| 1 | /*
|
|---|
| 2 | * MMX optimized DSP utils
|
|---|
| 3 | * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
|
|---|
| 4 | *
|
|---|
| 5 | * This library is free software; you can redistribute it and/or
|
|---|
| 6 | * modify it under the terms of the GNU Lesser General Public
|
|---|
| 7 | * License as published by the Free Software Foundation; either
|
|---|
| 8 | * version 2 of the License, or (at your option) any later version.
|
|---|
| 9 | *
|
|---|
| 10 | * This library 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 GNU
|
|---|
| 13 | * Lesser General Public License for more details.
|
|---|
| 14 | *
|
|---|
| 15 | * You should have received a copy of the GNU Lesser General Public
|
|---|
| 16 | * License along with this library; if not, write to the Free Software
|
|---|
| 17 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|---|
| 18 | */
|
|---|
| 19 | #define TESTCPU_MAIN
|
|---|
| 20 | #include "avcodec.h"
|
|---|
| 21 | #include "dsputil.h"
|
|---|
| 22 | #include "mpegvideo.h"
|
|---|
| 23 | #include "mpeg12data.h"
|
|---|
| 24 | #include "mpeg4data.h"
|
|---|
| 25 | #include "../libavcodec/i386/cputest.c"
|
|---|
| 26 | #include "../libavcodec/i386/dsputil_mmx.c"
|
|---|
| 27 |
|
|---|
| 28 | #include "../libavcodec/i386/fdct_mmx.c"
|
|---|
| 29 | #include "../libavcodec/i386/idct_mmx.c"
|
|---|
| 30 | #include "../libavcodec/i386/motion_est_mmx.c"
|
|---|
| 31 | #include "../libavcodec/i386/simple_idct_mmx.c"
|
|---|
| 32 | #include "../libavcodec/dsputil.c"
|
|---|
| 33 | #include "../libavcodec/simple_idct.c"
|
|---|
| 34 | #include "../libavcodec/jfdctfst.c"
|
|---|
| 35 |
|
|---|
| 36 | #undef TESTCPU_MAIN
|
|---|
| 37 |
|
|---|
| 38 | #define PAD 0x10000
|
|---|
| 39 | /*
|
|---|
| 40 | * for testing speed of various routine - should be probably extended
|
|---|
| 41 | * for a general purpose regression test later
|
|---|
| 42 | *
|
|---|
| 43 | * currently only for i386 - FIXME
|
|---|
| 44 | */
|
|---|
| 45 |
|
|---|
| 46 | #define PIX_FUNC_C(a) \
|
|---|
| 47 | { #a "_c", a ## _c, 0 }, \
|
|---|
| 48 | { #a "_mmx", a ## _mmx, MM_MMX }, \
|
|---|
| 49 | { #a "_mmx2", a ## _mmx2, MM_MMXEXT | PAD }
|
|---|
| 50 |
|
|---|
| 51 | #define PIX_FUNC(a) \
|
|---|
| 52 | { #a "_mmx", a ## _mmx, MM_MMX }, \
|
|---|
| 53 | { #a "_3dnow", a ## _3dnow, MM_3DNOW }, \
|
|---|
| 54 | { #a "_mmx2", a ## _mmx2, MM_MMXEXT | PAD }
|
|---|
| 55 |
|
|---|
| 56 | #define PIX_FUNC_MMX(a) \
|
|---|
| 57 | { #a "_mmx", a ## _mmx, MM_MMX | PAD }
|
|---|
| 58 |
|
|---|
| 59 | /*
|
|---|
| 60 | PIX_FUNC_C(pix_abs16x16),
|
|---|
| 61 | PIX_FUNC_C(pix_abs16x16_x2),
|
|---|
| 62 | PIX_FUNC_C(pix_abs16x16_y2),
|
|---|
| 63 | PIX_FUNC_C(pix_abs16x16_xy2),
|
|---|
| 64 | PIX_FUNC_C(pix_abs8x8),
|
|---|
| 65 | PIX_FUNC_C(pix_abs8x8_x2),
|
|---|
| 66 | PIX_FUNC_C(pix_abs8x8_y2),
|
|---|
| 67 | PIX_FUNC_C(pix_abs8x8_xy2),
|
|---|
| 68 | */
|
|---|
| 69 |
|
|---|
| 70 | static const struct pix_func {
|
|---|
| 71 | char* name;
|
|---|
| 72 | op_pixels_func func;
|
|---|
| 73 | int mm_flags;
|
|---|
| 74 | } pix_func[] = {
|
|---|
| 75 |
|
|---|
| 76 | PIX_FUNC_MMX(put_pixels),
|
|---|
| 77 | //PIX_FUNC_MMX(get_pixels),
|
|---|
| 78 | //PIX_FUNC_MMX(put_pixels_clamped),
|
|---|
| 79 | #if 1
|
|---|
