Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update threshold of bn1d #49734

Merged
merged 2 commits into from
Jan 13, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
4 changes: 4 additions & 0 deletions paddle/phi/kernels/funcs/norm_utils.h
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,10 @@ limitations under the License. */

namespace phi {
namespace funcs {
#define CUDNN_PER_ACTIVATION_THRESHOLD 10240
#define CUDNN_SPATIAL_THRESHOLD_TRAIN 880801
#define CUDNN_SPATIAL_THRESHOLD_EVAL 65535

inline void ExtractNCWHD(const phi::DDim &dims,
const DataLayout &data_layout,
int *N,
Expand Down
5 changes: 1 addition & 4 deletions paddle/phi/kernels/gpu/batch_norm_grad_kernel.cu
Original file line number Diff line number Diff line change
Expand Up @@ -907,15 +907,12 @@ void BatchNormGradRawKernel(const Context &ctx,
#else
}
// CUDNN only support small batch size
// const size_t CUDNN_PER_ACTIVATION_THRESHOLD = 131070;
const size_t CUDNN_PER_ACTIVATION_THRESHOLD = 10240;
const size_t CUDNN_SPATIAL_THRESHOLD = 880801;
bool use_native_nhwc =
d_x ? (x_dims.size() == 4 && compute_format == DataLayout::kNHWC)
: false;
const bool use_native_kernel =
((x_dims.size() == 2 && N >= CUDNN_PER_ACTIVATION_THRESHOLD) ||
(x_dims.size() == 3 && N >= CUDNN_SPATIAL_THRESHOLD));
(x_dims.size() == 3 && N >= CUDNN_SPATIAL_THRESHOLD_TRAIN));
if (use_native_nhwc || (d_x && d_scale && d_bias)) {
if (use_native_kernel || use_native_nhwc) {
if (x_dims.size() == 2 || use_native_nhwc) {
Expand Down
7 changes: 2 additions & 5 deletions paddle/phi/kernels/gpu/batch_norm_kernel.cu
Original file line number Diff line number Diff line change
Expand Up @@ -722,9 +722,6 @@ void BatchNormKernel(const Context &ctx,

auto handle = ctx.cudnn_handle();

const size_t CUDNN_PER_ACTIVATION_THRESHOLD = 10240;
const size_t CUDNN_SPATIAL_THRESHOLD = 880801;

// Now, depending on whether we are running test or not, we have two paths.
// It is training mode when it's not reference AND not using pre-trained
// model.
Expand Down Expand Up @@ -829,7 +826,7 @@ void BatchNormKernel(const Context &ctx,
#else
const bool use_native_kernel =
(x_dims.size() == 2 ||
(x_dims.size() == 3 && N >= CUDNN_SPATIAL_THRESHOLD));
(x_dims.size() == 3 && N >= CUDNN_SPATIAL_THRESHOLD_EVAL));
if (use_native_kernel) {
const int block_size = 256;
const int grid_size = (N * C * H * W * D + block_size - 1) / block_size;
Expand Down Expand Up @@ -1005,7 +1002,7 @@ void BatchNormKernel(const Context &ctx,
// const size_t CUDNN_PER_ACTIVATION_THRESHOLD = 131070;
const bool use_native_kernel =
((x_dims.size() == 2 && N >= CUDNN_PER_ACTIVATION_THRESHOLD) ||
(x_dims.size() == 3 && N >= CUDNN_SPATIAL_THRESHOLD));
(x_dims.size() == 3 && N >= CUDNN_SPATIAL_THRESHOLD_TRAIN));
if (use_native_kernel) {
dim3 block;
dim3 grid;
Expand Down