onnxruntime/cmake/patches/composable_kernel/Fix_Clang_Build.patch

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 3e1174ec..f8795475 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -36,6 +36,17 @@ set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 message("CMAKE_CXX_COMPILER_ID: ${CMAKE_CXX_COMPILER_ID}")
 
+add_compile_options(
+    -xhip
+    --offload-arch=gfx908
+    --offload-arch=gfx90a
+    -O3 # otherwise, "Illegal instruction detected" for gfx908
+    "SHELL:-mllvm -amdgpu-early-inline-all=true"
+    "SHELL:-mllvm -amdgpu-function-calls=false" # otherwise, "local memory (65920) exceeds limit (65536) in function"
+    -fhip-new-launch-api
+)
+
+
 ## OpenMP
 if(CMAKE_CXX_COMPILER_ID MATCHES "Clang")
 	# workaround issue hipcc in rocm3.5 cannot find openmp
@@ -245,9 +256,6 @@ rocm_package_setup_component(tests
 )
 
 add_subdirectory(library)
-add_subdirectory(example)
-add_subdirectory(test)
-add_subdirectory(profiler)
 
 #Create an interface target for the include only files and call it "composablekernels"
 include(CMakePackageConfigHelpers)
@@ -273,11 +281,3 @@ rocm_install(FILES
 
 set(CPACK_RESOURCE_FILE_LICENSE "${CMAKE_CURRENT_SOURCE_DIR}/LICENSE")
 set(CPACK_RPM_PACKAGE_LICENSE "MIT")
-
-rocm_create_package(
-    NAME composablekernel
-    DESCRIPTION "High Performance Composable Kernel for AMD GPUs"
-    MAINTAINER "MIOpen Kernels Dev Team <dl.MIOpen@amd.com>"
-    LDCONFIG
-    HEADER_ONLY
-)
diff --git a/include/ck/ck.hpp b/include/ck/ck.hpp
index fcaec592..8ea06421 100644
--- a/include/ck/ck.hpp
+++ b/include/ck/ck.hpp
@@ -126,7 +126,9 @@
 #define CK_EXPERIMENTAL_USE_MEMCPY_FOR_BIT_CAST 1
 
 // experimental feature: optimize for inter-wave scheduling policy
+#ifndef CK_EXPERIMENTAL_INTER_WAVE_SCHEDULING
 #define CK_EXPERIMENTAL_INTER_WAVE_SCHEDULING 0
+#endif
 #define CK_EXPERIMENTAL_INTER_WAVE_SCHEDULING_MAC_CLUSTERS 1
 
 // hack: have underlying assumption that need to be satsified, otherwise it's a bug