AuroraMiddleware/skia2
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Update Metal matrix intrinsic polyfills to allow half.

Browse Source 
The inverse, outerProduct and matrixCompMult polyfill functions in Metal
were written assuming that all float matrices would use the `float`
type. They now use a template so that `half` matrices will work too.

Change-Id: I7696c8ad1e4aaffbd71c56b9245485e74cd96c5a
Bug: skia:12339
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/463338
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This commit is contained in:
John Stiles 2021-10-25 19:51:13 -04:00 committed by SkCQ
parent 643bd0fc8f
commit b37100de7d
5 changed files with 76 additions and 72 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

75
src/sksl/codegen/SkSLMetalCodeGenerator.cpp
View File

@ -394,46 +394,49 @@ void MetalCodeGenerator::writeFunctionCall(const FunctionCall& c) {
} }
static constexpr char kInverse2x2[] = R"( static constexpr char kInverse2x2[] = R"(
float2x2 float2x2_inverse(float2x2 m) { template <typename T>
return float2x2(m[1][1], -m[0][1], -m[1][0], m[0][0]) * (1/determinant(m)); matrix<T, 2, 2> mat2_inverse(matrix<T, 2, 2> m) {
return matrix<T, 2, 2>(m[1][1], -m[0][1], -m[1][0], m[0][0]) * (1/determinant(m));
} }
)"; )";
static constexpr char kInverse3x3[] = R"( static constexpr char kInverse3x3[] = R"(
float3x3 float3x3_inverse(float3x3 m) { template <typename T>
float a00 = m[0][0], a01 = m[0][1], a02 = m[0][2]; matrix<T, 3, 3> mat3_inverse(matrix<T, 3, 3> m) {
float a10 = m[1][0], a11 = m[1][1], a12 = m[1][2]; T a00 = m[0][0], a01 = m[0][1], a02 = m[0][2];
float a20 = m[2][0], a21 = m[2][1], a22 = m[2][2]; T a10 = m[1][0], a11 = m[1][1], a12 = m[1][2];
float b01 = a22*a11 - a12*a21; T a20 = m[2][0], a21 = m[2][1], a22 = m[2][2];
float b11 = -a22*a10 + a12*a20; T b01 = a22*a11 - a12*a21;
float b21 = a21*a10 - a11*a20; T b11 = -a22*a10 + a12*a20;
float det = a00*b01 + a01*b11 + a02*b21; T b21 = a21*a10 - a11*a20;
return float3x3(b01, (-a22*a01 + a02*a21), ( a12*a01 - a02*a11), T det = a00*b01 + a01*b11 + a02*b21;
return matrix<T, 3, 3>(b01, (-a22*a01 + a02*a21), ( a12*a01 - a02*a11),
b11, ( a22*a00 - a02*a20), (-a12*a00 + a02*a10), b11, ( a22*a00 - a02*a20), (-a12*a00 + a02*a10),
b21, (-a21*a00 + a01*a20), ( a11*a00 - a01*a10)) * (1/det); b21, (-a21*a00 + a01*a20), ( a11*a00 - a01*a10)) * (1/det);
} }
)"; )";
static constexpr char kInverse4x4[] = R"( static constexpr char kInverse4x4[] = R"(
float4x4 float4x4_inverse(float4x4 m) { template <typename T>
float a00 = m[0][0], a01 = m[0][1], a02 = m[0][2], a03 = m[0][3]; matrix<T, 4, 4> mat4_inverse(matrix<T, 4, 4> m) {
float a10 = m[1][0], a11 = m[1][1], a12 = m[1][2], a13 = m[1][3]; T a00 = m[0][0], a01 = m[0][1], a02 = m[0][2], a03 = m[0][3];
float a20 = m[2][0], a21 = m[2][1], a22 = m[2][2], a23 = m[2][3]; T a10 = m[1][0], a11 = m[1][1], a12 = m[1][2], a13 = m[1][3];
float a30 = m[3][0], a31 = m[3][1], a32 = m[3][2], a33 = m[3][3]; T a20 = m[2][0], a21 = m[2][1], a22 = m[2][2], a23 = m[2][3];
float b00 = a00*a11 - a01*a10; T a30 = m[3][0], a31 = m[3][1], a32 = m[3][2], a33 = m[3][3];
