fdct_ia64.s

wince下的xvidcore开发库,可用于MP4等视频播放开发

// ******************************************************************************
// *                                                                            *
// *  This file is part of XviD, a free MPEG-4 video encoder/decoder            *
// *                                                                            *
// *                                                                            *
// *  XviD is free software; you can redistribute it and/or modify it           *
// *  under the terms of the GNU General Public License as published by         *
// *  the Free Software Foundation; either version 2 of the License, or         *
// *  (at your option) any later version.                                       *
// *                                                                            *
// *  XviD is distributed in the hope that it will be useful, but               *
// *  WITHOUT ANY WARRANTY; without even the implied warranty of                *
// *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the             *
// *  GNU General Public License for more details.                              *
// *                                                                            *
// *  You should have received a copy of the GNU General Public License         *
// *  along with this program; if not, write to the Free Software               *
// *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA  *
// *                                                                            *
// ******************************************************************************
//
// ******************************************************************************
// *                                                                            *
// *  fdct_ia64.s, IA-64 optimized forward DCT                                  *
// *                                                                            *
// *  Completed version provided by Intel at AppNote AP-922                     *
// *  http://developer.intel.com/software/products/college/ia32/strmsimd/       *
// *  Copyright (C) 1999 Intel Corporation,                                     *
// *                                                                            *
// *  This version was implemented during an IA-64 practical training at        *
// *  the University of Karlsruhe (http://i44w3.info.uni-karlsruhe.de/)         *
// *  Copyright (C) 2002 - Stephan Krause, Ingo-Marc Weber, Daniel Kallfass     *
// *                                                                            *
// *  For more information visit the XviD homepage: http://www.xvid.org         *
// *                                                                            *
// ******************************************************************************
//
// ******************************************************************************
// *                                                                            *
// *  Revision history:                                                         *
// *                                                                            *
// *  24.07.2002 Initial Version                                                *
// *                                                                            *
// ******************************************************************************


// This is a fast precise implementation of 8x8 Discrete Cosine Transform
// published in Intel Application Note 922 from 1999 and optimized for IA-64.
//
// An unoptimized "straight forward" version can be found at the end of this file.


        .pred.safe_across_calls p1-p5,p16-p63
.text
        .align 16
        .global fdct_ia64#
        .proc fdct_ia64#
fdct_ia64:
        .prologue
        alloc r14 = ar.pfs, 1, 56, 0, 0
                        // Save constants
        mov r31 = 0x32ec                // c0 = tan(1pi/16)
        mov r30 = 0x6a0a                // c1 = tan(2pi/16)
        mov r29 = 0xab0e                // c2 = tan(3pi/16)
        mov r28 = 0xb505                // g4 = cos(4pi/16)
        mov r27 = 0xd4db                // g3 = cos(3pi/16)
        mov r26 = 0xec83                // g2 = cos(2pi/16)
        mov r25 = 0xfb15                // g1 = cos(1pi/16)
        mov r24 = 0x0002                // correction bit for descaling
        mov r23 = 0x0004                // correction bit for descaling

                // Load Matrix into registers
        
        add loc0 = r0, r32
        add loc2 = 16, r32
        add loc4 = 32, r32
        add loc6 = 48, r32
        add loc8 = 64, r32
        add loc10 = 80, r32
        add loc12 = 96, r32
        add loc14 = 112, r32
        add loc1 = 8, r32
        add loc3 = 24, r32
        add loc5 = 40, r32
        add loc7 = 56, r32
        add loc9 = 72, r32
        add loc11 = 88, r32
        add loc13 = 104, r32
        add loc15 = 120, r32
        ;;
        ld8 loc16 = [loc0]
        ld8 loc17 = [loc2]
        ld8 loc18 = [loc4]
        ld8 loc19 = [loc6]
        ld8 loc20 = [loc8]
        ld8 loc21 = [loc10]
        ld8 loc22 = [loc12]
        ld8 loc23 = [loc14]
        ld8 loc24 = [loc1]
        ld8 loc25 = [loc3]
        ld8 loc26 = [loc5]
        ld8 loc27 = [loc7]
        mux2 r26 = r26, 0x00
        ld8 loc28 = [loc9]
        mux2 r31 = r31, 0x00
        mux2 r25 = r25, 0x00
        ld8 loc29 = [loc11]
        mux2 r30 = r30, 0x00
        mux2 r29 = r29, 0x00
        ld8 loc30 = [loc13]
        mux2 r28 = r28, 0x00
        mux2 r27 = r27, 0x00
        ld8 loc31 = [loc15]
        mux2 r24 = r24, 0x00
        mux2 r23 = r23, 0x00
        ;;
        pshl2 loc16 = loc16, 3
        pshl2 loc17 = loc17, 3
        pshl2 loc18 = loc18, 3
        pshl2 loc19 = loc19, 3
        pshl2 loc20 = loc20, 3
        pshl2 loc21 = loc21, 3
        pshl2 loc22 = loc22, 3
        pshl2 loc23 = loc23, 3
        ;;
        pshl2 loc24 = loc24, 3
                
