Skip to content

Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.

20.5 KB
Raw Blame
package mlx

// #include "generated.h"
import "C"

import (
	"reflect"
	"unsafe"
)

// Nvfp4MaxProduct is the product of the maximum E4M3 and E2M1 values. Every
// globalScale below is in MLX's representation: a checkpoint's multiplier m
// (ModelOpt's weight_scale_2) is held as m*Nvfp4MaxProduct, so Nvfp4MaxProduct
// itself is the identity. Wrappers that hand the scale to MLX pass it through;
// wrappers that apply it themselves divide it back out.
const Nvfp4MaxProduct = 448 * 6

// globalScaleFactor converts a stored global scale into the multiplier an
// output needs.
func globalScaleFactor(scale *Array) *Array {
	return DivScalar(scale, Nvfp4MaxProduct)
}

// identityGlobalScale leaves a value unchanged.
func identityGlobalScale() *Array {
	return FromValue(float32(Nvfp4MaxProduct))
}

// scaleAndCast applies a global scale to an output and casts back, fusing the
// divide, multiply and cast into one kernel. Eagerly this is several: MLX
// promotes binary operands with an astype, so a bf16 input times a float32
// scale round-trips through a full-size float32 intermediate.
var scaleAndCast = Compile2(
	"GlobalScaleOutput",
	func(out, scale *Array) *Array {
		return Mul(out, globalScaleFactor(scale)).AsType(out.DType())
	},
	Shapeless(),
)

// Quantization operations

func Quantize(w *Array, groupSize, bits int, mode string) (weights, scales, biases *Array) {
	cMode := C.CString(mode)
	defer C.free(unsafe.Pointer(cMode))
	optGroupSize := C.mlx_optional_int{value: C.int(groupSize), has_value: true}
	optBits := C.mlx_optional_int{value: C.int(bits), has_value: true}
	res := mlxCheck(C.mlx_vector_array_new())
	defer freeVectorArray(res)
	var globalScale C.mlx_array
	mlxCheck(C.mlx_quantize(&res, w.ctx, optGroupSize, optBits, cMode, globalScale, DefaultStream().ctx))

	vecSize := int(mlxCheck(C.mlx_vector_array_size(res)))
	w0 := New("QUANTIZE_W")
	mlxCheck(C.mlx_vector_array_get(&w0.ctx, res, 0))
	w1 := New("QUANTIZE_S")
	mlxCheck(C.mlx_vector_array_get(&w1.ctx, res, 1))
	if vecSize >= 3 {
		w2 := New("QUANTIZE_B")
		mlxCheck(C.mlx_vector_array_get(&w2.ctx, res, 2))
		return w0, w1, w2
	}
	return w0, w1, nil
}

func FromFP8(x *Array, dtype DType) *Array {
	out := New("FROM_FP8")
	mlxCheck(C.mlx_from_fp8(&out.ctx, x.ctx, C.mlx_dtype(dtype), DefaultStream().ctx))
	return out
}

func ToFP8(x *Array) *Array {
	out := New("TO_FP8")
	mlxCheck(C.mlx_to_fp8(&out.ctx, x.ctx, DefaultStream().ctx))
	return out
}

// Dequantize applies globalScale itself rather than forwarding it: MLX's own
// argument accepts only a scalar, and callers pass per-expert banks.
func Dequantize(w, scales, biases *Array, groupSize, bits int, mode string, globalScale *Array) *Array {
	cMode := C.CString(mode)
	defer C.free(unsafe.Pointer(cMode))
	optGroupSize := C.mlx_optional_int{value: C.int(groupSize), has_value: true}
	optBits := C.mlx_optional_int{value: C.int(bits), has_value: true}
	optDtype := C.mlx_optional_dtype{has_value: false}

	var b C.mlx_array
	if biases != nil {
		b = biases.ctx
	}

	out := New("DEQUANTIZE")
	var noGlobalScale C.mlx_array
	mlxCheck(C.mlx_dequantize(&out.ctx, w.ctx, scales.ctx, b, optGroupSize, optBits, cMode, noGlobalScale, optDtype, DefaultStream().ctx))
	if globalScale != nil {
		gs := globalScale
		for gs.NumDims() < out.NumDims() {
			gs = ExpandDims(gs, -1)
		}
		out = scaleAndCast(out, gs)
	}
	return out
}

