rm triton-to-linalg pass

CMakeLists.txt

@@ -23,7 +23,7 @@ add_subdirectory(test)
 add_subdirectory(tools/triton-shared-opt)
 
 if (TRITON_SHARED_BUILD_CPU_BACKEND)
-    add_triton_plugin(TritonShared ${CMAKE_CURRENT_SOURCE_DIR}/triton_shared.cc LINK_LIBS TritonSharedAnalysis TritonToLinalg TritonTilingExtIR)
+    add_triton_plugin(TritonShared ${CMAKE_CURRENT_SOURCE_DIR}/triton_shared.cc LINK_LIBS TritonSharedAnalysis TritonTilingExtIR)
     target_link_libraries(TritonShared PRIVATE Python3::Module pybind11::headers)
 endif()
 

README.md

@@ -95,7 +95,7 @@ As part of the conversion process, there are three important analyses:
 
 ### Conversion strategy
 
-We introduce the `TritonToLinalg` pass that converts the `triton` dialect to the `linalg` dialect on *tensors*. This means the resulting IR is fully compatible with `linalg` tiling and fusion transformation passes. As mentioned in the `Pointer analysis`'s description, we do however have to deal with memref instructions at the load and store boundaries and have to convert them to tensors using `bufferization.to_tensor`. Here's a simple example of what the IR looks like:
+We introduce the `TritonToLinalgExperimental` pass that converts the `triton` dialect to the `linalg` dialect on *tensors*. This means the resulting IR is fully compatible with `linalg` tiling and fusion transformation passes. As mentioned in the `Pointer analysis`'s description, we do however have to deal with memref instructions at the load and store boundaries and have to convert them to tensors using `bufferization.to_tensor`. Here's a simple example of what the IR looks like:
 
 ```mlir
 tt.func @kernel(%afloat : !tt.ptr<bf16>, %res : !tt.ptr<bf16>) {

include/triton-shared/Conversion/CMakeLists.txt

@@ -1,4 +1,3 @@
-add_subdirectory(TritonToLinalg)
 add_subdirectory(TritonToLinalgExperimental)
 add_subdirectory(TritonToStructured)
 add_subdirectory(TritonArithToLinalg)

lib/Conversion/CMakeLists.txt

@@ -1,4 +1,3 @@
-add_subdirectory(TritonToLinalg)
 add_subdirectory(TritonToLinalgExperimental)
 add_subdirectory(TritonToStructured)
 add_subdirectory(TritonToUnstructured)