Simd

clay.simd.Simd (extern class)

Batched geometry kernels operating on raw pointers (native C++).

Every function is a leaf call: no allocation, no callback into haxe, so no GC collection can happen while the provided pointers are in use. Callers must acquire pointers right before the call, keep the owning haxe objects alive on the stack for the duration of the call, and never cache raw pointers across calls.

The vectorized implementations (SSE2 on x86, NEON on ARM) are picked at compile time, with a plain C++ fallback. Define linc_simd_force_scalar to force the plain C++ path everywhere (useful to measure batching gains without vectorization).

Static Members

mode(): Int

Which implementation was compiled in: 0 = scalar, 1 = SSE2, 2 = NEON

Returns
Int

transformAffineIndexedF32(dst: cpp.RawPointer<cpp.Float32>, dstStrideFloats: Int, verts: cpp.RawPointer<cpp.Float32>, vertStrideFloats: Int, indices: cpp.RawPointer<Int>, indexCount: Int, a: cpp.Float32, b: cpp.Float32, c: cpp.Float32, d: cpp.Float32, tx: cpp.Float32, ty: cpp.Float32, z: cpp.Float32, textureSlot: cpp.Float32, writeSlot: Bool): Void

Transforms indexCount points gathered from verts (32-bit floats) via indices with the 2x3 affine matrix (a,b,c,d,tx,ty), writing sequential strided vertices [x y z (slot)] into dst.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`dstStrideFloats`	Int
`verts`	cpp.RawPointer<cpp.Float32>
`vertStrideFloats`	Int
`indices`	cpp.RawPointer<Int>
`indexCount`	Int
`a`	cpp.Float32
`b`	cpp.Float32
`c`	cpp.Float32
`d`	cpp.Float32
`tx`	cpp.Float32
`ty`	cpp.Float32
`z`	cpp.Float32
`textureSlot`	cpp.Float32
`writeSlot`	Bool

transformAffineIndexedF64(dst: cpp.RawPointer<cpp.Float32>, dstStrideFloats: Int, verts: cpp.RawPointer<Float>, vertStrideFloats: Int, indices: cpp.RawPointer<Int>, indexCount: Int, a: cpp.Float32, b: cpp.Float32, c: cpp.Float32, d: cpp.Float32, tx: cpp.Float32, ty: cpp.Float32, z: cpp.Float32, textureSlot: cpp.Float32, writeSlot: Bool): Void

Same as transformAffineIndexedF32 with a 64-bit float source (haxe Array<Float> holds 64-bit values on this target).

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`dstStrideFloats`	Int
`verts`	cpp.RawPointer<Float>
`vertStrideFloats`	Int
`indices`	cpp.RawPointer<Int>
`indexCount`	Int
`a`	cpp.Float32
`b`	cpp.Float32
`c`	cpp.Float32
`d`	cpp.Float32
`tx`	cpp.Float32
`ty`	cpp.Float32
`z`	cpp.Float32
`textureSlot`	cpp.Float32
`writeSlot`	Bool

transformAffineIndexedAttrsF32(dst: cpp.RawPointer<cpp.Float32>, dstStrideFloats: Int, verts: cpp.RawPointer<cpp.Float32>, vertStrideFloats: Int, indices: cpp.RawPointer<Int>, indexCount: Int, a: cpp.Float32, b: cpp.Float32, c: cpp.Float32, d: cpp.Float32, tx: cpp.Float32, ty: cpp.Float32, z: cpp.Float32, textureSlot: cpp.Float32, writeSlot: Bool, srcAttrCount: Int, dstAttrCount: Int): Void

Same as transformAffineIndexedF32 for meshes carrying custom float attributes interleaved with positions ([x y attr0..attrN] source, [x y z (slot) attr0..attrM] destination, zero-padded).

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`dstStrideFloats`	Int
`verts`	cpp.RawPointer<cpp.Float32>
`vertStrideFloats`	Int
`indices`	cpp.RawPointer<Int>
`indexCount`	Int
`a`	cpp.Float32
`b`	cpp.Float32
`c`	cpp.Float32
`d`	cpp.Float32
`tx`	cpp.Float32
`ty`	cpp.Float32
`z`	cpp.Float32
`textureSlot`	cpp.Float32
`writeSlot`	Bool
`srcAttrCount`	Int
`dstAttrCount`	Int

transformAffineIndexedAttrsF64(dst: cpp.RawPointer<cpp.Float32>, dstStrideFloats: Int, verts: cpp.RawPointer<Float>, vertStrideFloats: Int, indices: cpp.RawPointer<Int>, indexCount: Int, a: cpp.Float32, b: cpp.Float32, c: cpp.Float32, d: cpp.Float32, tx: cpp.Float32, ty: cpp.Float32, z: cpp.Float32, textureSlot: cpp.Float32, writeSlot: Bool, srcAttrCount: Int, dstAttrCount: Int): Void