// QuantizedMatmul applies globalScale to the output; MLX's quantized_matmul
// has none of its own.
func QuantizedMatmul(x, w, scales, biases *Array, transpose bool, groupSize, bits int, mode string, globalScale *Array) *Array {
	cMode := C.CString(mode)
	defer C.free(unsafe.Pointer(cMode))
	optGroupSize := C.mlx_optional_int{value: C.int(groupSize), has_value: true}
	optBits := C.mlx_optional_int{value: C.int(bits), has_value: true}

	var b C.mlx_array
	if biases != nil {
		b = biases.ctx
	}

	out := New("QUANTIZED_MATMUL")
	mlxCheck(C.mlx_quantized_matmul(&out.ctx, x.ctx, w.ctx, scales.ctx, b, C.bool(transpose), optGroupSize, optBits, cMode, DefaultStream().ctx))
	if globalScale != nil {
		out = scaleAndCast(out, globalScale)
	}
	return out
}

// GatherQMM multiplies x against the quantized weight banks selected by
// lhsIndices and rhsIndices, with one globalScale entry per expert. Metal
// applies it inside the gather kernel, other backends scale the gathered
// output rows in the wrapper.
func GatherQMM(x, w, scales *Array, biases, lhsIndices, rhsIndices *Array, transpose bool, groupSize, bits int, mode string, globalScale *Array, sortedIndices bool) *Array {
	cMode := C.CString(mode)
	defer C.free(unsafe.Pointer(cMode))
	optGroupSize := C.mlx_optional_int{value: C.int(groupSize), has_value: true}
	optBits := C.mlx_optional_int{value: C.int(bits), has_value: true}

	var b, lhs, rhs, gs C.mlx_array
	if biases != nil {
		b = biases.ctx
	}
	if lhsIndices != nil {
		lhs = lhsIndices.ctx
	}
	if rhsIndices != nil {
		rhs = rhsIndices.ctx
	}
	// The wrapper fallback needs rhs indices to map output rows to experts;
	// without them the native path reports the unsupported combination.
	applyWrapperScale := globalScale != nil && !MetalIsAvailable() && rhsIndices != nil
	if globalScale != nil && !applyWrapperScale {
		gs = globalScale.ctx
	}

	out := New("GATHER_QMM")
	mlxCheck(C.mlx_gather_qmm(&out.ctx, x.ctx, w.ctx, scales.ctx, b, lhs, rhs, C.bool(transpose), optGroupSize, optBits, cMode, gs, C.bool(sortedIndices), DefaultStream().ctx))
	if applyWrapperScale {
		out = mulGatherQMMGlobalScale(out, globalScale, rhsIndices)
	}
	return out
}

// mulGatherQMMGlobalScale scales gathered output rows by the per-expert
// global scale.
func mulGatherQMMGlobalScale(out, globalScale, rhsIndices *Array) *Array {
	// The indices address the weight's flattened batch dimensions, so the bank
	// has to be flat too before taking from it.
	perExpert := Reshape(globalScale, int32(globalScale.Size()))
	scaleRows := Take(perExpert, rhsIndices, 0)
	// out is the broadcast index shape plus the two matmul axes, and rhsIndices
	// may be shorter, so pad left before appending those axes.
	for scaleRows.NumDims() < out.NumDims()-2 {
		scaleRows = ExpandDims(scaleRows, 0)
	}
	scaleRows = ExpandDims(ExpandDims(scaleRows, -1), -1)
	return scaleAndCast(out, scaleRows)
}

