ffmpeg/libavcodec/x86/h264_chromamc_10bit.asm

   1 ;*****************************************************************************
   2 ;* MMX/SSE2/AVX-optimized 10-bit H.264 chroma MC code
   3 ;*****************************************************************************
   4 ;* Copyright (C) 2005-2011 x264 project
   5 ;*
   6 ;* Authors: Daniel Kang <daniel.d.kang@gmail.com>
   7 ;*
   8 ;* This file is part of FFmpeg.
   9 ;*
  10 ;* FFmpeg is free software; you can redistribute it and/or
  11 ;* modify it under the terms of the GNU Lesser General Public
  12 ;* License as published by the Free Software Foundation; either
  13 ;* version 2.1 of the License, or (at your option) any later version.
  14 ;*
  15 ;* FFmpeg is distributed in the hope that it will be useful,
  16 ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
  17 ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18 ;* Lesser General Public License for more details.
  19 ;*
  20 ;* You should have received a copy of the GNU Lesser General Public
  21 ;* License along with FFmpeg; if not, write to the Free Software
  22 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  23 ;******************************************************************************
  24
  25 %include "libavutil/x86/x86util.asm"
  26
  27 SECTION_RODATA
  28
  29 cextern pw_4
  30 cextern pw_8
  31 cextern pw_32
  32 cextern pw_64
  33
  34 SECTION .text
  35
  36
  37 %macro MV0_PIXELS_MC8 0
  38     lea           r4, [r2*3   ]
  39     lea           r5, [r2*4   ]
  40 .next4rows:
  41     movu          m0, [r1     ]
  42     movu          m1, [r1+r2  ]
  43     CHROMAMC_AVG  m0, [r0     ]
  44     CHROMAMC_AVG  m1, [r0+r2  ]
  45     mova   [r0     ], m0
  46     mova   [r0+r2  ], m1
  47     movu          m0, [r1+r2*2]
  48     movu          m1, [r1+r4  ]
  49     CHROMAMC_AVG  m0, [r0+r2*2]
  50     CHROMAMC_AVG  m1, [r0+r4  ]
  51     mova   [r0+r2*2], m0
  52     mova   [r0+r4  ], m1
  53     add           r1, r5
  54     add           r0, r5
  55     sub          r3d, 4
  56     jne .next4rows
  57 %endmacro
  58
  59 ;-----------------------------------------------------------------------------
  60 ; void ff_put/avg_h264_chroma_mc8(pixel *dst, pixel *src, int stride, int h,
  61 ;                                 int mx, int my)
  62 ;-----------------------------------------------------------------------------
  63 %macro CHROMA_MC8 1
  64 cglobal %1_h264_chroma_mc8_10, 6,7,8
  65     movsxdifnidn  r2, r2d
  66     mov          r6d, r5d
  67     or           r6d, r4d
  68     jne .at_least_one_non_zero
  69     ; mx == 0 AND my == 0 - no filter needed
  70     MV0_PIXELS_MC8
  71     REP_RET
  72
  73 .at_least_one_non_zero:
  74     mov          r6d, 2
  75     test         r5d, r5d
  76     je .x_interpolation
  77     mov           r6, r2        ; dxy = x ? 1 : stride
  78     test         r4d, r4d
  79     jne .xy_interpolation
  80 .x_interpolation:
  81     ; mx == 0 XOR my == 0 - 1 dimensional filter only
  82     or           r4d, r5d       ; x + y
  83     movd          m5, r4d
  84     mova          m4, [pw_8]
  85     mova          m6, [pw_4]    ; mm6 = rnd >> 3
  86     SPLATW        m5, m5        ; mm5 = B = x
  87     psubw         m4, m5        ; mm4 = A = 8-x
  88
  89 .next1drow:
  90     movu          m0, [r1   ]   ; mm0 = src[0..7]
  91     movu          m2, [r1+r6]   ; mm2 = src[1..8]
  92
  93     pmullw        m0, m4        ; mm0 = A * src[0..7]
  94     pmullw        m2, m5        ; mm2 = B * src[1..8]
  95
  96     paddw         m0, m6
  97     paddw         m0, m2
  98     psrlw         m0, 3
  99     CHROMAMC_AVG  m0, [r0]
 100     mova        [r0], m0        ; dst[0..7] = (A * src[0..7] + B * src[1..8] + (rnd >> 3)) >> 3
 101
 102     add           r0, r2
 103     add           r1, r2
 104     dec           r3d
 105     jne .next1drow
 106     REP_RET
 107
 108 .xy_interpolation: ; general case, bilinear
 109     movd          m4, r4m         ; x
 110     movd          m6, r5m         ; y
 111
 112     SPLATW        m4, m4          ; mm4 = x words
 113     SPLATW        m6, m6          ; mm6 = y words
 114     psllw         m5, m4, 3       ; mm5 = 8x
 115     pmullw        m4, m6          ; mm4 = x * y
 116     psllw         m6, 3           ; mm6 = 8y
 117     paddw         m1, m5, m6      ; mm7 = 8x+8y
 118     mova          m7, m4          ; DD = x * y
 119     psubw         m5, m4          ; mm5 = B = 8x - xy
 120     psubw         m6, m4          ; mm6 = C = 8y - xy
 121     paddw         m4, [pw_64]
 122     psubw         m4, m1          ; mm4 = A = xy - (8x+8y) + 64
