY35j+6SPZkjzZtyH/ncxLS60KuGJjFh5kSTc=";
             })
             (pkgs.fetchFromGitHub {
               owner = "m-labs";
               repo = "artiq";
-              rev = "28c9de3e251daa89a8c9fd79d5ab64a3ec03bac6";
+              rev = "923ca3377d42c815f979983134ec549dc39d3ca0";
-              sha256 = "sha256-vAvpbHc5B+1wtG8zqN7j9dQE1ON+i22v+uqA+tw6Gak=";
+              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/demo/min_artiq.py

@@ -206,7 +206,7 @@ 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

@@ -15,9 +15,10 @@ use pyo3::{
 use nac3core::{
     codegen::{
         expr::{destructure_range, gen_call},
-        irrt::call_ndarray_calc_size,
+        irrt::ndarray::call_ndarray_calc_size,
         llvm_intrinsics::{call_int_smax, call_memcpy_generic, call_stackrestore, call_stacksave},
         stmt::{gen_block, gen_for_callback_incrementing, gen_if_callback, gen_with},
+        type_aligned_alloca,
         types::{NDArrayType, ProxyType},
         values::{
             ArrayLikeIndexer, ArrayLikeValue, ArraySliceValue, ListValue, NDArrayValue, ProxyValue,
@@ -642,27 +643,12 @@ fn format_rpc_ret<'ctx>(
             // (4 + 4 * ndims) bytes with 8-byte alignment
             let sizeof_dims =
                 ctx.builder.build_int_mul(ndarray.load_ndims(ctx), llvm_usize_sizeof, "").unwrap();
-            let unaligned_buffer_size =
+            let buffer_size =
                 ctx.builder.build_int_add(sizeof_dims, llvm_pdata_sizeof, "").unwrap();
-            let buffer_size = round_up(ctx, unaligned_buffer_size, llvm_usize.const_int(8, false));
 
             let stackptr = call_stacksave(ctx, None);
-            // Just to be absolutely sure, alloca in [i8 x 8] slices to force 8-byte alignment
+            let buffer =
-            let buffer = ctx
+                type_aligned_alloca(generator, ctx, llvm_i8_8, buffer_size, Some("rpc.buffer"));
-                .builder
-                .build_array_alloca(
-                    llvm_i8_8,
-                    ctx.builder
-                        .build_int_unsigned_div(buffer_size, llvm_usize.const_int(8, false), "")
-                        .unwrap(),
-                    "rpc.buffer",
-                )
-                .unwrap();
-            let buffer = ctx
-                .builder
-                .build_bit_cast(buffer, llvm_pi8, "")
-                .map(BasicValueEnum::into_pointer_value)
-                .unwrap();
             let buffer = ArraySliceValue::from_ptr_val(buffer, buffer_size, None);
 
             // The first call to `rpc_recv` reads the top-level ndarray object: [pdata, shape]
@@ -735,7 +721,9 @@ fn format_rpc_ret<'ctx>(
                 );
             }
 
-            ndarray.create_data(ctx, llvm_elem_ty, num_elements);
+            unsafe {
+                ndarray.create_data(generator, ctx, num_elements);
+            }
 
             let ndarray_data = ndarray.data().base_ptr(ctx, generator);
             let ndarray_data_i8 =
@@ -1376,6 +1364,7 @@ fn polymorphic_print<'ctx>(
                 let val = NDArrayValue::from_pointer_value(
                     value.into_pointer_value(),
                     llvm_elem_ty,
+                    None,
                     llvm_usize,
                     None,
                 );

nac3artiq/src/lib.rs

@@ -24,7 +24,7 @@ use parking_lot::{Mutex, RwLock};
 use pyo3::{
     create_exception, exceptions,
     prelude::*,
-    types::{PyBytes, PyDict, PyNone, PySet},
+    types::{PyBytes, PyDict, PySet},
 };
 use tempfile::{self, TempDir};
 
@@ -142,32 +142,14 @@ impl Nac3 {
         module: &PyObject,
         registered_class_ids: &HashSet<u64>,
     ) -> PyResult<()> {
-        let (module_name, source_file, source) =
+        let (module_name, source_file) = Python::with_gil(|py| -> PyResult<(String, String)> {
-            Python::with_gil(|py| -> PyResult<(String, String, String)> {
             let module: &PyAny = module.extract(py)?;
-                let source_file = module.getattr("__file__");
+            Ok((module.getattr("__name__")?.extract()?, module.getattr("__file__")?.extract()?))
-                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 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}")))?;
 
@@ -1090,12 +1072,7 @@ impl Nac3 {
         })
     }
 
-    fn analyze(
+    fn analyze(&mut self, functions: &PySet, classes: &PySet) -> PyResult<()> {
-        &mut self,
-        functions: &PySet,
-        classes: &PySet,
-        content_modules: &PySet,
-    ) -> PyResult<()> {
         let (modules, class_ids) =
             Python::with_gil(|py| -> PyResult<(HashMap<u64, PyObject>, HashSet<u64>)> {
                 let mut modules: HashMap<u64, PyObject> = HashMap::new();
@@ -1105,22 +1082,14 @@ impl Nac3 {
                 let getmodule_fn = PyModule::import(py, "inspect")?.getattr("getmodule")?;
 
                 for function in functions {
-                    let module: PyObject = getmodule_fn.call1((function,))?.extract()?;
+                    let module = getmodule_fn.call1((function,))?.extract()?;
-                    if !module.is_none(py) {
                     modules.insert(id_fn.call1((&module,))?.extract()?, module);
                 }
-                }
                 for class in classes {
-                    let module: PyObject = getmodule_fn.call1((class,))?.extract()?;
+                    let module = 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()?;
-                    modules.insert(id_fn.call1((&module,))?.extract()?, module.into());
-                }
                 Ok((modules, class_ids))
             })?;
 

nac3core/irrt/irrt.cpp

@@ -3,3 +3,5 @@
 #include "irrt/math.hpp"
 #include "irrt/ndarray.hpp"
 #include "irrt/slice.hpp"
+#include "irrt/ndarray/basic.hpp"
+#include "irrt/ndarray/def.hpp"

