From fdd5ad9a542e78090893153b53ab4b56db915b6a Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Sun, 8 Oct 2023 12:05:15 +0300 Subject: [PATCH] metal : do not use mul_mm kernels when ne00 < 64 --- ggml-metal.m | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/ggml-metal.m b/ggml-metal.m index 92956ed97..a1b0cd233 100644 --- a/ggml-metal.m +++ b/ggml-metal.m @@ -1040,7 +1040,7 @@ void ggml_metal_graph_compute( !ggml_is_transposed(src0) && !ggml_is_transposed(src1) && src1t == GGML_TYPE_F32 && - ne00 % 32 == 0 && + ne00 % 32 == 0 && ne00 >= 64 && ne11 > ne11_mm_min) { //printf("matrix: ne00 = %6d, ne01 = %6d, ne02 = %6d, ne11 = %6d, ne12 = %6d\n", ne00, ne01, ne02, ne11, ne12); switch (src0->type) {