ROCm Port update

* use hipblas based on cublas
* Update Makefile for the Cuda kernels
* Expand arch list and make it overrideable
* Fix multi GPU on multiple amd architectures with rocblas_initialize() (ggml-org#5)
* add hipBLAS to README
* new build arg LLAMA_CUDA_MMQ_Y
* fix half2 decomposition
* Add intrinsics polyfills for AMD
* AMD assembly optimized __dp4a
* Allow overriding CC_TURING
* use "ROCm" instead of "CUDA"
* ignore all build dirs
* Add Dockerfiles
* fix llama-bench
* fix -nommq help for non CUDA/HIP

---------

Co-Authored-By: YellowRoseCx <80486540+YellowRoseCx@users.noreply.github.com>
Co-Authored-By: ardfork <134447697+ardfork@users.noreply.github.com>
Co-Authored-By: funnbot <22226942+funnbot@users.noreply.github.com>
Co-Authored-By: Engininja2 <139037756+Engininja2@users.noreply.github.com>
Co-Authored-By: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
Co-Authored-By: jammm <2500920+jammm@users.noreply.github.com>
Co-Authored-By: jdecourval <7315817+jdecourval@users.noreply.github.com>

Author: 7 people

.gitignore

@@ -12,20 +12,7 @@
 .vs/
 .vscode/
 
-build/
-build-em/
-build-debug/
-build-release/
-build-ci-debug/
-build-ci-release/
-build-static/
-build-cublas/
-build-opencl/
-build-metal/
-build-mpi/
-build-no-accel/
-build-sanitize-addr/
-build-sanitize-thread/
+build*/
 out/
 tmp/
 

CMakeLists.txt

@@ -137,16 +137,18 @@ if (LLAMA_HIPBLAS)
 
     find_package(hip)
     find_package(hipblas)
+    find_package(rocblas)
 
     if (${hipblas_FOUND} AND ${hip_FOUND})
         message(STATUS "HIP and hipBLAS found")
         add_compile_definitions(GGML_USE_HIPBLAS GGML_USE_CUBLAS)
         add_library(ggml-rocm OBJECT ggml-cuda.cu ggml-cuda.h)
         target_compile_definitions(ggml-rocm PRIVATE GGML_CUDA_DMMV_X=${LLAMA_CUDA_DMMV_X})
-        target_compile_definitions(ggml-rocm PRIVATE GGML_CUDA_DMMV_Y=${LLAMA_CUDA_DMMV_Y})
+        target_compile_definitions(ggml-rocm PRIVATE GGML_CUDA_MMV_Y=${LLAMA_CUDA_MMV_Y})
         target_compile_definitions(ggml-rocm PRIVATE K_QUANTS_PER_ITERATION=${LLAMA_CUDA_KQUANTS_ITER})
+        target_compile_definitions(ggml-rocm PRIVATE CC_TURING=1000000000)
         set_source_files_properties(ggml-cuda.cu PROPERTIES LANGUAGE CXX)
-        target_link_libraries(ggml-rocm PRIVATE hip::device PUBLIC hip::host roc::hipblas)
+        target_link_libraries(ggml-rocm PRIVATE hip::device PUBLIC hip::host roc::rocblas roc::hipblas)
 
         if (LLAMA_STATIC)
             message(FATAL_ERROR "Static linking not supported for HIP/ROCm")
@@ -175,6 +177,9 @@ if (LLAMA_HIPBLAS)
         message(STATUS "HIP and hipBLAS found")
         add_compile_definitions(GGML_USE_HIPBLAS GGML_USE_CUBLAS)
         add_library(ggml-rocm OBJECT ggml-cuda.cu ggml-cuda.h)
+        if (LLAMA_CUDA_FORCE_DMMV)
+            target_compile_definitions(ggml-rocm PRIVATE GGML_CUDA_FORCE_DMMV)
+        endif()
         target_compile_definitions(ggml-rocm PRIVATE GGML_CUDA_DMMV_X=${LLAMA_CUDA_DMMV_X})
         target_compile_definitions(ggml-rocm PRIVATE GGML_CUDA_MMV_Y=${LLAMA_CUDA_MMV_Y})
         target_compile_definitions(ggml-rocm PRIVATE K_QUANTS_PER_ITERATION=${LLAMA_CUDA_KQUANTS_ITER})

Makefile

@@ -194,14 +194,17 @@ ggml_v2-cuda-legacy.o: otherarch/ggml_v2-cuda-legacy.cu otherarch/ggml_v2-cuda-l
 endif # LLAMA_CUBLAS
 
