cuda arch flag for cublaslt (#701)

adding gb200 cuda arch flag for cublaslt compilation

superbench/benchmarks/micro_benchmarks/cublaslt_gemm/CMakeLists.txt

@@ -9,6 +9,10 @@ find_package(CUDAToolkit QUIET)
 if(CUDAToolkit_FOUND AND NOT CUDAToolkit_VERSION VERSION_LESS 11.8)
     set(CMAKE_CUDA_STANDARD 17)
     include(../cuda_common.cmake)
+    set(CUDA_ARCH_LIST "80;86;90")
+    if(CUDAToolkit_VERSION VERSION_GREATER_EQUAL "12.8.0")
+        list(APPEND CUDA_ARCH_LIST "100")
+    endif()
 
     add_library(cublaslt_utils SHARED cublaslt_utils.cc)
     target_link_libraries(cublaslt_utils CUDA::cublas CUDA::cublasLt)
@@ -17,6 +21,7 @@ if(CUDAToolkit_FOUND AND NOT CUDAToolkit_VERSION VERSION_LESS 11.8)
 
     add_executable(cublaslt_gemm cublaslt_gemm.cu)
     target_link_libraries(cublaslt_gemm cublaslt_utils)
-    set_target_properties(cublaslt_gemm PROPERTIES CUDA_ARCHITECTURES "80;86;90")
+    set_target_properties(cublaslt_gemm PROPERTIES CUDA_ARCHITECTURES "${CUDA_ARCH_LIST}")
+
     install(TARGETS cublaslt_gemm RUNTIME DESTINATION bin)
 endif()