FFmpeg: libavcodec/mips/dsputil

00001 /*
00002  * Copyright (c) 2012
00003  *      MIPS Technologies, Inc., California.
00004  *
00005  * Redistribution and use in source and binary forms, with or without
00006  * modification, are permitted provided that the following conditions
00007  * are met:
00008  * 1. Redistributions of source code must retain the above copyright
00009  *    notice, this list of conditions and the following disclaimer.
00010  * 2. Redistributions in binary form must reproduce the above copyright
00011  *    notice, this list of conditions and the following disclaimer in the
00012  *    documentation and/or other materials provided with the distribution.
00013  * 3. Neither the name of the MIPS Technologies, Inc., nor the names of is
00014  *    contributors may be used to endorse or promote products derived from
00015  *    this software without specific prior written permission.
00016  *
00017  * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND
00018  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
00019  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
00020  * ARE DISCLAIMED.  IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE
00021  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
00022  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
00023  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
00024  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
00025  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
00026  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
00027  * SUCH DAMAGE.
00028  *
00029  * Author:  Zoran Lukic (zoranl@mips.com)
00030  *
00031  * This file is part of FFmpeg.
00032  *
00033  * FFmpeg is free software; you can redistribute it and/or
00034  * modify it under the terms of the GNU Lesser General Public
00035  * License as published by the Free Software Foundation; either
00036  * version 2.1 of the License, or (at your option) any later version.
00037  *
00038  * FFmpeg is distributed in the hope that it will be useful,
00039  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00040  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00041  * Lesser General Public License for more details.
00042  *
00043  * You should have received a copy of the GNU Lesser General Public
00044  * License along with FFmpeg; if not, write to the Free Software
00045  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00046  */
00047 #include "config.h"
00048 #include "libavcodec/dsputil.h"
00049 
00050 #if HAVE_INLINE_ASM
00051 static void vector_fmul_window_mips(float *dst, const float *src0,
00052         const float *src1, const float *win, int len)
00053 {
00054     int i, j;
00055     /*
00056      * variables used in inline assembler
00057      */
00058     float * dst_i, * dst_j, * dst_i2, * dst_j2;
00059     float temp, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
00060 
00061     dst  += len;
00062     win  += len;
00063     src0 += len;
00064 
00065     for (i = -len, j = len - 1; i < 0; i += 8, j -= 8) {
00066 
00067         dst_i = dst + i;
00068         dst_j = dst + j;
00069 
00070         dst_i2 = dst + i + 4;
00071         dst_j2 = dst + j - 4;
00072 
00073         __asm__ volatile (
00074             "mul.s   %[temp],   %[s1],       %[wi]            \n\t"
00075             "mul.s   %[temp1],  %[s1],       %[wj]            \n\t"
00076             "mul.s   %[temp2],  %[s11],      %[wi1]           \n\t"
00077             "mul.s   %[temp3],  %[s11],      %[wj1]           \n\t"
00078 
00079             "msub.s  %[temp],   %[temp],     %[s0],  %[wj]    \n\t"
00080             "madd.s  %[temp1],  %[temp1],    %[s0],  %[wi]    \n\t"
00081             "msub.s  %[temp2],  %[temp2],    %[s01], %[wj1]   \n\t"
00082             "madd.s  %[temp3],  %[temp3],    %[s01], %[wi1]   \n\t"
00083 
00084             "swc1    %[temp],   0(%[dst_i])                   \n\t" /* dst[i] = s0*wj - s1*wi; */
00085             "swc1    %[temp1],  0(%[dst_j])                   \n\t" /* dst[j] = s0*wi + s1*wj; */
00086             "swc1    %[temp2],  4(%[dst_i])                   \n\t" /* dst[i+1] = s01*wj1 - s11*wi1; */
00087             "swc1    %[temp3], -4(%[dst_j])                   \n\t" /* dst[j-1] = s01*wi1 + s11*wj1; */
00088 
00089             "mul.s   %[temp4],  %[s12],      %[wi2]           \n\t"
00090             "mul.s   %[temp5],  %[s12],      %[wj2]           \n\t"
00091             "mul.s   %[temp6],  %[s13],      %[wi3]           \n\t"
00092             "mul.s   %[temp7],  %[s13],      %[wj3]           \n\t"
00093 
00094             "msub.s  %[temp4],  %[temp4],    %[s02], %[wj2]   \n\t"