nac3core/irrt/irrt/ndarray.hpp

@@ -2,6 +2,8 @@
 
 #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) {

nac3core/irrt/irrt/ndarray/basic.hpp

@@ -0,0 +1,342 @@
+#pragma once
+
+#include "irrt/debug.hpp"
+#include "irrt/exception.hpp"
+#include "irrt/int_types.hpp"
+#include "irrt/ndarray/def.hpp"
+
+namespace {
+namespace ndarray {
+namespace basic {
+/**
+ * @brief 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) {
+            raise_exception(SizeT, EXN_VALUE_ERROR,
+                            "negative dimensions are not allowed; axis {0} "
+                            "has dimension {1}",
+                            axis, shape[axis], NO_PARAM);
+        }
+    }
+}
+
+/**
+ * @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) {
+        // There is no corresponding NumPy error message like this.
+        raise_exception(SizeT, EXN_VALUE_ERROR, "Cannot write output of ndims {0} to an ndarray with ndims {1}",
+                        output_ndims, ndarray_ndims, NO_PARAM);
+    }
+
+    for (SizeT axis = 0; axis < ndarray_ndims; axis++) {
+        if (ndarray_shape[axis] != output_shape[axis]) {
+            // There is no corresponding NumPy error message like this.
+            raise_exception(SizeT, EXN_VALUE_ERROR,
+                            "Mismatched dimensions on axis {0}, output has "
+                            "dimension {1}, but destination ndarray has dimension {2}.",
+                            axis, output_shape[axis], ndarray_shape[axis]);
+        }
+    }
+}
+
+/**
+ * @brief Return the number of elements of an ndarray given its shape.
+ *
+ * @param ndims Number of dimensions in `shape`
+ * @param shape The shape of the ndarray
+ */
+template<typename SizeT>
+SizeT calc_size_from_shape(SizeT ndims, const SizeT* shape) {
+    SizeT size = 1;
+    for (SizeT axis = 0; axis < ndims; axis++)
+        size *= shape[axis];
+    return size;
+}
+
+/**
+ * @brief Compute the array indices of the `nth` (0-based) element of an ndarray given only its shape.
+ *
+ * @param ndims Number of elements in `shape` and `indices`
+ * @param shape The shape of the ndarray
+ * @param indices The returned indices indexing the ndarray with shape `shape`.
+ * @param nth The index of the element of interest.
+ */
+template<typename SizeT>
+void set_indices_by_nth(SizeT ndims, const SizeT* shape, SizeT* indices, SizeT nth) {
+    for (SizeT i = 0; i < ndims; i++) {
+        SizeT axis = ndims - i - 1;
+        SizeT dim = shape[axis];
+
+        indices[axis] = nth % dim;
+        nth /= dim;
+    }
+}
+
+/**
+ * @brief Return the number of elements of an `ndarray`
+ *
+ * This function corresponds to `<an_ndarray>.size`
+ */
+template<typename SizeT>
+SizeT size(const NDArray<SizeT>* ndarray) {
+    return calc_size_from_shape(ndarray->ndims, ndarray->shape);
+}
+
+/**
+ * @brief Return of the number of its content of an `ndarray`.
+ *
+ * This function corresponds to `<an_ndarray>.nbytes`.
+ */
+template<typename SizeT>
+SizeT nbytes(const NDArray<SizeT>* ndarray) {
+    return size(ndarray) * ndarray->itemsize;
+}
+
+/**
+ * @brief Get the `len()` of an ndarray, and asserts that `ndarray` is a sized object.
+ *
+ * This function corresponds to `<an_ndarray>.__len__`.
+ *
+ * @param dst_length The length.
+ */
+template<typename SizeT>
+SizeT len(const NDArray<SizeT>* ndarray) {
+    if (ndarray->ndims != 0) {
+        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
+ */
+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
+
+    // From
+    // https://github.com/numpy/numpy/blob/df256d0d2f3bc6833699529824781c58f9c6e697/numpy/core/src/multiarray/flagsobject.c#L95C1-L99C45:
+    //
+    // The traditional rule is that for an array to be flagged as C contiguous,
+    // the following must hold:
+    //
+    // strides[-1] == itemsize
+    // strides[i] == shape[i+1] * strides[i + 1]
+    // [...]
+    // According to these rules, a 0- or 1-dimensional array is either both
+    // C- and F-contiguous, or neither; and an array with 2+ dimensions
+    // can be C- or F- contiguous, or neither, but not both. Though there
+    // there are exceptions for arrays with zero or one item, in the first
+    // case the check is relaxed up to and including the first dimension
+    // with shape[i] == 0. In the second case `strides == itemsize` will
+    // can be true for all dimensions and both flags are set.
+
+    if (ndarray->ndims == 0) {
+        return true;
+    }
+
+    if (ndarray->strides[ndarray->ndims - 1] != ndarray->itemsize) {
+        return false;
+    }
+
+    for (SizeT i = 1; i < ndarray->ndims; i++) {
+        SizeT axis_i = ndarray->ndims - i - 1;
+        if (ndarray->strides[axis_i] != ndarray->shape[axis_i + 1] * ndarray->strides[axis_i + 1]) {
+            return false;
+        }
+    }
+
+    return true;
+}
+
+/**
+ * @brief Return the pointer to the element indexed by `indices` along the ndarray's axes.
+ *
+ * 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;
+    for (SizeT dim_i = 0; dim_i < ndarray->ndims; dim_i++)
+        element = static_cast<uint8_t*>(element) + indices[dim_i] * ndarray->strides[dim_i];
+    return element;
+}
+
+/**
+ * @brief Return the pointer to the nth (0-based) element of `ndarray` in flattened view.
+ *
+ * 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++) {
+        SizeT axis = ndarray->ndims - i - 1;
+        SizeT dim = ndarray->shape[axis];
+        element = static_cast<uint8_t*>(element) + ndarray->strides[axis] * (nth % dim);
+        nth /= dim;
+    }
+    return element;
+}
+
+/**
+ * @brief Update the strides of an ndarray given an ndarray `shape` to be contiguous.
+ *
+ * You might want to read https://ajcr.net/stride-guide-part-1/.
+ */
+template<typename SizeT>
+void set_strides_by_shape(NDArray<SizeT>* ndarray) {
+    SizeT stride_product = 1;
+    for (SizeT i = 0; i < ndarray->ndims; i++) {
+        SizeT axis = ndarray->ndims - i - 1;
+        ndarray->strides[axis] = stride_product * ndarray->itemsize;
+        stride_product *= ndarray->shape[axis];
+    }
+}
+
+/**
+ * @brief Set an element in `ndarray`.
+ *
+ * @param pelement Pointer to the element in `ndarray` to be set.
+ * @param pvalue Pointer to the value `pelement` will be set to.
+ */
+template<typename SizeT>
+void set_pelement_value(NDArray<SizeT>* ndarray, void* pelement, const void* pvalue) {
+    __builtin_memcpy(pelement, pvalue, ndarray->itemsize);
+}
+
+/**
+ * @brief Copy data from one ndarray to another of the exact same size and itemsize.
+ *
+ * Both ndarrays will be viewed in their flatten views when copying the elements.
+ */
+template<typename SizeT>
+void copy_data(const NDArray<SizeT>* src_ndarray, NDArray<SizeT>* dst_ndarray) {
+    // TODO: Make this faster with memcpy 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++) {
+        auto src_element = ndarray::basic::get_nth_pelement(src_ndarray, i);
+        auto dst_element = ndarray::basic::get_nth_pelement(dst_ndarray, i);
+        ndarray::basic::set_pelement_value(dst_ndarray, dst_element, src_element);
+    }
+}
+}  // namespace basic
+}  // namespace ndarray
+}  // namespace
+
+extern "C" {
+using namespace ndarray::basic;
+
+void __nac3_ndarray_util_assert_shape_no_negative(int32_t ndims, int32_t* shape) {
+    assert_shape_no_negative(ndims, shape);
+}
+
+void __nac3_ndarray_util_assert_shape_no_negative64(int64_t ndims, int64_t* shape) {
+    assert_shape_no_negative(ndims, shape);
+}
+
+void __nac3_ndarray_util_assert_output_shape_same(int32_t ndarray_ndims,
+                                                  const int32_t* ndarray_shape,
+                                                  int32_t output_ndims,
+                                                  const int32_t* output_shape) {
+    assert_output_shape_same(ndarray_ndims, ndarray_shape, output_ndims, output_shape);
+}
+
+void __nac3_ndarray_util_assert_output_shape_same64(int64_t ndarray_ndims,
+                                                    const int64_t* ndarray_shape,
+                                                    int64_t output_ndims,
+                                                    const int64_t* output_shape) {
+    assert_output_shape_same(ndarray_ndims, ndarray_shape, output_ndims, output_shape);
+}
+
+uint32_t __nac3_ndarray_size(NDArray<int32_t>* ndarray) {
+    return size(ndarray);
+}
+
+uint64_t __nac3_ndarray_size64(NDArray<int64_t>* ndarray) {
+    return size(ndarray);
+}
+
+uint32_t __nac3_ndarray_nbytes(NDArray<int32_t>* ndarray) {
+    return nbytes(ndarray);
+}
+
+uint64_t __nac3_ndarray_nbytes64(NDArray<int64_t>* ndarray) {
+    return nbytes(ndarray);
+}
+
+int32_t __nac3_ndarray_len(NDArray<int32_t>* ndarray) {
+    return len(ndarray);
+}
+
+int64_t __nac3_ndarray_len64(NDArray<int64_t>* ndarray) {
+    return len(ndarray);
+}
+
+bool __nac3_ndarray_is_c_contiguous(NDArray<int32_t>* ndarray) {
+    return is_c_contiguous(ndarray);
+}
+
+bool __nac3_ndarray_is_c_contiguous64(NDArray<int64_t>* ndarray) {
+    return is_c_contiguous(ndarray);
+}
+
+void* __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);
+}
+
+void* __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);
+}
+
+void __nac3_ndarray_set_strides_by_shape(NDArray<int32_t>* ndarray) {
+    set_strides_by_shape(ndarray);
+}
+
+void __nac3_ndarray_set_strides_by_shape64(NDArray<int64_t>* ndarray) {
+    set_strides_by_shape(ndarray);
+}
+
+void __nac3_ndarray_copy_data(NDArray<int32_t>* src_ndarray, NDArray<int32_t>* dst_ndarray) {
+    copy_data(src_ndarray, dst_ndarray);
+}
+
+void __nac3_ndarray_copy_data64(NDArray<int64_t>* src_ndarray, NDArray<int64_t>* dst_ndarray) {
+    copy_data(src_ndarray, dst_ndarray);
+}
+}

nac3core/irrt/irrt/ndarray/def.hpp

@@ -0,0 +1,51 @@
+#pragma once
+
+#include "irrt/int_types.hpp"
+
+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.
+ */
+template<typename SizeT>
+struct NDArray {
+    /**
+     * @brief The number of bytes of a single element in `data`.
+     */
+    SizeT itemsize;
+
+    /**
+     * @brief The number of dimensions of this shape.
+     */
+    SizeT ndims;
+
+    /**
+     * @brief The NDArray shape, with length equal to `ndims`.
+     *
+     * Note that it may contain 0.
+     */
+    SizeT* shape;
+
+    /**
+     * @brief Array strides, with length equal to `ndims`
+     *
+     * The stride values are in units of bytes, not number of elements.
+     *
+     * Note that `strides` can have negative values or contain 0.
+     */
+    SizeT* strides;
+
+    /**
+     * @brief The underlying data this `ndarray` is pointing to.
+     */
+    void* data;
+};
+}  // namespace

nac3core/src/codegen/builtin_fns.rs

@@ -74,6 +74,7 @@ pub fn call_len<'ctx, G: CodeGenerator + ?Sized>(
                 let arg = NDArrayValue::from_pointer_value(
                     arg.into_pointer_value(),
                     ctx.get_llvm_type(generator, elem_ty),
+                    None,
                     llvm_usize,
                     None,
                 );
@@ -153,7 +154,7 @@ pub fn call_int32<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ctx.primitives.int32,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_int32(generator, ctx, (elem_ty, val)),
             )?;
 
@@ -216,7 +217,7 @@ pub fn call_int64<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ctx.primitives.int64,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_int64(generator, ctx, (elem_ty, val)),
             )?;
 
@@ -295,7 +296,7 @@ pub fn call_uint32<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ctx.primitives.uint32,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_uint32(generator, ctx, (elem_ty, val)),
             )?;
 
@@ -363,7 +364,7 @@ pub fn call_uint64<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ctx.primitives.uint64,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_uint64(generator, ctx, (elem_ty, val)),
             )?;
 
@@ -430,7 +431,7 @@ pub fn call_float<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ctx.primitives.float,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_float(generator, ctx, (elem_ty, val)),
             )?;
 
@@ -477,7 +478,7 @@ pub fn call_round<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ret_elem_ty,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_round(generator, ctx, (elem_ty, val), ret_elem_ty),
             )?;
 
@@ -518,7 +519,7 @@ pub fn call_numpy_round<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ctx.primitives.float,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_numpy_round(generator, ctx, (elem_ty, val)),
             )?;
 
@@ -584,7 +585,7 @@ pub fn call_bool<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ctx.primitives.bool,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| {
                     let elem = call_bool(generator, ctx, (elem_ty, val))?;
 
@@ -639,7 +640,7 @@ pub fn call_floor<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ret_elem_ty,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_floor(generator, ctx, (elem_ty, val), ret_elem_ty),
             )?;
 
@@ -690,7 +691,7 @@ pub fn call_ceil<'ctx, G: CodeGenerator + ?Sized>(
                 ctx,
                 ret_elem_ty,
                 None,
-                NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None),
                 |generator, ctx, val| call_ceil(generator, ctx, (elem_ty, val), ret_elem_ty),
             )?;
 
@@ -921,8 +922,9 @@ pub fn call_numpy_max_min<'ctx, G: CodeGenerator + ?Sized>(
             let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, a_ty);
             let llvm_elem_ty = ctx.get_llvm_type(generator, elem_ty);
 
