Merge branch 'develop' of https://github.com/ROCm/composable_kernel into update_cka8w8

example/ck_tile/01_fmha/generate.py

@@ -47,6 +47,9 @@ def list_blobs(output_file : Optional[str], api_list : List[str], kernel_filter
     assert output_file is not None
     file_path = Path(output_file)
 
+    # create an empty file / drop its contents if it exists
+    open(file_path, "w").close()
+
     for api in api_list:
         handler = handlers[api][HandlerId.LIST_BLOBS]
         handler(file_path, kernel_filter, receipt, mask_impl)

example/ck_tile/02_layernorm2d/generate.py

@@ -559,7 +559,7 @@ float layernorm2d_fwd(layernorm2d_fwd_traits t,
         w_p = Path(self.working_path)
         list_p = w_p / 'layernorm2d_fwd_blobs.txt'
         blobs = self.get_blobs()
-        with list_p.open('a') as list_f:
+        with list_p.open('w') as list_f:
             # api related file
             list_f.write(str(w_p / (self.name_api + ".cpp"))  + "\n")
             list_f.write(str(w_p / (self.name_common_header + ".hpp"))  + "\n")

library/src/tensor_operation_instance/gpu/mha/CMakeLists.txt

@@ -27,11 +27,6 @@ rocm_install(FILES ${MHA_HEADERS} DESTINATION include/ck_tile/ops)
 # headers for building lib
 file(COPY ${MHA_HEADERS} DESTINATION ${FMHA_CPP_FOLDER})
 
-# Delete the blob file if it exists to avoid append of old content.
-if(EXISTS ${FMHA_CPP_FOLDER}/blob_list.txt)
-    file(REMOVE ${FMHA_CPP_FOLDER}/blob_list.txt)
-endif()
-
 set(FMHA_KNOWN_APIS "fwd,fwd_splitkv,fwd_appendkv,bwd")
 
 # generate a list of kernels, but not actually emit files at config stage