metal : add/mul/div use general kernel when src1 not cont

YellowRoseCx · Dec 9, 2023 · 7e2006b · 7e2006b
1 parent 06dfde3
commit 7e2006b
Showing 1 changed file with 3 additions and 4 deletions.
diff --git a/ggml-metal.m b/ggml-metal.m
@@ -1003,14 +1003,13 @@ void ggml_metal_graph_compute(
                     case GGML_OP_MUL:
                     case GGML_OP_DIV:
                         {
-                            GGML_ASSERT(ggml_is_contiguous(src0));
-                            GGML_ASSERT(ggml_is_contiguous(src1));
-
                             bool bcast_row = false;
 
                             int64_t nb = ne00;
 
-                            if (ggml_nelements(src1) == ne10 && ne00 % 4 == 0) {
+                            if (ggml_nelements(src1) == ne10 && ggml_is_contiguous(src1) && ne00 % 4 == 0) {
+                                GGML_ASSERT(ggml_is_contiguous(src0));
+
                                 // src1 is a row
                                 GGML_ASSERT(ne11 == 1);