[PATCH] D84252: [amdgpu] Run SROA after loop unrolling.

[Michael Liao via Phabricator via llvm-commits]

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

- There are promotable `alloca`s after loop unrolling.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D84252

Files:
  llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
  llvm/test/CodeGen/AMDGPU/opt-pipeline.ll


Index: llvm/test/CodeGen/AMDGPU/opt-pipeline.ll
===================================================================
--- llvm/test/CodeGen/AMDGPU/opt-pipeline.ll
+++ llvm/test/CodeGen/AMDGPU/opt-pipeline.ll
@@ -163,7 +163,9 @@
 ; GCN-O1-NEXT:         Recognize loop idioms
 ; GCN-O1-NEXT:         Delete dead loops
 ; GCN-O1-NEXT:         Unroll loops
+; GCN-O1-NEXT:       SROA
 ; GCN-O1-NEXT:       Phi Values Analysis
+; GCN-O1-NEXT:       Function Alias Analysis Results
 ; GCN-O1-NEXT:       Memory Dependence Analysis
 ; GCN-O1-NEXT:       MemCpy Optimization
 ; GCN-O1-NEXT:       Sparse Conditional Constant Propagation
@@ -481,6 +483,8 @@
 ; GCN-O2-NEXT:         Recognize loop idioms
 ; GCN-O2-NEXT:         Delete dead loops
 ; GCN-O2-NEXT:         Unroll loops
+; GCN-O2-NEXT:       SROA
+; GCN-O2-NEXT:       Function Alias Analysis Results
 ; GCN-O2-NEXT:       MergedLoadStoreMotion
 ; GCN-O2-NEXT:       Phi Values Analysis
 ; GCN-O2-NEXT:       Function Alias Analysis Results
@@ -839,6 +843,8 @@
 ; GCN-O3-NEXT:         Recognize loop idioms
 ; GCN-O3-NEXT:         Delete dead loops
 ; GCN-O3-NEXT:         Unroll loops
+; GCN-O3-NEXT:       SROA
+; GCN-O3-NEXT:       Function Alias Analysis Results
 ; GCN-O3-NEXT:       MergedLoadStoreMotion
 ; GCN-O3-NEXT:       Phi Values Analysis
 ; GCN-O3-NEXT:       Function Alias Analysis Results
Index: llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -485,6 +485,14 @@
       if (EnableOpt)
         PM.add(createAMDGPUPromoteAllocaToVector());
   });
+
+  Builder.addExtension(
+      PassManagerBuilder::EP_LoopOptimizerEnd,
+      [](const PassManagerBuilder &, legacy::PassManagerBase &PM) {
+        // Add SROA after loop unrolling as the promotable patterns exposed
+        // after small loop is fully unrolled.
+        PM.add(createSROAPass());
+      });
 }
 
 //===----------------------------------------------------------------------===//


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D84252.279579.patch
Type: text/x-patch
Size: 2089 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200721/8a832eb4/attachment.bin>