Same as transformAffineIndexedAttrsF32 with a 64-bit float source.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`dstStrideFloats`	Int
`verts`	cpp.RawPointer<Float>
`vertStrideFloats`	Int
`indices`	cpp.RawPointer<Int>
`indexCount`	Int
`a`	cpp.Float32
`b`	cpp.Float32
`c`	cpp.Float32
`d`	cpp.Float32
`tx`	cpp.Float32
`ty`	cpp.Float32
`z`	cpp.Float32
`textureSlot`	cpp.Float32
`writeSlot`	Bool
`srcAttrCount`	Int
`dstAttrCount`	Int

transformQuadCorners(dst: cpp.RawPointer<cpp.Float32>, dstStrideFloats: Int, w: cpp.Float32, h: cpp.Float32, a: cpp.Float32, b: cpp.Float32, c: cpp.Float32, d: cpp.Float32, tx: cpp.Float32, ty: cpp.Float32, z: cpp.Float32, textureSlot: cpp.Float32, writeSlot: Bool, flipOrder: Bool): Void

Transforms the 4 corners of a (w,h) rectangle with the 2x3 affine matrix and writes 4 strided vertices [x y z (slot)] into dst. Corner order: flipOrder=true -> br, bl, tl, tr ; false -> tl, tr, br, bl.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`dstStrideFloats`	Int
`w`	cpp.Float32
`h`	cpp.Float32
`a`	cpp.Float32
`b`	cpp.Float32
`c`	cpp.Float32
`d`	cpp.Float32
`tx`	cpp.Float32
`ty`	cpp.Float32
`z`	cpp.Float32
`textureSlot`	cpp.Float32
`writeSlot`	Bool
`flipOrder`	Bool

scaleUVIndexedF64(dst: cpp.RawPointer<cpp.Float32>, uvs: cpp.RawPointer<Float>, indices: cpp.RawPointer<Int>, indexCount: Int, uvFactorX: Float, uvFactorY: Float): Void

Gathers indexCount UV pairs from uvs (64-bit source, stride 2) via indices, multiplies by (uvFactorX, uvFactorY) in 64-bit and writes sequential 32-bit pairs into dst.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`uvs`	cpp.RawPointer<Float>
`indices`	cpp.RawPointer<Int>
`indexCount`	Int
`uvFactorX`	Float
`uvFactorY`	Float

premultiplyRGBA(dst: cpp.RawPointer<cpp.Float32>, src: cpp.RawPointer<cpp.Float32>, count: Int, globalAlpha: cpp.Float32, premultiply: Bool, zeroAlpha: Bool): Void

Reads count sequential RGBA colors from src, computes outA = globalAlpha * srcA, multiplies RGB by outA when premultiply is set, forces the stored alpha to 0 when zeroAlpha is set (RGB keep the premultiplied values), and writes sequential RGBA into dst.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`src`	cpp.RawPointer<cpp.Float32>
`count`	Int
`globalAlpha`	cpp.Float32
`premultiply`	Bool
`zeroAlpha`	Bool

premultiplyRGBAIndexed(dst: cpp.RawPointer<cpp.Float32>, src: cpp.RawPointer<cpp.Float32>, indices: cpp.RawPointer<Int>, indexCount: Int, globalAlpha: cpp.Float32, premultiply: Bool, zeroAlpha: Bool): Void

Same as premultiplyRGBA but colors are gathered from src via indices.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`src`	cpp.RawPointer<cpp.Float32>
`indices`	cpp.RawPointer<Int>
`indexCount`	Int
`globalAlpha`	cpp.Float32
`premultiply`	Bool
`zeroAlpha`	Bool

premultiplyARGB32(dst: cpp.RawPointer<cpp.Float32>, src: cpp.RawPointer<cpp.UInt32>, count: Int, globalAlpha: cpp.Float32, premultiply: Bool, zeroAlpha: Bool): Void

Same as premultiplyRGBA but the source colors are packed 32-bit integers laid out as 0xAARRGGBB, decoded to floats in [0,1].

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`src`	cpp.RawPointer<cpp.UInt32>
`count`	Int
`globalAlpha`	cpp.Float32
`premultiply`	Bool
`zeroAlpha`	Bool

premultiplyARGB32Indexed(dst: cpp.RawPointer<cpp.Float32>, src: cpp.RawPointer<cpp.UInt32>, indices: cpp.RawPointer<Int>, indexCount: Int, globalAlpha: cpp.Float32, premultiply: Bool, zeroAlpha: Bool): Void

Same as premultiplyARGB32 but colors are gathered from src via indices.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`src`	cpp.RawPointer<cpp.UInt32>
`indices`	cpp.RawPointer<Int>
`indexCount`	Int
`globalAlpha`	cpp.Float32
`premultiply`	Bool
`zeroAlpha`	Bool

fillColorRGBA(dst: cpp.RawPointer<cpp.Float32>, count: Int, r: cpp.Float32, g: cpp.Float32, b: cpp.Float32, a: cpp.Float32): Void

Writes count copies of the RGBA color into dst.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`count`	Int
`r`	cpp.Float32
`g`	cpp.Float32
`b`	cpp.Float32
`a`	cpp.Float32

fillFloats(dst: cpp.RawPointer<cpp.Float32>, count: Int, value: cpp.Float32): Void

Writes count copies of value into dst.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`count`	Int
`value`	cpp.Float32

fillIndicesSequentialU16(dst: cpp.RawPointer<cpp.UInt16>, start: Int, count: Int): Void

Writes count sequential indices start, start+1, ... into dst.