-            let n = NDArrayValue::from_pointer_value(n, llvm_elem_ty, llvm_usize, None);
+            let n = NDArrayValue::from_pointer_value(n, llvm_elem_ty, None, llvm_usize, None);
-            let n_sz = irrt::call_ndarray_calc_size(generator, ctx, &n.shape(), (None, None));
+            let n_sz =
+                irrt::ndarray::call_ndarray_calc_size(generator, ctx, &n.shape(), (None, None));
             if ctx.registry.llvm_options.opt_level == OptimizationLevel::None {
                 let n_sz_eqz = ctx
                     .builder
@@ -1135,7 +1137,7 @@ where
                 ctx,
                 ret_elem_ty,
                 None,
-                NDArrayValue::from_pointer_value(x, llvm_arg_elem_ty, llvm_usize, None),
+                NDArrayValue::from_pointer_value(x, llvm_arg_elem_ty, None, llvm_usize, None),
                 |generator, ctx, elem_val| {
                     helper_call_numpy_unary_elementwise(
                         generator,
@@ -1974,7 +1976,7 @@ pub fn call_np_linalg_cholesky<'ctx, G: CodeGenerator + ?Sized>(
             unsupported_type(ctx, FN_NAME, &[x1_ty]);
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
         let dim0 = unsafe {
             n1.shape()
                 .get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
@@ -2016,7 +2018,7 @@ pub fn call_np_linalg_qr<'ctx, G: CodeGenerator + ?Sized>(
             unimplemented!("{FN_NAME} operates on float type NdArrays only");
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
         let dim0 = unsafe {
             n1.shape()
                 .get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
@@ -2066,7 +2068,7 @@ pub fn call_np_linalg_svd<'ctx, G: CodeGenerator + ?Sized>(
             unsupported_type(ctx, FN_NAME, &[x1_ty]);
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
 
         let dim0 = unsafe {
             n1.shape()
@@ -2121,7 +2123,7 @@ pub fn call_np_linalg_inv<'ctx, G: CodeGenerator + ?Sized>(
             unsupported_type(ctx, FN_NAME, &[x1_ty]);
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
         let dim0 = unsafe {
             n1.shape()
                 .get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
@@ -2163,7 +2165,7 @@ pub fn call_np_linalg_pinv<'ctx, G: CodeGenerator + ?Sized>(
             unsupported_type(ctx, FN_NAME, &[x1_ty]);
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
 
         let dim0 = unsafe {
             n1.shape()
@@ -2206,7 +2208,7 @@ pub fn call_sp_linalg_lu<'ctx, G: CodeGenerator + ?Sized>(
             unsupported_type(ctx, FN_NAME, &[x1_ty]);
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
 
         let dim0 = unsafe {
             n1.shape()
@@ -2259,7 +2261,7 @@ pub fn call_np_linalg_matrix_power<'ctx, G: CodeGenerator + ?Sized>(
             unsupported_type(ctx, FN_NAME, &[x1_ty, x2_ty]);
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
         // Changing second parameter to a `NDArray` for uniformity in function call
         let n2_array = numpy::create_ndarray_const_shape(
             generator,
@@ -2354,7 +2356,7 @@ pub fn call_sp_linalg_schur<'ctx, G: CodeGenerator + ?Sized>(
             unsupported_type(ctx, FN_NAME, &[x1_ty]);
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
 
         let dim0 = unsafe {
             n1.shape()
@@ -2397,7 +2399,7 @@ pub fn call_sp_linalg_hessenberg<'ctx, G: CodeGenerator + ?Sized>(
             unsupported_type(ctx, FN_NAME, &[x1_ty]);
         };
 
-        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, n1_elem_ty, None, llvm_usize, None);
 
         let dim0 = unsafe {
             n1.shape()

nac3core/src/codegen/expr.rs

@@ -32,7 +32,7 @@ use super::{
         gen_for_callback_incrementing, gen_if_callback, gen_if_else_expr_callback, gen_raise,
         gen_var,
     },
-    types::{ListType, ProxyType},
+    types::{ListType, NDArrayType, ProxyType},
     values::{
         ArrayLikeIndexer, ArrayLikeValue, ListValue, NDArrayValue, ProxyValue, RangeValue,
         TypedArrayLikeAccessor, UntypedArrayLikeAccessor,
@@ -43,7 +43,7 @@ use crate::{
     symbol_resolver::{SymbolValue, ValueEnum},
     toplevel::{
         helper::PrimDef,
-        numpy::{make_ndarray_ty, unpack_ndarray_var_tys},
+        numpy::unpack_ndarray_var_tys,
         DefinitionId, TopLevelDef,
     },
     typecheck::{
@@ -1570,12 +1570,14 @@ pub fn gen_binop_expr_with_values<'ctx, G: CodeGenerator>(
             let left_val = NDArrayValue::from_pointer_value(
                 left_val.into_pointer_value(),
                 llvm_ndarray_dtype1,
+                None,
                 llvm_usize,
                 None,
             );
             let right_val = NDArrayValue::from_pointer_value(
                 right_val.into_pointer_value(),
                 llvm_ndarray_dtype2,
+                None,
                 llvm_usize,
                 None,
             );
@@ -1631,6 +1633,7 @@ pub fn gen_binop_expr_with_values<'ctx, G: CodeGenerator>(
             let ndarray_val = NDArrayValue::from_pointer_value(
                 if is_ndarray1 { left_val } else { right_val }.into_pointer_value(),
                 llvm_ndarray_dtype,
+                None,
                 llvm_usize,
                 None,
             );
@@ -1828,6 +1831,7 @@ pub fn gen_unaryop_expr_with_values<'ctx, G: CodeGenerator>(
         let val = NDArrayValue::from_pointer_value(
             val.into_pointer_value(),
             llvm_ndarray_dtype,
+            None,
             llvm_usize,
             None,
         );
@@ -1926,6 +1930,7 @@ pub fn gen_cmpop_expr_with_values<'ctx, G: CodeGenerator>(
                 let left_val = NDArrayValue::from_pointer_value(
                     lhs.into_pointer_value(),
                     llvm_ndarray_dtype1,
+                    None,
                     llvm_usize,
                     None,
                 );
@@ -2590,14 +2595,6 @@ fn gen_ndarray_subscript_expr<'ctx, G: CodeGenerator>(
         _ => 1,
     };
 
-    let ndarray_ndims_ty = ctx.unifier.get_fresh_literal(
-        ndims.iter().map(|v| SymbolValue::U64(v - subscripted_dims)).collect(),
-        None,
-    );
-    let ndarray_ty =
-        make_ndarray_ty(&mut ctx.unifier, &ctx.primitives, Some(ty), Some(ndarray_ndims_ty));
-    let llvm_pndarray_t = ctx.get_llvm_type(generator, ndarray_ty).into_pointer_type();
-    let llvm_ndarray_t = llvm_pndarray_t.get_element_type().into_struct_type();
     let llvm_ndarray_data_t = ctx.get_llvm_type(generator, ty).as_basic_type_enum();
     let sizeof_elem = llvm_ndarray_data_t.size_of().unwrap();
 
@@ -2792,25 +2789,15 @@ fn gen_ndarray_subscript_expr<'ctx, G: CodeGenerator>(
 
                 let Some(index_addr) = make_indices_arr(generator, ctx)? else { return Ok(None) };
 
+                let num_dims = v.load_ndims(ctx);
+                let num_dims = ctx.builder
+                    .build_int_sub(num_dims, llvm_usize.const_int(1, false), "")
+                    .unwrap();
+
                 // Create a new array, remove the top dimension from the dimension-size-list, and copy the
                 // elements over
-                let subscripted_ndarray =
+                let ndarray = NDArrayType::new(generator, ctx.ctx, llvm_ndarray_data_t)
-                    generator.gen_var_alloc(ctx, llvm_ndarray_t.into(), None)?;
+                    .construct_uninitialized(generator, ctx, num_dims, None);
-                let ndarray = NDArrayValue::from_pointer_value(
-                    subscripted_ndarray,
-                    llvm_ndarray_data_t,
-                    llvm_usize,
-                    None,
-                );
-
-                let num_dims = v.load_ndims(ctx);
-                ndarray.store_ndims(
-                    ctx,
-                    generator,
-                    ctx.builder
-                        .build_int_sub(num_dims, llvm_usize.const_int(1, false), "")
-                        .unwrap(),
-                );
 
                 let ndarray_num_dims = ndarray.load_ndims(ctx);
                 ndarray.create_shape(ctx, llvm_usize, ndarray_num_dims);
@@ -2842,7 +2829,7 @@ fn gen_ndarray_subscript_expr<'ctx, G: CodeGenerator>(
                     llvm_i1.const_zero(),
                 );
 
-                let ndarray_num_elems = call_ndarray_calc_size(
+                let ndarray_num_elems = ndarray::call_ndarray_calc_size(
                     generator,
                     ctx,
                     &ndarray.shape().as_slice_value(ctx, generator),
@@ -2852,7 +2839,9 @@ fn gen_ndarray_subscript_expr<'ctx, G: CodeGenerator>(
                     .builder
                     .build_int_z_extend_or_bit_cast(ndarray_num_elems, sizeof_elem.get_type(), "")
                     .unwrap();
-                ndarray.create_data(ctx, llvm_ndarray_data_t, ndarray_num_elems);
+                unsafe {
+                    ndarray.create_data(generator, ctx, ndarray_num_elems);
+                }
 
                 let v_data_src_ptr = v.data().ptr_offset(ctx, generator, &index_addr, None);
                 call_memcpy_generic(
@@ -3547,7 +3536,7 @@ pub fn gen_expr<'ctx, G: CodeGenerator>(
                     } else {
                         return Ok(None);
                     };
-                    let v = NDArrayValue::from_pointer_value(v, llvm_ty, usize, None);
+                    let v = NDArrayValue::from_pointer_value(v, llvm_ty, None, usize, None);
 
                     return gen_ndarray_subscript_expr(generator, ctx, *ty, *ndims, v, slice);
                 }
@@ -3598,3 +3587,90 @@ pub fn gen_expr<'ctx, G: CodeGenerator>(
         _ => unimplemented!(),
     }))
 }
+
+/// Creates a function in the current module and inserts a `call` instruction into the LLVM IR.
+pub fn create_fn_and_call<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    fn_name: &str,
+    ret_type: Option<BasicTypeEnum<'ctx>>,
+    (params, is_var_args): (&[BasicTypeEnum<'ctx>], bool),
+    args: &[BasicValueEnum<'ctx>],
+    call_value_name: Option<&str>,
+    configure: Option<&dyn Fn(&FunctionValue<'ctx>)>,
+) -> Option<BasicValueEnum<'ctx>> {
+    let intrinsic_fn = ctx.module.get_function(fn_name).unwrap_or_else(|| {
+        let params = params.iter().copied().map(BasicTypeEnum::into).collect_vec();
+        let fn_type = if let Some(ret_type) = ret_type {
+            ret_type.fn_type(params.as_slice(), is_var_args)
+        } else {
+            ctx.ctx.void_type().fn_type(params.as_slice(), is_var_args)
+        };
+
+        ctx.module.add_function(fn_name, fn_type, None)
+    });
+
+    if let Some(configure) = configure {
+        configure(&intrinsic_fn);
+    }
+
+    let args = args.iter().copied().map(BasicValueEnum::into).collect_vec();
+    ctx.builder
+        .build_call(intrinsic_fn, args.as_slice(), call_value_name.unwrap_or_default())
+        .map(CallSiteValue::try_as_basic_value)
+        .map(Either::left)
+        .unwrap()
+}
+
+/// Creates a function in the current module and inserts a `call` instruction into the LLVM IR.
+///
+/// This is a wrapper around [`create_fn_and_call`] for non-vararg function. This function allows
+/// parameters and arguments to be specified as tuples to better indicate the expected type and
+/// actual value of each parameter-argument pair of the call.
+pub fn create_and_call_function<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    fn_name: &str,
+    ret_type: Option<BasicTypeEnum<'ctx>>,
+    params: &[(BasicTypeEnum<'ctx>, BasicValueEnum<'ctx>)],
+    value_name: Option<&str>,
+    configure: Option<&dyn Fn(&FunctionValue<'ctx>)>,
+) -> Option<BasicValueEnum<'ctx>> {
+    let param_tys = params.iter().map(|(ty, _)| ty).copied().map(BasicTypeEnum::into).collect_vec();
+    let arg_values =
+        params.iter().map(|(_, value)| value).copied().map(BasicValueEnum::into).collect_vec();
+
+    create_fn_and_call(
+        ctx,
+        fn_name,
+        ret_type,
+        (param_tys.as_slice(), false),
+        arg_values.as_slice(),
+        value_name,
+        configure,
+    )
+}
+
+/// Creates a function in the current module and inserts a `call` instruction into the LLVM IR.
+///
+/// This is a wrapper around [`create_fn_and_call`] for non-vararg function. This function allows
+/// only arguments to be specified and performs inference for the parameter types using
+/// [`BasicValueEnum::get_type`].
+pub fn infer_and_call_function<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    fn_name: &str,
+    ret_type: Option<BasicTypeEnum<'ctx>>,
+    args: &[BasicValueEnum<'ctx>],
+    value_name: Option<&str>,
+    configure: Option<&dyn Fn(&FunctionValue<'ctx>)>,
+) -> Option<BasicValueEnum<'ctx>> {
+    let param_tys = args.iter().map(BasicValueEnum::get_type).collect_vec();
+
+    create_fn_and_call(
+        ctx,
+        fn_name,
+        ret_type,
+        (param_tys.as_slice(), false),
+        args,
+        value_name,
+        configure,
+    )
+}