                // *******************
                // column-DTC 1st half
                // *******************
                
        psub2 loc37 = loc17, loc22      // t5 = x1 - x6
        pshl2 loc25 = loc25, 3
        pshl2 loc26 = loc26, 3
        psub2 loc38 = loc18, loc21      // t6 = x2 - x5
        pshl2 loc27 = loc27, 3
        pshl2 loc28 = loc28, 3
        ;;
        padd2 loc32 = loc16, loc23      // t0 = x0 + x7
        pshl2 loc29 = loc29, 3
        pshl2 loc30 = loc30, 3
        padd2 loc33 = loc17, loc22      // t1 = x1 + x6
        padd2 loc40 = loc37, loc38      // buf0 = t5 + t6
        psub2 loc41 = loc37, loc38      // buf1 = t5 - t6
        ;;
        padd2 loc34 = loc18, loc21      // t2 = x2 + x5
        pshl2 loc31 = loc31, 3
        padd2 loc35 = loc19, loc20      // t3 = x3 + x4
        psub2 loc36 = loc16, loc23      // t4 = x0 - x7
        pmpyshr2 loc37 = loc40, r28, 16 // t5 = buf0 * g4
        pmpyshr2 loc38 = loc41, r28, 16 // t6 = buf1 * g4
        ;;
        psub2 loc39 = loc19, loc20      // t7 = x3 - x4
        padd2 loc37 = loc37, loc40      // t5 = t5 + buf1
        padd2 loc38 = loc38, loc41      // t6 = t6 + buf2

        padd2 loc16 = loc32, loc35      // x0 = t0 + t3
        padd2 loc17 = loc33, loc34      // x1 = t1 + t2
        psub2 loc18 = loc32, loc35      // x2 = t0 - t3
        ;;
        psub2 loc19 = loc33, loc34      // x3 = t1 - t2
        padd2 loc20 = loc36, loc37      // x4 = t4 + t5
        padd2 loc21 = loc38, loc39      // x5 = t6 + t7
        psub2 loc22 = loc36, loc37      // x6 = t4 - t5
        psub2 loc23 = loc38, loc39      // x7 = t6 - t7
        ;;
        pmpyshr2 loc42 = loc18, r30, 16 // buf2 = x2 * c1
        padd2 loc32 = loc16, loc17      // t0 = x0 + x1
        pmpyshr2 loc43 = loc19, r30, 16 // buf3 = x3 * c1
        pmpyshr2 loc44 = loc20, r31, 16 // buf4 = x4 * c0
        psub2 loc33 = loc16, loc17      // t1 = x0 - x1
        pmpyshr2 loc46 = loc22, r29, 16 // buf6 = x6 * c2
        pmpyshr2 loc47 = loc23, r29, 16 // buf7 = x7 * c2
        ;;
        padd2 loc46 = loc46, loc22      // buf6 = buf6 + x6
        pmpyshr2 loc45 = loc21, r31, 16 // buf5 = x5 * c0
        padd2 loc47 = loc47, loc23      // buf7 = buf7 + x7
        padd2 loc34 = loc18, loc43      // t2 = x2 + (x3 * c1)
        psub2 loc35 = loc42, loc19      // t3 = (c1 * x2) - x3
        psub2 loc37 = loc44, loc21      // t5 = (c1 * x4) - x5
        ;;
        padd2 loc36 = loc20, loc45      // t4 = x4 + (x5 * c1)
        padd2 loc38 = loc22, loc47      // t6 = x6 + (x7 * c1)
        pmpyshr2 loc16 = loc32, r28, 16 // x0 = t0 * g4
        pmpyshr2 loc17 = loc33, r28, 16 // x1 = t1 * g4
        psub2 loc39 = loc46, loc23      // t7 = (c1 * x6) - x7
        ;;
        padd2 loc48 = loc16, loc32      // y0 = x0 + t0
        pmpyshr2 loc18 = loc34, r26, 16 // x2 = t2 * g2
        pmpyshr2 loc19 = loc35, r26, 16 // x3 = t3 * g2
        padd2 loc52 = loc17, loc33      // y4 = x1 + t1
        pmpyshr2 loc20 = loc36, r25, 16 // x4 = t4 * g1
        pmpyshr2 loc21 = loc37, r25, 16 // x5 = t5 * g1
        ;;
        padd2 loc50 = loc18, loc34      // y2 = x2 + t2
        pmpyshr2 loc22 = loc38, r27, 16 // x6 = t6 * g3
        pmpyshr2 loc23 = loc39, r27, 16 // x7 = t7 * g3
        padd2 loc55 = loc21, loc37      // y7 = x5 + t5
        padd2 loc49 = loc20, loc36      // y1 = x4 + t4
        padd2 loc54 = loc19, loc35      // y6 = x3 + t3
        ;;
        padd2 loc51 = loc22, loc38      // y3 = x6 + t6
        padd2 loc53 = loc23, loc39      // y5 = x7 + t7
        