// Missing tensor ops

func Arange(start, stop, step float64, dtype DType) *Array {
	out := New("ARANGE")
	mlxCheck(C.mlx_arange(&out.ctx, C.double(start), C.double(stop), C.double(step), C.mlx_dtype(dtype), DefaultStream().ctx))
	return out
}

func Tile(a *Array, reps []int32) *Array {
	cReps := make([]C.int, len(reps))
	for i, r := range reps {
		cReps[i] = C.int(r)
	}
	out := New("TILE")
	mlxCheck(C.mlx_tile(&out.ctx, a.ctx, unsafe.SliceData(cReps), C.size_t(len(reps)), DefaultStream().ctx))
	return out
}

// BroadcastTo broadcasts an array to the given shape.
func BroadcastTo(a *Array, shape ...int32) *Array {
	cShape := make([]C.int, len(shape))
	for i, s := range shape {
		cShape[i] = C.int(s)
	}
	out := New("BROADCAST_TO")
	mlxCheck(C.mlx_broadcast_to(&out.ctx, a.ctx, unsafe.SliceData(cShape), C.size_t(len(shape)), DefaultStream().ctx))
	return out
}

func Tri(n, m int32, k int) *Array {
	out := New("TRI")
	mlxCheck(C.mlx_tri(&out.ctx, C.int(n), C.int(m), C.int(k), C.mlx_dtype(DTypeFloat32), DefaultStream().ctx))
	return out
}

func Where(condition, a, b *Array) *Array {
	out := New("WHERE")
	mlxCheck(C.mlx_where(&out.ctx, condition.ctx, a.ctx, b.ctx, DefaultStream().ctx))
	return out
}

func Conv1d(x, weight *Array, bias *Array, stride, padding, dilation, groups int32) *Array {
	out := New("CONV1D")
	mlxCheck(C.mlx_conv1d(
		&out.ctx,
		x.ctx,
		weight.ctx,
		C.int(stride),
		C.int(padding),
		C.int(dilation),
		C.int(groups),
		DefaultStream().ctx,
	))
	if bias != nil {
		out = Add(out, bias)
	}
	return out
}

func Contiguous(a *Array, allowColMajor bool) *Array {
	out := New("CONTIGUOUS")
	mlxCheck(C.mlx_contiguous(&out.ctx, a.ctx, C.bool(allowColMajor), DefaultStream().ctx))
	return out
}

// Conv2d performs 2D convolution: x [N,H,W,C_in], weight [C_out,kH,kW,C_in].
// MLX uses NHWC layout.
func Conv2d(x, weight *Array, strideH, strideW, padH, padW, dilationH, dilationW, groups int32) *Array {
	out := New("CONV2D")
	mlxCheck(C.mlx_conv2d(
		&out.ctx,
		x.ctx,
		weight.ctx,
		C.int(strideH), C.int(strideW),
		C.int(padH), C.int(padW),
		C.int(dilationH), C.int(dilationW),
		C.int(groups),
		DefaultStream().ctx,
	))
	return out
}

// Pad pads array a along the given axes with specified low/high pad sizes.
// mode should be "constant", "edge", or "reflect".
func Pad(a *Array, axes []int, lowPad, highPad []int, padValue *Array, mode string) *Array {
	cAxes := make([]C.int, len(axes))
	cLow := make([]C.int, len(lowPad))
	cHigh := make([]C.int, len(highPad))
	for i := range axes {
		cAxes[i] = C.int(axes[i])
		cLow[i] = C.int(lowPad[i])
		cHigh[i] = C.int(highPad[i])
	}
	cMode := C.CString(mode)
	defer C.free(unsafe.Pointer(cMode))
	out := New("PAD")
	mlxCheck(C.mlx_pad(
		&out.ctx,
		a.ctx,
		unsafe.SliceData(cAxes), C.size_t(len(cAxes)),
		unsafe.SliceData(cLow), C.size_t(len(cLow)),
		unsafe.SliceData(cHigh), C.size_t(len(cHigh)),
		padValue.ctx,
		cMode,
		DefaultStream().ctx,
	))
	return out
}

