cmake : enable separable compilation for CUDA

ggerganov · Dec 7, 2023 · 49020c0 · 49020c0
1 parent 06b5c62
commit 49020c0
Showing 1 changed file with 4 additions and 0 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -279,6 +279,10 @@ if (LLAMA_CUBLAS)
         if (LLAMA_CUDA_FORCE_MMQ)
             add_compile_definitions(GGML_CUDA_FORCE_MMQ)
         endif()
+
+        # required for dynamic parallelism
+        set(CMAKE_CUDA_SEPARABLE_COMPILATION ON)
+
         add_compile_definitions(GGML_CUDA_DMMV_X=${LLAMA_CUDA_DMMV_X})
         add_compile_definitions(GGML_CUDA_MMV_Y=${LLAMA_CUDA_MMV_Y})
         if (DEFINED LLAMA_CUDA_DMMV_Y)