libavcodec/x86/h263_loopfilter.asm

   1 ;******************************************************************************
   2 ;* MMX-optimized H.263 loop filter
   3 ;*
   4 ;* This file is part of Libav.
   5 ;*
   6 ;* Libav 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 ;* Libav 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 Libav; if not, write to the Free Software
  18 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19 ;******************************************************************************
  20
  21 %include "libavutil/x86/x86util.asm"
  22
  23 SECTION_RODATA
  24 cextern pb_FC
  25 cextern h263_loop_filter_strength
  26
  27 SECTION .text
  28
  29 %macro H263_LOOP_FILTER 5
  30     pxor         m7, m7
  31     mova         m0, [%1]
  32     mova         m1, [%1]
  33     mova         m2, [%4]
  34     mova         m3, [%4]
  35     punpcklbw    m0, m7
  36     punpckhbw    m1, m7
  37     punpcklbw    m2, m7
  38     punpckhbw    m3, m7
  39     psubw        m0, m2
  40     psubw        m1, m3
  41     mova         m2, [%2]
  42     mova         m3, [%2]
  43     mova         m4, [%3]
  44     mova         m5, [%3]
  45     punpcklbw    m2, m7
  46     punpckhbw    m3, m7
  47     punpcklbw    m4, m7
  48     punpckhbw    m5, m7
  49     psubw        m4, m2
  50     psubw        m5, m3
  51     psllw        m4, 2
  52     psllw        m5, 2
  53     paddw        m4, m0
  54     paddw        m5, m1
  55     pxor         m6, m6
  56     pcmpgtw      m6, m4
  57     pcmpgtw      m7, m5
  58     pxor         m4, m6
  59     pxor         m5, m7
  60     psubw        m4, m6
  61     psubw        m5, m7
  62     psrlw        m4, 3
  63     psrlw        m5, 3
  64     packuswb     m4, m5
  65     packsswb     m6, m7
  66     pxor         m7, m7
  67     movd         m2, %5
  68     punpcklbw    m2, m2
  69     punpcklbw    m2, m2
  70     punpcklbw    m2, m2
  71     psubusb      m2, m4
  72     mova         m3, m2
  73     psubusb      m3, m4
  74     psubb        m2, m3
  75     mova         m3, [%2]
  76     mova         m4, [%3]
  77     pxor         m3, m6
  78     pxor         m4, m6
  79     paddusb      m3, m2
  80     psubusb      m4, m2
  81     pxor         m3, m6
  82     pxor         m4, m6
  83     paddusb      m2, m2
  84     packsswb     m0, m1
  85     pcmpgtb      m7, m0
  86     pxor         m0, m7
  87     psubb        m0, m7
  88     mova         m1, m0
  89     psubusb      m0, m2
  90     psubb        m1, m0
  91     pand         m1, [pb_FC]
  92     psrlw        m1, 2
  93     pxor         m1, m7
  94     psubb        m1, m7
  95     mova         m5, [%1]
  96     mova         m6, [%4]
  97     psubb        m5, m1
  98     paddb        m6, m1
  99 %endmacro
 100
 101 INIT_MMX mmx
 102 ; void ff_h263_v_loop_filter_mmx(uint8_t *src, int stride, int qscale)
 103 cglobal h263_v_loop_filter, 3,5
 104     movsxdifnidn r1, r1d
 105     movsxdifnidn r2, r2d
 106
 107     lea          r4, [h263_loop_filter_strength]
 108     movzx       r3d, BYTE [r4+r2]
 109     movsx        r2, r3b
 110     shl          r2, 1
 111
 112     mov          r3, r0
 113     sub          r3, r1
 114     mov          r4, r3
 115     sub          r4, r1
 116     H263_LOOP_FILTER r4, r3, r0, r0+r1, r2d
 117
 118     mova       [r3], m3
 119     mova       [r0], m4
 120     mova       [r4], m5
 121     mova    [r0+r1], m6
 122     RET
 123
 124 %macro TRANSPOSE4X4 2
 125     movd      m0, [%1]
 126     movd      m1, [%1+r1]
 127     movd      m2, [%1+r1*2]
 128     movd      m3, [%1+r3]
 129     punpcklbw m0, m1
 130     punpcklbw m2, m3
 131     mova      m1, m0
 132     punpcklwd m0, m2
 133     punpckhwd m1, m2
 134     movd [%2+ 0], m0
 135     punpckhdq m0, m0
 136     movd [%2+ 8], m0
 137     movd [%2+16], m1
 138     punpckhdq m1, m1
 139     movd [%2+24], m1
 140 %endmacro
 141
 142
 143 ; void ff_h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale)
 144 INIT_MMX mmx
 145 cglobal h263_h_loop_filter, 3,5,0,32
 146     movsxdifnidn r1, r1d
 147     movsxdifnidn r2, r2d
 148
 149     lea          r4, [h263_loop_filter_strength]
 150     movzx       r3d, BYTE [r4+r2]
 151     movsx        r2, r3b
 152     shl          r2, 1
 153
 154     sub          r0, 2
 155     lea          r3, [r1*3]
 156
 157     TRANSPOSE4X4 r0, rsp
 158     lea          r4, [r0+r1*4]
 159     TRANSPOSE4X4 r4, rsp+4
 160
 161     H263_LOOP_FILTER rsp, rsp+8, rsp+16, rsp+24, r2d
 162
 163     mova         m1, m5
 164     mova         m0, m4
 165     punpcklbw    m5, m3
 166     punpcklbw    m4, m6
 167     punpckhbw    m1, m3
 168     punpckhbw    m0, m6
 169     mova         m3, m5
 170     mova         m6, m1
 171     punpcklwd    m5, m4
 172     punpcklwd    m1, m0
 173     punpckhwd    m3, m4
 174     punpckhwd    m6, m0
 175     movd       [r0], m5
 176     punpckhdq    m5, m5
 177     movd  [r0+r1*1], m5
 178     movd  [r0+r1*2], m3
 179     punpckhdq    m3, m3
 180     movd    [r0+r3], m3
 181     movd       [r4], m1
 182     punpckhdq    m1, m1
 183     movd  [r4+r1*1], m1
 184     movd  [r4+r1*2], m6
 185     punpckhdq    m6, m6
 186     movd    [r4+r3], m6
 187     RET