// PadConstant pads with zeros along the given axes.
func PadConstant(a *Array, axes []int, lowPad, highPad []int) *Array {
	zero := NewScalarArray(float32(0))
	return Pad(a, axes, lowPad, highPad, zero, "constant")
}

func DepthwiseConv1d(x, weight *Array, bias *Array) *Array {
	groups := int32(x.Dim(x.NumDims() - 1))
	return Conv1d(x, weight, bias, 1, 0, 1, groups)
}

// Maximum returns element-wise maximum of two arrays.
func Maximum(a, b *Array) *Array {
	out := New("MAXIMUM")
	mlxCheck(C.mlx_maximum(&out.ctx, a.ctx, b.ctx, DefaultStream().ctx))
	return out
}

// Minimum returns element-wise minimum of two arrays.
func Minimum(a, b *Array) *Array {
	out := New("MINIMUM")
	mlxCheck(C.mlx_minimum(&out.ctx, a.ctx, b.ctx, DefaultStream().ctx))
	return out
}

// Softplus computes log(1 + exp(x)) using logaddexp for numerical stability.
func Softplus(a *Array) *Array {
	return Logaddexp(a, Zeros(a.DType(), a.Dims()...))
}

// ReLU computes max(0, x).
func ReLU(a *Array) *Array {
	return Maximum(a, NewScalarArray(float32(0)))
}

// GLU applies Gated Linear Unit: splits x along last dim into two halves,
// returns first * sigmoid(second).
func GLU(a *Array) *Array {
	lastDim := a.NumDims() - 1
	halfSize := a.Dim(lastDim) / 2
	first := SliceStartStop(a,
		make([]int32, lastDim+1), // all zeros for start
		appendDims(a, lastDim, int32(halfSize)),
	)
	second := SliceStartStop(a,
		appendDimsStart(a, lastDim, int32(halfSize)),
		appendDims(a, lastDim, int32(a.Dim(lastDim))),
	)
	return first.Multiply(second.Sigmoid())
}

// helper: builds stop array for SliceStartStop where the target axis = val
func appendDims(a *Array, targetAxis int, val int32) []int32 {
	n := a.NumDims()
	out := make([]int32, n)
	for i := range n {
		if i == targetAxis {
			out[i] = val
		} else {
			out[i] = int32(a.Dim(i))
		}
	}
	return out
}

// helper: builds start array for SliceStartStop where the target axis = val
func appendDimsStart(a *Array, targetAxis int, val int32) []int32 {
	n := a.NumDims()
	out := make([]int32, n)
	for i := range n {
		if i == targetAxis {
			out[i] = val
		}
	}
	return out
}

// Clamp clamps array values to [min, max].
func Clamp(a *Array, minVal, maxVal float32) *Array {
	return Minimum(Maximum(a, NewScalarArray(minVal)), NewScalarArray(maxVal))
}

// Convenience wrappers (function-style for the model code)

func Stack(arrays []*Array, axis int) *Array {
	vectorData := make([]C.mlx_array, len(arrays))
	for i := range arrays {
		vectorData[i] = arrays[i].ctx
	}
	vector := mlxCheck(C.mlx_vector_array_new_data(unsafe.SliceData(vectorData), C.size_t(len(vectorData))))
	defer freeVectorArray(vector)

	out := New("STACK")
	mlxCheck(C.mlx_stack_axis(&out.ctx, vector, C.int(axis), DefaultStream().ctx))
	return out
}

func Neg(a *Array) *Array {
	return a.Negative()
}

func Sum(a *Array, axis int, keepDims bool) *Array {
	return a.SumAxis(axis, keepDims)
}

func Argsort(a *Array, axis int) *Array {
	return a.ArgsortAxis(axis)
}

func Take(a *Array, indices *Array, axis int) *Array {
	return a.TakeAxis(indices, axis)
}