float b01 = a00*a12 - a02*a10; T b00 = a00*a11 - a01*a10;
float b02 = a00*a13 - a03*a10; T b01 = a00*a12 - a02*a10;
float b03 = a01*a12 - a02*a11; T b02 = a00*a13 - a03*a10;
float b04 = a01*a13 - a03*a11; T b03 = a01*a12 - a02*a11;
float b05 = a02*a13 - a03*a12; T b04 = a01*a13 - a03*a11;
float b06 = a20*a31 - a21*a30; T b05 = a02*a13 - a03*a12;
float b07 = a20*a32 - a22*a30; T b06 = a20*a31 - a21*a30;
float b08 = a20*a33 - a23*a30; T b07 = a20*a32 - a22*a30;
float b09 = a21*a32 - a22*a31; T b08 = a20*a33 - a23*a30;
float b10 = a21*a33 - a23*a31; T b09 = a21*a32 - a22*a31;
float b11 = a22*a33 - a23*a32; T b10 = a21*a33 - a23*a31;
float det = b00*b11 - b01*b10 + b02*b09 + b03*b08 - b04*b07 + b05*b06; T b11 = a22*a33 - a23*a32;
return float4x4(a11*b11 - a12*b10 + a13*b09, T det = b00*b11 - b01*b10 + b02*b09 + b03*b08 - b04*b07 + b05*b06;
return matrix<T, 4, 4>(a11*b11 - a12*b10 + a13*b09,
a02*b10 - a01*b11 - a03*b09, a02*b10 - a01*b11 - a03*b09,
a31*b05 - a32*b04 + a33*b03, a31*b05 - a32*b04 + a33*b03,
a22*b04 - a21*b05 - a23*b03, a22*b04 - a21*b05 - a23*b03,
@ -458,7 +461,7 @@ String MetalCodeGenerator::getInversePolyfill(const ExpressionArray& arguments)
const Type& type = arguments.front()->type(); const Type& type = arguments.front()->type();
if (type.isMatrix() && type.rows() == type.columns()) { if (type.isMatrix() && type.rows() == type.columns()) {
// Inject the correct polyfill based on the matrix size. // Inject the correct polyfill based on the matrix size.
String name = this->typeName(type) + "_inverse"; auto name = String::printf("mat%d_inverse", type.columns());
auto [iter, didInsert] = fWrittenIntrinsics.insert(name); auto [iter, didInsert] = fWrittenIntrinsics.insert(name);
if (didInsert) { if (didInsert) {
switch (type.rows()) { switch (type.rows()) {
@ -482,8 +485,8 @@ String MetalCodeGenerator::getInversePolyfill(const ExpressionArray& arguments)
void MetalCodeGenerator::writeMatrixCompMult() { void MetalCodeGenerator::writeMatrixCompMult() {
static constexpr char kMatrixCompMult[] = R"( static constexpr char kMatrixCompMult[] = R"(
template <int C, int R> template <typename T, int C, int R>
matrix<float, C, R> matrixCompMult(matrix<float, C, R> a, const matrix<float, C, R> b) { matrix<T, C, R> matrixCompMult(matrix<T, C, R> a, const matrix<T, C, R> b) {
for (int c = 0; c < C; ++c) { for (int c = 0; c < C; ++c) {
a[c] *= b[c]; a[c] *= b[c];
} }
@ -500,9 +503,9 @@ matrix<float, C, R> matrixCompMult(matrix<float, C, R> a, const matrix<float, C,
void MetalCodeGenerator::writeOuterProduct() { void MetalCodeGenerator::writeOuterProduct() {
static constexpr char kOuterProduct[] = R"( static constexpr char kOuterProduct[] = R"(
template <int C, int R> template <typename T, int C, int R>
matrix<float, C, R> outerProduct(const vec<float, R> a, const vec<float, C> b) { matrix<T, C, R> outerProduct(const vec<T, R> a, const vec<T, C> b) {
matrix<float, C, R> result; matrix<T, C, R> result;
for (int c = 0; c < C; ++c) { for (int c = 0; c < C; ++c) {
result[c] = a * b[c]; result[c] = a * b[c];
} }

