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

Fix a comment about LDSM layout. (#514)

Browse Source 
Co-authored-by: peisun1115 <peis@google.com>
This commit is contained in:
Pei Sun 2022-06-04 20:04:00 -07:00 committed by GitHub
parent a29dfb1c63
commit c3881d097e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
include/cutlass/gemm/warp/mma_tensor_op_tile_iterator.h
View File

@ -260,8 +260,8 @@ public:
else if (Policy::LdsmShape::kContiguous == 2 &&
kOperand == Operand::kA) {
// Matrix multiply 16816 A
// Q0 Q2
// Q1 Q3
// Q0 Q1
// Q2 Q3
partition_contiguous_idx = ((lane_in_quad_pair >> 2) ^ (i >> 1));
access_contiguous_idx =
(((quad_pair & 1) + ((i & 1) << 1)) ^ lane_in_quad);
@ -269,8 +269,8 @@ public:
} else if (Policy::LdsmShape::kContiguous == 2 &&
kOperand == Operand::kB) {
// Matrix multiply 16816 B
// Q0 Q1
// Q2 Q3
// Q0 Q2
// Q1 Q3
partition_contiguous_idx = ((lane_in_quad_pair >> 2) ^ (i >> 1));
access_contiguous_idx = ((quad_quad + ((i & 1) << 1)) ^ lane_in_quad);
access_strided_idx = lane_in_quad_quad;