Vulkan row split matrix multiply

ggml/CMakeLists.txt

@@ -222,6 +222,7 @@ option(GGML_VULKAN_MEMORY_DEBUG             "ggml: enable Vulkan memory debug ou
 option(GGML_VULKAN_SHADER_DEBUG_INFO        "ggml: enable Vulkan shader debug info"           OFF)
 option(GGML_VULKAN_VALIDATE                 "ggml: enable Vulkan validation"                  OFF)
 option(GGML_VULKAN_RUN_TESTS                "ggml: run Vulkan tests"                          OFF)
+option(GGML_VULKAN_ENABLE_LAYER_PARALLELISM "ggml: enable Vulkan layer parallelism"           OFF)
 option(GGML_WEBGPU                          "ggml: use WebGPU"                                OFF)
 option(GGML_WEBGPU_DEBUG                    "ggml: enable WebGPU debug output"                OFF)
 option(GGML_WEBGPU_CPU_PROFILE              "ggml: enable WebGPU profiling (CPU)"             OFF)

ggml/include/ggml-vulkan.h

@@ -22,6 +22,11 @@ GGML_BACKEND_API ggml_backend_buffer_type_t ggml_backend_vk_buffer_type(size_t d
 // pinned host buffer for use with the CPU backend for faster copies between CPU and GPU
 GGML_BACKEND_API ggml_backend_buffer_type_t ggml_backend_vk_host_buffer_type(void);
 
+#ifdef GGML_VULKAN_ENABLE_LAYER_PARALLELISM
+// split buffer type for distributing a tensor across multiple devices
+GGML_BACKEND_API ggml_backend_buffer_type_t ggml_backend_vk_split_buffer_type(int main_device, const float * tensor_split);
+#endif
+
 GGML_BACKEND_API ggml_backend_reg_t ggml_backend_vk_reg(void);
 
 #ifdef  __cplusplus

ggml/src/ggml-vulkan/CMakeLists.txt

@@ -120,6 +120,10 @@ if (Vulkan_FOUND)
         add_compile_definitions(GGML_VULKAN_RUN_TESTS)
     endif()
 
+    if (GGML_VULKAN_ENABLE_LAYER_PARALLELISM)
+        add_compile_definitions(GGML_VULKAN_ENABLE_LAYER_PARALLELISM)
+    endif()
+
     # Set up toolchain for host compilation whether cross-compiling or not
     if (CMAKE_CROSSCOMPILING)
         if (GGML_VULKAN_SHADERS_GEN_TOOLCHAIN)