annotate x86/vp56dsp_init.c @ 12492:58a960d6e34c libavcodec

Rename h264_idct_sse2.asm to h264_idct.asm; move inline IDCT asm from h264dsp_mmx.c to h264_idct.asm (as yasm code). Because the loops are now coded in asm instead of C, this is (depending on the function) up to 50% faster for cases where gcc didn't do a great job at looping. Since h264_idct_add8() is now faster than the manual loop setup in h264.c, in-asm idct calling can now be enabled for chroma as well (see r16207). For MMX, this is 5% faster. For SSE2 (which isn't done for chroma if h264.c does the looping), this makes it up to 50% faster. Speed gain overall is ~0.5-1.0%.
author rbultje
date Tue, 14 Sep 2010 13:36:26 +0000
parents 9fef0a8ddd63
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
12417
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
1 /*
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
2 * VP6 MMX/SSE2 optimizations
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
3 * Copyright (C) 2009 Sebastien Lucas <sebastien.lucas@gmail.com>
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
4 * Copyright (C) 2009 Zuxy Meng <zuxy.meng@gmail.com>
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
5 *
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
6 * This file is part of FFmpeg.
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
7 *
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
8 * FFmpeg is free software; you can redistribute it and/or
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
9 * modify it under the terms of the GNU Lesser General Public
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
10 * License as published by the Free Software Foundation; either
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
11 * version 2.1 of the License, or (at your option) any later version.
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
12 *
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
13 * FFmpeg is distributed in the hope that it will be useful,
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
16 * Lesser General Public License for more details.
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
17 *
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
18 * You should have received a copy of the GNU Lesser General Public
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
19 * License along with FFmpeg; if not, write to the Free Software
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
21 */
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
22
12475
9fef0a8ddd63 Move mm_support() from libavcodec to libavutil, make it a public
stefano
parents: 12456
diff changeset
23 #include "libavutil/cpu.h"
12417
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
24 #include "libavutil/x86_cpu.h"
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
25 #include "libavcodec/dsputil.h"
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
26 #include "libavcodec/vp56dsp.h"
12418
e17840120b80 Move vp6_filter_diag4() x86 SIMD code from inline ASM to YASM. This should
rbultje
parents: 12417
diff changeset
27
e17840120b80 Move vp6_filter_diag4() x86 SIMD code from inline ASM to YASM. This should
rbultje
parents: 12417
diff changeset
28 void ff_vp6_filter_diag4_mmx(uint8_t *dst, uint8_t *src, int stride,
e17840120b80 Move vp6_filter_diag4() x86 SIMD code from inline ASM to YASM. This should
rbultje
parents: 12417
diff changeset
29 const int16_t *h_weights,const int16_t *v_weights);
e17840120b80 Move vp6_filter_diag4() x86 SIMD code from inline ASM to YASM. This should
rbultje
parents: 12417
diff changeset
30 void ff_vp6_filter_diag4_sse2(uint8_t *dst, uint8_t *src, int stride,
e17840120b80 Move vp6_filter_diag4() x86 SIMD code from inline ASM to YASM. This should
rbultje
parents: 12417
diff changeset
31 const int16_t *h_weights,const int16_t *v_weights);
12417
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
32
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
33 av_cold void ff_vp56dsp_init_x86(VP56DSPContext* c, enum CodecID codec)
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
34 {
12418
e17840120b80 Move vp6_filter_diag4() x86 SIMD code from inline ASM to YASM. This should
rbultje
parents: 12417
diff changeset
35 #if HAVE_YASM
12475
9fef0a8ddd63 Move mm_support() from libavcodec to libavutil, make it a public
stefano
parents: 12456
diff changeset
36 int mm_flags = av_get_cpu_flags();
12417
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
37
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
38 if (CONFIG_VP6_DECODER && codec == CODEC_ID_VP6) {
12456
a5ddb39627fd Rename FF_MM_ symbols related to CPU features flags as AV_CPU_FLAG_
stefano
parents: 12418
diff changeset
39 if (mm_flags & AV_CPU_FLAG_MMX) {
12417
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
40 c->vp6_filter_diag4 = ff_vp6_filter_diag4_mmx;
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
41 }
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
42
12456
a5ddb39627fd Rename FF_MM_ symbols related to CPU features flags as AV_CPU_FLAG_
stefano
parents: 12418
diff changeset
43 if (mm_flags & AV_CPU_FLAG_SSE2) {
12417
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
44 c->vp6_filter_diag4 = ff_vp6_filter_diag4_sse2;
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
45 }
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
46 }
12418
e17840120b80 Move vp6_filter_diag4() x86 SIMD code from inline ASM to YASM. This should
rbultje
parents: 12417
diff changeset
47 #endif
12417
9f06475db098 Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
rbultje
parents:
diff changeset
48 }