func RSqrt(a *Array) *Array {
	out := New("RSQRT")
	mlxCheck(C.mlx_rsqrt(&out.ctx, a.ctx, DefaultStream().ctx))
	return out
}

func Mean(a *Array, axis int, keepDims bool) *Array {
	out := New("MEAN_AXIS")
	mlxCheck(C.mlx_mean_axis(&out.ctx, a.ctx, C.int(axis), C.bool(keepDims), DefaultStream().ctx))
	return out
}

func Argpartition(a *Array, kth int, axis int) *Array {
	return a.ArgpartitionAxis(kth, axis)
}

func TakeAlongAxis(a, indices *Array, axis int) *Array {
	return a.TakeAlongAxis(indices, axis)
}

// Function-style wrappers for model code.

func Add(a, b *Array) *Array {
	return a.Add(b)
}

func Sub(a, b *Array) *Array {
	return a.Subtract(b)
}

func Mul(a, b *Array) *Array {
	return a.Multiply(b)
}

func Div(a, b *Array) *Array {
	return a.Divide(b)
}

func Matmul(a, b *Array) *Array {
	return a.Matmul(b)
}

func Reshape(a *Array, shape ...int32) *Array {
	axes := make([]int, len(shape))
	for i, s := range shape {
		axes[i] = int(s)
	}
	return a.Reshape(axes...)
}

func Transpose(a *Array, axes ...int) *Array {
	return a.Transpose(axes...)
}

func ExpandDims(a *Array, axis int) *Array {
	return a.ExpandDims(axis)
}

func Squeeze(a *Array, axis int) *Array {
	return a.Squeeze(axis)
}

func Flatten(a *Array) *Array {
	return a.Flatten(0, -1)
}

func Concatenate(arrays []*Array, axis int) *Array {
	if len(arrays) == 0 {
		return nil
	}
	if len(arrays) == 1 {
		return arrays[0].Clone()
	}
	return arrays[0].Concatenate(axis, arrays[1:]...)
}

func SliceStartStop(a *Array, start, stop []int32) *Array {
	n := len(start)
	cStart := make([]C.int, n)
	cStop := make([]C.int, n)
	cStrides := make([]C.int, n)
	for i := range n {
		cStart[i] = C.int(start[i])
		cStop[i] = C.int(stop[i])
		cStrides[i] = 1
	}
	out := New("SLICE")
	mlxCheck(C.mlx_slice(&out.ctx, a.ctx, unsafe.SliceData(cStart), C.size_t(n), unsafe.SliceData(cStop), C.size_t(n), unsafe.SliceData(cStrides), C.size_t(n), DefaultStream().ctx))
	return out
}

func GatherMM(a, b *Array, lhsIndices, rhsIndices *Array, sortedIndices bool) *Array {
	if lhsIndices == nil {
		lhsIndices = New("")
	}
	if rhsIndices == nil {
		rhsIndices = New("")
	}
	return a.GatherMM(b, lhsIndices, rhsIndices, sortedIndices)
}

// RoPEWithBase applies rotary position embeddings to x. offsets is an
// int32 array of shape [B] giving each batch row's starting position;
// the kernel applies positions offsets[b] + 0..T-1 per row.
func RoPEWithBase(x *Array, dims int, traditional bool, base, scale float32, offsets *Array) *Array {
	return RoPEWithFreqs(x, dims, traditional, base, scale, offsets, nil)
}

