Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

Limit workspace for cudnnGet results #14326

Merged
merged 2 commits into from
Mar 6, 2019
Merged

Limit workspace for cudnnGet results #14326

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4d411f4
Limit workspace for cudnnGet results
ptrendx Mar 1, 2019
7ef11e6
Retrigger CI
ptrendx Mar 5, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions src/operator/nn/cudnn/cudnn_convolution-inl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -750,7 +750,7 @@ class CuDNNConvolutionOp {
i = 0;
while (i < nalgo
&& (fwd_algo[i].status != CUDNN_STATUS_SUCCESS
|| (param_.cudnn_tune.value() == conv::kLimited
|| (param_.cudnn_tune.value() != conv::kFastest
&& fwd_algo[i].memory > workspace_byte))) {
++i;
min_memory_needs =
Expand Down Expand Up @@ -787,7 +787,7 @@ class CuDNNConvolutionOp {
i = 0;
while (i < nalgo
&& (bwd_filter_algo[i].status != CUDNN_STATUS_SUCCESS
|| (param_.cudnn_tune.value() == conv::kLimited
|| (param_.cudnn_tune.value() != conv::kFastest
&& bwd_filter_algo[i].memory > workspace_byte))) {
++i;
min_memory_needs = (i == 0) ?
Expand Down Expand Up @@ -825,7 +825,7 @@ class CuDNNConvolutionOp {
i = 0;
while (i < nalgo
&& (bwd_data_algo[i].status != CUDNN_STATUS_SUCCESS
|| (param_.cudnn_tune.value() == conv::kLimited
|| (param_.cudnn_tune.value() != conv::kFastest
&& bwd_data_algo[i].memory > workspace_byte))) {
++i;
min_memory_needs = (i == 0) ?
Expand Down Expand Up @@ -924,7 +924,7 @@ class CuDNNConvolutionOp {
#if CUDNN_MAJOR >= 7
(!enforce_determinism || result.determinism == cudnnDeterminism_t::CUDNN_DETERMINISTIC) &&
#endif
(param_.cudnn_tune.value() != conv::kLimited || result.memory <= workspace_byte)) {
(param_.cudnn_tune.value() == conv::kFastest || result.memory <= workspace_byte)) {
algo->Set(result.algo, algo_is_tensor_core);
return;
}
Expand Down