From bda44ec484e45eb0ba8c78e67179dc2f9ce19368 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Daniel=20Gr=C3=A4fe?= <Daniel.Alm@ForumD.net>
Date: Sun, 21 Jan 2024 10:28:08 +0100
Subject: [PATCH] Use a custom batch size threshold for using a batch-specific
 kernel.

---
 ggml/src/ggml-metal.m | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ggml/src/ggml-metal.m b/ggml/src/ggml-metal.m
index 48b8131312a3ef..f8e97fde6d93d6 100644
--- a/ggml/src/ggml-metal.m
+++ b/ggml/src/ggml-metal.m
@@ -1543,7 +1543,7 @@ static enum ggml_status ggml_metal_graph_compute(
 
                         // find the break-even point where the matrix-matrix kernel becomes more efficient compared
                         // to the matrix-vector kernel
-                        int ne11_mm_min = 1;
+                        int ne11_mm_min = 6;
 
 #if 0
                         // the numbers below are measured on M2 Ultra for 7B and 13B models