diff --git a/mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp b/mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp
index 0a6fb6451f700..dd16ec4b73e9f 100644
--- a/mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp
+++ b/mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp
@@ -455,9 +455,7 @@ void mlir::populateGpuToROCDLConversionPatterns(
   // TODO: Add alignment for workgroup memory
   patterns.add<GPUDynamicSharedMemoryOpLowering>(converter);
 
-  patterns
-      .add<GPUShuffleOpLowering, GPULaneIdOpToROCDL, GPUSubgroupSizeOpToROCDL>(
-          converter);
+  patterns.add<GPUShuffleOpLowering, GPULaneIdOpToROCDL>(converter);
   patterns.add<GPUSubgroupSizeOpToROCDL>(converter, chipset);
 
   populateMathToROCDLConversionPatterns(converter, patterns);