Name	Type
`dst`	cpp.RawPointer<cpp.UInt16>
`start`	Int
`count`	Int

putQuadIndicesU16(dst: cpp.RawPointer<cpp.UInt16>, base: Int): Void

Writes the 6 indices of a quad (two triangles): base, base+1, base+2, base, base+2, base+3.

Name	Type
`dst`	cpp.RawPointer<cpp.UInt16>
`base`	Int

putQuadWireframeIndicesU16(dst: cpp.RawPointer<cpp.UInt16>, base: Int): Void

Writes the 12 line indices of a quad wireframe: (b,b+1),(b+1,b+2),(b+2,b),(b,b+2),(b+2,b+3),(b+3,b).

Name	Type
`dst`	cpp.RawPointer<cpp.UInt16>
`base`	Int

fillQuadAttrsF64(dst: cpp.RawPointer<cpp.Float32>, dstStrideFloats: Int, attrOffset: Int, attrs: cpp.RawPointer<Float>, srcCount: Int, dstCount: Int): Void

Writes the same attribute values on the 4 corners of a quad, at attrOffset floats inside each vertex of stride dstStrideFloats, reading srcCount values from attrs (may be null when 0) and zero-padding up to dstCount.

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`dstStrideFloats`	Int
`attrOffset`	Int
`attrs`	cpp.RawPointer<Float>
`srcCount`	Int
`dstCount`	Int

emitQuad(pos: cpp.RawPointer<cpp.Float32>, posStrideFloats: Int, idx: cpp.RawPointer<cpp.UInt16>, indexBase: Int, color: cpp.RawPointer<cpp.Float32>, uv: cpp.RawPointer<cpp.Float32>, w: cpp.Float32, h: cpp.Float32, a: cpp.Float32, b: cpp.Float32, c: cpp.Float32, d: cpp.Float32, tx: cpp.Float32, ty: cpp.Float32, z: cpp.Float32, textureSlot: cpp.Float32, writeSlot: Bool, flipOrder: Bool, wireframe: Bool, colR: cpp.Float32, colG: cpp.Float32, colB: cpp.Float32, colA: cpp.Float32, u0: cpp.Float32, v0: cpp.Float32, u1: cpp.Float32, v1: cpp.Float32, u2: cpp.Float32, v2: cpp.Float32, u3: cpp.Float32, v3: cpp.Float32): Void

Emits a complete quad in a single call: indices (6 triangle indices, or 12 line indices when wireframe is set), 4 transformed corners, 4 identical colors and 4 uv pairs. One call per quad keeps the per-call overhead negligible compared to the quad's small workload.

Name	Type
`pos`	cpp.RawPointer<cpp.Float32>
`posStrideFloats`	Int
`idx`	cpp.RawPointer<cpp.UInt16>
`indexBase`	Int
`color`	cpp.RawPointer<cpp.Float32>
`uv`	cpp.RawPointer<cpp.Float32>
`w`	cpp.Float32
`h`	cpp.Float32
`a`	cpp.Float32
`b`	cpp.Float32
`c`	cpp.Float32
`d`	cpp.Float32
`tx`	cpp.Float32
`ty`	cpp.Float32
`z`	cpp.Float32
`textureSlot`	cpp.Float32
`writeSlot`	Bool
`flipOrder`	Bool
`wireframe`	Bool
`colR`	cpp.Float32
`colG`	cpp.Float32
`colB`	cpp.Float32
`colA`	cpp.Float32
`u0`	cpp.Float32
`v0`	cpp.Float32
`u1`	cpp.Float32
`v1`	cpp.Float32
`u2`	cpp.Float32
`v2`	cpp.Float32
`u3`	cpp.Float32
`v3`	cpp.Float32

storeUV4(dst: cpp.RawPointer<cpp.Float32>, u0: cpp.Float32, v0: cpp.Float32, u1: cpp.Float32, v1: cpp.Float32, u2: cpp.Float32, v2: cpp.Float32, u3: cpp.Float32, v3: cpp.Float32): Void

Writes 4 finished UV pairs into dst (8 contiguous floats).

Name	Type
`dst`	cpp.RawPointer<cpp.Float32>
`u0`	cpp.Float32
`v0`	cpp.Float32
`u1`	cpp.Float32
`v1`	cpp.Float32
`u2`	cpp.Float32
`v2`	cpp.Float32
`u3`	cpp.Float32
`v3`	cpp.Float32

Metadata

Name	Parameters
`:build`	clay.simd.Linc.touch()
`:build`	clay.simd.Linc.xml("simd", "../simd")
`:include`	"linc_simd.h"

Simd

Static Members #

Metadata #

Static Members

Metadata