21
tests/sksl/intrinsics/Inverse.metal
View File

@ -44,15 +44,16 @@ thread bool operator!=(const float4x4 left, const float4x4 right) {
return !(left == right); return !(left == right);
} }
float3x3 float3x3_inverse(float3x3 m) { template <typename T>
float a00 = m[0][0], a01 = m[0][1], a02 = m[0][2]; matrix<T, 3, 3> mat3_inverse(matrix<T, 3, 3> m) {
float a10 = m[1][0], a11 = m[1][1], a12 = m[1][2]; T a00 = m[0][0], a01 = m[0][1], a02 = m[0][2];
float a20 = m[2][0], a21 = m[2][1], a22 = m[2][2]; T a10 = m[1][0], a11 = m[1][1], a12 = m[1][2];
float b01 = a22*a11 - a12*a21; T a20 = m[2][0], a21 = m[2][1], a22 = m[2][2];
float b11 = -a22*a10 + a12*a20; T b01 = a22*a11 - a12*a21;
float b21 = a21*a10 - a11*a20; T b11 = -a22*a10 + a12*a20;
float det = a00*b01 + a01*b11 + a02*b21; T b21 = a21*a10 - a11*a20;
return float3x3(b01, (-a22*a01 + a02*a21), ( a12*a01 - a02*a11), T det = a00*b01 + a01*b11 + a02*b21;
return matrix<T, 3, 3>(b01, (-a22*a01 + a02*a21), ( a12*a01 - a02*a11),
b11, ( a22*a00 - a02*a20), (-a12*a00 + a02*a10), b11, ( a22*a00 - a02*a20), (-a12*a00 + a02*a10),
b21, (-a21*a00 + a01*a20), ( a11*a00 - a01*a10)) * (1/det); b21, (-a21*a00 + a01*a20), ( a11*a00 - a01*a10)) * (1/det);
} }
@ -62,6 +63,6 @@ fragment Outputs fragmentMain(Inputs _in [[stage_in]], constant Uniforms& _unifo
float2x2 inv2x2 = float2x2(float2(-2.0, 1.0), float2(1.5, -0.5)); float2x2 inv2x2 = float2x2(float2(-2.0, 1.0), float2(1.5, -0.5));
float3x3 inv3x3 = float3x3(float3(-24.0, 18.0, 5.0), float3(20.0, -15.0, -4.0), float3(-5.0, 4.0, 1.0)); float3x3 inv3x3 = float3x3(float3(-24.0, 18.0, 5.0), float3(20.0, -15.0, -4.0), float3(-5.0, 4.0, 1.0));
float4x4 inv4x4 = float4x4(float4(-2.0, -0.5, 1.0, 0.5), float4(1.0, 0.5, 0.0, -0.5), float4(-8.0, -1.0, 2.0, 2.0), float4(3.0, 0.5, -1.0, -0.5)); float4x4 inv4x4 = float4x4(float4(-2.0, -0.5, 1.0, 0.5), float4(1.0, 0.5, 0.0, -0.5), float4(-8.0, -1.0, 2.0, 2.0), float4(3.0, 0.5, -1.0, -0.5));
_out.sk_FragColor = ((float2x2(float2(-2.0, 1.0), float2(1.5, -0.5)) == inv2x2 && float3x3(float3(-24.0, 18.0, 5.0), float3(20.0, -15.0, -4.0), float3(-5.0, 4.0, 1.0)) == inv3x3) && float4x4(float4(-2.0, -0.5, 1.0, 0.5), float4(1.0, 0.5, 0.0, -0.5), float4(-8.0, -1.0, 2.0, 2.0), float4(3.0, 0.5, -1.0, -0.5)) == inv4x4) && float3x3_inverse(float3x3(float3(1.0, 2.0, 3.0), float3(4.0, 5.0, 6.0), float3(7.0, 8.0, 9.0))) != inv3x3 ? _uniforms.colorGreen : _uniforms.colorRed; _out.sk_FragColor = ((float2x2(float2(-2.0, 1.0), float2(1.5, -0.5)) == inv2x2 && float3x3(float3(-24.0, 18.0, 5.0), float3(20.0, -15.0, -4.0), float3(-5.0, 4.0, 1.0)) == inv3x3) && float4x4(float4(-2.0, -0.5, 1.0, 0.5), float4(1.0, 0.5, 0.0, -0.5), float4(-8.0, -1.0, 2.0, 2.0), float4(3.0, 0.5, -1.0, -0.5)) == inv4x4) && mat3_inverse(float3x3(float3(1.0, 2.0, 3.0), float3(4.0, 5.0, 6.0), float3(7.0, 8.0, 9.0))) != inv3x3 ? _uniforms.colorGreen : _uniforms.colorRed;
return _out; return _out;
} }

