diff options
| -rw-r--r-- | CMakeLists.txt | 3 | ||||
| -rw-r--r-- | ggml/CMakeLists.txt | 3 | ||||
| -rw-r--r-- | ggml/src/ggml.c | 5 |
3 files changed, 6 insertions, 5 deletions
diff --git a/CMakeLists.txt b/CMakeLists.txt index 0723e071..cb59656e 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -7,6 +7,9 @@ set(CMAKE_WARN_UNUSED_CLI YES) set(CMAKE_EXPORT_COMPILE_COMMANDS ON) +set(CMAKE_CXX_STANDARD 17) +set(CMAKE_CXX_STANDARD_REQUIRED true) + set(CMAKE_CUDA_USE_RESPONSE_FILE_FOR_INCLUDES 0) set(CMAKE_CUDA_USE_RESPONSE_FILE_FOR_LIBRARIES 0) set(CMAKE_CUDA_USE_RESPONSE_FILE_FOR_OBJECTS 0) diff --git a/ggml/CMakeLists.txt b/ggml/CMakeLists.txt index 1bb3d1ee..90b37d5b 100644 --- a/ggml/CMakeLists.txt +++ b/ggml/CMakeLists.txt @@ -165,9 +165,6 @@ option(GGML_BUILD_EXAMPLES "ggml: build examples" ${GGML_STANDALONE}) set(CMAKE_C_STANDARD 11) set(CMAKE_C_STANDARD_REQUIRED true) -set(CMAKE_CXX_STANDARD 17) -set(CMAKE_CXX_STANDARD_REQUIRED true) - set(THREADS_PREFER_PTHREAD_FLAG ON) find_package(Threads REQUIRED) diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c index 8f3d8f3f..86cc669e 100644 --- a/ggml/src/ggml.c +++ b/ggml/src/ggml.c @@ -13459,6 +13459,8 @@ UseGgmlGemm1:; } } + ggml_barrier(params->shared); + #if IK_PRINT_TIMING int64_t t2 = ggml_time_us(); if (ith == 0) printf("quantize(%s): %d us\n", dst->name, (int)(t2 - t1)); @@ -13468,11 +13470,10 @@ UseGgmlGemm1:; wdata -= GGML_MAX_NAME; memcpy(wdata, src1->name, GGML_MAX_NAME); // Every thread starts at ith, so the first unprocessed chunk is nth. This save a bit of coordination right at the start. - atomic_store(¶ms->shared->current_chunk, nth); + //atomic_store(¶ms->shared->current_chunk, nth); } AlreadyQunatized:; - ggml_barrier(params->shared); } const void * wdata = (src1->type == vec_dot_type) ? src1->data |