Note: This issue was copied from ggml-org#10180
Original Author: @ggerganov
Original Issue Number: ggml-org#10180
Created: 2024-11-05T07:12:48Z
As per recent discussions (e.g. ggml-org#10144 (review)), we should split the large ggml-cpu.c implementation into smaller modules - similar to how the CUDA backend is organized. We should utilize C++11 C++ to reduce code duplication.
Note: This issue was copied from ggml-org#10180
Original Author: @ggerganov
Original Issue Number: ggml-org#10180
Created: 2024-11-05T07:12:48Z
As per recent discussions (e.g. ggml-org#10144 (review)), we should split the large
ggml-cpu.cimplementation into smaller modules - similar to how the CUDA backend is organized. We should utilizeC++11C++ to reduce code duplication.