yum-mirror/slang
Making it easier to work with shaders
git clone https://git.yummers.dev/yum-mirror/slang
f55f669d1
master
1//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-vk -render-feature cooperative-vector -output-using-type 2 3// These platforms don't support these operations into structured buffers 4//DISABLE_TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-dx12 -render-feature cooperative-vector -dx12-experimental -output-using-type -profile cs_6_9 -Xslang... -Xdxc -Vd -X. 5//DISABLE_TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-cpu -output-using-type 6 7// CHECK: type: half 8// CHECK-NEXT: 112.000000 9// CHECK-NEXT: 2.000000 10// CHECK-NEXT: 3.000000 11// CHECK-NEXT: 4.000000 12// CHECK-NEXT: 5.000000 13// CHECK-NEXT: 6.000000 14// CHECK-NEXT: 7.000000 15// CHECK-NEXT: 8.000000 16// CHECK-NEXT: 2.000000 17// CHECK-NEXT: 4.000000 18// CHECK-NEXT: 6.000000 19// CHECK-NEXT: 8.000000 20// CHECK-NEXT: 10.000000 21// CHECK-NEXT: 12.000000 22// CHECK-NEXT: 14.000000 23// CHECK-NEXT: 16.000000 24// CHECK-NEXT: 3.000000 25// CHECK-NEXT: 6.000000 26// CHECK-NEXT: 9.000000 27// CHECK-NEXT: 12.000000 28// CHECK-NEXT: 15.000000 29// CHECK-NEXT: 18.000000 30// CHECK-NEXT: 21.000000 31// CHECK-NEXT: 24.000000 32// CHECK-NEXT: 4.000000 33// CHECK-NEXT: 8.000000 34// CHECK-NEXT: 12.000000 35// CHECK-NEXT: 16.000000 36// CHECK-NEXT: 20.000000 37// CHECK-NEXT: 24.000000 38// CHECK-NEXT: 28.000000 39// CHECK-NEXT: 32.000000 40 41//TEST_INPUT:ubuffer(data=[ 0x40003C00 0x44004200 ], stride=4),name=inputA 42// [1,2,3,4] 43StructuredBuffer<half> inputA; 44 45//TEST_INPUT:ubuffer(data=[ 0x40003C00 0x44004200 0x46004500 0x48004700 ], stride=4),name=inputB 46// [1,2,3,4,5,6,7,8] 47StructuredBuffer<half> inputB; 48 49//TEST_INPUT:ubuffer(data=[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0], stride=4):out,name=output 50RWStructuredBuffer<half> output; 51 52[numthreads(1, 1, 1)] 53void computeMain() 54{ 55 let vecA = coopVecLoad<4>(inputA); 56 let vecB = coopVecLoad<8>(inputB); 57 58 output[0] = half(111); 59 60 coopVecOuterProductAccumulate( 61 vecA, 62 vecB, 63 output, 64 0, 65 32, 66 CoopVecMatrixLayout::TrainingOptimal, 67 CoopVecComponentType::Float16, 68 ); 69}