[Intel] Remove ViewOpToLLVM (#3579)

There is no need to specialize `ViewOpToLLVM` conversion anymore, we can
reuse the common version.

Signed-off-by: Whitney Tsang <[email protected]>

Author: whitneywhtsang

test/Conversion/intel/tritongpu_to_gen.mlir

@@ -623,6 +623,8 @@ module attributes {"ttg.num-ctas" = 1 : i32, "ttg.num-warps" = 4 : i32} {
     // CHECK-NEXT: llvm.mlir.constant(1 : i32) : i32
     // CHECK-NEXT: llvm.mlir.constant(32 : i32) : i32
     // CHECK-NEXT: llvm.mlir.constant(512 : i32) : i32
+    // CHECK-NEXT: llvm.add
+    // CHECK-NEXT: llvm.add
     // CHECK-NEXT: llvm.mlir.constant(0 : i32) : i32
     // CHECK-NEXT: llvm.mul
     // CHECK-NEXT: llvm.add

third_party/intel/lib/TritonIntelGPUToLLVM/CMakeLists.txt

@@ -24,7 +24,6 @@ add_triton_library(TritonIntelGPUToLLVM
     TypeConverter.cpp
     UpcastMXFPToLLVM.cpp
     Utility.cpp
-    ViewOpToLLVM.cpp
 
     DEPENDS
     TritonIntelGPUConversionPassIncGen

third_party/intel/lib/TritonIntelGPUToLLVM/PatternTritonGPUOpToLLVM.h

@@ -29,10 +29,6 @@ void populateConvertLayoutOpToLLVMPatterns(LLVMTypeConverter &typeConverter,
                                            RewritePatternSet &patterns,
                                            PatternBenefit benefit);
 
-void populateViewOpToLLVMPatterns(LLVMTypeConverter &typeConverter,
-                                  RewritePatternSet &patterns,
-                                  PatternBenefit benefit);
-
 void populateDotOpToLLVMPatterns(
     TritonIntelGPUToLLVMTypeConverter &typeConverter,
     RewritePatternSet &patterns, PatternBenefit benefit);

third_party/intel/lib/TritonIntelGPUToLLVM/PipelineManager.h

@@ -234,7 +234,8 @@ class TritonGPUToLLVMPipelineManager {
                                                  targetInfo, benefit);
       mlir::triton::populateGatherOpToLLVMPatterns(typeConverter, patterns,
                                                    targetInfo, benefit);
-      intel::populateViewOpToLLVMPatterns(typeConverter, patterns, benefit);
+      mlir::triton::populateViewOpToLLVMPatterns(typeConverter, patterns,
+                                                 benefit);
 
       intel::populateTensorPtrOpsToLLVMPatterns(typeConverter, patterns,
                                                 benefit);