leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 26 Packages Projects Releases Wiki Activity

CUTLASS 3.2 (#1024)

Browse Source 
* CUTLASS 3.2
This commit is contained in:
ANIKET SHIVAM
2023-08-07 20:50:32 -04:00
committed by GitHub
parent a0d787b746
commit 4575443d44
392 changed files with 47559 additions and 7940 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/unit/gemm/threadblock/mma_multistage.cu
-2
View File
@@ -2838,7 +2838,6 @@ TEST(SM80_gemm_threadblock_crosswise,
}
////////////////////////////////////////////////////////////////////////////////
TEST(SM80_gemm_threadblock_crosswise,
tensor_op_64x64x1024_64x64x1024_16x8x256_3stage) {
using ElementA = cutlass::uint1b_t;
@@ -3328,7 +3327,6 @@ TEST(SM80_gemm_threadblock_crosswise,
}
////////////////////////////////////////////////////////////////////////////////
TEST(SM80_gemm_threadblock_congruous,
tensor_op_64x64x16_32x64x16_8x8x4_3stage) {
using ElementA = double;
test/unit/gemm/threadblock/mma_multistage_sparse_testbed.h
-3
View File
@@ -296,7 +296,6 @@ struct SparseTestbed {
} else if (init_A == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_A.host_view());
} else {
// TODO: Implement the rest
return false;
}
@@ -322,7 +321,6 @@ struct SparseTestbed {
} else if (init_B == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_B.host_view());
} else {
// TODO: Implement the rest
return false;
}
@@ -339,7 +337,6 @@ struct SparseTestbed {
cutlass::reference::host::TensorFill(matrix_E.host_view(),
(ElementE)(content));
} else {
// TODO: Implement the rest
return false;
}
test/unit/gemm/threadblock/mma_multistage_testbed.h
-2
View File
@@ -253,7 +253,6 @@ struct Testbed {
} else if (init_A == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_A.host_view());
} else {
// TODO: Implement the rest
return false;
}
@@ -279,7 +278,6 @@ struct Testbed {
} else if (init_B == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_B.host_view());
} else {
// TODO: Implement the rest
return false;
}
test/unit/gemm/threadblock/mma_multistage_testbed_slicedk.h
-2
View File
@@ -244,7 +244,6 @@ struct Testbed {
} else if (init_A == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_A.host_view());
} else {
// TODO: Implement the rest
return false;
}
@@ -270,7 +269,6 @@ struct Testbed {
} else if (init_B == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_B.host_view());
} else {
// TODO: Implement the rest
return false;
}
test/unit/gemm/threadblock/mma_pipelined_sm75.cu
-1
View File
@@ -1793,7 +1793,6 @@ TEST(SM75_gemm_threadblock_interleaved,
}
////////////////////////////////////////////////////////////////////////////////
TEST(SM75_gemm_threadblock_crosswise, tensor_op_64x64x512_64x64x512_8x8x128) {
using ElementA = cutlass::uint1b_t;
using LayoutA = cutlass::layout::RowMajor;
test/unit/gemm/threadblock/mma_pipelined_testbed.h
-2
View File
@@ -262,7 +262,6 @@ struct Testbed {
} else if (init_A == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_A.host_view());
} else {
// TODO: Implement the rest
return false;
}
@@ -288,7 +287,6 @@ struct Testbed {
} else if (init_B == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_B.host_view());
} else {
// TODO: Implement the rest
return false;
}
test/unit/gemm/threadblock/mma_pipelined_testbed_slicedk.h
-2
View File
@@ -250,7 +250,6 @@ struct Testbed {
} else if (init_A == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_A.host_view());
} else {
// TODO: Implement the rest
return false;
}
@@ -276,7 +275,6 @@ struct Testbed {
} else if (init_B == cutlass::Distribution::Identity) {
cutlass::reference::host::TensorFillIdentity(matrix_B.host_view());
} else {
// TODO: Implement the rest
return false;
}
test/unit/gemm/threadblock/mma_pipelined_wmma_sm75.cu
+1 -1
View File
@@ -262,7 +262,6 @@ TEST(SM75_gemm_threadblock_wmma_tensor_op_row_col_col_s4, 64x64x64_64x64x64_8x8x
problem_size.k(), alpha, beta)
.run(grid, block);
}
TEST(SM75_gemm_threadblock_wmma_tensor_op_row_col_row_b1, 64x64x512_64x64x512_8x8x128) {
using ElementA = cutlass::uint1b_t;
using LayoutA = cutlass::layout::RowMajor;
@@ -332,6 +331,7 @@ TEST(SM75_gemm_threadblock_wmma_tensor_op_row_col_col_b1, 64x64x512_64x64x512_8x
problem_size.k(), alpha, beta)
.run(grid, block);
}
#endif //CUTLASS_SUBBYTE_INTEGER_MATRIX_MULTIPLY_ENABLED
#endif //CUTLASS_ARCH_WMMA_SM75_ENABLED
test/unit/gemm/threadblock/mma_planar_complex_testbed.h
-2
View File
@@ -232,7 +232,6 @@ struct TestbedPlanarComplex {
} else if (init_A == cutlass::Distribution::Identity) {
//cutlass::reference::host::TensorFillIdentity(matrix_A.host_view());
} else {
// TODO: Implement the rest
return false;
}
@@ -270,7 +269,6 @@ struct TestbedPlanarComplex {
//cutlass::reference::host::TensorFillIdentity(matrix_B.host_view());
} else {
// TODO: Implement the rest
return false;
}
test/unit/gemm/threadblock/mma_singlestage_wmma_sm75.cu
-1
View File
@@ -262,7 +262,6 @@ TEST(SM75_gemm_threadblock_singlestage_wmma_tensor_op_row_col_col_s4, 64x64x64_6
problem_size.k(), alpha, beta)
.run(grid, block);
}
TEST(SM75_gemm_threadblock_singlestage_wmma_tensor_op_row_col_row_b1, 64x64x512_64x64x512_8x8x128) {
using ElementA = cutlass::uint1b_t;
using LayoutA = cutlass::layout::RowMajor;