Rename getRawLoad to getRawAsync

[rdspring1]

This PR renames getRawLoad to getRawAsync in ParallelDimensionMap. Recently, CircularBufferLoopStage changed LoadWarp to AsyncWarp. This function corresponds with LoadWarp and needs to be updated accordingly.

[github-actions]

Review updated until commit e937d1f

Description

• Rename getRawLoad to getRawAsync in ParallelDimensionMap

• Update method calls in CudaKernelGenerator

• Update method documentation in parallel_dimension_map.h

---

Changes walkthrough 📝

Relevant files

Enhancement

codegen.cpp Update `getRawLoad` to `getRawAsync`                                          csrc/codegen.cpp Updated getRawLoad to getRawAsync in genLoadBlockDim method +3/-3      parallel_dimension_map.cpp Rename `getRawLoad` to `getRawAsync`                                          csrc/parallel_dimension_map.cpp Renamed getRawLoad to getRawAsync +1/-1      parallel_dimension_map.h Rename `getRawLoad` to `getRawAsync` and update documentation csrc/parallel_dimension_map.h Renamed getRawLoad to getRawAsync Updated method documentation to reflect AsyncWarp usage +4/-4

---

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🧪 No relevant tests

⚡ Recommended focus areas for review Incorrect Implementation The implementation of getRawAsync seems to be incorrect. It returns a padded value when warp specialization is enabled, which is different from the behavior of the old getRawLoad method. This change might lead to incorrect behavior in the code. Val* ParallelDimensionMap::getRawAsync(ParallelType pt) const { if (warp_specialized_types_.count(pt)) { return IrBuilder::create<Val>( getWarpSpecializationPaddedVal(pt), DataType::Index); } Documentation Mismatch The documentation for getRawAsync does not match its implementation. The comment suggests that the method returns a padded value for warp specialization, but the implementation returns a padded value regardless of warp specialization. //! 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.

[rdspring1]

!test