[PATCH] D83675: AMDGPUPerfHintAnalysis: reduce indirect and large stride weights

[Jay Foad via Phabricator via llvm-commits]

foad created this revision.
foad added reviewers: rampitec, nhaehnle, arsenm.
Herald added subscribers: llvm-commits, kerbowa, hiraditya, t-tye, tpr, dstuttard, yaxunl, wdng, jvesely, kzhuravl.
Herald added a project: LLVM.

Reduce the default weight for indirect and large stride accesses. Many
graphics shaders are no more than a couple of thousand instructions in
total, so with the old default weight of 1000, even a single indirect or
large stride access would be enough to push them over the 50% threshold.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D83675

Files:
  llvm/lib/Target/AMDGPU/AMDGPUPerfHintAnalysis.cpp


Index: llvm/lib/Target/AMDGPU/AMDGPUPerfHintAnalysis.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUPerfHintAnalysis.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUPerfHintAnalysis.cpp
@@ -43,11 +43,11 @@
                     cl::desc("Kernel limit wave threshold in %"));
 
 static cl::opt<unsigned>
-    IAWeight("amdgpu-indirect-access-weight", cl::init(1000), cl::Hidden,
+    IAWeight("amdgpu-indirect-access-weight", cl::init(100), cl::Hidden,
              cl::desc("Indirect access memory instruction weight"));
 
 static cl::opt<unsigned>
-    LSWeight("amdgpu-large-stride-weight", cl::init(1000), cl::Hidden,
+    LSWeight("amdgpu-large-stride-weight", cl::init(100), cl::Hidden,
              cl::desc("Large stride memory access weight"));
 
 static cl::opt<unsigned>


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D83675.277403.patch
Type: text/x-patch
Size: 829 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200713/992aca2e/attachment.bin>