leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 29 Packages Projects Releases Wiki Activity

v4.3 tag release update. (#2789)

Browse Source
This commit is contained in:
Junkai-Wu
2025-11-20 20:49:44 -05:00
committed by GitHub
parent 406e078b29
commit 8cd5bef43a
225 changed files with 23229 additions and 2813 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/util/include/cutlass/util/reference/device/gemm_complex.h
+11 -6
View File
@@ -108,7 +108,9 @@ __global__ void GemmComplex(
tensor_a.add_pointer_offset(batch_idx * batch_stride_A);
tensor_b.add_pointer_offset(batch_idx * batch_stride_B);
tensor_c.add_pointer_offset(batch_idx * batch_stride_C);
if(tensor_c.data()) {
tensor_c.add_pointer_offset(batch_idx * batch_stride_C);
}
tensor_d.add_pointer_offset(batch_idx * batch_stride_D);
for (; batch_idx < batch_count; batch_idx += gridDim.z) {
@@ -163,17 +165,20 @@ __global__ void GemmComplex(
MatrixCoord coord = MatrixCoord(row, col);
if (row < M && col < N) {
tensor_d.at(coord) = convert_op(
alpha * ScalarType(accum[i][j]) +
beta * ScalarType(tensor_c.at(coord)));
ScalarType epilog = alpha * ScalarType(accum[i][j]);
if(tensor_c.data()) {
epilog += beta * ScalarType(tensor_c.at(coord));
}
tensor_d.at(coord) = convert_op(epilog);
}
}
}
tensor_a.add_pointer_offset(batch_stride_A * gridDim.z);
tensor_b.add_pointer_offset(batch_stride_B * gridDim.z);
tensor_c.add_pointer_offset(batch_stride_C * gridDim.z);
if(tensor_c.data()) {
tensor_c.add_pointer_offset(batch_stride_C * gridDim.z);
}
tensor_d.add_pointer_offset(batch_stride_D * gridDim.z);
} // for (batch_idx)
tools/util/include/cutlass/util/reference/host/gemm_complex.h
+8 -5
View File
@@ -154,10 +154,11 @@ void GemmComplex(
MatrixCoord coord = MatrixCoord(row, col);
if (row < M && col < N) {
tensor_d.at(coord) = convert_op(
alpha * ScalarType(accum[i][j]) +
beta * ScalarType(tensor_c.at(coord)));
ScalarType epilog = alpha * ScalarType(accum[i][j]);
if(tensor_c.data()) {
epilog += beta * ScalarType(tensor_c.at(coord));
}
tensor_d.at(coord) = convert_op(epilog);
}
}
}
@@ -167,7 +168,9 @@ void GemmComplex(
tensor_a.add_pointer_offset(batch_stride_A);
tensor_b.add_pointer_offset(batch_stride_B);
tensor_c.add_pointer_offset(batch_stride_C);
if(tensor_c.data()) {
tensor_c.add_pointer_offset(batch_stride_C);
}
tensor_d.add_pointer_offset(batch_stride_D);
} // for (batch_idx)