From 8d25450d3c750a80cdd0cfde3b132def855148fc Mon Sep 17 00:00:00 2001
From: Sun Yimin <emmansun@users.noreply.github.com>
Date: Tue, 12 Nov 2024 10:34:52 +0800
Subject: [PATCH] sm3: try s390x transpose matrix optimization

---
 sm3/sm3blocks_s390x.s | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sm3/sm3blocks_s390x.s b/sm3/sm3blocks_s390x.s
index 8eeb1a7..cd1a930 100644
--- a/sm3/sm3blocks_s390x.s
+++ b/sm3/sm3blocks_s390x.s
@@ -47,7 +47,7 @@ GLOBL mask<>(SB), RODATA, $64
 	VPERM T2, T3, M0, TMP1; \
 	VPERM T0, T1, M1, TMP2; \
 	VPERM T2, T3, M1, TMP3; \
-	VPERM TMP0, TMP1, M2, T0; \
+	VPDI $0x2, TMP0, TMP1, T0; \
 	VPERM TMP0, TMP1, M3, T1; \
 	VPERM TMP2, TMP3, M2, T2; \
 	VPERM TMP2, TMP3, M3, T3