i965_drv_video/shaders/h264/mc/interpolate_C_4x4.asm

   1 /*\r
   2  * Interpolation kernel for chrominance 4x4 motion compensation\r
   3  * Copyright © <2010>, Intel Corporation.\r
   4  *\r
   5  * This program is licensed under the terms and conditions of the\r
   6  * Eclipse Public License (EPL), version 1.0.  The full text of the EPL is at\r
   7  * http://www.opensource.org/licenses/eclipse-1.0.php.\r
   8  *\r
   9  */\r
  10 //      Kernel name: Interpolate_C_4x4_Func.asm\r
  11 //\r
  12 //      Interpolation kernel for chrominance 4x4 motion compensation\r
  13 //\r
  14 //  $Revision: 8 $\r
  15 //  $Date: 10/09/06 4:00p $\r
  16 //\r
  17 \r
  18 \r
  19 //#if !defined(__Interpolate_C_4x4_Func__)              // Make sure this is only included once\r
  20 //#define __Interpolate_C_4x4_Func__\r
  21 \r
  22 \r
  23 INTERLABEL(Interpolate_C_4x4_Func):\r
  24 \r
  25 \r
  26         // (8-xFrac) and (8-yFrac)\r
  27     add (2)             gW0<1>:w                        gMVX_FRACC<2;2,1>:w                             -0x08:w\r
  28 \r
  29         // Compute the GRF address of the starting position of the reference area\r
  30     mov (1)             pREF0:w                         nOFFSET_REFC:w          {NoDDClr}\r
  31         mov (1)         pREF1:uw                        nOFFSET_REFC+16:w       {NoDDChk,NoDDClr}\r
  32         mov (1)         pRESULT:uw                      gpINTPC:uw                      {NoDDChk}\r
  33 \r
  34         // gCOEFA = (8-xFrac)*(8-yFrac)\r
  35     // gCOEFB = xFrac*(8-yFrac)  \r
  36     // gCOEFC = (8-xFrac)*yFrac\r
  37     // gCOEFD = xFrac*yFrac \r
  38     mul (1)             gCOEFD:w                gMVX_FRACC:w                                    gMVY_FRACC:w    {NoDDClr}\r
  39     mul (1)             gCOEFA:w                        -gW0:w                                                  -gW1:uw         {NoDDClr,NoDDChk}\r
  40     mul (1)             gCOEFB:w                        gMVX_FRACC:w                                    -gW1:uw         {NoDDClr,NoDDChk}\r
  41     mul (1)             gCOEFC:w                    -gW0:w                                                      gMVY_FRACC:w {NoDDChk}\r
  42 \r
  43         add (2)         gW0<1>:uw                       pREF0<2;2,1>:uw                                 16:w\r
  44 \r
  45     // (8-xFrac)*(8-yFrac)*A\r
  46     // ---------------------\r
  47     mul (16)    acc0<1>:uw                      r[pREF0,0]<16;8,1>:ub                   gCOEFA:uw\r
  48     mul (16)    acc1<1>:uw                      r[pREF0,nGRFWIB]<16;8,1>:ub             gCOEFA:uw\r
  49         \r
  50     // xFrac*(8-yFrac)*B\r
  51     // -------------------\r
  52     mac (16)    acc0<1>:uw          r[pREF0,2]<16;8,1>:ub                       gCOEFB:uw\r
  53     mac (16)    acc1<1>:uw          r[pREF0,nGRFWIB+2]<16;8,1>:ub       gCOEFB:uw\r
  54 \r
  55     // (8-xFrac)*yFrac*C\r
  56     // -------------------\r
  57     mov (2)             pREF0<1>:uw                     gW0<2;2,1>:uw\r
  58     mac (16)    acc0<1>:uw          r[pREF0,0]<8,1>:ub                          gCOEFC:uw\r
  59     mac (16)    acc1<1>:uw          r[pREF0,nGRFWIB]<8,1>:ub            gCOEFC:uw\r
  60             \r
  61     // xFrac*yFrac*D\r
  62     // -----------------\r
  63     mac (16)    r[pRESULT]<1>:uw        r[pREF0,2]<8,1>:ub                              gCOEFD:uw\r
  64     mac (16)    r[pRESULT,GRFWIB]<1>:uw r[pREF0,nGRFWIB+2]<8,1>:ub gCOEFD:uw {SecHalf}\r
  65 \r
  66    \r
  67 //#endif        // !defined(__Interpolate_C_4x4_Func__)\r