// RoPEWithFreqs applies RoPE with optional custom frequencies.
// When freqs is non-nil, it is used instead of computing from base.
// Note: MLX takes reciprocal(freqs) internally to get inv_freq, so pass
// the actual frequencies (base^(2i/dim)), not the inverse frequencies.
func RoPEWithFreqs(x *Array, dims int, traditional bool, base, scale float32, offsets *Array, freqs *Array) *Array {
	var freqsCtx C.mlx_array
	var optBase C.mlx_optional_float
	if freqs != nil {
		freqsCtx = freqs.ctx
		optBase = C.mlx_optional_float{has_value: C.bool(false)}
	} else {
		empty := New("")
		freqsCtx = empty.ctx
		optBase = C.mlx_optional_float{
			value:     C.float(base),
			has_value: C.bool(func() bool { return base != 0 }()),
		}
	}
	out := New("FAST_ROPE")
	mlxCheck(C.mlx_fast_rope_dynamic(
		&out.ctx,
		x.ctx,
		C.int(dims),
		C.bool(traditional),
		optBase,
		C.float(scale),
		offsets.ctx,
		freqsCtx,
		DefaultStream().ctx,
	))
	return out
}

func Sigmoid(a *Array) *Array {
	return a.Sigmoid()
}

func Erf(a *Array) *Array {
	out := New("ERF")
	mlxCheck(C.mlx_erf(&out.ctx, a.ctx, DefaultStream().ctx))
	return out
}

func Exp(a *Array) *Array {
	out := New("EXP")
	mlxCheck(C.mlx_exp(&out.ctx, a.ctx, DefaultStream().ctx))
	return out
}

func Log(a *Array) *Array {
	out := New("LOG")
	mlxCheck(C.mlx_log(&out.ctx, a.ctx, DefaultStream().ctx))
	return out
}

func Sin(a *Array) *Array {
	out := New("SIN")
	mlxCheck(C.mlx_sin(&out.ctx, a.ctx, DefaultStream().ctx))
	return out
}

func Cos(a *Array) *Array {
	out := New("COS")
	mlxCheck(C.mlx_cos(&out.ctx, a.ctx, DefaultStream().ctx))
	return out
}

func erf(a *Array) *Array {
	out := New("ERF")
	mlxCheck(C.mlx_erf(&out.ctx, a.ctx, DefaultStream().ctx))
	return out
}

func Clip(a, aMin, aMax *Array) *Array {
	out := New("CLIP")
	mlxCheck(C.mlx_clip(&out.ctx, a.ctx, aMin.ctx, aMax.ctx, DefaultStream().ctx))
	return out
}

func Logaddexp(a, b *Array) *Array {
	out := New("LOGADDEXP")
	mlxCheck(C.mlx_logaddexp(&out.ctx, a.ctx, b.ctx, DefaultStream().ctx))
	return out
}

func SoftmaxAxis(a *Array, axis int, precise bool) *Array {
	out := New("SOFTMAX_AXIS")
	mlxCheck(C.mlx_softmax_axis(&out.ctx, a.ctx, C.int(axis), C.bool(precise), DefaultStream().ctx))
	return out
}

func LayerNormFn(x, weight, bias *Array, eps float32) *Array {
	out := New("FAST_LAYERNORM")
	var w, b C.mlx_array
	if weight != nil {
		w = weight.ctx
	}
	if bias != nil {
		b = bias.ctx
	}
	mlxCheck(C.mlx_fast_layer_norm(&out.ctx, x.ctx, w, b, C.float(eps), DefaultStream().ctx))
	return out
}

func RMSNormFn(x, weight *Array, eps float32) *Array {
	out := New("FAST_RMSNORM")
	var w C.mlx_array
	if weight != nil {
		w = weight.ctx
	}
	mlxCheck(C.mlx_fast_rms_norm(&out.ctx, x.ctx, w, C.float(eps), DefaultStream().ctx))
	return out
}

func AddMM(c, a, b *Array, alpha, beta float32) *Array {
	return c.Addmm(a, b, alpha, beta)
}

