3 * This file is part of FFmpeg.
5 * FFmpeg is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * FFmpeg is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License along
16 * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
22 #include "libavutil/common.h"
23 #include "libavutil/intreadwrite.h"
24 #include "libavutil/mem.h"
26 #include "libswscale/rgb2rgb.h"
30 #define randomize_buffers(buf, size) \
33 for (j = 0; j < size; j+=4) \
34 AV_WN32(buf + j, rnd()); \
37 static const uint8_t width[] = {12, 16, 20, 32, 36, 128};
38 static const struct {uint8_t w, h, s;} planes[] = {
39 {12,16,12}, {16,16,16}, {20,23,25}, {32,18,48}, {8,128,16}, {128,128,128}
42 #define MAX_STRIDE 128
43 #define MAX_HEIGHT 128
45 static void check_shuffle_bytes(void * func, const char * report)
48 LOCAL_ALIGNED_32(uint8_t, src0, [MAX_STRIDE]);
49 LOCAL_ALIGNED_32(uint8_t, src1, [MAX_STRIDE]);
50 LOCAL_ALIGNED_32(uint8_t, dst0, [MAX_STRIDE]);
51 LOCAL_ALIGNED_32(uint8_t, dst1, [MAX_STRIDE]);
53 declare_func_emms(AV_CPU_FLAG_MMX, void, const uint8_t *src, uint8_t *dst, int src_size);
55 memset(dst0, 0, MAX_STRIDE);
56 memset(dst1, 0, MAX_STRIDE);
57 randomize_buffers(src0, MAX_STRIDE);
58 memcpy(src1, src0, MAX_STRIDE);
60 if (check_func(func, "%s", report)) {
61 for (i = 0; i < 6; i ++) {
62 call_ref(src0, dst0, width[i]);
63 call_new(src1, dst1, width[i]);
64 if (memcmp(dst0, dst1, MAX_STRIDE))
67 bench_new(src0, dst0, width[5]);
71 static void check_uyvy_to_422p(void)
75 LOCAL_ALIGNED_32(uint8_t, src0, [MAX_STRIDE * MAX_HEIGHT * 2]);
76 LOCAL_ALIGNED_32(uint8_t, src1, [MAX_STRIDE * MAX_HEIGHT * 2]);
77 LOCAL_ALIGNED_32(uint8_t, dst_y_0, [MAX_STRIDE * MAX_HEIGHT]);
78 LOCAL_ALIGNED_32(uint8_t, dst_y_1, [MAX_STRIDE * MAX_HEIGHT]);
79 LOCAL_ALIGNED_32(uint8_t, dst_u_0, [(MAX_STRIDE/2) * MAX_HEIGHT]);
80 LOCAL_ALIGNED_32(uint8_t, dst_u_1, [(MAX_STRIDE/2) * MAX_HEIGHT]);
81 LOCAL_ALIGNED_32(uint8_t, dst_v_0, [(MAX_STRIDE/2) * MAX_HEIGHT]);
82 LOCAL_ALIGNED_32(uint8_t, dst_v_1, [(MAX_STRIDE/2) * MAX_HEIGHT]);
84 declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
85 const uint8_t *src, int width, int height,
86 int lumStride, int chromStride, int srcStride);
88 randomize_buffers(src0, MAX_STRIDE * MAX_HEIGHT * 2);
89 memcpy(src1, src0, MAX_STRIDE * MAX_HEIGHT * 2);
91 if (check_func(uyvytoyuv422, "uyvytoyuv422")) {
92 for (i = 0; i < 6; i ++) {
93 memset(dst_y_0, 0, MAX_STRIDE * MAX_HEIGHT);
94 memset(dst_y_1, 0, MAX_STRIDE * MAX_HEIGHT);
95 memset(dst_u_0, 0, (MAX_STRIDE/2) * MAX_HEIGHT);
96 memset(dst_u_1, 0, (MAX_STRIDE/2) * MAX_HEIGHT);
97 memset(dst_v_0, 0, (MAX_STRIDE/2) * MAX_HEIGHT);
98 memset(dst_v_1, 0, (MAX_STRIDE/2) * MAX_HEIGHT);
100 call_ref(dst_y_0, dst_u_0, dst_v_0, src0, planes[i].w, planes[i].h,
101 MAX_STRIDE, MAX_STRIDE / 2, planes[i].s);
102 call_new(dst_y_1, dst_u_1, dst_v_1, src1, planes[i].w, planes[i].h,
103 MAX_STRIDE, MAX_STRIDE / 2, planes[i].s);
104 if (memcmp(dst_y_0, dst_y_1, MAX_STRIDE * MAX_HEIGHT) ||
105 memcmp(dst_u_0, dst_u_1, (MAX_STRIDE/2) * MAX_HEIGHT) ||
106 memcmp(dst_v_0, dst_v_1, (MAX_STRIDE/2) * MAX_HEIGHT))
109 bench_new(dst_y_1, dst_u_1, dst_v_1, src1, planes[5].w, planes[5].h,
110 MAX_STRIDE, MAX_STRIDE / 2, planes[5].s);
114 static void check_interleave_bytes(void)
116 LOCAL_ALIGNED_16(uint8_t, src0_buf, [MAX_STRIDE*MAX_HEIGHT+1]);
117 LOCAL_ALIGNED_16(uint8_t, src1_buf, [MAX_STRIDE*MAX_HEIGHT+1]);
118 LOCAL_ALIGNED_16(uint8_t, dst0_buf, [2*MAX_STRIDE*MAX_HEIGHT+2]);
119 LOCAL_ALIGNED_16(uint8_t, dst1_buf, [2*MAX_STRIDE*MAX_HEIGHT+2]);
120 // Intentionally using unaligned buffers, as this function doesn't have
121 // any alignment requirements.
