CUTLASS 3.1 (#915)

Co-authored-by: Aniket Shivam <ashivam@nvidia.com>
This commit is contained in:
ANIKET SHIVAM
2023-04-14 23:19:34 -04:00
committed by GitHub
co-authored by Aniket Shivam
parent 9b8166e3f0
commit d572cc1aab
482 changed files with 37175 additions and 16410 deletions
@@ -391,14 +391,16 @@ Status RankKOperationProfiler::initialize_workspace(
static_cast<library::RankKDescription const &>(operation->description());
if (options.execution_mode != ExecutionMode::kDryRun) {
int seed_shift = 0;
rank_k_workspace_.A = device_context.allocate_tensor(
options,
"A",
operation_desc.A.element,
operation_desc.A.layout,
{int(problem_.n), int(problem_.k)},
{int(problem_.lda)}
{int(problem_.lda)},
1, // batch_count
seed_shift++
);
rank_k_workspace_.C = device_context.allocate_tensor(
@@ -408,7 +410,8 @@ Status RankKOperationProfiler::initialize_workspace(
operation_desc.C.layout,
{int(problem_.n), int(problem_.n)},
{int(problem_.ldc)},
1 // batch_count = 1, default
1, // batch_count
seed_shift++
);
rank_k_workspace_.Computed = device_context.allocate_tensor(