Skip to content

[Fix] change Flatten to Fuse #56719

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 4 commits into from
Sep 4, 2023
Merged

[Fix] change Flatten to Fuse #56719

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c26458f
[Fix] change Flatten to Fuse
BiynXu Aug 28, 2023
ed990d3
[Fix] change Flatten to Fuse in injective
BiynXu Aug 29, 2023
72b3ef8
Merge branch 'develop' into change_Flatten_in_the_elementwise_schedul…
BiynXu Aug 31, 2023
8c11926
[Fix] group norm unittest timeout
BiynXu Aug 31, 2023
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
16 changes: 8 additions & 8 deletions paddle/cinn/hlir/pe/ir_schedule_pe.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,15 +46,15 @@ void IRElementwiseSchedule(ir::IRSchedule &ir_sch, // NOLINT
<< ir_sch.GetModule().GetExprs().at(0);
if (target == common::DefaultNVGPUTarget()) {
auto blocks = ir_sch.GetAllBlocks();
ir_sch.FlattenLoops(ir_sch.GetLoops(blocks[0]), true);
std::vector<ir::Expr> loops = ir_sch.GetLoops(blocks[0]);
ir::Expr loop = ir_sch.Fuse(loops);

auto loops = ir_sch.GetLoops(blocks[0]);
auto size = std::accumulate(
output_shape.begin(), output_shape.end(), 1, std::multiplies<int>());
if (size <= target.max_num_threads()) {
ir_sch.Bind(loops[0], "threadIdx.x");
ir_sch.Bind(loop, "threadIdx.x");
} else {
auto splited = ir_sch.Split(loops[0], {-1, target.max_num_threads()});
auto splited = ir_sch.Split(loop, {-1, target.max_num_threads()});
ir_sch.Bind(splited[0], "blockIdx.x");
ir_sch.Bind(splited[1], "threadIdx.x");
}
Expand All @@ -74,15 +74,15 @@ void IRInjectiveSchedule(ir::IRSchedule &ir_sch, // NOLINT
<< ir_sch.GetModule().GetExprs().at(0);
if (target == common::DefaultNVGPUTarget()) {
auto blocks = ir_sch.GetAllBlocks();
ir_sch.FlattenLoops(ir_sch.GetLoops(blocks[0]), false);
std::vector<ir::Expr> loops = ir_sch.GetLoops(blocks[0]);
ir::Expr loop = ir_sch.Fuse(loops);

auto loops = ir_sch.GetLoops(blocks[0]);
auto size = std::accumulate(
output_shape.begin(), output_shape.end(), 1, std::multiplies<int>());
if (size <= target.max_num_threads()) {
ir_sch.Bind(loops[0], "threadIdx.x");
ir_sch.Bind(loop, "threadIdx.x");
} else {
auto splited = ir_sch.Split(loops[0], {-1, target.max_num_threads()});
auto splited = ir_sch.Split(loop, {-1, target.max_num_threads()});
ir_sch.Bind(splited[0], "blockIdx.x");
ir_sch.Bind(splited[1], "threadIdx.x");
}
Expand Down
2 changes: 1 addition & 1 deletion test/legacy_test/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1057,7 +1057,7 @@ set_tests_properties(
PROPERTIES TIMEOUT 120)
set_tests_properties(test_conv_nn_grad PROPERTIES TIMEOUT 120)
set_tests_properties(test_program_prune_backward PROPERTIES TIMEOUT 120)
set_tests_properties(test_group_norm_op PROPERTIES TIMEOUT 300)
set_tests_properties(test_group_norm_op PROPERTIES TIMEOUT 1000)
set_tests_properties(test_imperative_optimizer PROPERTIES TIMEOUT 250)
set_tests_properties(test_imperative_optimizer_v2 PROPERTIES TIMEOUT 250)
set_tests_properties(test_pool2d_op PROPERTIES TIMEOUT 120)
Expand Down