Q6_K_R4 (#130)

* Adding q6_k_r4

* q6_k_r4: 1st functional AVX2 version

* q6_k_r4: AVX2 and simple Zen4

"Simple" as in processing 4 instead of 8 rows at once.
On Zen4 we get PP-512(LLaMA-3.1-8B) = 238.3 t/s vs
195.2 t/s for Q6_K. TG-128 @ 1 thread is 7.94 t/s
vs 5.38 t/s for Q6_K.

* q6_k_r4: 1st NEON version

PP-512(LLaMA-3.1-8B) = 78 t/s vs 57.6 t/s for q6_K.
TG-128 is slightly lower rthan q6_K for low number of threads,
becomes very slightly better at 8 threads.

* q6_k_r4: slightly faster NEON

PP-512(LLaMA-3.1-8B) = 83.25 t/s

* q6_k_r4: slightly faster Zen4

238.3 t/s -> 243.2 t/s

---------

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

1 files changed, 22 insertions, 0 deletions

diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index fadda3e3..b92c2352 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -927,6 +927,19 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
         .nrows                    = 1,
         .row_meta_size            = 0,
     },
+    [GGML_TYPE_Q6_K_R4] = {
+        .type_name                = "q6_k_r4",
+        .blck_size                = QK_K,
+        .type_size                = sizeof(block_q6_K),
+        .is_quantized             = true,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q6_k_r4,
+        .from_float               = quantize_row_q6_k_r4,
+        .from_float_ref           = (ggml_from_float_t) quantize_row_q6_k_r4_ref,
+        .vec_dot                  = vec_dot_q6_k_r4_q8_k,
+        .vec_dot_type             = GGML_TYPE_Q8_K,
+        .nrows                    = 1,
+        .row_meta_size            = 0,
+    },
     [GGML_TYPE_IQ2_XXS] = {
         .type_name                = "iq2_xxs",
         .blck_size                = QK_K,
@@ -4036,6 +4049,7 @@ enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype) {
         case GGML_FTYPE_MOSTLY_Q4_K_R4:       wtype = GGML_TYPE_Q4_K_R4;  break;
         case GGML_FTYPE_MOSTLY_Q5_K:          wtype = GGML_TYPE_Q5_K;     break;
         case GGML_FTYPE_MOSTLY_Q6_K:          wtype = GGML_TYPE_Q6_K;     break;
+        case GGML_FTYPE_MOSTLY_Q6_K_R4:       wtype = GGML_TYPE_Q6_K_R4;  break;
         case GGML_FTYPE_MOSTLY_IQ2_XXS:       wtype = GGML_TYPE_IQ2_XXS;  break;
         case GGML_FTYPE_MOSTLY_IQ2_XS:        wtype = GGML_TYPE_IQ2_XS;   break;
         case GGML_FTYPE_MOSTLY_IQ3_XXS:       wtype = GGML_TYPE_IQ3_XXS;  break;
@@ -10567,6 +10581,7 @@ static void ggml_compute_forward_add(
         case GGML_TYPE_Q4_K_R4:
         case GGML_TYPE_Q5_K:
         case GGML_TYPE_Q6_K:
+        case GGML_TYPE_Q6_K_R4:
         case GGML_TYPE_IQ2_XXS:
         case GGML_TYPE_IQ2_XS:
         case GGML_TYPE_IQ3_XXS:
@@ -11017,6 +11032,7 @@ static void ggml_compute_forward_add1(
         case GGML_TYPE_Q4_K_R4:
         case GGML_TYPE_Q5_K:
         case GGML_TYPE_Q6_K:
+        case GGML_TYPE_Q6_K_R4:
         case GGML_TYPE_IQ2_XXS:
         case GGML_TYPE_IQ2_XS:
         case GGML_TYPE_IQ3_XXS:
@@ -11164,6 +11180,7 @@ static void ggml_compute_forward_acc(
         case GGML_TYPE_Q4_K_R4:
         case GGML_TYPE_Q5_K:
         case GGML_TYPE_Q6_K:
+        case GGML_TYPE_Q6_K_R4:
         case GGML_TYPE_IQ2_XXS:
         case GGML_TYPE_IQ2_XS:
         case GGML_TYPE_IQ3_XXS:
@@ -14357,6 +14374,7 @@ static void ggml_compute_forward_out_prod(
         case GGML_TYPE_Q4_K_R4:
         case GGML_TYPE_Q5_K:
         case GGML_TYPE_Q6_K:
+        case GGML_TYPE_Q6_K_R4:
         case GGML_TYPE_IQ2_XXS:
         case GGML_TYPE_IQ2_XS:
         case GGML_TYPE_IQ3_XXS:
@@ -14744,6 +14762,7 @@ static void ggml_compute_forward_set(
         case GGML_TYPE_Q4_K_R4:
         case GGML_TYPE_Q5_K:
         case GGML_TYPE_Q6_K:
+        case GGML_TYPE_Q6_K_R4:
         case GGML_TYPE_IQ2_XXS:
         case GGML_TYPE_IQ2_XS:
         case GGML_TYPE_IQ3_XXS:
@@ -15025,6 +15044,7 @@ static void ggml_compute_forward_get_rows(
         case GGML_TYPE_Q4_K_R4:
         case GGML_TYPE_Q5_K:
         case GGML_TYPE_Q6_K:
+        case GGML_TYPE_Q6_K_R4:
         case GGML_TYPE_IQ2_XXS:
         case GGML_TYPE_IQ2_XS:
         case GGML_TYPE_IQ3_XXS:
@@ -15633,6 +15653,7 @@ static void ggml_compute_forward_clamp(
         case GGML_TYPE_Q4_K_R4:
         case GGML_TYPE_Q5_K:
         case GGML_TYPE_Q6_K:
+        case GGML_TYPE_Q6_K_R4:
         case GGML_TYPE_IQ2_XXS:
         case GGML_TYPE_IQ2_XS:
         case GGML_TYPE_IQ3_XXS:
@@ -22469,6 +22490,7 @@ size_t ggml_quantize_chunk(
         case GGML_TYPE_Q4_K_R4: result = quantize_q4_k_r4(src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break;
         case GGML_TYPE_Q5_K:    result = quantize_q5_K(src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break;
         case GGML_TYPE_Q6_K:    result = quantize_q6_K(src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break;
+        case GGML_TYPE_Q6_K_R4: result = quantize_q6_k_r4(src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break;
         case GGML_TYPE_IQ2_XXS: result = quantize_iq2_xxs(src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break;
         case GGML_TYPE_IQ2_XS:  result = quantize_iq2_xs (src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break;
         case GGML_TYPE_IQ3_XXS: result = quantize_iq3_xxs(src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break;