                //divide by 4
                
        padd2 loc48 = loc48, r24
        padd2 loc49 = loc49, r24
        padd2 loc50 = loc50, r24
        padd2 loc52 = loc52, r24
        ;;
        padd2 loc51 = loc51, r24
        pshr2 loc48 = loc48, 2
        padd2 loc53 = loc53, r24
        pshr2 loc49 = loc49, 2
        padd2 loc54 = loc54, r24
        pshr2 loc50 = loc50, 2
        padd2 loc55 = loc55, r24
        pshr2 loc52 = loc52, 2
        ;;
        pshr2 loc51 = loc51, 2
        pshr2 loc53 = loc53, 2
        pshr2 loc54 = loc54, 2
        pshr2 loc55 = loc55, 2


                // *******************
                // column-DTC 2nd half
                // *******************

        psub2 loc37 = loc25, loc30      // t5 = x1.2 - x6.2
        psub2 loc38 = loc26, loc29      // t6 = x2.2 - x5.2
        padd2 loc32 = loc24, loc31      // t0 = x0.2 + x7.2
        padd2 loc33 = loc25, loc30      // t1 = x1.2 + x6.2
        ;;
        padd2 loc34 = loc26, loc29      // t2 = x2.2 + x5.2
        psub2 loc41 = loc37, loc38      // buf1 = t5 - t6
        padd2 loc40 = loc37, loc38      // buf0 = t5 + t6
        padd2 loc35 = loc27, loc28      // t3 = x3.2 + x4.2
        ;;
        psub2 loc36 = loc24, loc31      // t4 = x0.2 - x7.2
        pmpyshr2 loc37 = loc40, r28, 16 // t5 = buf0 * g4
        pmpyshr2 loc38 = loc41, r28, 16 // t6 = buf1 * g4
        ;;
        psub2 loc39 = loc27, loc28      // t7 = x3.2 - x4.2
        padd2 loc37 = loc37, loc40      // t5 = t5 + buf1
        padd2 loc38 = loc38, loc41      // t6 = t6 + buf2

        padd2 loc16 = loc32, loc35      // x0 = t0 + t3
        padd2 loc17 = loc33, loc34      // x1 = t1 + t2
        psub2 loc18 = loc32, loc35      // x2 = t0 - t3
        ;;
        psub2 loc19 = loc33, loc34      // x3 = t1 - t2
        padd2 loc20 = loc36, loc37      // x4 = t4 + t5
        padd2 loc21 = loc38, loc39      // x5 = t6 + t7
        psub2 loc22 = loc36, loc37      // x6 = t4 - t5
        psub2 loc23 = loc38, loc39      // x7 = t6 - t7
        ;;
        pmpyshr2 loc42 = loc18, r30, 16 // buf2 = x2 * c1
        padd2 loc32 = loc16, loc17      // t0 = x0 + x1
        pmpyshr2 loc43 = loc19, r30, 16 // buf3 = x3 * c1
        pmpyshr2 loc44 = loc20, r31, 16 // buf4 = x4 * c0
        psub2 loc33 = loc16, loc17      // t1 = x0 - x1
        pmpyshr2 loc45 = loc21, r31, 16 // buf5 = x5 * c0
        pmpyshr2 loc46 = loc22, r29, 16 // buf6 = x6 * c2
        pmpyshr2 loc47 = loc23, r29, 16 // buf7 = x7 * c2
        ;;
        padd2 loc34 = loc18, loc43      // t2 = x2 + buf3
        padd2 loc46 = loc46, loc22      // buf6 = buf6 + x6
        padd2 loc47 = loc47, loc23      // buf7 = buf7 + x7
        psub2 loc35 = loc42, loc19      // t3 = buf2 - x3
        padd2 loc36 = loc20, loc45      // t4 = x4 + buf5
        pmpyshr2 loc16 = loc32, r28, 16 // x0 = t0 * g4
        ;;
        psub2 loc37 = loc44, loc21      // t5 = buf4 - x5
        padd2 loc38 = loc22, loc47      // t6 = x6 + buf7
        psub2 loc39 = loc46, loc23      // t7 = buf6 - x7
        pmpyshr2 loc17 = loc33, r28, 16 // x1 = t1 * g4
        ;;
        padd2 loc40 = loc16, loc32      // y0.2 = x0 + t0
        pmpyshr2 loc18 = loc34, r26, 16 // x2 = t2 * g2
        pmpyshr2 loc19 = loc35, r26, 16 // x3 = t3 * g2
        padd2 loc44 = loc17, loc33      // y4.2 = x1 + t1
        pmpyshr2 loc20 = loc36, r25, 16 // x4 = t4 * g1
        pmpyshr2 loc21 = loc37, r25, 16 // x5 = t5 * g1
        ;;
        padd2 loc42 = loc18, loc34      // y2.2 = x2 + t2
        pmpyshr2 loc22 = loc38, r27, 16 // x6 = t6 * g3
        pmpyshr2 loc23 = loc39, r27, 16 // x7 = t7 * g3
        padd2 loc47 = loc21, loc37      // y7.2 = x5 + t5
        padd2 loc41 = loc20, loc36      // y1.2 = x4 + t4
        padd2 loc46 = loc19, loc35      // y6.2 = x3 + t3
        ;;
        padd2 loc43 = loc22, loc38      // y3.2 = x6 + t6