Skip to content

Rename getRawLoad to getRawAsync #4333

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
rdspring1 merged 1 commit into from
Apr 30, 2025
Merged

Rename getRawLoad to getRawAsync #4333

Show file tree
Hide file tree
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
6 changes: 3 additions & 3 deletions csrc/codegen.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1276,9 +1276,9 @@ class CudaKernelGenerator : private kir::ConstIrVisitor {
std::string genLoadBlockDim() {
std::stringstream ss;
const auto& pdim_map = kernel_->summary().parallel_dimension_map;
Val* tidx = pdim_map.getRawLoad(ParallelType::TIDx);
Val* tidy = pdim_map.getRawLoad(ParallelType::TIDy);
Val* tidz = pdim_map.getRawLoad(ParallelType::TIDz);
Val* tidx = pdim_map.getRawAsync(ParallelType::TIDx);
Val* tidy = pdim_map.getRawAsync(ParallelType::TIDy);
Val* tidz = pdim_map.getRawAsync(ParallelType::TIDz);
int64_t num_threads = tidx->value().as<int64_t>() +
tidy->value().as<int64_t>() + tidz->value().as<int64_t>();
NVF_ERROR(
Expand Down
2 changes: 1 addition & 1 deletion csrc/parallel_dimension_map.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -275,7 +275,7 @@ Val* ParallelDimensionMap::getRawCompute(ParallelType pt) const {
return raw;
}

Val* ParallelDimensionMap::getRawLoad(ParallelType pt) const {
Val* ParallelDimensionMap::getRawAsync(ParallelType pt) const {
if (warp_specialized_types_.count(pt)) {
return IrBuilder::create<Val>(
getWarpSpecializationPaddedVal(pt), DataType::Index);
Expand Down
8 changes: 4 additions & 4 deletions csrc/parallel_dimension_map.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,10 +50,10 @@ class ParallelDimensionMap {
Val* getRawCompute(ParallelType pt) const;

//! Get the "load" parallel dimension on the given ParallelType. In case
//! of no warp specialization, this is the same as getRaw(pt). If we are doing
//! warp specialization on pt, the result is 1, because the last of pt is used
//! for loading circular buffer tensors.
Val* getRawLoad(ParallelType pt) const;
//! of without warp specialization, this is the same as getRaw(pt). For warp
//! warp specialization on pt, the result is padded_value. The last part of
//! pt is used for AsyncWarp warp group.
Val* getRawAsync(ParallelType pt) const;

//! The padded val ensures that CTA has 128 threads for the AsyncWarp. This
//! function returns the padded val for the warp specialized ParallelType.
Expand Down