[CPU] Optimize vector transfers for pack/unpack before vector lowering (

iree-org#15465) Fixes iree-org#15349
nithinsubbiah · Nov 13, 2023 · 41a8152 · 41a8152
1 parent dc506b8
commit 41a8152
Showing 1 changed file with 3 additions and 0 deletions.
diff --git a/compiler/src/iree/compiler/Codegen/LLVMCPU/Passes.cpp b/compiler/src/iree/compiler/Codegen/LLVMCPU/Passes.cpp
@@ -646,6 +646,9 @@ void addCPUDataTilingPipeline(OpPassManager &passManager,
   addBufferizePasses(nestedModulePM);
 
   {
+    // Fold unit dims before vector lowering.
+    nestedModulePM.addNestedPass<func::FuncOp>(
+        createOptimizeVectorTransferPass(/*flatten=*/false));
     LLVMCPUVectorLoweringPassOptions options;
     options.splitVectorTransfersTo = "linalg-copy";
     nestedModulePM.addNestedPass<func::FuncOp>(