ffmpeg/libswresample/x86/rematrix.asm

   1 ;******************************************************************************
   2 ;* Copyright (c) 2012 Michael Niedermayer
   3 ;*
   4 ;* This file is part of FFmpeg.
   5 ;*
   6 ;* FFmpeg 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 ;* FFmpeg 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 FFmpeg; 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
  24 SECTION_RODATA 32
  25 dw1: times 8  dd 1
  26 w1 : times 16 dw 1
  27
  28 SECTION .text
  29
  30 %macro MIX2_FLT 1
  31 cglobal mix_2_1_%1_float, 7, 7, 6, out, in1, in2, coeffp, index1, index2, len
  32 %ifidn %1, a
  33     test in1q, mmsize-1
  34         jne mix_2_1_float_u_int %+ SUFFIX
  35     test in2q, mmsize-1
  36         jne mix_2_1_float_u_int %+ SUFFIX
  37     test outq, mmsize-1
  38         jne mix_2_1_float_u_int %+ SUFFIX
  39 %else
  40 mix_2_1_float_u_int %+ SUFFIX
  41 %endif
  42     VBROADCASTSS m4, [coeffpq + 4*index1q]
  43     VBROADCASTSS m5, [coeffpq + 4*index2q]
  44     shl lend    , 2
  45     add in1q    , lenq
  46     add in2q    , lenq
  47     add outq    , lenq
  48     neg lenq
  49 .next:
  50 %ifidn %1, a
  51     mulps        m0, m4, [in1q + lenq         ]
  52     mulps        m1, m5, [in2q + lenq         ]
  53     mulps        m2, m4, [in1q + lenq + mmsize]
  54     mulps        m3, m5, [in2q + lenq + mmsize]
  55 %else
  56     movu         m0, [in1q + lenq         ]
  57     movu         m1, [in2q + lenq         ]
  58     movu         m2, [in1q + lenq + mmsize]
  59     movu         m3, [in2q + lenq + mmsize]
  60     mulps        m0, m0, m4
  61     mulps        m1, m1, m5
  62     mulps        m2, m2, m4
  63     mulps        m3, m3, m5
  64 %endif
  65     addps        m0, m0, m1
  66     addps        m2, m2, m3
  67     mov%1  [outq + lenq         ], m0
  68     mov%1  [outq + lenq + mmsize], m2
  69     add        lenq, mmsize*2
  70         jl .next
  71     REP_RET
  72 %endmacro
  73
  74 %macro MIX1_FLT 1
  75 cglobal mix_1_1_%1_float, 5, 5, 3, out, in, coeffp, index, len
  76 %ifidn %1, a
  77     test inq, mmsize-1
  78         jne mix_1_1_float_u_int %+ SUFFIX
  79     test outq, mmsize-1
  80         jne mix_1_1_float_u_int %+ SUFFIX
  81 %else
  82 mix_1_1_float_u_int %+ SUFFIX
  83 %endif
  84     VBROADCASTSS m2, [coeffpq + 4*indexq]
  85     shl lenq    , 2
  86     add inq     , lenq
  87     add outq    , lenq
  88     neg lenq
  89 .next:
  90 %ifidn %1, a
  91     mulps        m0, m2, [inq + lenq         ]
  92     mulps        m1, m2, [inq + lenq + mmsize]
  93 %else
  94     movu         m0, [inq + lenq         ]
  95     movu         m1, [inq + lenq + mmsize]
  96     mulps        m0, m0, m2
  97     mulps        m1, m1, m2
  98 %endif
  99     mov%1  [outq + lenq         ], m0
 100     mov%1  [outq + lenq + mmsize], m1
 101     add        lenq, mmsize*2
 102         jl .next
 103     REP_RET
 104 %endmacro
 105
 106 %macro MIX1_INT16 1
 107 cglobal mix_1_1_%1_int16, 5, 5, 6, out, in, coeffp, index, len
 108 %ifidn %1, a
 109     test inq, mmsize-1
 110         jne mix_1_1_int16_u_int %+ SUFFIX
 111     test outq, mmsize-1
 112         jne mix_1_1_int16_u_int %+ SUFFIX
 113 %else
 114 mix_1_1_int16_u_int %+ SUFFIX
 115 %endif
 116     movd   m4, [coeffpq + 4*indexq]
 117     SPLATW m5, m4
