]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
ggml : AVX2 support for Q4_0_8_8 (llama/8713)
authorSrihari-mcw <redacted>
Wed, 4 Sep 2024 16:51:22 +0000 (22:21 +0530)
committerGeorgi Gerganov <redacted>
Tue, 24 Sep 2024 16:45:08 +0000 (19:45 +0300)
commit01e214a1d7c3c75ab1bfac9d2d8438ff52fa6569
treeff555c7f54f31b289df43efb3d6eb4f5ec375826
parent1cecfe6a02909d797ae08c8ee3a5a1f28cb542f0
ggml : AVX2 support for Q4_0_8_8 (llama/8713)

* Add AVX2 based implementations for quantize_q8_0_4x8, ggml_gemv_q4_0_8x8_q8_0 and ggml_gemm_q4_0_8x8_q8_0 functions

* Update code to fix issues occuring due to non alignment of elements to be processed as multiple of 16 in MSVC

* Update comments and indentation

* Make updates to reduce number of load instructions
ggml/src/ggml-aarch64.c