00095             "madd.s  %[temp5],  %[temp5],    %[s02], %[wi2]   \n\t"
00096             "msub.s  %[temp6],  %[temp6],    %[s03], %[wj3]   \n\t"
00097             "madd.s  %[temp7],  %[temp7],    %[s03], %[wi3]   \n\t"
00098 
00099             "swc1    %[temp4],  8(%[dst_i])                   \n\t" /* dst[i+2] = s02*wj2 - s12*wi2; */
00100             "swc1    %[temp5], -8(%[dst_j])                   \n\t" /* dst[j-2] = s02*wi2 + s12*wj2; */
00101             "swc1    %[temp6],  12(%[dst_i])                  \n\t" /* dst[i+2] = s03*wj3 - s13*wi3; */
00102             "swc1    %[temp7], -12(%[dst_j])                  \n\t" /* dst[j-3] = s03*wi3 + s13*wj3; */
00103             : [temp]"=&f"(temp),  [temp1]"=&f"(temp1), [temp2]"=&f"(temp2),
00104               [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5),
00105               [temp6]"=&f"(temp6), [temp7]"=&f"(temp7)
00106             : [dst_j]"r"(dst_j),     [dst_i]"r" (dst_i),
00107               [s0] "f"(src0[i]),     [wj] "f"(win[j]),     [s1] "f"(src1[j]),
00108               [wi] "f"(win[i]),      [s01]"f"(src0[i + 1]),[wj1]"f"(win[j - 1]),
00109               [s11]"f"(src1[j - 1]), [wi1]"f"(win[i + 1]), [s02]"f"(src0[i + 2]),
00110               [wj2]"f"(win[j - 2]),  [s12]"f"(src1[j - 2]),[wi2]"f"(win[i + 2]),
00111               [s03]"f"(src0[i + 3]), [wj3]"f"(win[j - 3]), [s13]"f"(src1[j - 3]),
00112               [wi3]"f"(win[i + 3])
00113             : "memory"
00114         );
00115 
00116         __asm__ volatile (
00117             "mul.s  %[temp],   %[s1],       %[wi]            \n\t"
00118             "mul.s  %[temp1],  %[s1],       %[wj]            \n\t"
00119             "mul.s  %[temp2],  %[s11],      %[wi1]           \n\t"
00120             "mul.s  %[temp3],  %[s11],      %[wj1]           \n\t"
00121 
00122             "msub.s %[temp],   %[temp],     %[s0],  %[wj]    \n\t"
00123             "madd.s %[temp1],  %[temp1],    %[s0],  %[wi]    \n\t"
00124             "msub.s %[temp2],  %[temp2],    %[s01], %[wj1]   \n\t"
00125             "madd.s %[temp3],  %[temp3],    %[s01], %[wi1]   \n\t"
00126 
00127             "swc1   %[temp],   0(%[dst_i2])                  \n\t" /* dst[i] = s0*wj - s1*wi; */
00128             "swc1   %[temp1],  0(%[dst_j2])                  \n\t" /* dst[j] = s0*wi + s1*wj; */
00129             "swc1   %[temp2],  4(%[dst_i2])                  \n\t" /* dst[i+1] = s01*wj1 - s11*wi1; */
00130             "swc1   %[temp3], -4(%[dst_j2])                  \n\t" /* dst[j-1] = s01*wi1 + s11*wj1; */
00131 
00132             "mul.s  %[temp4],  %[s12],      %[wi2]           \n\t"
00133             "mul.s  %[temp5],  %[s12],      %[wj2]           \n\t"
00134             "mul.s  %[temp6],  %[s13],      %[wi3]           \n\t"
00135             "mul.s  %[temp7],  %[s13],      %[wj3]           \n\t"
00136 
00137             "msub.s %[temp4],  %[temp4],    %[s02], %[wj2]   \n\t"
00138             "madd.s %[temp5],  %[temp5],    %[s02], %[wi2]   \n\t"
00139             "msub.s %[temp6],  %[temp6],    %[s03], %[wj3]   \n\t"
00140             "madd.s %[temp7],  %[temp7],    %[s03], %[wi3]   \n\t"
00141 
00142             "swc1   %[temp4],  8(%[dst_i2])                  \n\t" /* dst[i+2] = s02*wj2 - s12*wi2; */
00143             "swc1   %[temp5], -8(%[dst_j2])                  \n\t" /* dst[j-2] = s02*wi2 + s12*wj2; */
00144             "swc1   %[temp6],  12(%[dst_i2])                 \n\t" /* dst[i+2] = s03*wj3 - s13*wi3; */
00145             "swc1   %[temp7], -12(%[dst_j2])                 \n\t" /* dst[j-3] = s03*wi3 + s13*wj3; */
00146             : [temp]"=&f"(temp),
00147               [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), [temp3]"=&f"(temp3),
00148               [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), [temp6]"=&f"(temp6),
00149               [temp7]  "=&f" (temp7)
00150             : [dst_j2]"r"(dst_j2),   [dst_i2]"r"(dst_i2),
00151               [s0] "f"(src0[i + 4]), [wj] "f"(win[j - 4]), [s1] "f"(src1[j - 4]),
00152               [wi] "f"(win[i + 4]),  [s01]"f"(src0[i + 5]),[wj1]"f"(win[j - 5]),
00153               [s11]"f"(src1[j - 5]), [wi1]"f"(win[i + 5]), [s02]"f"(src0[i + 6]),
00154               [wj2]"f"(win[j - 6]),  [s12]"f"(src1[j - 6]),[wi2]"f"(win[i + 6]),
00155               [s03]"f"(src0[i + 7]), [wj3]"f"(win[j - 7]), [s13]"f"(src1[j - 7]),
00156               [wi3]"f"(win[i + 7])
00157             : "memory"
00158         );
00159     }
00160 }
00161 #endif /* HAVE_INLINE_ASM */
00162 
00163 av_cold void ff_dsputil_init_mips( DSPContext* c, AVCodecContext *avctx )
00164 {
00165 #if HAVE_INLINE_ASM
00166     c->vector_fmul_window = vector_fmul_window_mips;
00167 #endif
00168 }
libavcodec/mips/dsputil_mips.c