| 80 | PIX_FUNC(put_pixels_x2),
|
|---|
| 81 | PIX_FUNC(put_pixels_y2),
|
|---|
| 82 | PIX_FUNC_MMX(put_pixels_xy2),
|
|---|
| 83 |
|
|---|
| 84 | PIX_FUNC(put_no_rnd_pixels_x2),
|
|---|
| 85 | PIX_FUNC(put_no_rnd_pixels_y2),
|
|---|
| 86 | PIX_FUNC_MMX(put_no_rnd_pixels_xy2),
|
|---|
| 87 |
|
|---|
| 88 | PIX_FUNC(avg_pixels),
|
|---|
| 89 | PIX_FUNC(avg_pixels_x2),
|
|---|
| 90 | PIX_FUNC(avg_pixels_y2),
|
|---|
| 91 | PIX_FUNC(avg_pixels_xy2),
|
|---|
| 92 |
|
|---|
| 93 | PIX_FUNC_MMX(avg_no_rnd_pixels),
|
|---|
| 94 | PIX_FUNC_MMX(avg_no_rnd_pixels_x2),
|
|---|
| 95 | PIX_FUNC_MMX(avg_no_rnd_pixels_y2),
|
|---|
| 96 | PIX_FUNC_MMX(avg_no_rnd_pixels_xy2),
|
|---|
| 97 | #endif
|
|---|
| 98 | { 0, 0 }
|
|---|
| 99 | };
|
|---|
| 100 |
|
|---|
| 101 | static inline long long rdtsc()
|
|---|
| 102 | {
|
|---|
| 103 | long long l;
|
|---|
| 104 | asm volatile( "rdtsc\n\t"
|
|---|
| 105 | : "=A" (l)
|
|---|
| 106 | );
|
|---|
| 107 | return l;
|
|---|
| 108 | }
|
|---|
| 109 |
|
|---|
| 110 | static test_speed(int step)
|
|---|
| 111 | {
|
|---|
| 112 | const struct pix_func* pix = pix_func;
|
|---|
| 113 | const int linesize = 720;
|
|---|
| 114 | char empty[32768];
|
|---|
| 115 | char* bu =(char*)(((long)empty + 32) & ~0xf);
|
|---|
| 116 |
|
|---|
| 117 | int sum = 0;
|
|---|
| 118 |
|
|---|
| 119 | while (pix->name)
|
|---|
| 120 | {
|
|---|
| 121 | int i;
|
|---|
| 122 | uint64_t te, ts;
|
|---|
| 123 | op_pixels_func func = pix->func;
|
|---|
| 124 | char* im = bu;
|
|---|
| 125 |
|
|---|
| 126 | if (pix->mm_flags & mm_flags)
|
|---|
| 127 | {
|
|---|
| 128 | printf("%30s... ", pix->name);
|
|---|
| 129 | fflush(stdout);
|
|---|
| 130 | ts = rdtsc();
|
|---|
| 131 | for(i=0; i<100000; i++){
|
|---|
| 132 | func(im, im + 1000, linesize, 16);
|
|---|
| 133 | im += step;
|
|---|
| 134 | if (im > bu + 20000)
|
|---|
| 135 | im = bu;
|
|---|
| 136 | }
|
|---|
| 137 | te = rdtsc();
|
|---|
| 138 | emms();
|
|---|
| 139 | printf("% 9d\n", (int)(te - ts));
|
|---|
| 140 | sum += (te - ts) / 100000;
|
|---|
| 141 | if (pix->mm_flags & PAD)
|
|---|
| 142 | puts("");
|
|---|
| 143 | }
|
|---|
| 144 | pix++;
|
|---|
| 145 | }
|
|---|
| 146 |
|
|---|
| 147 | printf("Total sum: %d\n", sum);
|
|---|
| 148 | }
|
|---|
| 149 |
|
|---|
| 150 | int main(int argc, char* argv[])
|
|---|
| 151 | {
|
|---|
| 152 | int step = 16;
|
|---|
| 153 |
|
|---|
| 154 | if (argc > 1)
|
|---|
| 155 | {
|
|---|
| 156 | // something simple for now
|
|---|
| 157 | if (argc > 2 && (strcmp("-s", argv[1]) == 0
|
|---|
| 158 | || strcmp("-step", argv[1]) == 0))
|
|---|
| 159 | step = atoi(argv[2]);
|
|---|
| 160 | }
|
|---|
| 161 |
|
|---|
| 162 | mm_flags = mm_support();
|
|---|
| 163 | printf("%s: detected CPU flags:", argv[0]);
|
|---|
| 164 | if (mm_flags & MM_MMX)
|
|---|
| 165 | printf(" mmx");
|
|---|
| 166 | if (mm_flags & MM_MMXEXT)
|
|---|
| 167 | printf(" mmxext");
|
|---|
| 168 | if (mm_flags & MM_3DNOW)
|
|---|
| 169 | printf(" 3dnow");
|
|---|
| 170 | if (mm_flags & MM_SSE)
|
|---|
| 171 | printf(" sse");
|
|---|
| 172 | if (mm_flags & MM_SSE2)
|
|---|
| 173 | printf(" sse2");
|
|---|
| 174 | printf("\n");
|
|---|
| 175 |
|
|---|
| 176 | printf("Using step: %d\n", step);
|
|---|
| 177 | test_speed(step);
|
|---|
| 178 | }
|
|---|