122 uint8_t *src0 = src0_buf + 1;
123 uint8_t *src1 = src1_buf + 1;
124 uint8_t *dst0 = dst0_buf + 2;
125 uint8_t *dst1 = dst1_buf + 2;
127 declare_func_emms(AV_CPU_FLAG_MMX, void, const uint8_t *, const uint8_t *,
128 uint8_t *, int, int, int, int, int);
130 randomize_buffers(src0, MAX_STRIDE * MAX_HEIGHT);
131 randomize_buffers(src1, MAX_STRIDE * MAX_HEIGHT);
133 if (check_func(interleaveBytes, "interleave_bytes")) {
134 for (int i = 0; i <= 16; i++) {
135 // Try all widths [1,16], and try one random width.
137 int w = i > 0 ? i : (1 + (rnd() % (MAX_STRIDE-2)));
138 int h = 1 + (rnd() % (MAX_HEIGHT-2));
140 int src0_offset = 0, src0_stride = MAX_STRIDE;
141 int src1_offset = 0, src1_stride = MAX_STRIDE;
142 int dst_offset = 0, dst_stride = 2 * MAX_STRIDE;
144 memset(dst0, 0, 2 * MAX_STRIDE * MAX_HEIGHT);
145 memset(dst1, 0, 2 * MAX_STRIDE * MAX_HEIGHT);
147 // Try different combinations of negative strides
149 src0_offset = (h-1)*src0_stride;
150 src0_stride = -src0_stride;
153 src1_offset = (h-1)*src1_stride;
154 src1_stride = -src1_stride;
157 dst_offset = (h-1)*dst_stride;
158 dst_stride = -dst_stride;
161 call_ref(src0 + src0_offset, src1 + src1_offset, dst0 + dst_offset,
162 w, h, src0_stride, src1_stride, dst_stride);
163 call_new(src0 + src0_offset, src1 + src1_offset, dst1 + dst_offset,
164 w, h, src0_stride, src1_stride, dst_stride);
165 // Check a one pixel-pair edge around the destination area,
166 // to catch overwrites past the end.
167 checkasm_check(uint8_t, dst0, 2*MAX_STRIDE, dst1, 2*MAX_STRIDE,
168 2 * w + 2, h + 1, "dst");
171 bench_new(src0, src1, dst1, 127, MAX_HEIGHT,
172 MAX_STRIDE, MAX_STRIDE, 2*MAX_STRIDE);
174 if (check_func(interleaveBytes, "interleave_bytes_aligned")) {
175 // Bench the function in a more typical case, with aligned
176 // buffers and widths.
177 bench_new(src0_buf, src1_buf, dst1_buf, 128, MAX_HEIGHT,
178 MAX_STRIDE, MAX_STRIDE, 2*MAX_STRIDE);
182 void checkasm_check_sw_rgb(void)
184 ff_sws_rgb2rgb_init();
186 check_shuffle_bytes(shuffle_bytes_2103, "shuffle_bytes_2103");
187 report("shuffle_bytes_2103");
189 check_shuffle_bytes(shuffle_bytes_0321, "shuffle_bytes_0321");
190 report("shuffle_bytes_0321");
192 check_shuffle_bytes(shuffle_bytes_1230, "shuffle_bytes_1230");
193 report("shuffle_bytes_1230");
195 check_shuffle_bytes(shuffle_bytes_3012, "shuffle_bytes_3012");
196 report("shuffle_bytes_3012");
198 check_shuffle_bytes(shuffle_bytes_3210, "shuffle_bytes_3210");
199 report("shuffle_bytes_3210");
201 check_uyvy_to_422p();
202 report("uyvytoyuv422");
204 check_interleave_bytes();
205 report("interleave_bytes");