LCOV - code coverage report
Current view: top level - src/libavcodec/x86 - vp9dsp_init_16bpp.c (source / functions) Hit Total Coverage
Test: coverage.info Lines: 51 54 94.4 %
Date: 2017-01-22 02:20:28 Functions: 1 1 100.0 %

          Line data    Source code
       1             : /*
       2             :  * VP9 SIMD optimizations
       3             :  *
       4             :  * Copyright (c) 2013 Ronald S. Bultje <rsbultje gmail com>
       5             :  *
       6             :  * This file is part of FFmpeg.
       7             :  *
       8             :  * FFmpeg is free software; you can redistribute it and/or
       9             :  * modify it under the terms of the GNU Lesser General Public
      10             :  * License as published by the Free Software Foundation; either
      11             :  * version 2.1 of the License, or (at your option) any later version.
      12             :  *
      13             :  * FFmpeg is distributed in the hope that it will be useful,
      14             :  * but WITHOUT ANY WARRANTY; without even the implied warranty of
      15             :  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
      16             :  * Lesser General Public License for more details.
      17             :  *
      18             :  * You should have received a copy of the GNU Lesser General Public
      19             :  * License along with FFmpeg; if not, write to the Free Software
      20             :  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
      21             :  */
      22             : 
      23             : #include "libavutil/attributes.h"
      24             : #include "libavutil/cpu.h"
      25             : #include "libavutil/mem.h"
      26             : #include "libavutil/x86/cpu.h"
      27             : #include "libavcodec/vp9dsp.h"
      28             : #include "libavcodec/x86/vp9dsp_init.h"
      29             : 
      30             : #if HAVE_YASM
      31             : 
      32             : decl_fpel_func(put,   8,    , mmx);
      33             : decl_fpel_func(avg,   8, _16, mmxext);
      34             : decl_fpel_func(put,  16,    , sse);
      35             : decl_fpel_func(put,  32,    , sse);
      36             : decl_fpel_func(put,  64,    , sse);
      37             : decl_fpel_func(put, 128,    , sse);
      38             : decl_fpel_func(avg,  16, _16, sse2);
      39             : decl_fpel_func(avg,  32, _16, sse2);
      40             : decl_fpel_func(avg,  64, _16, sse2);
      41             : decl_fpel_func(avg, 128, _16, sse2);
      42             : decl_fpel_func(put,  32,    , avx);
      43             : decl_fpel_func(put,  64,    , avx);
      44             : decl_fpel_func(put, 128,    , avx);
      45             : decl_fpel_func(avg,  32, _16, avx2);
      46             : decl_fpel_func(avg,  64, _16, avx2);
      47             : decl_fpel_func(avg, 128, _16, avx2);
      48             : 
      49             : decl_ipred_fns(v,       16, mmx,    sse);
      50             : decl_ipred_fns(h,       16, mmxext, sse2);
      51             : decl_ipred_fns(dc,      16, mmxext, sse2);
      52             : decl_ipred_fns(dc_top,  16, mmxext, sse2);
      53             : decl_ipred_fns(dc_left, 16, mmxext, sse2);
      54             : 
      55             : #define decl_ipred_dir_funcs(type) \
      56             : decl_ipred_fns(type, 16, sse2,  sse2); \
      57             : decl_ipred_fns(type, 16, ssse3, ssse3); \
      58             : decl_ipred_fns(type, 16, avx,   avx)
      59             : 
      60             : decl_ipred_dir_funcs(dl);
      61             : decl_ipred_dir_funcs(dr);
      62             : decl_ipred_dir_funcs(vl);
      63             : decl_ipred_dir_funcs(vr);
      64             : decl_ipred_dir_funcs(hu);
      65             : decl_ipred_dir_funcs(hd);
      66             : #endif /* HAVE_YASM */
      67             : 
      68         126 : av_cold void ff_vp9dsp_init_16bpp_x86(VP9DSPContext *dsp)
      69             : {
      70             : #if HAVE_YASM
      71         126 :     int cpu_flags = av_get_cpu_flags();
      72             : 
