Kawrakow 154e0d75fc Merge mainline llama.cpp (#3) ...

* Merging mainline - WIP

* Merging mainline - WIP

AVX2 and CUDA appear to work.
CUDA performance seems slightly (~1-2%) lower as it is so often
the case with llama.cpp/ggml after some "improvements" have been made.

* Merging mainline - fix Metal

* Remove check

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>

2024-07-27 07:55:01 +02:00

..

common.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_add.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_addrow.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_cpy_f16_f16.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_cpy_f16_f32.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_cpy_f32_f16.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_cpy_f32_f32.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_diagmask.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_gelu.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_getrows_f16.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_getrows_f32.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_getrows_q4_0.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_getrows_q4_1.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_getrows_q6_k.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_getrows.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul_mat_f16.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul_mat_mat_f32.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul_mat_q4_0.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul_mat_q4_1.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul_mat_q6_k.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul_mat_q8_0.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul_mv_q_n_pre.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul_mv_q_n.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_mul.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_norm.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_relu.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_rmsnorm.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_rope_f16.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_rope_f32.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_scale_8.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_scale.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_silu.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

op_softmax.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

rope_common.comp

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00