leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 27 Packages Projects Releases Wiki Activity

CUTLASS 2.3.0 final.

Browse Source
This commit is contained in:
akerr
2020-09-25 10:34:46 -07:00
parent c53f3339bb
commit 37a8f9e598
8 changed files with 15 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cutlass/epilogue/threadblock/predicated_tile_iterator.h
+8 -8
View File
@@ -372,12 +372,11 @@ public:
bool guard = row_guard && mask_.predicates[column];
cutlass::arch::global_store<AccessType, sizeof(AccessType)>(
frag_ptr[frag_row_idx * ThreadMap::Iterations::kColumn +
column],
(void *)&memory_pointer[column * ThreadMap::Delta::kColumn /
kElementsPerAccess],
guard);
if (guard) {
memory_pointer[column * ThreadMap::Delta::kColumn / kElementsPerAccess] =
frag_ptr[frag_row_idx * ThreadMap::Iterations::kColumn + column];
}
}
if (row + 1 < ThreadMap::Iterations::kRow) {
@@ -691,8 +690,9 @@ public:
bool guard = col_guard && mask_.predicates[iteration_contiguous_];
cutlass::arch::global_store<AccessType, sizeof(AccessType)>(
*frag_ptr, (void *)memory_pointer, guard);
if (guard) {
*memory_pointer = *frag_ptr;
}
}
/// Overrides the internal iteration index