Skip to content

Optimize the overhead for all_elementwise case in BroadcastKernel. #57247

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.

Sign up for GitHub

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

Merged
merged 1 commit into from
Sep 13, 2023
Merged

Optimize the overhead for all_elementwise case in BroadcastKernel. #57247

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
9 changes: 7 additions & 2 deletions paddle/phi/kernels/funcs/broadcast_function.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,14 +74,13 @@ struct BroadcastTypeClassifier {
void InitBroadcastConfigs(const std::vector<const DenseTensor *> &ins,
std::vector<DenseTensor *> *outs,
int axis) {
#ifdef PADDLE_WITH_XPU_KP
const auto dims_simplifier =
BroadcastDimsSimplifier(ins, (*outs)[0]->dims(), axis);
if (VLOG_IS_ON(6)) {
DimsSimplifiedLogger<int64_t>::Log(
ins, outs, dims_simplifier, "BroadcastKernel");
}

#ifdef PADDLE_WITH_XPU_KP
configs[0] = kps::details::BroadcastConfig(dims_simplifier.out_dims,
dims_simplifier.in_dims[0],
dims_simplifier.in_dims[1],
Expand All @@ -92,6 +91,12 @@ struct BroadcastTypeClassifier {
dims_simplifier.rank);
#else
if (!all_elementwise) {
const auto dims_simplifier =
BroadcastDimsSimplifier(ins, (*outs)[0]->dims(), axis);
if (VLOG_IS_ON(6)) {
DimsSimplifiedLogger<int64_t>::Log(
ins, outs, dims_simplifier, "BroadcastKernel");
}
for (int i = 0; i < Arity; ++i) {
// if data shape is[m, n], then you should set data_dim = {n, m}
// eg: out's shape [3, 45, 1]. then out_dims = {1, 45, 3}
Expand Down