libavcodec/x86/vp6dsp.asm

   1 ;******************************************************************************
   2 ;* MMX/SSE2-optimized functions for the VP6 decoder
   3 ;* Copyright (C) 2009  Sebastien Lucas <sebastien.lucas@gmail.com>
   4 ;* Copyright (C) 2009  Zuxy Meng <zuxy.meng@gmail.com>
   5 ;*
   6 ;* This file is part of Libav.
   7 ;*
   8 ;* Libav 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 ;* Libav 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 Libav; 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/x86/x86util.asm"
  24
  25 cextern pw_64
  26
  27 SECTION .text
  28
  29 %macro DIAG4 6
  30 %if mmsize == 8
  31     movq          m0, [%1+%2]
  32     movq          m1, [%1+%3]
  33     movq          m3, m0
  34     movq          m4, m1
  35     punpcklbw     m0, m7
  36     punpcklbw     m1, m7
  37     punpckhbw     m3, m7
  38     punpckhbw     m4, m7
  39     pmullw        m0, [rsp+8*11] ; src[x-8 ] * biweight [0]
  40     pmullw        m1, [rsp+8*12] ; src[x   ] * biweight [1]
  41     pmullw        m3, [rsp+8*11] ; src[x-8 ] * biweight [0]
  42     pmullw        m4, [rsp+8*12] ; src[x   ] * biweight [1]
  43     paddw         m0, m1
  44     paddw         m3, m4
  45     movq          m1, [%1+%4]
  46     movq          m2, [%1+%5]
  47     movq          m4, m1
  48     movq          m5, m2
  49     punpcklbw     m1, m7
  50     punpcklbw     m2, m7
  51     punpckhbw     m4, m7
  52     punpckhbw     m5, m7
  53     pmullw        m1, [rsp+8*13] ; src[x+8 ] * biweight [2]
  54     pmullw        m2, [rsp+8*14] ; src[x+16] * biweight [3]
  55     pmullw        m4, [rsp+8*13] ; src[x+8 ] * biweight [2]
  56     pmullw        m5, [rsp+8*14] ; src[x+16] * biweight [3]
  57     paddw         m1, m2
  58     paddw         m4, m5
  59     paddsw        m0, m1
  60     paddsw        m3, m4
  61     paddsw        m0, m6         ; Add 64
  62     paddsw        m3, m6         ; Add 64
  63     psraw         m0, 7
  64     psraw         m3, 7
  65     packuswb      m0, m3
  66     movq        [%6], m0
  67 %else ; mmsize == 16
  68     movq          m0, [%1+%2]
  69     movq          m1, [%1+%3]
  70     punpcklbw     m0, m7
  71     punpcklbw     m1, m7
  72     pmullw        m0, m4         ; src[x-8 ] * biweight [0]
  73     pmullw        m1, m5         ; src[x   ] * biweight [1]
  74     paddw         m0, m1
  75     movq          m1, [%1+%4]
  76     movq          m2, [%1+%5]
  77     punpcklbw     m1, m7
  78     punpcklbw     m2, m7
  79     pmullw        m1, m6         ; src[x+8 ] * biweight [2]
  80     pmullw        m2, m3         ; src[x+16] * biweight [3]
  81     paddw         m1, m2
  82     paddsw        m0, m1
  83     paddsw        m0, [pw_64]    ; Add 64
  84     psraw         m0, 7
  85     packuswb      m0, m0
  86     movq        [%6], m0
  87 %endif ; mmsize == 8/16
  88 %endmacro
  89
  90 %macro SPLAT4REGS 0
  91 %if mmsize == 8
  92     movq         m5, m3
  93     punpcklwd    m3, m3
  94     movq         m4, m3
  95     punpckldq    m3, m3
  96     punpckhdq    m4, m4
  97     punpckhwd    m5, m5
  98     movq         m2, m5
  99     punpckhdq    m2, m2
 100     punpckldq    m5, m5
 101     movq [rsp+8*11], m3
 102     movq [rsp+8*12], m4
 103     movq [rsp+8*13], m5
 104     movq [rsp+8*14], m2
 105 %else ; mmsize == 16
 106     pshuflw      m4, m3, 0x0
 107     pshuflw      m5, m3, 0x55
 108     pshuflw      m6, m3, 0xAA
 109     pshuflw      m3, m3, 0xFF
 110     punpcklqdq   m4, m4
 111     punpcklqdq   m5, m5
 112     punpcklqdq   m6, m6
 113     punpcklqdq   m3, m3
 114 %endif ; mmsize == 8/16
 115 %endmacro
 116
 117 %macro vp6_filter_diag4 0
 118 ; void ff_vp6_filter_diag4_<opt>(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
 119 ;                                const int16_t h_weight[4], const int16_t v_weights[4])
 120 cglobal vp6_filter_diag4, 5, 7, 8
 121     mov          r5, rsp         ; backup stack pointer
 122     and         rsp, ~(mmsize-1) ; align stack
 123 %if mmsize == 16
 124     sub         rsp, 8*11
 125 %else
 126     sub         rsp, 8*15
 127     movq         m6, [pw_64]
 128 %endif
 129
 130     sub          r1, r2
 131
 132     pxor         m7, m7
 133     movq         m3, [r3]
 134     SPLAT4REGS
 135
 136     mov          r3, rsp
 137     mov          r6, 11
 138 .nextrow:
 139     DIAG4        r1, -1, 0, 1, 2, r3
 140     add          r3, 8
 141     add          r1, r2
 142     dec          r6
 143     jnz .nextrow
 144
 145     movq         m3, [r4]
 146     SPLAT4REGS
 147
 148     lea          r3, [rsp+8]
 149     mov          r6, 8
 150 .nextcol:
 151     DIAG4        r3, -8, 0, 8, 16, r0
 152     add          r3, 8
 153     add          r0, r2
 154     dec          r6
 155     jnz .nextcol
 156
 157     mov         rsp, r5          ; restore stack pointer
 158     RET
 159 %endmacro
 160
 161 %if ARCH_X86_32
 162 INIT_MMX mmx
 163 vp6_filter_diag4
 164 %endif
 165
 166 INIT_XMM sse2
 167 vp6_filter_diag4