add fast matmul iquants (llama/22504)

2026-10-10 00:05:41 +02:00 · 2026-05-01 13:07:35 +03:00
parent 66392cf1a2
commit d74c56862b
3 changed files with 443 additions and 1 deletions
@@ -1422,7 +1422,7 @@ static webgpu_encoded_op ggml_webgpu_mul_mat(webgpu_context & ctx,
                case GGML_TYPE_IQ3_S:
                case GGML_TYPE_IQ4_NL:
                case GGML_TYPE_IQ4_XS:
-                    use_fast = is_vec;
+                    use_fast = true;
                    break;
                default:
                    break;