metal : add build system support for embedded metal library (#5604)

* add build support for embedded metal library

* Update Makefile

---------

Co-authored-by: Haoxiang Fei <redacted>
Co-authored-by: Georgi Gerganov <redacted>

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 168b133f4e1c1e2af564298ecf8c8684ea993cde..3c4629001a38d747e36487591bed3e91065579e2 100644 (file)
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -110,6 +110,7 @@ option(LLAMA_VULKAN_RUN_TESTS                "llama: run Vulkan tests"
 option(LLAMA_METAL                           "llama: use Metal"                                 ${LLAMA_METAL_DEFAULT})
 option(LLAMA_METAL_NDEBUG                    "llama: disable Metal debugging"                   OFF)
 option(LLAMA_METAL_SHADER_DEBUG              "llama: compile Metal with -fno-fast-math"         OFF)
+option(LLAMA_METAL_EMBED_LIBRARY             "llama: embed Metal library"                       OFF)
 option(LLAMA_KOMPUTE                         "llama: use Kompute"                               OFF)
 option(LLAMA_MPI                             "llama: use MPI"                                   OFF)
 option(LLAMA_QKK_64                          "llama: use super-block size of 64 for k-quants"   OFF)
@@ -201,6 +202,29 @@ if (LLAMA_METAL)
     # copy ggml-metal.metal to bin directory
     configure_file(ggml-metal.metal ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/ggml-metal.metal COPYONLY)
 
+    if (LLAMA_METAL_EMBED_LIBRARY)
+        enable_language(ASM)
+        add_compile_definitions(GGML_METAL_EMBED_LIBRARY)
+
+        set(METALLIB_SOURCE "${CMAKE_SOURCE_DIR}/ggml-metal.metal")
+        file(MAKE_DIRECTORY "${CMAKE_BINARY_DIR}/autogenerated")
+        set(EMBED_METALLIB_ASSEMBLY "${CMAKE_BINARY_DIR}/autogenerated/ggml-embed-metallib.s")
+
+        add_custom_command(
+            OUTPUT ${EMBED_METALLIB_ASSEMBLY}
+            COMMAND echo ".section __DATA,__ggml_metallib" > ${EMBED_METALLIB_ASSEMBLY}
+            COMMAND echo ".globl _ggml_metallib_start" >> ${EMBED_METALLIB_ASSEMBLY}
+            COMMAND echo "_ggml_metallib_start:" >> ${EMBED_METALLIB_ASSEMBLY}
+            COMMAND echo ".incbin \\\"${METALLIB_SOURCE}\\\"" >> ${EMBED_METALLIB_ASSEMBLY}
+            COMMAND echo ".globl _ggml_metallib_end" >> ${EMBED_METALLIB_ASSEMBLY}
+            COMMAND echo "_ggml_metallib_end:" >> ${EMBED_METALLIB_ASSEMBLY}
+            DEPENDS ${METALLIB_SOURCE}
+            COMMENT "Generate assembly for embedded Metal library"
+        )
+
+        set(GGML_SOURCES_METAL ${GGML_SOURCES_METAL} ${EMBED_METALLIB_ASSEMBLY})
+    endif()
+
     if (LLAMA_METAL_SHADER_DEBUG)
         # custom command to do the following:
         #   xcrun -sdk macosx metal    -fno-fast-math -c ggml-metal.metal -o ggml-metal.air

diff --git a/Makefile b/Makefile
index db5df1b328560f2b9a222ebc2bc9d1de75b5bfdc..211a08d7ff186e8a20b3bec5c57dcd46194bea91 100644 (file)
--- a/Makefile
+++ b/Makefile
@@ -533,11 +533,29 @@ ifdef LLAMA_METAL
 ifdef LLAMA_METAL_NDEBUG
        MK_CPPFLAGS += -DGGML_METAL_NDEBUG
 endif
+ifdef LLAMA_METAL_EMBED_LIBRARY
+       MK_CPPFLAGS += -DGGML_METAL_EMBED_LIBRARY
+       OBJS        += ggml-metal-embed.o
+endif
 endif # LLAMA_METAL
 
 ifdef LLAMA_METAL
 ggml-metal.o: ggml-metal.m ggml-metal.h
        $(CC) $(CFLAGS) -c $< -o $@
+
+ifdef LLAMA_METAL_EMBED_LIBRARY
+ggml-metal-embed.o: ggml-metal.metal
+       @echo "Embedding Metal library"
+       $(eval TEMP_ASSEMBLY=$(shell mktemp))
+       @echo ".section __DATA, __ggml_metallib" > $(TEMP_ASSEMBLY)
+       @echo ".globl _ggml_metallib_start" >> $(TEMP_ASSEMBLY)
+       @echo "_ggml_metallib_start:" >> $(TEMP_ASSEMBLY)
+       @echo ".incbin \"$<\"" >> $(TEMP_ASSEMBLY)
+       @echo ".globl _ggml_metallib_end" >> $(TEMP_ASSEMBLY)
+       @echo "_ggml_metallib_end:" >> $(TEMP_ASSEMBLY)
+       @$(AS) $(TEMP_ASSEMBLY) -o $@
+       @rm -f ${TEMP_ASSEMBLY}
+endif
 endif # LLAMA_METAL
 
 ifdef LLAMA_MPI