metal : small-batch mat-mul kernels (ggerganov#10581)

* metal : small-batch mat-mul kernels ggml-ci * metal : add rest of types ggml-ci * metal : final adjustments ggml-ci * metal : add comments ggml-ci
3Simplex · Dec 3, 2024 · 0115df2 · 0115df2
1 parent 515d4e5
commit 0115df2
Show file tree

Hide file tree

Showing 4 changed files with 709 additions and 29 deletions.
diff --git a/ggml/src/ggml-metal/ggml-metal-impl.h b/ggml/src/ggml-metal/ggml-metal-impl.h
@@ -192,6 +192,30 @@ typedef struct {
     int16_t  r3;
 } ggml_metal_kargs_mul_mv;
 
+typedef struct {
+    int32_t  ne00;
+    int32_t  ne01;
+    int32_t  ne02;
+    uint64_t nb00;
+    uint64_t nb01;
+    uint64_t nb02;
+    uint64_t nb03;
+    int32_t  ne10;
+    int32_t  ne11;
+    int32_t  ne12;
+    uint64_t nb10;
+    uint64_t nb11;
+    uint64_t nb12;
+    uint64_t nb13;
+    int32_t  ne0;
+    int32_t  ne1;
+    int16_t  r2;
+    int16_t  r3;
+    int16_t  nsg;
+    int16_t  nxpsg;
+    int16_t  r1ptg;
+} ggml_metal_kargs_mul_mv_ext;
+
 typedef struct {
     int32_t  nei0;
     int32_t  nei1;