// Scalar helpers

// scalarWithDtype creates a scalar array matching the dtype of a.
// Matching dtype is important for graph fusion and avoiding implicit casts.
func scalarWithDtype(s float32, a *Array) C.mlx_array {
	dtype := a.DType()
	f32 := mlxCheck(C.mlx_array_new_float(C.float(s)))
	if dtype == DTypeFloat32 {
		return f32
	}
	defer freeArray(f32)
	casted := mlxCheck(C.mlx_array_new())
	mlxCheck(C.mlx_astype(&casted, f32, C.mlx_dtype(dtype), DefaultStream().ctx))
	return casted
}

func AddScalar(a *Array, s float32) *Array {
	scalar := scalarWithDtype(s, a)
	defer freeArray(scalar)
	out := New("ADD_SCALAR")
	mlxCheck(C.mlx_add(&out.ctx, a.ctx, scalar, DefaultStream().ctx))
	return out
}

func MulScalar(a *Array, s float32) *Array {
	scalar := scalarWithDtype(s, a)
	defer freeArray(scalar)
	out := New("MUL_SCALAR")
	mlxCheck(C.mlx_multiply(&out.ctx, a.ctx, scalar, DefaultStream().ctx))
	return out
}

func DivScalar(a *Array, s float32) *Array {
	scalar := scalarWithDtype(s, a)
	defer freeArray(scalar)
	out := New("DIV_SCALAR")
	mlxCheck(C.mlx_divide(&out.ctx, a.ctx, scalar, DefaultStream().ctx))
	return out
}

func FloorDivideScalar(a *Array, s int32) *Array {
	scalar := FromValue(int(s))
	return a.FloorDivide(scalar)
}

// Array constructors

func NewArrayInt32(data []int32, shape []int32) *Array {
	cShape := make([]C.int, len(shape))
	for i, s := range shape {
		cShape[i] = C.int(s)
	}
	out := New("NEW_ARRAY_INT32")
	out.ctx = mlxCheck(C.mlx_array_new_data(unsafe.Pointer(&data[0]), unsafe.SliceData(cShape), C.int(len(shape)), C.mlx_dtype(DTypeInt32)))
	return out
}

func NewScalarArray(value float32) *Array {
	out := New("SCALAR")
	out.ctx = mlxCheck(C.mlx_array_new_float32(C.float(value)))
	return out
}

func ZerosF32(shape []int32) *Array {
	return Zeros(DTypeFloat32, func() []int {
		ints := make([]int, len(shape))
		for i, s := range shape {
			ints[i] = int(s)
		}
		return ints
	}()...)
}

// Utility

func Collect(v any) []*Array {
	var arrays []*Array
	seen := make(map[uintptr]bool)
	collect(reflect.ValueOf(v), &arrays, seen)
	return arrays
}

func collect(v reflect.Value, arrays *[]*Array, seen map[uintptr]bool) {
	if !v.IsValid() {
		return
	}

	if v.Kind() == reflect.Pointer {
		if v.IsNil() {
			return
		}
		ptr := v.Pointer()
		if seen[ptr] {
			return
		}
		seen[ptr] = true

		if arr, ok := v.Interface().(*Array); ok {
			if arr != nil {
				*arrays = append(*arrays, arr)
			}
			return
		}
		collect(v.Elem(), arrays, seen)
		return
	}

	switch v.Kind() {
	case reflect.Struct:
		// Check if this struct IS an Array (not a pointer to one)
		if arr, ok := v.Addr().Interface().(*Array); ok {
			if arr != nil {
				*arrays = append(*arrays, arr)
			}
			return
		}
		for i := range v.NumField() {
			field := v.Field(i)
			if field.CanInterface() {
				collect(field, arrays, seen)
			}
		}
	case reflect.Slice:
		for i := range v.Len() {
			collect(v.Index(i), arrays, seen)
		}
	case reflect.Map:
		for _, key := range v.MapKeys() {
			collect(v.MapIndex(key), arrays, seen)
		}
	case reflect.Interface:
		if !v.IsNil() {
			collect(v.Elem(), arrays, seen)
		}
	}
}

func EnableCompile() {
	mlxCheck(C.mlx_enable_compile())
}

func DisableCompile() {
	mlxCheck(C.mlx_disable_compile())
}