This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU: Run LoadStoreVectorizer pass by default
ClosedPublic

Authored by arsenm on Apr 25 2016, 3:16 PM.

Details

Reviewers
tstellarAMD

Diff Detail

Event Timeline

arsenm updated this revision to Diff 54929.Apr 25 2016, 3:16 PM
arsenm retitled this revision from to AMDGPU: Run LoadStoreVectorizer pass.
arsenm updated this object.
arsenm added a reviewer: tstellarAMD.
arsenm added a subscriber: llvm-commits.
jlebar added a subscriber: jlebar.Jun 14 2016, 10:46 AM

BTW I think you need this patch in order to build with dynamic linking.

diff --git a/lib/Target/AMDGPU/LLVMBuild.txt b/lib/Target/AMDGPU/LLVMBuild.txt
index 711cf24..0b6c3c5 100644
--- a/lib/Target/AMDGPU/LLVMBuild.txt
+++ b/lib/Target/AMDGPU/LLVMBuild.txt
@@ -30,5 +30,5 @@ has_disassembler = 1
 type = Library
 name = AMDGPUCodeGen
 parent = AMDGPU
-required_libraries = Analysis AsmPrinter CodeGen Core IPO MC AMDGPUAsmParser AMDGPUAsmPrinter AMDGPUDesc AMDGPUInfo AMDGPUUtils Scalar SelectionDAG Support Target TransformUtils
+required_libraries = Analysis AsmPrinter CodeGen Core IPO MC AMDGPUAsmParser AMDGPUAsmPrinter AMDGPUDesc AMDGPUInfo AMDGPUUtils Scalar SelectionDAG Support Target TransformUtils Vectorize
 add_to_library_groups = AMDGPU
escha added a subscriber: escha.Jun 14 2016, 10:52 AM

Make sure to check that the alias analyses are set up properly in the TM; this bit me when I implemented this out of tree (i.e. confirm that the AA queries are succeeding).

arsenm updated this revision to Diff 69075.Aug 23 2016, 9:09 PM
arsenm retitled this revision from AMDGPU: Run LoadStoreVectorizer pass to AMDGPU: Run LoadStoreVectorizer pass by default.
tstellarAMD accepted this revision.Aug 25 2016, 2:30 PM
tstellarAMD edited edge metadata.
This revision is now accepted and ready to land.Aug 25 2016, 2:30 PM
arsenm closed this revision.Sep 9 2016, 3:38 PM

r281112