add no-offload-uniform-block flag for rocm5.7 and up (#838)

* add -fno-offload-uniform-block flag for rocm5.7 and up

* add a comment and compiler ticket number

CMakeLists.txt

@@ -89,6 +89,14 @@ else()
     message("Building CK for the following targets: ${AMDGPU_TARGETS}")
 endif()
 find_package(hip)
+# No assumption that HIP kernels are launched with uniform block size for backward compatibility
+# SWDEV-413293 and https://reviews.llvm.org/D155213
+math(EXPR hip_VERSION_FLAT "(${hip_VERSION_MAJOR} * 1000 + ${hip_VERSION_MINOR}) * 100000 + ${hip_VERSION_PATCH}")
+message("hip_version_flat=${hip_VERSION_FLAT}")
+if(${hip_VERSION_FLAT} GREATER 500700000)
+   message("Adding the fno-offload-uniform-block compiler flag")
+   add_compile_options(-fno-offload-uniform-block)
+endif()
 
 option(USE_BITINT_EXTENSION_INT4, "Whether to enable clang's BitInt extension to provide int4 data type." OFF)
 option(USE_OPT_NAVI3X, "Whether to enable LDS cumode and Wavefront32 mode for NAVI3X silicons." OFF)