Q5_0_R4 (#121)

* Adding q5_0_r4

We get PP-512(LLaMA-3.1-8B) = 256.7 t/s on a Ryzen-7950X.
We even get TG-128 improvement to 11.7 t/s from 11.1 t/s.

* q5_0_r4: NEON

We get PP-512(LLaMA-3.1-8B) = 99.6 t/s on M2-Max,
up from 71.0 t/s for Q5_0. The difference to mainline llama.cpp
is no longer funny: they get 26.5 t/s for Q5_0.

For TG, we are nor able to fully saturate memory bandwidth
and arrive at 22.1 t/s @ 8 threads. Mainline llama.cpp gets
20.6 t/s for Q5_0.

---------

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

1 files changed, 6 insertions, 0 deletions

diff --git a/ggml/src/iqk/iqk_quantize.h b/ggml/src/iqk/iqk_quantize.h
index 53caed4c..24c241a2 100644
--- a/ggml/src/iqk/iqk_quantize.h
+++ b/ggml/src/iqk/iqk_quantize.h
@@ -81,6 +81,12 @@ size_t quantize_q8_0_r4(const float * GGML_RESTRICT src, void * GGML_RESTRICT ds
 void   dequantize_row_q8_0_r4(const block_q8_0_x4  * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
 void   vec_dot_q8_0_r4_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
 
+void   quantize_row_q5_0_r4_ref(const float * GGML_RESTRICT x, block_q5_0_r4  * GGML_RESTRICT y, int64_t k);
+void   quantize_row_q5_0_r4(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
+size_t quantize_q5_0_r4(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
+void   dequantize_row_q5_0_r4(const block_q5_0_r4  * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
+void   vec_dot_q5_0_r4_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
+
 #ifdef __cplusplus
 }
 #endif