libavcodec/x86/pngdsp.asm

   1 ;******************************************************************************
   2 ;* x86 optimizations for PNG decoding
   3 ;*
   4 ;* Copyright (c) 2008 Loren Merritt <lorenm@u.washington.edu>
   5 ;* Copyright (c) 2012 Ronald S. Bultje <rsbultje@gmail.com>
   6 ;*
   7 ;* This file is part of Libav.
   8 ;*
   9 ;* Libav is free software; you can redistribute it and/or
  10 ;* modify it under the terms of the GNU Lesser General Public
  11 ;* License as published by the Free Software Foundation; either
  12 ;* version 2.1 of the License, or (at your option) any later version.
  13 ;*
  14 ;* Libav is distributed in the hope that it will be useful,
  15 ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
  16 ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17 ;* Lesser General Public License for more details.
  18 ;*
  19 ;* You should have received a copy of the GNU Lesser General Public
  20 ;* License along with Libav; if not, write to the Free Software
  21 ;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  22 ;******************************************************************************
  23
  24 %include "libavutil/x86/x86util.asm"
  25
  26 SECTION_RODATA
  27
  28 cextern pw_255
  29
  30 SECTION_TEXT
  31
  32 ; %1 = nr. of xmm registers used
  33 %macro ADD_BYTES_FN 1
  34 cglobal add_bytes_l2, 4, 6, %1, dst, src1, src2, wa, w, i
  35 %if ARCH_X86_64
  36     movsxd             waq, wad
  37 %endif
  38     xor                 iq, iq
  39
  40     ; vector loop
  41     mov                 wq, waq
  42     and                waq, ~(mmsize*2-1)
  43     jmp .end_v
  44 .loop_v:
  45     mova                m0, [src1q+iq]
  46     mova                m1, [src1q+iq+mmsize]
  47     paddb               m0, [src2q+iq]
  48     paddb               m1, [src2q+iq+mmsize]
  49     mova  [dstq+iq       ], m0
  50     mova  [dstq+iq+mmsize], m1
  51     add                 iq, mmsize*2
  52 .end_v:
  53     cmp                 iq, waq
  54     jl .loop_v
  55
  56 %if mmsize == 16
  57     ; vector loop
  58     mov                waq, wq
  59     and                waq, ~7
  60     jmp .end_l
  61 .loop_l:
  62     movq               mm0, [src1q+iq]
  63     paddb              mm0, [src2q+iq]
  64     movq  [dstq+iq       ], mm0
  65     add                 iq, 8
  66 .end_l:
  67     cmp                 iq, waq
  68     jl .loop_l
  69 %endif
  70
  71     ; scalar loop for leftover
  72     jmp .end_s
  73 .loop_s:
  74     mov                wab, [src1q+iq]
  75     add                wab, [src2q+iq]
  76     mov          [dstq+iq], wab
  77     inc                 iq
  78 .end_s:
  79     cmp                 iq, wq
  80     jl .loop_s
  81     REP_RET
  82 %endmacro
  83
  84 %if ARCH_X86_32
  85 INIT_MMX mmx
  86 ADD_BYTES_FN 0
  87 %endif
  88
  89 INIT_XMM sse2
  90 ADD_BYTES_FN 2
  91
  92 %macro ADD_PAETH_PRED_FN 1
  93 cglobal add_png_paeth_prediction, 5, 7, %1, dst, src, top, w, bpp, end, cntr
  94 %if ARCH_X86_64
  95     movsxd            bppq, bppd
  96     movsxd              wq, wd
  97 %endif
  98     lea               endq, [dstq+wq-(mmsize/2-1)]
  99     sub               topq, dstq
 100     sub               srcq, dstq
 101     sub               dstq, bppq
 102     pxor                m7, m7
 103
 104     PUSH              dstq
 105     lea              cntrq, [bppq-1]
 106     shr              cntrq, 2 + mmsize/16
 107 .bpp_loop:
 108     lea               dstq, [dstq+cntrq*(mmsize/2)]
 109     movh                m0, [dstq]
 110     movh                m1, [topq+dstq]
 111     punpcklbw           m0, m7
 112     punpcklbw           m1, m7
 113     add               dstq, bppq
 114 .loop:
 115     mova                m2, m1
 116     movh                m1, [topq+dstq]
 117     mova                m3, m2
 118     punpcklbw           m1, m7
 119     mova                m4, m2
 120     psubw               m3, m1
 121     psubw               m4, m0
 122     mova                m5, m3
 123     paddw               m5, m4
 124 %if cpuflag(ssse3)
 125     pabsw               m3, m3
 126     pabsw               m4, m4
 127     pabsw               m5, m5
 128 %else ; !cpuflag(ssse3)
 129     psubw               m7, m5
 130     pmaxsw              m5, m7
 131     pxor                m6, m6
 132     pxor                m7, m7
 133     psubw               m6, m3
 134     psubw               m7, m4
 135     pmaxsw              m3, m6
 136     pmaxsw              m4, m7
 137     pxor                m7, m7
 138 %endif ; cpuflag(ssse3)
 139     mova                m6, m4
 140     pminsw              m6, m5
 141     pcmpgtw             m3, m6
 142     pcmpgtw             m4, m5
 143     mova                m6, m4
 144     pand                m4, m3
 145     pandn               m6, m3
 146     pandn               m3, m0
 147     movh                m0, [srcq+dstq]
 148     pand                m6, m1
 149     pand                m2, m4
 150     punpcklbw           m0, m7
 151     paddw               m0, m6
 152     paddw               m3, m2
 153     paddw               m0, m3
 154     pand                m0, [pw_255]
 155     mova                m3, m0
 156     packuswb            m3, m3
 157     movh            [dstq], m3
 158     add               dstq, bppq
 159     cmp               dstq, endq
 160     jle .loop
 161
 162     mov               dstq, [rsp]
 163     dec              cntrq
 164     jge .bpp_loop
 165     POP               dstq
 166     RET
 167 %endmacro
 168
 169 INIT_MMX mmxext
 170 ADD_PAETH_PRED_FN 0
 171
 172 INIT_MMX ssse3
 173 ADD_PAETH_PRED_FN 0