ggml : minor naming changes (ggml-org#8433)

* ggml : minor naming changes

ggml-ci

* ggml : use PRId64 [no ci]

* ggml : revert FA K/Q names

Author: ggerganov

examples/quantize-stats/quantize-stats.cpp

@@ -154,7 +154,7 @@ static void test_roundtrip_on_chunk(
     }
 
     if (use_reference) {
-        qfns.from_float_reference(input_scratch, quantized_scratch, chunk_size);
+        qfns.from_float_ref(input_scratch, quantized_scratch, chunk_size);
     } else {
         qfns.from_float(input_scratch, quantized_scratch, chunk_size);
     }

ggml/include/ggml.h

@@ -708,9 +708,9 @@ extern "C" {
     GGML_API GGML_CALL size_t  ggml_nbytes      (const struct ggml_tensor * tensor);
     GGML_API           size_t  ggml_nbytes_pad  (const struct ggml_tensor * tensor); // same as ggml_nbytes() but padded to GGML_MEM_ALIGN
 
-    GGML_API GGML_CALL int    ggml_blck_size(enum ggml_type type);
-    GGML_API GGML_CALL size_t ggml_type_size(enum ggml_type type);             // size in bytes for all elements in a block
-    GGML_API GGML_CALL size_t ggml_row_size (enum ggml_type type, int64_t ne); // size in bytes for all elements in a row
+    GGML_API GGML_CALL int64_t ggml_blck_size(enum ggml_type type);
+    GGML_API GGML_CALL size_t  ggml_type_size(enum ggml_type type);             // size in bytes for all elements in a block
+    GGML_API GGML_CALL size_t  ggml_row_size (enum ggml_type type, int64_t ne); // size in bytes for all elements in a row
 
     GGML_DEPRECATED(
     GGML_API double ggml_type_sizef(enum ggml_type type), // ggml_type_size()/ggml_blck_size() as float
@@ -2404,20 +2404,31 @@ extern "C" {
 #endif
     typedef void (*ggml_to_float_t)  (const void  * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
     typedef void (*ggml_from_float_t)(const float * GGML_RESTRICT x, void  * GGML_RESTRICT y, int64_t k);
-    typedef void (*ggml_vec_dot_t)   (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT x, size_t bx,
-                                      const void * GGML_RESTRICT y, size_t by, int nrc);
+    typedef void (*ggml_from_float_to_mat_t)
+                                     (const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t nr, int64_t k, int64_t bs);
+    typedef void (*ggml_vec_dot_t)  (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT x, size_t bx,
+                                       const void * GGML_RESTRICT y, size_t by, int nrc);
+    typedef void (*ggml_gemv_t)     (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT x,
+                                       const void * GGML_RESTRICT y, int nr, int nc);
+    typedef void (*ggml_gemm_t)     (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT x,
+                                       const void * GGML_RESTRICT y, int nr, int nc);
 
     typedef struct {
-        const char      * type_name;
-        int               blck_size;
-        size_t            type_size;
-        bool              is_quantized;
-        ggml_to_float_t   to_float;
-        ggml_from_float_t from_float;
-        ggml_from_float_t from_float_reference;
-        ggml_vec_dot_t    vec_dot;
-        enum ggml_type    vec_dot_type;
-        int64_t           nrows; // number of rows to process simultaneously;
+        const char             * type_name;
+        int64_t                  blck_size;
+        int64_t                  blck_size_interleave; // interleave elements in blocks
+        size_t                   type_size;
+        bool                     is_quantized;
+        ggml_to_float_t          to_float;
+        ggml_from_float_t        from_float;
+        ggml_from_float_t        from_float_ref;
+        ggml_from_float_to_mat_t from_float_to_mat;
+        ggml_vec_dot_t           vec_dot;
+        enum ggml_type           vec_dot_type;
+        int64_t                  nrows; // number of rows to process simultaneously
+        int64_t                  ncols; // number of columns to process simultaneously
+        ggml_gemv_t              gemv;
+        ggml_gemm_t              gemm;
     } ggml_type_traits_t;
 
     GGML_API ggml_type_traits_t ggml_internal_get_type_traits(enum ggml_type type);