nac3core/src/codegen/irrt/mod.rs

@@ -13,12 +13,11 @@ use super::{CodeGenContext, CodeGenerator};
 use crate::{symbol_resolver::SymbolResolver, typecheck::typedef::Type};
 pub use list::*;
 pub use math::*;
-pub use ndarray::*;
 pub use slice::*;
 
 mod list;
 mod math;
-mod ndarray;
+pub mod ndarray;
 mod slice;
 
 #[must_use]
@@ -60,6 +59,27 @@ pub fn load_irrt<'ctx>(ctx: &'ctx Context, symbol_resolver: &dyn SymbolResolver)
     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: &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
+}
+
 /// NOTE: the output value of the end index of this function should be compared ***inclusively***,
 /// because python allows `a[2::-1]`, whose semantic is `[a[2], a[1], a[0]]`, which is equivalent to
 /// NO numeric slice in python.

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

@@ -0,0 +1,258 @@
+use inkwell::{
+    values::{BasicValueEnum, IntValue, PointerValue},
+    AddressSpace,
+};
+
+use crate::codegen::{
+    expr::create_and_call_function,
+    irrt::get_usize_dependent_function_name,
+    types::NDArrayType,
+    values::{NDArrayValue, ProxyValue},
+    CodeGenContext, CodeGenerator,
+};
+
+pub fn call_nac3_ndarray_util_assert_shape_no_negative<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut 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: &mut G,
+    ctx: &mut 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: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: NDArrayValue<'ctx>,
+) -> IntValue<'ctx> {
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_ndarray = NDArrayType::llvm_type(ctx.ctx, llvm_usize);
+
+    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: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: NDArrayValue<'ctx>,
+) -> IntValue<'ctx> {
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_ndarray = NDArrayType::llvm_type(ctx.ctx, llvm_usize);
+
+    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: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: NDArrayValue<'ctx>,
+) -> IntValue<'ctx> {
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_ndarray = NDArrayType::llvm_type(ctx.ctx, llvm_usize);
+
+    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: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: NDArrayValue<'ctx>,
+) -> IntValue<'ctx> {
+    let llvm_i1 = ctx.ctx.bool_type();
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_ndarray = NDArrayType::llvm_type(ctx.ctx, llvm_usize);
+
+    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: &mut G,
+    ctx: &mut 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 = NDArrayType::llvm_type(ctx.ctx, llvm_usize);
+
+    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: &mut G,
+    ctx: &mut 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 = NDArrayType::llvm_type(ctx.ctx, llvm_usize);
+
+    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: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: NDArrayValue<'ctx>,
+) {
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_ndarray = NDArrayType::llvm_type(ctx.ctx, llvm_usize);
+
+    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: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    src_ndarray: NDArrayValue<'ctx>,
+    dst_ndarray: NDArrayValue<'ctx>,
+) {
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_ndarray = NDArrayType::llvm_type(ctx.ctx, llvm_usize);
+
+    let name = get_usize_dependent_function_name(generator, ctx, "__nac3_ndarray_copy_data");
+
+    create_and_call_function(
+        ctx,
+        &name,
+        None,
+        &[
+            (llvm_ndarray.into(), src_ndarray.as_base_value().into()),
+            (llvm_ndarray.into(), dst_ndarray.as_base_value().into()),
+        ],
+        None,
+        None,
+    );
+}

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

@@ -15,6 +15,9 @@ use crate::codegen::{
     },
     CodeGenContext, CodeGenerator,
 };
+pub use basic::*;
+
+mod basic;
 
 /// Generates a call to `__nac3_ndarray_calc_size`. Returns an [`IntValue`] representing the
 /// calculated total size.

nac3core/src/codegen/llvm_intrinsics.rs

@@ -201,6 +201,52 @@ 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 (rather than 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_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:
@@ -343,3 +389,25 @@ 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

@@ -1119,3 +1119,106 @@ 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_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_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/numpy.rs

@@ -3,14 +3,18 @@ use inkwell::{
     values::{BasicValue, BasicValueEnum, IntValue, PointerValue},
     AddressSpace, IntPredicate, OptimizationLevel,
 };
+use itertools::Itertools;
 
 use nac3parser::ast::{Operator, StrRef};
 
 use super::{
     expr::gen_binop_expr_with_values,
     irrt::{
-        calculate_len_for_slice_range, call_ndarray_calc_broadcast,
+        calculate_len_for_slice_range,
-        call_ndarray_calc_broadcast_index, call_ndarray_calc_nd_indices, call_ndarray_calc_size,
+        ndarray::{
+            call_ndarray_calc_broadcast, call_ndarray_calc_broadcast_index,
+            call_ndarray_calc_nd_indices, call_ndarray_calc_size,
+        },
     },
     llvm_intrinsics::{self, call_memcpy_generic},
     macros::codegen_unreachable,
@@ -27,7 +31,7 @@ use crate::{
     symbol_resolver::ValueEnum,
     toplevel::{
         helper::{arraylike_flatten_element_type, PrimDef},
-        numpy::{make_ndarray_ty, unpack_ndarray_var_tys},
+        numpy::unpack_ndarray_var_tys,
         DefinitionId,
     },
     typecheck::{
@@ -37,25 +41,23 @@ use crate::{
 };
 
 /// Creates an uninitialized `NDArray` instance.
+#[deprecated = "Use NDArrayType::construct_uninitialized instead."]
 fn create_ndarray_uninitialized<'ctx, G: CodeGenerator + ?Sized>(
     generator: &mut G,
     ctx: &mut CodeGenContext<'ctx, '_>,
     elem_ty: Type,
 ) -> Result<NDArrayValue<'ctx>, String> {
     let llvm_elem_ty = ctx.get_llvm_type(generator, elem_ty);
-    let ndarray_ty = make_ndarray_ty(&mut ctx.unifier, &ctx.primitives, Some(elem_ty), None);
-
     let llvm_usize = generator.get_size_type(ctx.ctx);
 
-    let llvm_ndarray_t = ctx
+    let llvm_ndarray_t = NDArrayType::new(generator, ctx.ctx, llvm_elem_ty)
-        .get_llvm_type(generator, ndarray_ty)
+        .as_base_type()
-        .into_pointer_type()
         .get_element_type()
         .into_struct_type();
 
     let ndarray = generator.gen_var_alloc(ctx, llvm_ndarray_t.into(), None)?;
 
-    Ok(NDArrayValue::from_pointer_value(ndarray, llvm_elem_ty, llvm_usize, None))
+    Ok(NDArrayValue::from_pointer_value(ndarray, llvm_elem_ty, None, llvm_usize, None))
 }
 
 /// Creates an `NDArray` instance from a dynamic shape.
@@ -83,6 +85,7 @@ where
     ) -> Result<IntValue<'ctx>, String>,
 {
     let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_elem_ty = ctx.get_llvm_type(generator, elem_ty);
 
     // Assert that all dimensions are non-negative
     let shape_len = shape_len_fn(generator, ctx, shape)?;
@@ -122,10 +125,10 @@ where
         llvm_usize.const_int(1, false),
     )?;
 
-    let ndarray = create_ndarray_uninitialized(generator, ctx, elem_ty)?;
-
     let num_dims = shape_len_fn(generator, ctx, shape)?;
-    ndarray.store_ndims(ctx, generator, num_dims);
+
+    let ndarray = NDArrayType::new(generator, ctx.ctx, llvm_elem_ty)
+        .construct_uninitialized(generator, ctx, num_dims, None);
 
     let ndarray_num_dims = ndarray.load_ndims(ctx);
     ndarray.create_shape(ctx, llvm_usize, ndarray_num_dims);
@@ -189,28 +192,10 @@ pub fn create_ndarray_const_shape<'ctx, G: CodeGenerator + ?Sized>(
         // TODO: Disallow dim_sz > u32_MAX
     }
 
-    let ndarray = create_ndarray_uninitialized(generator, ctx, elem_ty)?;
+    let llvm_dtype = ctx.get_llvm_type(generator, elem_ty);
-
-    let num_dims = llvm_usize.const_int(shape.len() as u64, false);
-    ndarray.store_ndims(ctx, generator, num_dims);
-
-    let ndarray_num_dims = ndarray.load_ndims(ctx);
-    ndarray.create_shape(ctx, llvm_usize, ndarray_num_dims);
-
-    for (i, &shape_dim) in shape.iter().enumerate() {
-        let shape_dim = ctx.builder.build_int_z_extend(shape_dim, llvm_usize, "").unwrap();
-        let ndarray_dim = unsafe {
-            ndarray.shape().ptr_offset_unchecked(
-                ctx,
-                generator,
-                &llvm_usize.const_int(i as u64, true),
-                None,
-            )
-        };
-
-        ctx.builder.build_store(ndarray_dim, shape_dim).unwrap();
-    }
 
+    let ndarray = NDArrayType::new(generator, ctx.ctx, llvm_dtype)
+        .construct_dyn_shape(generator, ctx, shape, None);
     let ndarray = ndarray_init_data(generator, ctx, elem_ty, ndarray);
 
     Ok(ndarray)
@@ -232,7 +217,9 @@ fn ndarray_init_data<'ctx, G: CodeGenerator + ?Sized>(
         &ndarray.shape().as_slice_value(ctx, generator),
         (None, None),
     );
-    ndarray.create_data(ctx, llvm_ndarray_data_t, ndarray_num_elems);
+    unsafe {
+        ndarray.create_data(generator, ctx, ndarray_num_elems);
+    }
 
     ndarray
 }
@@ -338,20 +325,24 @@ fn call_ndarray_empty_impl<'ctx, G: CodeGenerator + ?Sized>(
             // Get the length/size of the tuple, which also happens to be the value of `ndims`.
             let ndims = shape_tuple.get_type().count_fields();
 