4
tests/sksl/intrinsics/MatrixCompMultES2.metal
View File

@ -19,8 +19,8 @@ thread bool operator!=(const float2x2 left, const float2x2 right);
thread bool operator==(const float3x3 left, const float3x3 right); thread bool operator==(const float3x3 left, const float3x3 right);
thread bool operator!=(const float3x3 left, const float3x3 right); thread bool operator!=(const float3x3 left, const float3x3 right);
template <int C, int R> template <typename T, int C, int R>
matrix<float, C, R> matrixCompMult(matrix<float, C, R> a, const matrix<float, C, R> b) { matrix<T, C, R> matrixCompMult(matrix<T, C, R> a, const matrix<T, C, R> b) {
for (int c = 0; c < C; ++c) { for (int c = 0; c < C; ++c) {
a[c] *= b[c]; a[c] *= b[c];
} }

4
tests/sksl/intrinsics/MatrixCompMultES3.metal
View File

@ -20,8 +20,8 @@ thread bool operator!=(const float4x2 left, const float4x2 right);
thread bool operator==(const float4x3 left, const float4x3 right); thread bool operator==(const float4x3 left, const float4x3 right);
thread bool operator!=(const float4x3 left, const float4x3 right); thread bool operator!=(const float4x3 left, const float4x3 right);
template <int C, int R> template <typename T, int C, int R>
matrix<float, C, R> matrixCompMult(matrix<float, C, R> a, const matrix<float, C, R> b) { matrix<T, C, R> matrixCompMult(matrix<T, C, R> a, const matrix<T, C, R> b) {
for (int c = 0; c < C; ++c) { for (int c = 0; c < C; ++c) {
a[c] *= b[c]; a[c] *= b[c];
} }

6
tests/sksl/intrinsics/OuterProduct.metal
View File

@ -39,9 +39,9 @@ thread bool operator!=(const float2x2 left, const float2x2 right) {
return !(left == right); return !(left == right);
} }
template <int C, int R> template <typename T, int C, int R>
matrix<float, C, R> outerProduct(const vec<float, R> a, const vec<float, C> b) { matrix<T, C, R> outerProduct(const vec<T, R> a, const vec<T, C> b) {
matrix<float, C, R> result; matrix<T, C, R> result;
for (int c = 0; c < C; ++c) { for (int c = 0; c < C; ++c) {
result[c] = a * b[c]; result[c] = a * b[c];
} }