CUDA: compress mode size

cuda 12.8 added the option to specify stronger compression for binaries.

Author: Green-Sky

ggml/src/ggml-cuda/CMakeLists.txt

@@ -98,6 +98,15 @@ if (CUDAToolkit_FOUND)
 
     set(CUDA_FLAGS -use_fast_math)
 
+    if (CUDAToolkit_VERSION VERSION_GREATER_EQUAL "12.8")
+        # Options are:
+        # - none (not recommended)
+        # - speed (default)
+        # - balance
+        # - size
+        list(APPEND CUDA_FLAGS -compress-mode=size)
+    endif()
+
     if (GGML_FATAL_WARNINGS)
         list(APPEND CUDA_FLAGS -Werror all-warnings)
     endif()