1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
|
From 19cdfbefcf600c878fd55c144c4dea7186fbc395 Mon Sep 17 00:00:00 2001
From: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Date: Sat, 22 Sep 2018 17:21:30 -0400
Subject: [PATCH] celt: Force stack alignment for functions using __m128
We are using an old GCC for Windows 32bit builds, and it assumes the
stack will be 16bytes aligned when dealing with __m128 which should be
aligned. As a side effect, when it's not, movaps will crash. This should
be fixed in newer version of GCC.
---
celt/x86/pitch_sse.c | 11 ++++++++++-
1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/celt/x86/pitch_sse.c b/celt/x86/pitch_sse.c
index 20e7312..3eee8da 100644
--- a/celt/x86/pitch_sse.c
+++ b/celt/x86/pitch_sse.c
@@ -40,6 +40,16 @@
#include <xmmintrin.h>
#include "arch.h"
+#ifdef _WIN32
+void xcorr_kernel_sse(const opus_val16 *x, const opus_val16 *y, opus_val32 sum[4], int len) __attribute__((force_align_arg_pointer));
+void dual_inner_prod_sse(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02,
+ int N, opus_val32 *xy1, opus_val32 *xy2) __attribute__((force_align_arg_pointer));
+opus_val32 celt_inner_prod_sse(const opus_val16 *x, const opus_val16 *y,
+ int N) __attribute__((force_align_arg_pointer));
+void comb_filter_const_sse(opus_val32 *y, opus_val32 *x, int T, int N,
+ opus_val16 g10, opus_val16 g11, opus_val16 g12) __attribute__((force_align_arg_pointer));
+#endif
+
void xcorr_kernel_sse(const opus_val16 *x, const opus_val16 *y, opus_val32 sum[4], int len)
{
int j;
@@ -75,7 +85,6 @@ void xcorr_kernel_sse(const opus_val16 *x, const opus_val16 *y, opus_val32 sum[4
_mm_storeu_ps(sum,_mm_add_ps(xsum1,xsum2));
}
-
void dual_inner_prod_sse(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02,
int N, opus_val32 *xy1, opus_val32 *xy2)
{
--
2.17.1
|
