AMDGPU: Implement hook for InferAddressSpaces
ClosedPublic
Actions

Authored by arsenm on Jan 26 2017, 10:10 AM.

Details

Reviewers

Summary

For now just port some of the existing NVPTX tests
and from an old HSAIL optimization pass which
approximately did the same thing.

Don't enable the pass yet until more testing is done.

LGTM.

This revision is now accepted and ready to land.Jan 30 2017, 4:01 AM

r293580

		Path
M		lib/Target/AMDGPU/AMDGPUSubtarget.h (8 lines)
M		lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h (12 lines)
A	M	test/Transforms/InferAddressSpaces/AMDGPU/basic.ll (122 lines)
A	M	test/Transforms/InferAddressSpaces/AMDGPU/infer-address-space.ll (176 lines)
A	M	test/Transforms/InferAddressSpaces/AMDGPU/old-pass-regressions.ll (143 lines)
A	M	test/Transforms/InferAddressSpaces/lit.local.cfg (3 lines)

Status	Author	Revision
Closed	arsenm	D29213 InferAddressSpaces: Handle icmp
Abandoned	arsenm	D29212 InferAddressSpaces: Handle a few intrinsics
Closed	arsenm	D29211 InferAddressSpaces: Support memory intrinsics
Closed	arsenm	D29190 InferAddressSpaces: Support atomics
Closed	arsenm	D29188 InferAddressSpaces: Don't replace volatile users
Closed	arsenm	D29187 AMDGPU: Implement hook for InferAddressSpaces
Closed	arsenm	D29166 NVPTX: Move InferAddressSpaces to generic code
Closed	arsenm	D29164 NVPTX: Refactor NVPTXInferAddressSpaces to check TTI