LCOV - code coverage report
Current view: top level - src/libavcodec/x86 - lossless_videoencdsp_init.c (source / functions) Hit Total Coverage
Test: coverage.info Lines: 6 18 33.3 %
Date: 2017-06-23 19:00:59 Functions: 1 2 50.0 %

          Line data    Source code
       1             : /*
       2             :  * SIMD-optimized lossless video encoding functions
       3             :  * Copyright (c) 2000, 2001 Fabrice Bellard
       4             :  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
       5             :  *
       6             :  * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
       7             :  *
       8             :  * This file is part of FFmpeg.
       9             :  *
      10             :  * FFmpeg is free software; you can redistribute it and/or
      11             :  * modify it under the terms of the GNU Lesser General Public
      12             :  * License as published by the Free Software Foundation; either
      13             :  * version 2.1 of the License, or (at your option) any later version.
      14             :  *
      15             :  * FFmpeg is distributed in the hope that it will be useful,
      16             :  * but WITHOUT ANY WARRANTY; without even the implied warranty of
      17             :  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
      18             :  * Lesser General Public License for more details.
      19             :  *
      20             :  * You should have received a copy of the GNU Lesser General Public
      21             :  * License along with FFmpeg; if not, write to the Free Software
      22             :  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
      23             :  */
      24             : 
      25             : #include "libavutil/attributes.h"
      26             : #include "libavutil/cpu.h"
      27             : #include "libavutil/x86/asm.h"
      28             : #include "libavutil/x86/cpu.h"
      29             : #include "libavcodec/lossless_videoencdsp.h"
      30             : #include "libavcodec/mathops.h"
      31             : 
      32             : void ff_diff_bytes_mmx(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
      33             :                        intptr_t w);
      34             : void ff_diff_bytes_sse2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
      35             :                         intptr_t w);
      36             : void ff_diff_bytes_avx2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
      37             :                         intptr_t w);
      38             : 
      39             : #if HAVE_INLINE_ASM
      40             : 
      41           0 : static void sub_median_pred_mmxext(uint8_t *dst, const uint8_t *src1,
      42             :                                    const uint8_t *src2, intptr_t w,
      43             :                                    int *left, int *left_top)
      44             : {
      45           0 :     x86_reg i = 0;
      46             :     uint8_t l, lt;
      47             : 
      48           0 :     __asm__ volatile (
      49             :         "movq  (%1, %0), %%mm0          \n\t" // LT
      50             :         "psllq $8, %%mm0                \n\t"
      51             :         "1:                             \n\t"
      52             :         "movq  (%1, %0), %%mm1          \n\t" // T
      53             :         "movq  -1(%2, %0), %%mm2        \n\t" // L
      54             :         "movq  (%2, %0), %%mm3          \n\t" // X
      55             :         "movq %%mm2, %%mm4              \n\t" // L
      56             :         "psubb %%mm0, %%mm2             \n\t"
      57             :         "paddb %%mm1, %%mm2             \n\t" // L + T - LT
      58             :         "movq %%mm4, %%mm5              \n\t" // L
      59             :         "pmaxub %%mm1, %%mm4            \n\t" // max(T, L)
      60             :         "pminub %%mm5, %%mm1            \n\t" // min(T, L)
      61             :         "pminub %%mm2, %%mm4            \n\t"
      62             :         "pmaxub %%mm1, %%mm4            \n\t"
      63             :         "psubb %%mm4, %%mm3             \n\t" // dst - pred
      64             :         "movq %%mm3, (%3, %0)           \n\t"
      65             :         "add $8, %0                     \n\t"
      66             :         "movq -1(%1, %0), %%mm0         \n\t" // LT
      67             :         "cmp %4, %0                     \n\t"
      68             :         " jb 1b                         \n\t"
      69             :         : "+r" (i)
      70             :         : "r" (src1), "r" (src2), "r" (dst), "r" ((x86_reg) w));
      71             : 
      72           0 :     l  = *left;
      73           0 :     lt = *left_top;
      74             : 
      75           0 :     dst[0] = src2[0] - mid_pred(l, src1[0], (l + src1[0] - lt) & 0xFF);
      76             : 
      77           0 :     *left_top = src1[w - 1];
      78           0 :     *left     = src2[w - 1];
      79           0 : }
      80             : 
      81             : #endif /* HAVE_INLINE_ASM */
      82             : 
      83         149 : av_cold void ff_llvidencdsp_init_x86(LLVidEncDSPContext *c)
      84             : {
      85         149 :     av_unused int cpu_flags = av_get_cpu_flags();
      86             : 
      87             :     if (ARCH_X86_32 && EXTERNAL_MMX(cpu_flags)) {
      88             :         c->diff_bytes = ff_diff_bytes_mmx;
      89             :     }
      90             : 
      91             : #if HAVE_INLINE_ASM
      92         149 :     if (INLINE_MMXEXT(cpu_flags)) {
      93           0 :         c->sub_median_pred = sub_median_pred_mmxext;
      94             :     }
      95             : #endif /* HAVE_INLINE_ASM */
      96             : 
      97         149 :     if (EXTERNAL_SSE2(cpu_flags)) {
      98           0 :         c->diff_bytes = ff_diff_bytes_sse2;
      99             :     }
     100             : 
     101         149 :     if (EXTERNAL_AVX2_FAST(cpu_flags)) {
     102           0 :         c->diff_bytes = ff_diff_bytes_avx2;
     103             :     }
     104         149 : }

Generated by: LCOV version 1.13