 ifdef LLAMA_HIPBLAS
-	ROCM_PATH  ?= /opt/rocm
+	ROCM_PATH	?= /opt/rocm
 	CC         := $(ROCM_PATH)/llvm/bin/clang
 	CXX        := $(ROCM_PATH)/llvm/bin/clang++
-	GPU_TARGETS = gfx803 gfx900 gfx906 gfx908 gfx90a gfx1030 gfx1100
+	GPU_TARGETS ?= gfx803 gfx900 gfx906 gfx908 gfx90a gfx1030 gfx1100
 	LLAMA_CUDA_DMMV_X ?= 128 
 	LLAMA_CUDA_MMV_Y ?= 2
 	LLAMA_CUDA_KQUANTS_ITER ?= 1
 	HIPFLAGS   += -DGGML_USE_HIPBLAS -DGGML_USE_CUBLAS $(shell $(ROCM_PATH)/bin/hipconfig -C)
+ifdef LLAMA_CUDA_FORCE_DMMV
+	HIPFLAGS 	+= -DGGML_CUDA_FORCE_DMMV
+endif # LLAMA_CUDA_FORCE_DMMV
 	HIPLDFLAGS    += -L$(ROCM_PATH)/lib -Wl,-rpath=$(ROCM_PATH)/lib -lhipblas -lamdhip64 -lrocblas
 	HIP_OBJS       += ggml-cuda.o ggml_v2-cuda.o ggml_v2-cuda-legacy.o
 ggml-cuda.o: HIPFLAGS += $(addprefix --offload-arch=,$(GPU_TARGETS)) \
@@ -228,6 +231,7 @@ ggml_v2-cuda-legacy.o: otherarch/ggml_v2-cuda-legacy.cu otherarch/ggml_v2-cuda-l
 endif # LLAMA_HIPBLAS
 
 
+
 ifdef LLAMA_METAL
 	CFLAGS   += -DGGML_USE_METAL -DGGML_METAL_NDEBUG
 	CXXFLAGS += -DGGML_USE_METAL

examples/common.cpp

@@ -597,11 +597,13 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
     fprintf(stdout, "                        number of layers to store in VRAM\n");
     fprintf(stdout, "  -ts SPLIT --tensor-split SPLIT\n");
     fprintf(stdout, "                        how to split tensors across multiple GPUs, comma-separated list of proportions, e.g. 3,1\n");
-    fprintf(stdout, "  -mg i, --main-gpu i   the GPU to use for scratch and small tensors\n" );
-    fprintf(stdout, "  -lv, --low-vram       don't allocate VRAM scratch buffer\n" );
-    fprintf(stdout, "  -mmq, --mul-mat-q     use experimental mul_mat_q CUDA kernels instead of cuBLAS. TEMP!!!\n" );
-    fprintf(stdout, "                        Reduces VRAM usage by 700/970/1430 MiB for 7b/13b/33b but prompt processing speed\n" );
-    fprintf(stdout, "                        is still suboptimal, especially q2_K, q3_K, q5_K, and q6_K.\n" );
+    fprintf(stdout, "  -mg i, --main-gpu i   the GPU to use for scratch and small tensors\n");
+    fprintf(stdout, "  -lv, --low-vram       don't allocate VRAM scratch buffer\n");
+#ifdef GGML_USE_CUBLAS
+    fprintf(stdout, "  -nommq, --no-mul-mat-q\n");
+    fprintf(stdout, "                        use " GGML_CUBLAS_NAME " instead of custom mul_mat_q " GGML_CUDA_NAME " kernels.\n");
+    fprintf(stdout, "                        Not recommended since this is both slower and uses more VRAM.\n");
+#endif // GGML_USE_CUBLAS
 #endif
     fprintf(stdout, "  --mtest               compute maximum memory usage\n");
     fprintf(stdout, "  --export              export the computation graph to 'llama.ggml'\n");

ggml-cuda.cu

@@ -4746,7 +4746,7 @@ void ggml_init_cublas() {
         CUDA_CHECK(cudaGetDeviceCount(&g_device_count));
         GGML_ASSERT(g_device_count <= GGML_CUDA_MAX_DEVICES);
         int64_t total_vram = 0;
-        fprintf(stderr, "%s: found %d CUDA devices:\n", __func__, g_device_count);
+        fprintf(stderr, "%s: found %d " GGML_CUDA_NAME " devices:\n", __func__, g_device_count);
         for (int id = 0; id < g_device_count; ++id) {
             cudaDeviceProp prop;
             CUDA_CHECK(cudaGetDeviceProperties(&prop, id));

ggml-cuda.h

@@ -2,6 +2,14 @@
 
 #include "ggml.h"
 
+#ifdef GGML_USE_HIPBLAS
+#define GGML_CUDA_NAME "ROCm"
+#define GGML_CUBLAS_NAME "hipBLAS"
+#else
+#define GGML_CUDA_NAME "CUDA"
+#define GGML_CUBLAS_NAME "cuBLAS"
+#endif
+
 #ifdef  __cplusplus
 extern "C" {
 #endif

llama.cpp

@@ -1187,7 +1187,7 @@ static void llama_model_load_internal(
     (void) main_gpu;
     (void) mul_mat_q;
 #if defined(GGML_USE_CUBLAS)
-    LLAMA_LOG_INFO("%s: using CUDA for GPU acceleration\n", __func__);
+    LLAMA_LOG_INFO("%s: using " GGML_CUDA_NAME " for GPU acceleration\n", __func__);
     ggml_cuda_set_main_device(main_gpu);
     ggml_cuda_set_mul_mat_q(mul_mat_q);
 #define LLAMA_BACKEND_OFFLOAD       GGML_BACKEND_GPU