 123
 124     movu          m0, [r1  ]      ; mm0 = src[0..7]
 125     movu          m1, [r1+2]      ; mm1 = src[1..8]
 126 .next2drow:
 127     add           r1, r2
 128
 129     pmullw        m2, m0, m4
 130     pmullw        m1, m5
 131     paddw         m2, m1          ; mm2 = A * src[0..7] + B * src[1..8]
 132
 133     movu          m0, [r1]
 134     movu          m1, [r1+2]
 135     pmullw        m3, m0, m6
 136     paddw         m2, m3          ; mm2 += C * src[0..7+strde]
 137     pmullw        m3, m1, m7
 138     paddw         m2, m3          ; mm2 += D * src[1..8+strde]
 139
 140     paddw         m2, [pw_32]
 141     psrlw         m2, 6
 142     CHROMAMC_AVG  m2, [r0]
 143     mova        [r0], m2          ; dst[0..7] = (mm2 + 32) >> 6
 144
 145     add           r0, r2
 146     dec          r3d
 147     jne .next2drow
 148     REP_RET
 149 %endmacro
 150
 151 ;-----------------------------------------------------------------------------
 152 ; void ff_put/avg_h264_chroma_mc4(pixel *dst, pixel *src, int stride, int h,
 153 ;                                 int mx, int my)
 154 ;-----------------------------------------------------------------------------
 155 ;TODO: xmm mc4
 156 %macro MC4_OP 2
 157     movq          %1, [r1  ]
 158     movq          m1, [r1+2]
 159     add           r1, r2
 160     pmullw        %1, m4
 161     pmullw        m1, m2
 162     paddw         m1, %1
 163     mova          %1, m1
 164
 165     pmullw        %2, m5
 166     pmullw        m1, m3
 167     paddw         %2, [pw_32]
 168     paddw         m1, %2
 169     psrlw         m1, 6
 170     CHROMAMC_AVG  m1, %2, [r0]
 171     movq        [r0], m1
 172     add           r0, r2
 173 %endmacro
 174
 175 %macro CHROMA_MC4 1
 176 cglobal %1_h264_chroma_mc4_10, 6,6,7
 177     movsxdifnidn  r2, r2d
 178     movd          m2, r4m         ; x
 179     movd          m3, r5m         ; y
 180     mova          m4, [pw_8]
 181     mova          m5, m4
 182     SPLATW        m2, m2
 183     SPLATW        m3, m3
 184     psubw         m4, m2
 185     psubw         m5, m3
 186
 187     movq          m0, [r1  ]
 188     movq          m6, [r1+2]
 189     add           r1, r2
 190     pmullw        m0, m4
 191     pmullw        m6, m2
 192     paddw         m6, m0
 193
 194 .next2rows:
 195     MC4_OP m0, m6
 196     MC4_OP m6, m0
 197     sub   r3d, 2
 198     jnz .next2rows
 199     REP_RET
 200 %endmacro
 201
 202 ;-----------------------------------------------------------------------------
 203 ; void ff_put/avg_h264_chroma_mc2(pixel *dst, pixel *src, int stride, int h,
 204 ;                                 int mx, int my)
 205 ;-----------------------------------------------------------------------------
 206 %macro CHROMA_MC2 1
 207 cglobal %1_h264_chroma_mc2_10, 6,7
 208     movsxdifnidn  r2, r2d
 209     mov          r6d, r4d
 210     shl          r4d, 16
 211     sub          r4d, r6d
 212     add          r4d, 8
 213     imul         r5d, r4d         ; x*y<<16 | y*(8-x)
 214     shl          r4d, 3
 215     sub          r4d, r5d         ; x*(8-y)<<16 | (8-x)*(8-y)
 216
 217     movd          m5, r4d
 218     movd          m6, r5d
 219     punpckldq     m5, m5          ; mm5 = {A,B,A,B}
 220     punpckldq     m6, m6          ; mm6 = {C,D,C,D}
 221     pxor          m7, m7
 222     pshufw        m2, [r1], 0x94    ; mm0 = src[0,1,1,2]
 223
 224 .nextrow:
 225     add           r1, r2
 226     movq          m1, m2
 227     pmaddwd       m1, m5          ; mm1 = A * src[0,1] + B * src[1,2]
 228     pshufw        m0, [r1], 0x94    ; mm0 = src[0,1,1,2]
 229     movq          m2, m0
 230     pmaddwd       m0, m6
 231     paddw         m1, [pw_32]
 232     paddw         m1, m0          ; mm1 += C * src[0,1] + D * src[1,2]
 233     psrlw         m1, 6
 234     packssdw      m1, m7
 235     CHROMAMC_AVG  m1, m3, [r0]
 236     movd        [r0], m1
 237     add           r0, r2
 238     dec          r3d
 239     jnz .nextrow
 240     REP_RET
 241 %endmacro
 242
 243 %macro NOTHING 2-3
 244 %endmacro
 245 %macro AVG 2-3
 246 %if %0==3
 247     movq          %2, %3
 248 %endif
 249     pavgw         %1, %2
 250 %endmacro
 251
 252 %define CHROMAMC_AVG  NOTHING
 253 INIT_XMM sse2
 254 CHROMA_MC8 put
 255 %if HAVE_AVX_EXTERNAL
 256 INIT_XMM avx
 257 CHROMA_MC8 put
 258 %endif
 259 INIT_MMX mmxext
 260 CHROMA_MC4 put
 261 CHROMA_MC2 put
 262
 263 %define CHROMAMC_AVG  AVG
 264 INIT_XMM sse2
 265 CHROMA_MC8 avg
 266 %if HAVE_AVX_EXTERNAL
 267 INIT_XMM avx
 268 CHROMA_MC8 avg
 269 %endif
 270 INIT_MMX mmxext
 271 CHROMA_MC4 avg
 272 CHROMA_MC2 avg