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
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
|
From af3ad8dc31da3072bd6b9ce3e44cd20e6253c50d Mon Sep 17 00:00:00 2001
From: "L. E. Segovia" <amy@centricular.com>
Date: Tue, 15 Sep 2026 17:50:56 -0300
Subject: [PATCH] audio-resampler-neon: handle Thumb1-only builds correctly
Fixes #5280
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/12490>
Upstream-Status: Backport [https://gitlab.freedesktop.org/gstreamer/gstreamer/-/commit/af3ad8dc31da3072bd6b9ce3e44cd20e6253c50d]
Cherry-picked to the 1.28 branch as 4f4fec2de8b8 (!12501), so this patch
can be dropped at 1.28.8.
Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
---
.../gst-libs/gst/audio/audio-resampler-neon.h | 48 ++++++++++++-------
meson.build | 3 +-
2 files changed, 31 insertions(+), 20 deletions(-)
--- a/gst-libs/gst/audio/audio-resampler-neon.h
+++ b/gst-libs/gst/audio/audio-resampler-neon.h
@@ -20,6 +20,18 @@
#include <gst/gstcpuid.h>
#include <stdint.h>
+#if defined(__thumb2__)
+/* Thumb-2: request the wide encoding explicitly. */
+#define ADD_W(x, y) "add.w " #x ", " #x ", %[" #y "]\n"
+#elif defined(__thumb__)
+/* Thumb-1 has no 3-operand ADD for high registers and no wide encodings. */
+#define ADD_W(x, y) "add " #x ", %[" #y "]\n"
+#else
+/* ARM (A32) has a single 32-bit ADD encoding and rejects the .w width
+ * suffix as invalid syntax, so emit a plain 3-operand ADD. */
+#define ADD_W(x, y) "add " #x ", " #x ", %[" #y "]\n"
+#endif
+
static inline void
inner_product_gint16_full_1_neon (gint16 * o, const gint16 * a,
const gint16 * b, gint len, const gint16 * icoeff, gint bstride)
@@ -133,11 +145,11 @@ inner_product_gint16_cubic_1_neon (gint1
"1:"
" mov r8, %[b]\n"
" vld1.16 {d16, d17}, [%[b]]!\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.16 {d18, d19}, [r8]\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.16 {d20, d21}, [r8]\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.16 {d22, d23}, [r8]\n"
" vld1.16 {d24, d25}, [%[a]]!\n"
" subs %[len], %[len], #8\n"
@@ -214,11 +226,11 @@ interpolate_gint16_cubic_neon (gpointer
"1:"
" mov r8, %[a]\n"
" vld1.16 {d16, d17}, [%[a]]!\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.16 {d18, d19}, [r8]\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.16 {d20, d21}, [r8]\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.16 {d22, d23}, [r8]\n"
" subs %[len], %[len], #8\n"
" vmull.s16 q0, d16, d24\n"
@@ -340,11 +352,11 @@ inner_product_gint32_cubic_1_neon (gint3
"1:"
" mov r8, %[b]\n"
" vld1.32 {d16, d17}, [%[b]]!\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.32 {d18, d19}, [r8]\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.32 {d20, d21}, [r8]\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.32 {d22, d23}, [r8]\n"
" vld1.32 {d24, d25}, [%[a]]!\n"
" subs %[len], %[len], #4\n"
@@ -426,11 +438,11 @@ interpolate_gint32_cubic_neon (gpointer
"1:"
" mov r8, %[a]\n"
" vld1.32 {d16, d17}, [%[a]]!\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.32 {d18, d19}, [r8]\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.32 {d20, d21}, [r8]\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.32 {d22, d23}, [r8]\n"
" subs %[len], %[len], #4\n"
" vmull.s32 q0, d16, d24\n"
@@ -545,11 +557,11 @@ inner_product_gfloat_cubic_1_neon (gfloa
"1:"
" mov r8, %[b]\n"
" vld1.32 {q8}, [%[b]]!\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.32 {q9}, [r8]\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.32 {q10}, [r8]\n"
- " add.w r8, r8, %[bstride]\n"
+ ADD_W(r8, bstride)
" vld1.32 {q11}, [r8]\n"
" vld1.32 {q12}, [%[a]]!\n"
" subs %[len], %[len], #4\n"
@@ -623,11 +635,11 @@ interpolate_gfloat_cubic_neon (gpointer
"1:"
" mov r8, %[a]\n"
" vld1.32 {q8}, [%[a]]!\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.32 {q9}, [r8]\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.32 {q10}, [r8]\n"
- " add.w r8, r8, %[astride]\n"
+ ADD_W(r8, astride)
" vld1.32 {q11}, [r8]\n"
" subs %[len], %[len], #4\n"
" vmul.f32 q0, q8, q12\n"
--- a/meson.build
+++ b/meson.build
@@ -457,10 +457,9 @@ if host_machine.cpu_family() == 'arm'
#include <arm_neon.h>
int32x4_t testfunc(int16_t *a, int16_t *b) {
asm volatile ("vmull.s16 q0, d0, d0" : : : "q0");
- asm volatile ("add.w r1, r0, r0" : : : "r0", "r1");
return vmull_s16(vld1_s16(a), vld1_s16(b));
}
-''', name : 'NEON with 32-bit wide Thumb support')
+''', name : 'NEON support')
core_conf.set('HAVE_ARM_NEON', true)
endif
endif
|