 118     psllq  m4, 32
 119     psrlq  m4, 48
 120     mova   m0, [w1]
 121     psllw  m0, m4
 122     psrlw  m0, 1
 123     punpcklwd m5, m0
 124     add lenq    , lenq
 125     add inq     , lenq
 126     add outq    , lenq
 127     neg lenq
 128 .next:
 129     mov%1        m0, [inq + lenq         ]
 130     mov%1        m2, [inq + lenq + mmsize]
 131     mova         m1, m0
 132     mova         m3, m2
 133     punpcklwd    m0, [w1]
 134     punpckhwd    m1, [w1]
 135     punpcklwd    m2, [w1]
 136     punpckhwd    m3, [w1]
 137     pmaddwd      m0, m5
 138     pmaddwd      m1, m5
 139     pmaddwd      m2, m5
 140     pmaddwd      m3, m5
 141     psrad        m0, m4
 142     psrad        m1, m4
 143     psrad        m2, m4
 144     psrad        m3, m4
 145     packssdw     m0, m1
 146     packssdw     m2, m3
 147     mov%1  [outq + lenq         ], m0
 148     mov%1  [outq + lenq + mmsize], m2
 149     add        lenq, mmsize*2
 150         jl .next
 151 %if mmsize == 8
 152     emms
 153     RET
 154 %else
 155     REP_RET
 156 %endif
 157 %endmacro
 158
 159 %macro MIX2_INT16 1
 160 cglobal mix_2_1_%1_int16, 7, 7, 8, out, in1, in2, coeffp, index1, index2, len
 161 %ifidn %1, a
 162     test in1q, mmsize-1
 163         jne mix_2_1_int16_u_int %+ SUFFIX
 164     test in2q, mmsize-1
 165         jne mix_2_1_int16_u_int %+ SUFFIX
 166     test outq, mmsize-1
 167         jne mix_2_1_int16_u_int %+ SUFFIX
 168 %else
 169 mix_2_1_int16_u_int %+ SUFFIX
 170 %endif
 171     movd   m4, [coeffpq + 4*index1q]
 172     movd   m6, [coeffpq + 4*index2q]
 173     SPLATW m5, m4
 174     SPLATW m6, m6
 175     psllq  m4, 32
 176     psrlq  m4, 48
 177     mova   m7, [dw1]
 178     pslld  m7, m4
 179     psrld  m7, 1
 180     punpcklwd m5, m6
 181     add lend    , lend
 182     add in1q    , lenq
 183     add in2q    , lenq
 184     add outq    , lenq
 185     neg lenq
 186 .next:
 187     mov%1        m0, [in1q + lenq         ]
 188     mov%1        m2, [in2q + lenq         ]
 189     mova         m1, m0
 190     punpcklwd    m0, m2
 191     punpckhwd    m1, m2
 192
 193     mov%1        m2, [in1q + lenq + mmsize]
 194     mov%1        m6, [in2q + lenq + mmsize]
 195     mova         m3, m2
 196     punpcklwd    m2, m6
 197     punpckhwd    m3, m6
 198
 199     pmaddwd      m0, m5
 200     pmaddwd      m1, m5
 201     pmaddwd      m2, m5
 202     pmaddwd      m3, m5
 203     paddd        m0, m7
 204     paddd        m1, m7
 205     paddd        m2, m7
 206     paddd        m3, m7
 207     psrad        m0, m4
 208     psrad        m1, m4
 209     psrad        m2, m4
 210     psrad        m3, m4
 211     packssdw     m0, m1
 212     packssdw     m2, m3
 213     mov%1  [outq + lenq         ], m0
 214     mov%1  [outq + lenq + mmsize], m2
 215     add        lenq, mmsize*2
 216         jl .next
 217 %if mmsize == 8
 218     emms
 219     RET
 220 %else
 221     REP_RET
 222 %endif
 223 %endmacro
 224
 225
 226 INIT_MMX mmx
 227 MIX1_INT16 u
 228 MIX1_INT16 a
 229 MIX2_INT16 u
 230 MIX2_INT16 a
 231
 232 INIT_XMM sse
 233 MIX2_FLT u
 234 MIX2_FLT a
 235 MIX1_FLT u
 236 MIX1_FLT a
 237
 238 INIT_XMM sse2
 239 MIX1_INT16 u
 240 MIX1_INT16 a
 241 MIX2_INT16 u
 242 MIX2_INT16 a
 243
 244 %if HAVE_AVX_EXTERNAL
 245 INIT_YMM avx
 246 MIX2_FLT u
 247 MIX2_FLT a
 248 MIX1_FLT u
 249 MIX1_FLT a
 250 %endif