-            let mut shape = Vec::with_capacity(ndims as usize);
+            let shape = (0..ndims)
-            for dim_i in 0..ndims {
+                .map(|dim_i| {
-                let dim = ctx
+                    ctx.builder
-                    .builder
                         .build_extract_value(shape_tuple, dim_i, format!("dim{dim_i}").as_str())
+                        .map(BasicValueEnum::into_int_value)
+                        .map(|v| {
+                            ctx.builder.build_int_z_extend_or_bit_cast(v, llvm_usize, "").unwrap()
+                        })
                         .unwrap()
-                    .into_int_value();
+                })
+                .collect_vec();
 
-                shape.push(dim);
-            }
             create_ndarray_const_shape(generator, ctx, elem_ty, shape.as_slice())
         }
         BasicValueEnum::IntValue(shape_int) => {
             // 3. A scalar int; e.g., `np.empty(3)`, this is functionally equivalent to `np.empty([3])`
+            let shape_int =
+                ctx.builder.build_int_z_extend_or_bit_cast(shape_int, llvm_usize, "").unwrap();
 
             create_ndarray_const_shape(generator, ctx, elem_ty, &[shape_int])
         }
@@ -505,6 +496,7 @@ where
         let lhs_val = NDArrayValue::from_pointer_value(
             lhs_val.into_pointer_value(),
             llvm_lhs_elem_ty,
+            None,
             llvm_usize,
             None,
         );
@@ -517,6 +509,7 @@ where
         let rhs_val = NDArrayValue::from_pointer_value(
             rhs_val.into_pointer_value(),
             llvm_rhs_elem_ty,
+            None,
             llvm_usize,
             None,
         );
@@ -532,6 +525,7 @@ where
             let lhs = NDArrayValue::from_pointer_value(
                 lhs_val.into_pointer_value(),
                 llvm_lhs_elem_ty,
+                None,
                 llvm_usize,
                 None,
             );
@@ -548,6 +542,7 @@ where
             let rhs = NDArrayValue::from_pointer_value(
                 rhs_val.into_pointer_value(),
                 llvm_rhs_elem_ty,
+                None,
                 llvm_usize,
                 None,
             );
@@ -706,7 +701,8 @@ fn llvm_arraylike_get_ndims<'ctx, G: CodeGenerator + ?Sized>(
         {
             let dtype = arraylike_flatten_element_type(&mut ctx.unifier, ty);
             let llvm_elem_ty = ctx.get_llvm_type(generator, dtype);
-            NDArrayValue::from_pointer_value(v, llvm_elem_ty, llvm_usize, None).load_ndims(ctx)
+            NDArrayValue::from_pointer_value(v, llvm_elem_ty, None, llvm_usize, None)
+                .load_ndims(ctx)
         }
 
         BasicValueEnum::PointerValue(v) if ListValue::is_representable(v, llvm_usize).is_ok() => {
@@ -856,7 +852,7 @@ fn call_ndarray_array_impl<'ctx, G: CodeGenerator + ?Sized>(
     // object is an NDArray instance - copy object unless copy=0 && ndmin < object.ndims
     if NDArrayValue::is_representable(object, llvm_usize).is_ok() {
         let llvm_elem_ty = ctx.get_llvm_type(generator, elem_ty);
-        let object = NDArrayValue::from_pointer_value(object, llvm_elem_ty, llvm_usize, None);
+        let object = NDArrayValue::from_pointer_value(object, llvm_elem_ty, None, llvm_usize, None);
 
         let ndarray = gen_if_else_expr_callback(
             generator,
@@ -932,6 +928,7 @@ fn call_ndarray_array_impl<'ctx, G: CodeGenerator + ?Sized>(
         return Ok(NDArrayValue::from_pointer_value(
             ndarray.map(BasicValueEnum::into_pointer_value).unwrap(),
             llvm_elem_ty,
+            None,
             llvm_usize,
             None,
         ));
@@ -1269,6 +1266,7 @@ pub fn ndarray_sliced_copy<'ctx, G: CodeGenerator + ?Sized>(
 ) -> Result<NDArrayValue<'ctx>, String> {
     let llvm_i32 = ctx.ctx.i32_type();
     let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_elem_ty = ctx.get_llvm_type(generator, elem_ty);
 
     let ndarray = if slices.is_empty() {
         create_ndarray_dyn_shape(
@@ -1282,8 +1280,8 @@ pub fn ndarray_sliced_copy<'ctx, G: CodeGenerator + ?Sized>(
             },
         )?
     } else {
-        let ndarray = create_ndarray_uninitialized(generator, ctx, elem_ty)?;
+        let ndarray = NDArrayType::new(generator, ctx.ctx, llvm_elem_ty)
-        ndarray.store_ndims(ctx, generator, this.load_ndims(ctx));
+            .construct_uninitialized(generator, ctx, this.load_ndims(ctx), None);
 
         let ndims = this.load_ndims(ctx);
         ndarray.create_shape(ctx, llvm_usize, ndims);
@@ -1465,6 +1463,7 @@ where
             let lhs_val = NDArrayValue::from_pointer_value(
                 lhs_val.into_pointer_value(),
                 llvm_lhs_elem_ty,
+                None,
                 llvm_usize,
                 None,
             );
@@ -1473,6 +1472,7 @@ where
             let rhs_val = NDArrayValue::from_pointer_value(
                 rhs_val.into_pointer_value(),
                 llvm_rhs_elem_ty,
+                None,
                 llvm_usize,
                 None,
             );
@@ -1499,6 +1499,7 @@ where
             let ndarray = NDArrayValue::from_pointer_value(
                 if lhs_scalar { rhs_val } else { lhs_val }.into_pointer_value(),
                 llvm_elem_ty,
+                None,
                 llvm_usize,
                 None,
             );
@@ -2061,6 +2062,7 @@ pub fn gen_ndarray_copy<'ctx>(
         NDArrayValue::from_pointer_value(
             this_arg.into_pointer_value(),
             llvm_elem_ty,
+            None,
             llvm_usize,
             None,
         ),
@@ -2098,7 +2100,7 @@ pub fn gen_ndarray_fill<'ctx>(
     ndarray_fill_flattened(
         generator,
         context,
-        NDArrayValue::from_pointer_value(this_arg, llvm_elem_ty, llvm_usize, None),
+        NDArrayValue::from_pointer_value(this_arg, llvm_elem_ty, None, llvm_usize, None),
         |generator, ctx, _| {
             let value = if value_arg.is_pointer_value() {
                 let llvm_i1 = ctx.ctx.bool_type();
@@ -2140,7 +2142,7 @@ pub fn ndarray_transpose<'ctx, G: CodeGenerator + ?Sized>(
     if let BasicValueEnum::PointerValue(n1) = x1 {
         let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
         let llvm_elem_ty = ctx.get_llvm_type(generator, elem_ty);
-        let n1 = NDArrayValue::from_pointer_value(n1, llvm_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, llvm_elem_ty, None, llvm_usize, None);
         let n_sz = call_ndarray_calc_size(generator, ctx, &n1.shape(), (None, None));
 
         // Dimensions are reversed in the transposed array
@@ -2260,7 +2262,7 @@ pub fn ndarray_reshape<'ctx, G: CodeGenerator + ?Sized>(
     if let BasicValueEnum::PointerValue(n1) = x1 {
         let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
         let llvm_elem_ty = ctx.get_llvm_type(generator, elem_ty);
-        let n1 = NDArrayValue::from_pointer_value(n1, llvm_elem_ty, llvm_usize, None);
+        let n1 = NDArrayValue::from_pointer_value(n1, llvm_elem_ty, None, llvm_usize, None);
         let n_sz = call_ndarray_calc_size(generator, ctx, &n1.shape(), (None, None));
 
         let acc = generator.gen_var_alloc(ctx, llvm_usize.into(), None)?;
@@ -2548,8 +2550,8 @@ pub fn ndarray_dot<'ctx, G: CodeGenerator + ?Sized>(
             let llvm_n1_data_ty = ctx.get_llvm_type(generator, n1_dtype);
             let llvm_n2_data_ty = ctx.get_llvm_type(generator, n2_dtype);
 
-            let n1 = NDArrayValue::from_pointer_value(n1, llvm_n1_data_ty, llvm_usize, None);
+            let n1 = NDArrayValue::from_pointer_value(n1, llvm_n1_data_ty, None, llvm_usize, None);
-            let n2 = NDArrayValue::from_pointer_value(n2, llvm_n2_data_ty, llvm_usize, None);
+            let n2 = NDArrayValue::from_pointer_value(n2, llvm_n2_data_ty, None, llvm_usize, None);
 
             let n1_sz = call_ndarray_calc_size(generator, ctx, &n1.shape(), (None, None));
             let n2_sz = call_ndarray_calc_size(generator, ctx, &n1.shape(), (None, None));

nac3core/src/codegen/types/ndarray.rs

@@ -1,5 +1,5 @@
 use inkwell::{
-    context::Context,
+    context::{AsContextRef, Context},
     types::{AnyTypeEnum, BasicType, BasicTypeEnum, IntType, PointerType},
     values::{IntValue, PointerValue},
     AddressSpace,
@@ -12,9 +12,13 @@ use super::{
     structure::{StructField, StructFields},
     ProxyType,
 };
-use crate::codegen::{
+use crate::{
-    values::{ArraySliceValue, NDArrayValue, ProxyValue},
+    codegen::{
+        values::{ArraySliceValue, NDArrayValue, ProxyValue, TypedArrayLikeMutator},
         {CodeGenContext, CodeGenerator},
+    },
+    toplevel::numpy::unpack_ndarray_var_tys,
+    typecheck::typedef::Type,
 };
 
 /// Proxy type for a `ndarray` type in LLVM.
@@ -27,10 +31,14 @@ pub struct NDArrayType<'ctx> {
 
 #[derive(PartialEq, Eq, Clone, Copy, StructFields)]
 pub struct NDArrayStructFields<'ctx> {
+    #[value_type(usize)]
+    pub itemsize: StructField<'ctx, IntValue<'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(usize.ptr_type(AddressSpace::default()))]
+    pub strides: StructField<'ctx, PointerValue<'ctx>>,
     #[value_type(i8_type().ptr_type(AddressSpace::default()))]
     pub data: StructField<'ctx, PointerValue<'ctx>>,
 }
@@ -41,70 +49,45 @@ impl<'ctx> NDArrayType<'ctx> {
         llvm_ty: PointerType<'ctx>,
         llvm_usize: IntType<'ctx>,
     ) -> Result<(), String> {
+        let ctx = llvm_ty.get_context();
+
+        let llvm_expected_ty = Self::fields(ctx, llvm_usize).into_vec();
+
         let llvm_ndarray_ty = llvm_ty.get_element_type();
         let AnyTypeEnum::StructType(llvm_ndarray_ty) = llvm_ndarray_ty else {
             return Err(format!("Expected struct type for `NDArray` type, got {llvm_ndarray_ty}"));
         };
-        if llvm_ndarray_ty.count_fields() != 3 {
+        if llvm_ndarray_ty.count_fields() != u32::try_from(llvm_expected_ty.len()).unwrap() {
             return Err(format!(
-                "Expected 3 fields in `NDArray`, got {}",
+                "Expected {} fields in `NDArray`, got {}",
+                llvm_expected_ty.len(),
                 llvm_ndarray_ty.count_fields()
             ));
         }
 
