[LLVMGPU] Turn on the vector distribution pipeline by default (#17291)
All known outstanding compilation failures with this pipeline have been
resolved, so now is a good time to turn it on by default to start
getting the predictable good performance for AMD gpu's w/ mma intrinsics
like CDNA and RDNA cards.
diff --git a/compiler/src/iree/compiler/Codegen/LLVMGPU/KernelConfig.cpp b/compiler/src/iree/compiler/Codegen/LLVMGPU/KernelConfig.cpp
index 1117801..d740b5d 100644
--- a/compiler/src/iree/compiler/Codegen/LLVMGPU/KernelConfig.cpp
+++ b/compiler/src/iree/compiler/Codegen/LLVMGPU/KernelConfig.cpp
@@ -43,7 +43,7 @@
llvm::cl::opt<bool> clGPUEnableVectorDistribution(
"iree-codegen-llvmgpu-use-vector-distribution",
llvm::cl::desc("enable the usage of the vector distribution pipeline"),
- llvm::cl::init(false));
+ llvm::cl::init(true));
llvm::cl::opt<bool> clGPUEnableTransformDialectJit(
"iree-codegen-llvmgpu-enable-transform-dialect-jit",
