Patch for llama.cpp multi gpu support

llama-metal-multi-gpu.patch

diff --git a/cmake/llama-config.cmake.in b/cmake/llama-config.cmake.in
index 5c55bc6b..69bc027f 100644
--- a/cmake/llama-config.cmake.in
+++ b/cmake/llama-config.cmake.in
@@ -46,6 +46,7 @@ set(GGML_VULKAN_PERF      @GGML_VULKAN_PERF@)
 set(GGML_VULKAN_VALIDATE  @GGML_VULKAN_VALIDATE@)
 set(GGML_VULKAN_RUN_TESTS @GGML_VULKAN_RUN_TESTS@)
 
+set(GGML_METAL_MULTIDEV @GGML_METAL_MULTIDEV@)
 set(GGML_METAL_USE_BF16 @GGML_METAL_USE_BF16@)
 set(GGML_METAL_NDEBUG   @GGML_METAL_NDEBUG@)
 set(GGML_METAL_SHADER_DEBUG  @GGML_METAL_SHADER_DEBUG@)
diff --git a/ggml/CMakeLists.txt b/ggml/CMakeLists.txt
index e33d9748..ad46e743 100644
--- a/ggml/CMakeLists.txt
+++ b/ggml/CMakeLists.txt
@@ -166,6 +166,7 @@ option(GGML_KOMPUTE                         "ggml: use Kompute"
 option(GGML_METAL                           "ggml: use Metal"                                 ${GGML_METAL_DEFAULT})
 option(GGML_METAL_USE_BF16                  "ggml: use bfloat if available"                   OFF)
 option(GGML_METAL_NDEBUG                    "ggml: disable Metal debugging"                   OFF)
+option(GGML_METAL_MULTIDEV                  "ggml: enable Metal multi-device selection"       OFF)
 option(GGML_METAL_SHADER_DEBUG              "ggml: compile Metal with -fno-fast-math"         OFF)
 option(GGML_METAL_EMBED_LIBRARY             "ggml: embed Metal library"                       ${GGML_METAL})
 set   (GGML_METAL_MACOSX_VERSION_MIN "" CACHE STRING
diff --git a/ggml/src/ggml-metal/CMakeLists.txt b/ggml/src/ggml-metal/CMakeLists.txt
index 1bad2720..cf67d6f0 100644
--- a/ggml/src/ggml-metal/CMakeLists.txt
+++ b/ggml/src/ggml-metal/CMakeLists.txt
@@ -18,6 +18,10 @@ if (GGML_METAL_NDEBUG)
     add_compile_definitions(GGML_METAL_NDEBUG)
 endif()
 
+if (GGML_METAL_MULTIDEV)
+       add_compile_definitions(GGML_METAL_MULTIDEV)
+endif()
+
 if (GGML_METAL_USE_BF16)
     add_compile_definitions(GGML_METAL_USE_BF16)
 endif()
diff --git a/ggml/src/ggml-metal/ggml-metal.m b/ggml/src/ggml-metal/ggml-metal.m
index 28f590f9..6e67274f 100644
--- a/ggml/src/ggml-metal/ggml-metal.m
+++ b/ggml/src/ggml-metal/ggml-metal.m
@@ -58,7 +58,14 @@ static id<MTLDevice> ggml_backend_metal_device_acq(struct ggml_backend_metal_dev
     assert(ctx != NULL);
 
     if (ctx->mtl_device == nil) {
+#if defined(GGML_METAL_MULTIDEV)
+       NSArray<id<MTLDevice>> *devices = MTLCopyAllDevices();
+        NSString *user_device_string = [[NSProcessInfo processInfo] environment][@"GGML_METAL_DEVICE"];
+       int user_device_int = [user_device_string intValue];
+        ctx->mtl_device = devices[user_device_int];
+#else
         ctx->mtl_device = MTLCreateSystemDefaultDevice();
+#endif
 
         ctx->has_simdgroup_reduction  = [ctx->mtl_device supportsFamily:MTLGPUFamilyApple7];
         ctx->has_simdgroup_reduction |= [ctx->mtl_device supportsFamily:MTLGPUFamilyMetal3_GGML];