diff options
Diffstat (limited to 'llama.cpp')
| -rw-r--r-- | llama.cpp | 16 |
1 files changed, 16 insertions, 0 deletions
@@ -11,6 +11,8 @@ # include "ggml-cuda.h" #elif defined(GGML_USE_CLBLAST) # include "ggml-opencl.h" +#elif defined(GGML_USE_SYCL) +# include "ggml-sycl.h" #endif #ifdef GGML_USE_METAL @@ -1278,6 +1280,8 @@ static ggml_backend_buffer_type_t llama_default_buffer_type_cpu(bool host_buffer if (host_buffer) { buft = ggml_backend_cuda_host_buffer_type(); } +#elif defined(GGML_USE_SYCL) + buft = ggml_backend_sycl_host_buffer_type(); #elif defined(GGML_USE_CPU_HBM) buft = ggml_backend_cpu_hbm_buffer_type(); #endif @@ -1297,6 +1301,8 @@ static ggml_backend_buffer_type_t llama_default_buffer_type_offload(int gpu) { buft = ggml_backend_metal_buffer_type(); #elif defined(GGML_USE_CUBLAS) buft = ggml_backend_cuda_buffer_type(gpu); +#elif defined(GGML_USE_SYCL) + buft = ggml_backend_sycl_buffer_type(gpu); #elif defined(GGML_USE_CLBLAST) buft = ggml_backend_opencl_buffer_type(); #endif @@ -10225,6 +10231,16 @@ struct llama_context * llama_new_context_with_model( } } } +#elif defined(GGML_USE_SYCL) + if (model->n_gpu_layers > 0) { + ggml_backend_t backend = ggml_backend_sycl_init(model->main_gpu); + if (backend == nullptr) { + LLAMA_LOG_ERROR("%s: failed to initialize SYCL%d backend\n", __func__, model->main_gpu); + llama_free(ctx); + return nullptr; + } + ctx->backends.push_back(backend); + } #endif ctx->backend_cpu = ggml_backend_cpu_init(); if (ctx->backend_cpu == nullptr) { |