-        let ndarray_ndims_ty = llvm_ndarray_ty.get_field_type_at_index(0).unwrap();
+        llvm_expected_ty
-        let Ok(ndarray_ndims_ty) = IntType::try_from(ndarray_ndims_ty) else {
+            .iter()
-            return Err(format!("Expected int type for `ndarray.0`, got {ndarray_ndims_ty}"));
+            .enumerate()
-        };
+            .map(|(i, expected_ty)| {
-        if ndarray_ndims_ty.get_bit_width() != llvm_usize.get_bit_width() {
+                (expected_ty.1, llvm_ndarray_ty.get_field_type_at_index(i as u32).unwrap())
-            return Err(format!(
+            })
-                "Expected {}-bit int type for `ndarray.0`, got {}-bit int",
+            .try_for_each(|(expected_ty, actual_ty)| {
-                llvm_usize.get_bit_width(),
+                if expected_ty == actual_ty {
-                ndarray_ndims_ty.get_bit_width()
+                    Ok(())
-            ));
+                } else {
-        }
+                    Err(format!("Expected {expected_ty} for `ndarray.data`, got {actual_ty}"))
-
-        let ndarray_dims_ty = llvm_ndarray_ty.get_field_type_at_index(1).unwrap();
-        let Ok(ndarray_pdims) = PointerType::try_from(ndarray_dims_ty) else {
-            return Err(format!("Expected pointer type for `ndarray.1`, got {ndarray_dims_ty}"));
-        };
-        let ndarray_dims = ndarray_pdims.get_element_type();
-        let Ok(ndarray_dims) = IntType::try_from(ndarray_dims) else {
-            return Err(format!(
-                "Expected pointer-to-int type for `ndarray.1`, got pointer-to-{ndarray_dims}"
-            ));
-        };
-        if ndarray_dims.get_bit_width() != llvm_usize.get_bit_width() {
-            return Err(format!(
-                "Expected pointer-to-{}-bit int type for `ndarray.1`, got pointer-to-{}-bit int",
-                llvm_usize.get_bit_width(),
-                ndarray_dims.get_bit_width()
-            ));
-        }
-
-        let ndarray_data_ty = llvm_ndarray_ty.get_field_type_at_index(2).unwrap();
-        let Ok(ndarray_pdata) = PointerType::try_from(ndarray_data_ty) else {
-            return Err(format!("Expected pointer type for `ndarray.2`, got {ndarray_data_ty}"));
-        };
-        let ndarray_data = ndarray_pdata.get_element_type();
-        let Ok(ndarray_data) = IntType::try_from(ndarray_data) else {
-            return Err(format!(
-                "Expected pointer-to-int type for `ndarray.2`, got pointer-to-{ndarray_data}"
-            ));
-        };
-        if ndarray_data.get_bit_width() != 8 {
-            return Err(format!(
-                "Expected pointer-to-8-bit int type for `ndarray.1`, got pointer-to-{}-bit int",
-                ndarray_data.get_bit_width()
-            ));
                 }
+            })?;
 
         Ok(())
     }
 
-    // TODO: Move this into e.g. StructProxyType
+    // TODO: Move this as a member of this Struct
     #[must_use]
-    fn fields(ctx: &'ctx Context, llvm_usize: IntType<'ctx>) -> NDArrayStructFields<'ctx> {
+    fn fields(
+        ctx: impl AsContextRef<'ctx>,
+        llvm_usize: IntType<'ctx>,
+    ) -> NDArrayStructFields<'ctx> {
         NDArrayStructFields::new(ctx, llvm_usize)
     }
 
@@ -112,7 +95,7 @@ impl<'ctx> NDArrayType<'ctx> {
     #[must_use]
     pub fn get_fields(
         &self,
-        ctx: &'ctx Context,
+        ctx: impl AsContextRef<'ctx>,
         llvm_usize: IntType<'ctx>,
     ) -> NDArrayStructFields<'ctx> {
         Self::fields(ctx, llvm_usize)
@@ -120,8 +103,8 @@ impl<'ctx> NDArrayType<'ctx> {
 
     /// Creates an LLVM type corresponding to the expected structure of an `NDArray`.
     #[must_use]
-    fn llvm_type(ctx: &'ctx Context, llvm_usize: IntType<'ctx>) -> PointerType<'ctx> {
+    pub fn llvm_type(ctx: &'ctx Context, llvm_usize: IntType<'ctx>) -> PointerType<'ctx> {
-        // struct NDArray { num_dims: size_t, dims: size_t*, data: i8* }
+        // struct NDArray { data: i8*, itemsize: size_t, ndims: size_t, shape: size_t*, strides: size_t* }
         //
         // * data    : Pointer to an array containing the array data
         // * itemsize: The size of each NDArray elements in bytes
@@ -147,6 +130,21 @@ impl<'ctx> NDArrayType<'ctx> {
         NDArrayType { ty: llvm_ndarray, dtype, llvm_usize }
     }
 
+    /// Creates an [`NDArrayType`] 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);
+
+        NDArrayType { ty: Self::llvm_type(ctx.ctx, llvm_usize), dtype: llvm_dtype, llvm_usize }
+    }
+
     /// Creates an [`NDArrayType`] from a [`PointerType`] representing an `NDArray`.
     #[must_use]
     pub fn from_type(
@@ -165,7 +163,7 @@ impl<'ctx> NDArrayType<'ctx> {
         self.as_base_type()
             .get_element_type()
             .into_struct_type()
-            .get_field_type_at_index(0)
+            .get_field_type_at_index(1)
             .map(BasicTypeEnum::into_int_type)
             .unwrap()
     }
@@ -175,6 +173,119 @@ impl<'ctx> NDArrayType<'ctx> {
     pub fn element_type(&self) -> BasicTypeEnum<'ctx> {
         self.dtype
     }
+
+    /// 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.
+    #[must_use]
+    pub fn construct_uninitialized<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        // ndims: u64,
+        ndims: IntValue<'ctx>,
+        name: Option<&'ctx str>,
+    ) -> <Self as ProxyType<'ctx>>::Value {
+        let ndarray = self.new_value(generator, ctx, name);
+
+        let itemsize =
+            ctx.builder.build_int_cast(self.dtype.size_of().unwrap(), self.llvm_usize, "").unwrap();
+        ndarray.store_itemsize(ctx, generator, itemsize);
+
+        ndarray.store_ndims(ctx, generator, ndims);
+
+        ndarray.create_shape(ctx, self.llvm_usize, ndims);
+        ndarray.create_strides(ctx, self.llvm_usize, ndims);
+
+        ndarray
+    }
+
+    /// Convenience function. Allocate an [`NDArrayObject`] with a statically known shape.
+    ///
+    /// The returned [`NDArrayObject`]'s `data` and `strides` are uninitialized.
+    #[must_use]
+    pub fn construct_const_shape<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        shape: &[u64],
+        name: Option<&'ctx str>,
+    ) -> <Self as ProxyType<'ctx>>::Value {
+        let llvm_usize = generator.get_size_type(ctx.ctx);
+
+        let ndarray = self.construct_uninitialized(
+            generator,
+            ctx,
+            llvm_usize.const_int(shape.len() as u64, false),
+            name,
+        );
+
+        // Write shape
+        let ndarray_shape = ndarray.shape();
+        for (i, dim) in shape.iter().enumerate() {
+            let dim = self.llvm_usize.const_int(*dim, false);
+            unsafe {
+                ndarray_shape.set_typed_unchecked(
+                    ctx,
+                    generator,
+                    &self.llvm_usize.const_int(i as u64, false),
+                    dim,
+                );
+            }
+        }
+
+        ndarray
+    }
+
+    /// Convenience function. Allocate an [`NDArrayObject`] with a dynamically known shape.
+    ///
+    /// The returned [`NDArrayObject`]'s `data` and `strides` are uninitialized.
+    #[must_use]
+    pub fn construct_dyn_shape<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        shape: &[IntValue<'ctx>],
+        name: Option<&'ctx str>,
+    ) -> <Self as ProxyType<'ctx>>::Value {
+        let llvm_usize = generator.get_size_type(ctx.ctx);
+
+        let ndarray = self.construct_uninitialized(
+            generator,
+            ctx,
+            llvm_usize.const_int(shape.len() as u64, false),
+            name,
+        );
+
+        // Write shape
+        let ndarray_shape = ndarray.shape();
+        for (i, dim) in shape.iter().enumerate() {
+            assert_eq!(
+                dim.get_type(),
+                self.llvm_usize,
+                "Expected {} but got {}",
+                self.llvm_usize.print_to_string(),
+                dim.get_type().print_to_string()
+            );
+            unsafe {
+                ndarray_shape.set_typed_unchecked(
+                    ctx,
+                    generator,
+                    &self.llvm_usize.const_int(i as u64, false),
+                    *dim,
+                );
+            }
+        }
+
+        ndarray
+    }
 }
 
 impl<'ctx> ProxyType<'ctx> for NDArrayType<'ctx> {
@@ -243,7 +354,7 @@ impl<'ctx> ProxyType<'ctx> for NDArrayType<'ctx> {
     ) -> Self::Value {
         debug_assert_eq!(value.get_type(), self.as_base_type());
 
-        NDArrayValue::from_pointer_value(value, self.dtype, self.llvm_usize, name)
+        NDArrayValue::from_pointer_value(value, self.dtype, None, self.llvm_usize, name)
     }
 
     fn as_base_type(&self) -> Self::Base {

nac3core/src/codegen/types/structure.rs

@@ -145,7 +145,7 @@ where
     }
 
     /// Sets the value of this field for a given `obj`.
-    pub fn set_from_value(&self, obj: StructValue<'ctx>, value: Value) {
+    pub fn set_for_value(&self, obj: StructValue<'ctx>, value: Value) {
         obj.set_field_at_index(self.index, value);
     }
 

nac3core/src/codegen/values/ndarray.rs

@@ -9,10 +9,11 @@ use super::{
     UntypedArrayLikeAccessor, UntypedArrayLikeMutator,
 };
 use crate::codegen::{
-    irrt::{call_ndarray_calc_size, call_ndarray_flatten_index},
+    irrt,
-    llvm_intrinsics::call_int_umin,
+    llvm_intrinsics::{call_int_umin, call_memcpy_generic_array},
     stmt::gen_for_callback_incrementing,
-    types::NDArrayType,
+    type_aligned_alloca,
+    types::{structure::StructField, NDArrayType},
     CodeGenContext, CodeGenerator,
 };
 
