| 1 | /* |
| 2 | * Copyright (c) 2009 David Conrad <lessen42@gmail.com> |
| 3 | * |
| 4 | * This file is part of FFmpeg. |
| 5 | * |
| 6 | * FFmpeg is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU Lesser General Public |
| 8 | * License as published by the Free Software Foundation; either |
| 9 | * version 2.1 of the License, or (at your option) any later version. |
| 10 | * |
| 11 | * FFmpeg is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 14 | * Lesser General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU Lesser General Public |
| 17 | * License along with FFmpeg; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 19 | */ |
| 20 | |
| 21 | #include <stdint.h> |
| 22 | |
| 23 | #include "libavutil/attributes.h" |
| 24 | #include "libavutil/cpu.h" |
| 25 | #include "libavutil/x86/cpu.h" |
| 26 | #include "libavutil/x86/asm.h" |
| 27 | #include "libavcodec/avcodec.h" |
| 28 | #include "libavcodec/vp3dsp.h" |
| 29 | #include "config.h" |
| 30 | |
| 31 | void ff_vp3_idct_put_mmx(uint8_t *dest, int line_size, int16_t *block); |
| 32 | void ff_vp3_idct_add_mmx(uint8_t *dest, int line_size, int16_t *block); |
| 33 | |
| 34 | void ff_vp3_idct_put_sse2(uint8_t *dest, int line_size, int16_t *block); |
| 35 | void ff_vp3_idct_add_sse2(uint8_t *dest, int line_size, int16_t *block); |
| 36 | |
| 37 | void ff_vp3_idct_dc_add_mmxext(uint8_t *dest, int line_size, |
| 38 | int16_t *block); |
| 39 | |
| 40 | void ff_vp3_v_loop_filter_mmxext(uint8_t *src, int stride, |
| 41 | int *bounding_values); |
| 42 | void ff_vp3_h_loop_filter_mmxext(uint8_t *src, int stride, |
| 43 | int *bounding_values); |
| 44 | |
| 45 | #if HAVE_MMX_INLINE |
| 46 | |
| 47 | #define MOVQ_BFE(regd) \ |
| 48 | __asm__ volatile ( \ |
| 49 | "pcmpeqd %%"#regd", %%"#regd" \n\t" \ |
| 50 | "paddb %%"#regd", %%"#regd" \n\t" ::) |
| 51 | |
| 52 | #define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \ |
| 53 | "movq "#rega", "#regr" \n\t" \ |
| 54 | "movq "#regc", "#regp" \n\t" \ |
| 55 | "pand "#regb", "#regr" \n\t" \ |
| 56 | "pand "#regd", "#regp" \n\t" \ |
| 57 | "pxor "#rega", "#regb" \n\t" \ |
| 58 | "pxor "#regc", "#regd" \n\t" \ |
| 59 | "pand %%mm6, "#regb" \n\t" \ |
| 60 | "pand %%mm6, "#regd" \n\t" \ |
| 61 | "psrlq $1, "#regb" \n\t" \ |
| 62 | "psrlq $1, "#regd" \n\t" \ |
| 63 | "paddb "#regb", "#regr" \n\t" \ |
| 64 | "paddb "#regd", "#regp" \n\t" |
| 65 | |
| 66 | #if HAVE_6REGS |
| 67 | static void put_vp_no_rnd_pixels8_l2_mmx(uint8_t *dst, const uint8_t *a, const uint8_t *b, ptrdiff_t stride, int h) |
| 68 | { |
| 69 | // START_TIMER |
| 70 | MOVQ_BFE(mm6); |
| 71 | __asm__ volatile( |
| 72 | "1: \n\t" |
| 73 | "movq (%1), %%mm0 \n\t" |
| 74 | "movq (%2), %%mm1 \n\t" |
| 75 | "movq (%1,%4), %%mm2 \n\t" |
| 76 | "movq (%2,%4), %%mm3 \n\t" |
| 77 | PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) |
| 78 | "movq %%mm4, (%3) \n\t" |
| 79 | "movq %%mm5, (%3,%4) \n\t" |
| 80 | |
| 81 | "movq (%1,%4,2), %%mm0 \n\t" |
| 82 | "movq (%2,%4,2), %%mm1 \n\t" |
| 83 | "movq (%1,%5), %%mm2 \n\t" |
| 84 | "movq (%2,%5), %%mm3 \n\t" |
| 85 | "lea (%1,%4,4), %1 \n\t" |
| 86 | "lea (%2,%4,4), %2 \n\t" |
| 87 | PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) |
| 88 | "movq %%mm4, (%3,%4,2) \n\t" |
| 89 | "movq %%mm5, (%3,%5) \n\t" |
| 90 | "lea (%3,%4,4), %3 \n\t" |
| 91 | "subl $4, %0 \n\t" |
| 92 | "jnz 1b \n\t" |
| 93 | :"+r"(h), "+r"(a), "+r"(b), "+r"(dst) |
| 94 | :"r"((x86_reg)stride), "r"((x86_reg)3L*stride) |
| 95 | :"memory"); |
| 96 | // STOP_TIMER("put_vp_no_rnd_pixels8_l2_mmx") |
| 97 | } |
| 98 | #endif /*HAVE_6REGS */ |
| 99 | #endif /* HAVE_MMX_INLINE */ |
| 100 | |
| 101 | av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags) |
| 102 | { |
| 103 | int cpu_flags = av_get_cpu_flags(); |
| 104 | |
| 105 | #if HAVE_6REGS && HAVE_MMX_INLINE |
| 106 | c->put_no_rnd_pixels_l2 = put_vp_no_rnd_pixels8_l2_mmx; |
| 107 | #endif /* HAVE_6REGS && HAVE_MMX_INLINE */ |
| 108 | |
| 109 | #if ARCH_X86_32 |
| 110 | if (EXTERNAL_MMX(cpu_flags)) { |
| 111 | c->idct_put = ff_vp3_idct_put_mmx; |
| 112 | c->idct_add = ff_vp3_idct_add_mmx; |
| 113 | } |
| 114 | #endif |
| 115 | |
| 116 | if (EXTERNAL_MMXEXT(cpu_flags)) { |
| 117 | c->idct_dc_add = ff_vp3_idct_dc_add_mmxext; |
| 118 | |
| 119 | if (!(flags & CODEC_FLAG_BITEXACT)) { |
| 120 | c->v_loop_filter = ff_vp3_v_loop_filter_mmxext; |
| 121 | c->h_loop_filter = ff_vp3_h_loop_filter_mmxext; |
| 122 | } |
| 123 | } |
| 124 | |
| 125 | if (EXTERNAL_SSE2(cpu_flags)) { |
| 126 | c->idct_put = ff_vp3_idct_put_sse2; |
| 127 | c->idct_add = ff_vp3_idct_add_sse2; |
| 128 | } |
| 129 | } |