      73         126 :     if (EXTERNAL_MMX(cpu_flags)) {
      74         100 :         init_fpel_func(4, 0,   8, put, , mmx);
      75         100 :         init_ipred_func(v, VERT, 4, 16, mmx);
      76             :     }
      77             : 
      78         126 :     if (EXTERNAL_MMXEXT(cpu_flags)) {
      79          90 :         init_fpel_func(4, 1,   8, avg, _16, mmxext);
      80          90 :         init_ipred_func(h, HOR, 4, 16, mmxext);
      81          90 :         init_ipred_func(dc, DC, 4, 16, mmxext);
      82          90 :         init_ipred_func(dc_top,  TOP_DC,  4, 16, mmxext);
      83          90 :         init_ipred_func(dc_left, LEFT_DC, 4, 16, mmxext);
      84             :     }
      85             : 
      86         126 :     if (EXTERNAL_SSE(cpu_flags)) {
      87          80 :         init_fpel_func(3, 0,  16, put, , sse);
      88          80 :         init_fpel_func(2, 0,  32, put, , sse);
      89          80 :         init_fpel_func(1, 0,  64, put, , sse);
      90          80 :         init_fpel_func(0, 0, 128, put, , sse);
      91          80 :         init_8_16_32_ipred_funcs(v, VERT, 16, sse);
      92             :     }
      93             : 
      94         126 :     if (EXTERNAL_SSE2(cpu_flags)) {
      95          70 :         init_fpel_func(3, 1,  16, avg, _16, sse2);
      96          70 :         init_fpel_func(2, 1,  32, avg, _16, sse2);
      97          70 :         init_fpel_func(1, 1,  64, avg, _16, sse2);
      98          70 :         init_fpel_func(0, 1, 128, avg, _16, sse2);
      99          70 :         init_8_16_32_ipred_funcs(h, HOR, 16, sse2);
     100          70 :         init_8_16_32_ipred_funcs(dc, DC, 16, sse2);
     101          70 :         init_8_16_32_ipred_funcs(dc_top,  TOP_DC,  16, sse2);
     102          70 :         init_8_16_32_ipred_funcs(dc_left, LEFT_DC, 16, sse2);
     103          70 :         init_ipred_funcs(dl, DIAG_DOWN_LEFT, 16, sse2);
     104          70 :         init_ipred_funcs(dr, DIAG_DOWN_RIGHT, 16, sse2);
     105          70 :         init_ipred_funcs(vl, VERT_LEFT, 16, sse2);
     106          70 :         init_ipred_funcs(vr, VERT_RIGHT, 16, sse2);
     107          70 :         init_ipred_funcs(hu, HOR_UP, 16, sse2);
     108          70 :         init_ipred_funcs(hd, HOR_DOWN, 16, sse2);
     109             :     }
     110             : 
     111         126 :     if (EXTERNAL_SSSE3(cpu_flags)) {
     112          50 :         init_ipred_funcs(dl, DIAG_DOWN_LEFT, 16, ssse3);
     113          50 :         init_ipred_funcs(dr, DIAG_DOWN_RIGHT, 16, ssse3);
     114          50 :         init_ipred_funcs(vl, VERT_LEFT, 16, ssse3);
     115          50 :         init_ipred_funcs(vr, VERT_RIGHT, 16, ssse3);
     116          50 :         init_ipred_funcs(hu, HOR_UP, 16, ssse3);
     117          50 :         init_ipred_funcs(hd, HOR_DOWN, 16, ssse3);
     118             :     }
     119             : 
     120         126 :     if (EXTERNAL_AVX_FAST(cpu_flags)) {
     121          10 :         init_fpel_func(2, 0,  32, put, , avx);
     122          10 :         init_fpel_func(1, 0,  64, put, , avx);
     123          10 :         init_fpel_func(0, 0, 128, put, , avx);
     124          10 :         init_ipred_funcs(dl, DIAG_DOWN_LEFT, 16, avx);
     125          10 :         init_ipred_funcs(dr, DIAG_DOWN_RIGHT, 16, avx);
     126          10 :         init_ipred_funcs(vl, VERT_LEFT, 16, avx);
     127          10 :         init_ipred_funcs(vr, VERT_RIGHT, 16, avx);
     128          10 :         init_ipred_funcs(hu, HOR_UP, 16, avx);
     129          10 :         init_ipred_funcs(hd, HOR_DOWN, 16, avx);
     130             :     }
     131             : 
     132         126 :     if (EXTERNAL_AVX2_FAST(cpu_flags)) {
     133           0 :         init_fpel_func(2, 1,  32, avg, _16, avx2);
     134           0 :         init_fpel_func(1, 1,  64, avg, _16, avx2);
     135           0 :         init_fpel_func(0, 1, 128, avg, _16, avx2);
     136             :     }
     137             : 
     138             : #endif /* HAVE_YASM */
     139         126 : }

Generated by: LCOV version 1.12