@@ -21,6 +22,7 @@ use crate::codegen::{
 pub struct NDArrayValue<'ctx> {
     value: PointerValue<'ctx>,
     dtype: BasicTypeEnum<'ctx>,
+    ndims: Option<u64>,
     llvm_usize: IntType<'ctx>,
     name: Option<&'ctx str>,
 }
@@ -40,12 +42,13 @@ impl<'ctx> NDArrayValue<'ctx> {
     pub fn from_pointer_value(
         ptr: PointerValue<'ctx>,
         dtype: BasicTypeEnum<'ctx>,
+        ndims: Option<u64>,
         llvm_usize: IntType<'ctx>,
         name: Option<&'ctx str>,
     ) -> Self {
         debug_assert!(Self::is_representable(ptr, llvm_usize).is_ok());
 
-        NDArrayValue { value: ptr, dtype, llvm_usize, name }
+        NDArrayValue { value: ptr, dtype, ndims, llvm_usize, name }
     }
 
     /// Returns the pointer to the field storing the number of dimensions of this `NDArray`.
@@ -75,6 +78,29 @@ impl<'ctx> NDArrayValue<'ctx> {
         ctx.builder.build_load(pndims, "").map(BasicValueEnum::into_int_value).unwrap()
     }
 
+    fn itemsize(&self, ctx: &CodeGenContext<'ctx, '_>) -> StructField<'ctx, IntValue<'ctx>> {
+        self.get_type()
+            .get_fields(ctx.ctx, self.llvm_usize)
+            .itemsize
+    }
+
+    /// Stores the size of each element `itemsize` into this instance.
+    pub fn store_itemsize<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        generator: &G,
+        itemsize: IntValue<'ctx>,
+    ) {
+        debug_assert_eq!(itemsize.get_type(), generator.get_size_type(ctx.ctx));
+
+        self.itemsize(ctx).set(ctx, self.value, itemsize, self.name);
+    }
+
+    /// Returns the size of each element of this `NDArray` as a value.
+    pub fn load_itemsize(&self, ctx: &CodeGenContext<'ctx, '_>) -> IntValue<'ctx> {
+        self.itemsize(ctx).get(ctx, self.value, self.name)
+    }
+
     /// Returns the double-indirection pointer to the `shape` array, as if by calling
     /// `getelementptr` on the field.
     fn ptr_to_shape(&self, ctx: &CodeGenContext<'ctx, '_>) -> PointerValue<'ctx> {
@@ -105,6 +131,36 @@ impl<'ctx> NDArrayValue<'ctx> {
         NDArrayShapeProxy(self)
     }
 
+    /// Returns the double-indirection pointer to the `stride` array, as if by calling
+    /// `getelementptr` on the field.
+    fn ptr_to_strides(&self, ctx: &CodeGenContext<'ctx, '_>) -> PointerValue<'ctx> {
+        self.get_type()
+            .get_fields(ctx.ctx, self.llvm_usize)
+            .strides
+            .ptr_by_gep(ctx, self.value, self.name)
+    }
+
+    /// Stores the array of dimension sizes `dims` into this instance.
+    fn store_strides(&self, ctx: &CodeGenContext<'ctx, '_>, dims: PointerValue<'ctx>) {
+        ctx.builder.build_store(self.ptr_to_strides(ctx), dims).unwrap();
+    }
+
+    /// Convenience method for creating a new array storing the stride with the given `size`.
+    pub fn create_strides(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        llvm_usize: IntType<'ctx>,
+        size: IntValue<'ctx>,
+    ) {
+        self.store_strides(ctx, ctx.builder.build_array_alloca(llvm_usize, size, "").unwrap());
+    }
+
+    /// Returns a proxy object to the field storing the stride of each dimension of this `NDArray`.
+    #[must_use]
+    pub fn strides(&self) -> NDArrayStridesProxy<'ctx, '_> {
+        NDArrayStridesProxy(self)
+    }
+
     /// Returns the double-indirection pointer to the `data` array, as if by calling `getelementptr`
     /// on the field.
     pub fn ptr_to_data(&self, ctx: &CodeGenContext<'ctx, '_>) -> PointerValue<'ctx> {
@@ -125,21 +181,29 @@ impl<'ctx> NDArrayValue<'ctx> {
 
     /// Convenience method for creating a new array storing data elements with the given element
     /// type `elem_ty` and `size`.
-    pub fn create_data(
+    ///
+    /// The data buffer will be allocated on the stack, and is considered to be owned by this ndarray instance.
+    ///
+    /// # Safety
+    ///
+    /// `shape` and `itemsize` of the ndarray must be initialized.
+    pub unsafe fn create_data<G: CodeGenerator + ?Sized>(
         &self,
-        ctx: &CodeGenContext<'ctx, '_>,
+        generator: &mut G,
-        elem_ty: BasicTypeEnum<'ctx>,
+        ctx: &mut CodeGenContext<'ctx, '_>,
         size: IntValue<'ctx>,
     ) {
+        // let itemsize =
+        //     ctx.builder.build_int_cast(self.load_itemsize(ctx), size.get_type(), "").unwrap();
         let itemsize =
-            ctx.builder.build_int_cast(elem_ty.size_of().unwrap(), size.get_type(), "").unwrap();
+            ctx.builder.build_int_cast(self.dtype.size_of().unwrap(), size.get_type(), "").unwrap();
         let nbytes = ctx.builder.build_int_mul(size, itemsize, "").unwrap();
+        // let nbytes = self.nbytes(generator, ctx);
 
-        // TODO: What about alignment?
+        let data = type_aligned_alloca(generator, ctx, self.dtype, nbytes, None);
-        self.store_data(
+        self.store_data(ctx, data);
-            ctx,
+
-            ctx.builder.build_array_alloca(ctx.ctx.i8_type(), nbytes, "").unwrap(),
+        // self.set_strides_contiguous(generator, ctx);
-        );
     }
 
     /// Returns a proxy object to the field storing the data of this `NDArray`.
@@ -147,6 +211,133 @@ impl<'ctx> NDArrayValue<'ctx> {
     pub fn data(&self) -> NDArrayDataProxy<'ctx, '_> {
         NDArrayDataProxy(self)
     }
+
+    /// Copy shape dimensions from an array.
+    pub fn copy_shape_from_array<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        shape: PointerValue<'ctx>,
+    ) {
+        let num_items = self.load_ndims(ctx);
+
+        call_memcpy_generic_array(
+            ctx,
+            self.shape().base_ptr(ctx, generator),
+            shape,
+            num_items,
+            ctx.ctx.bool_type().const_zero(),
+        );
+    }
+
+    /// 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: NDArrayValue<'ctx>,
+    ) {
+        assert_eq!(self.ndims, src_ndarray.ndims);
+        let src_shape = src_ndarray.shape().base_ptr(ctx, generator);
+        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: PointerValue<'ctx>,
+    ) {
+        let num_items = self.load_ndims(ctx);
+
+        call_memcpy_generic_array(
+            ctx,
+            self.strides().base_ptr(ctx, generator),
+            strides,
+            num_items,
+            ctx.ctx.bool_type().const_zero(),
+        );
+    }
+
+    /// 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: NDArrayValue<'ctx>,
+    ) {
+        assert_eq!(self.ndims, src_ndarray.ndims);
+        let src_strides = src_ndarray.strides().base_ptr(ctx, generator);
+        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, '_>,
+    ) -> IntValue<'ctx> {
+        irrt::ndarray::call_nac3_ndarray_size(generator, ctx, *self)
+    }
+
+    /// Get the `ndarray.nbytes` of this ndarray.
+    pub fn nbytes<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> IntValue<'ctx> {
+        irrt::ndarray::call_nac3_ndarray_nbytes(generator, ctx, *self)
+    }
+
+    /// Get the `len()` of this ndarray.
+    pub fn len<G: CodeGenerator + ?Sized>(
+        &self,
+        generator: &mut G,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> IntValue<'ctx> {
+        irrt::ndarray::call_nac3_ndarray_len(generator, ctx, *self)
+    }
+
+    /// 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, '_>,
+    ) -> IntValue<'ctx> {
+        irrt::ndarray::call_nac3_ndarray_is_c_contiguous(generator, ctx, *self)
+    }
+
+    /// 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, '_>,
+    ) {
+        irrt::ndarray::call_nac3_ndarray_set_strides_by_shape(generator, ctx, *self);
+    }
+
+    /// 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: NDArrayValue<'ctx>,
+    ) {
+        assert_eq!(self.dtype, src.dtype, "self and src dtype should match");
+        irrt::ndarray::call_nac3_ndarray_copy_data(generator, ctx, src, *self);
+    }
 }
 
 impl<'ctx> ProxyValue<'ctx> for NDArrayValue<'ctx> {
@@ -168,103 +359,6 @@ impl<'ctx> From<NDArrayValue<'ctx>> for PointerValue<'ctx> {
     }
 }
 
