From be67ea8c7ca22d4803b3828c183af1d47a3932cb Mon Sep 17 00:00:00 2001
From: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date: Mon, 27 Jun 2022 23:27:48 -0400
Subject: [PATCH] Fixes wrap-around in silk_inner_prod16_sse4_1()
Thanks Tim
---
silk/fixed/x86/vector_ops_FIX_sse4_1.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/silk/fixed/x86/vector_ops_FIX_sse4_1.c b/silk/fixed/x86/vector_ops_FIX_sse4_1.c
index 0cfb08d90..a46289bbc 100644
--- a/silk/fixed/x86/vector_ops_FIX_sse4_1.c
+++ b/silk/fixed/x86/vector_ops_FIX_sse4_1.c
@@ -36,6 +36,7 @@
#include "SigProc_FIX.h"
#include "pitch.h"
+#include "celt/x86/x86cpu.h"
opus_int64 silk_inner_prod16_sse4_1(
const opus_int16 *inVec1, /* I input vector 1 */
@@ -78,7 +79,7 @@ opus_int64 silk_inner_prod16_sse4_1(
_mm_storel_epi64( (__m128i *)&sum, acc1 );
for( ; i < len; i++ ) {
- sum = silk_SMLABB( sum, inVec1[ i ], inVec2[ i ] );
+ sum = silk_SMLALBB( sum, inVec1[ i ], inVec2[ i ] );
}
#ifdef OPUS_CHECK_ASM
--
GitLab