-/// Proxy type for accessing the `dims` array of an `NDArray` instance in LLVM.
-#[derive(Copy, Clone)]
-pub struct NDArrayShapeProxy<'ctx, 'a>(&'a NDArrayValue<'ctx>);
-
-impl<'ctx> ArrayLikeValue<'ctx> for NDArrayShapeProxy<'ctx, '_> {
-    fn element_type<G: CodeGenerator + ?Sized>(
-        &self,
-        ctx: &CodeGenContext<'ctx, '_>,
-        generator: &G,
-    ) -> AnyTypeEnum<'ctx> {
-        self.0.shape().base_ptr(ctx, generator).get_type().get_element_type()
-    }
-
-    fn base_ptr<G: CodeGenerator + ?Sized>(
-        &self,
-        ctx: &CodeGenContext<'ctx, '_>,
-        _: &G,
-    ) -> PointerValue<'ctx> {
-        let var_name = self.0.name.map(|v| format!("{v}.data")).unwrap_or_default();
-
-        ctx.builder
-            .build_load(self.0.ptr_to_shape(ctx), var_name.as_str())
-            .map(BasicValueEnum::into_pointer_value)
-            .unwrap()
-    }
-
-    fn size<G: CodeGenerator + ?Sized>(
-        &self,
-        ctx: &CodeGenContext<'ctx, '_>,
-        _: &G,
-    ) -> IntValue<'ctx> {
-        self.0.load_ndims(ctx)
-    }
-}
-
-impl<'ctx> ArrayLikeIndexer<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {
-    unsafe fn ptr_offset_unchecked<G: CodeGenerator + ?Sized>(
-        &self,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        generator: &mut G,
-        idx: &IntValue<'ctx>,
-        name: Option<&str>,
-    ) -> PointerValue<'ctx> {
-        let var_name = name.map(|v| format!("{v}.addr")).unwrap_or_default();
-
-        unsafe {
-            ctx.builder
-                .build_in_bounds_gep(self.base_ptr(ctx, generator), &[*idx], var_name.as_str())
-                .unwrap()
-        }
-    }
-
-    fn ptr_offset<G: CodeGenerator + ?Sized>(
-        &self,
-        ctx: &mut CodeGenContext<'ctx, '_>,
-        generator: &mut G,
-        idx: &IntValue<'ctx>,
-        name: Option<&str>,
-    ) -> PointerValue<'ctx> {
-        let size = self.size(ctx, generator);
-        let in_range = ctx.builder.build_int_compare(IntPredicate::ULT, *idx, size, "").unwrap();
-        ctx.make_assert(
-            generator,
-            in_range,
-            "0:IndexError",
-            "index {0} is out of bounds for axis 0 with size {1}",
-            [Some(*idx), Some(self.0.load_ndims(ctx)), None],
-            ctx.current_loc,
-        );
-
-        unsafe { self.ptr_offset_unchecked(ctx, generator, idx, name) }
-    }
-}
-
-impl<'ctx> UntypedArrayLikeAccessor<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {}
-impl<'ctx> UntypedArrayLikeMutator<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {}
-
-impl<'ctx> TypedArrayLikeAccessor<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {
-    fn downcast_to_type(
-        &self,
-        _: &mut CodeGenContext<'ctx, '_>,
-        value: BasicValueEnum<'ctx>,
-    ) -> IntValue<'ctx> {
-        value.into_int_value()
-    }
-}
-
-impl<'ctx> TypedArrayLikeMutator<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {
-    fn upcast_from_type(
-        &self,
-        _: &mut CodeGenContext<'ctx, '_>,
-        value: IntValue<'ctx>,
-    ) -> BasicValueEnum<'ctx> {
-        value.into()
-    }
-}
-
 /// Proxy type for accessing the `data` array of an `NDArray` instance in LLVM.
 #[derive(Copy, Clone)]
 pub struct NDArrayDataProxy<'ctx, 'a>(&'a NDArrayValue<'ctx>);
@@ -296,7 +390,12 @@ impl<'ctx> ArrayLikeValue<'ctx> for NDArrayDataProxy<'ctx, '_> {
         ctx: &CodeGenContext<'ctx, '_>,
         generator: &G,
     ) -> IntValue<'ctx> {
-        call_ndarray_calc_size(generator, ctx, &self.as_slice_value(ctx, generator), (None, None))
+        irrt::ndarray::call_ndarray_calc_size(
+            generator,
+            ctx,
+            &self.as_slice_value(ctx, generator),
+            (None, None),
+        )
     }
 }
 
@@ -405,7 +504,7 @@ impl<'ctx, Index: UntypedArrayLikeAccessor<'ctx>> ArrayLikeIndexer<'ctx, Index>
             indices_elem_ty.get_bit_width()
         );
 
-        let index = call_ndarray_flatten_index(generator, ctx, *self.0, indices);
+        let index = irrt::ndarray::call_ndarray_flatten_index(generator, ctx, *self.0, indices);
         let sizeof_elem = ctx
             .builder
             .build_int_truncate_or_bit_cast(
@@ -521,3 +620,197 @@ impl<'ctx, Index: UntypedArrayLikeAccessor<'ctx>> UntypedArrayLikeMutator<'ctx,
     for NDArrayDataProxy<'ctx, '_>
 {
 }
+
+/// Proxy type for accessing the `dims` array of an `NDArray` instance in LLVM.
+#[derive(Copy, Clone)]
+pub struct NDArrayShapeProxy<'ctx, 'a>(&'a NDArrayValue<'ctx>);
+
+impl<'ctx> ArrayLikeValue<'ctx> for NDArrayShapeProxy<'ctx, '_> {
+    fn element_type<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        generator: &G,
+    ) -> AnyTypeEnum<'ctx> {
+        self.0.shape().base_ptr(ctx, generator).get_type().get_element_type()
+    }
+
+    fn base_ptr<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        _: &G,
+    ) -> PointerValue<'ctx> {
+        let var_name = self.0.name.map(|v| format!("{v}.data")).unwrap_or_default();
+
+        ctx.builder
+            .build_load(self.0.ptr_to_shape(ctx), var_name.as_str())
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap()
+    }
+
+    fn size<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        _: &G,
+    ) -> IntValue<'ctx> {
+        self.0.load_ndims(ctx)
+    }
+}
+
+impl<'ctx> ArrayLikeIndexer<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {
+    unsafe fn ptr_offset_unchecked<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        generator: &mut G,
+        idx: &IntValue<'ctx>,
+        name: Option<&str>,
+    ) -> PointerValue<'ctx> {
+        let var_name = name.map(|v| format!("{v}.addr")).unwrap_or_default();
+
+        unsafe {
+            ctx.builder
+                .build_in_bounds_gep(self.base_ptr(ctx, generator), &[*idx], var_name.as_str())
+                .unwrap()
+        }
+    }
+
+    fn ptr_offset<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        generator: &mut G,
+        idx: &IntValue<'ctx>,
+        name: Option<&str>,
+    ) -> PointerValue<'ctx> {
+        let size = self.size(ctx, generator);
+        let in_range = ctx.builder.build_int_compare(IntPredicate::ULT, *idx, size, "").unwrap();
+        ctx.make_assert(
+            generator,
+            in_range,
+            "0:IndexError",
+            "index {0} is out of bounds for axis 0 with size {1}",
+            [Some(*idx), Some(self.0.load_ndims(ctx)), None],
+            ctx.current_loc,
+        );
+
+        unsafe { self.ptr_offset_unchecked(ctx, generator, idx, name) }
+    }
+}
+
+impl<'ctx> UntypedArrayLikeAccessor<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {}
+impl<'ctx> UntypedArrayLikeMutator<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {}
+
+impl<'ctx> TypedArrayLikeAccessor<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {
+    fn downcast_to_type(
+        &self,
+        _: &mut CodeGenContext<'ctx, '_>,
+        value: BasicValueEnum<'ctx>,
+    ) -> IntValue<'ctx> {
+        value.into_int_value()
+    }
+}
+
+impl<'ctx> TypedArrayLikeMutator<'ctx, IntValue<'ctx>> for NDArrayShapeProxy<'ctx, '_> {
+    fn upcast_from_type(
+        &self,
+        _: &mut CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> BasicValueEnum<'ctx> {
+        value.into()
+    }
+}
+
+/// Proxy type for accessing the `dims` array of an `NDArray` instance in LLVM.
+#[derive(Copy, Clone)]
+pub struct NDArrayStridesProxy<'ctx, 'a>(&'a NDArrayValue<'ctx>);
+
+impl<'ctx> ArrayLikeValue<'ctx> for NDArrayStridesProxy<'ctx, '_> {
+    fn element_type<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        generator: &G,
+    ) -> AnyTypeEnum<'ctx> {
+        self.0.strides().base_ptr(ctx, generator).get_type().get_element_type()
+    }
+
+    fn base_ptr<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        _: &G,
+    ) -> PointerValue<'ctx> {
+        let var_name = self.0.name.map(|v| format!("{v}.strides")).unwrap_or_default();
+
+        ctx.builder
+            .build_load(self.0.ptr_to_strides(ctx), var_name.as_str())
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap()
+    }
+
+    fn size<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        _: &G,
+    ) -> IntValue<'ctx> {
+        self.0.load_ndims(ctx)
+    }
+}
+
+impl<'ctx> ArrayLikeIndexer<'ctx, IntValue<'ctx>> for NDArrayStridesProxy<'ctx, '_> {
+    unsafe fn ptr_offset_unchecked<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        generator: &mut G,
+        idx: &IntValue<'ctx>,
+        name: Option<&str>,
+    ) -> PointerValue<'ctx> {
+        let var_name = name.map(|v| format!("{v}.addr")).unwrap_or_default();
+
+        unsafe {
+            ctx.builder
+                .build_in_bounds_gep(self.base_ptr(ctx, generator), &[*idx], var_name.as_str())
+                .unwrap()
+        }
+    }
+
+    fn ptr_offset<G: CodeGenerator + ?Sized>(
+        &self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        generator: &mut G,
+        idx: &IntValue<'ctx>,
+        name: Option<&str>,
+    ) -> PointerValue<'ctx> {
+        let size = self.size(ctx, generator);
+        let in_range = ctx.builder.build_int_compare(IntPredicate::ULT, *idx, size, "").unwrap();
+        ctx.make_assert(
+            generator,
+            in_range,
+            "0:IndexError",
+            "index {0} is out of bounds for axis 0 with size {1}",
+            [Some(*idx), Some(self.0.load_ndims(ctx)), None],
+            ctx.current_loc,
+        );
+
+        unsafe { self.ptr_offset_unchecked(ctx, generator, idx, name) }
+    }
+}
+
+impl<'ctx> UntypedArrayLikeAccessor<'ctx, IntValue<'ctx>> for NDArrayStridesProxy<'ctx, '_> {}
+impl<'ctx> UntypedArrayLikeMutator<'ctx, IntValue<'ctx>> for NDArrayStridesProxy<'ctx, '_> {}
+
+impl<'ctx> TypedArrayLikeAccessor<'ctx, IntValue<'ctx>> for NDArrayStridesProxy<'ctx, '_> {
+    fn downcast_to_type(
+        &self,
+        _: &mut CodeGenContext<'ctx, '_>,
+        value: BasicValueEnum<'ctx>,
+    ) -> IntValue<'ctx> {
+        value.into_int_value()
+    }
+}
+
+impl<'ctx> TypedArrayLikeMutator<'ctx, IntValue<'ctx>> for NDArrayStridesProxy<'ctx, '_> {
+    fn upcast_from_type(
+        &self,
+        _: &mut CodeGenContext<'ctx, '_>,
+        value: IntValue<'ctx>,
+    ) -> BasicValueEnum<'ctx> {
+        value.into()
+    }
+}

nac3standalone/demo/src/ndarray.py

@@ -1759,14 +1759,14 @@ def run() -> int32:
     test_ndarray_reshape()
 
     test_ndarray_dot()
-    test_ndarray_cholesky()
+    # test_ndarray_cholesky()
-    test_ndarray_qr()
+    # test_ndarray_qr()
-    test_ndarray_svd()
+    # test_ndarray_svd()
-    test_ndarray_linalg_inv()
+    # test_ndarray_linalg_inv()
-    test_ndarray_pinv()
+    # test_ndarray_pinv()
-    test_ndarray_matrix_power()
+    # test_ndarray_matrix_power()
-    test_ndarray_det()
+    # test_ndarray_det()
-    test_ndarray_lu()
+    # test_ndarray_lu()
-    test_ndarray_schur()
+    # test_ndarray_schur()
-    test_ndarray_hessenberg()
+    # test_ndarray_hessenberg()
     return 0