libavcodec/x86/mlpdsp.c

   1 /*
   2  * MLP DSP functions x86-optimized
   3  * Copyright (c) 2009 Ramiro Polla
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 #include "libavutil/x86_cpu.h"
  23 #include "libavcodec/dsputil.h"
  24 #include "libavcodec/mlp.h"
  25
  26 #if HAVE_7REGS && HAVE_TEN_OPERANDS
  27
  28 extern void ff_mlp_firorder_8;
  29 extern void ff_mlp_firorder_7;
  30 extern void ff_mlp_firorder_6;
  31 extern void ff_mlp_firorder_5;
  32 extern void ff_mlp_firorder_4;
  33 extern void ff_mlp_firorder_3;
  34 extern void ff_mlp_firorder_2;
  35 extern void ff_mlp_firorder_1;
  36 extern void ff_mlp_firorder_0;
  37
  38 extern void ff_mlp_iirorder_4;
  39 extern void ff_mlp_iirorder_3;
  40 extern void ff_mlp_iirorder_2;
  41 extern void ff_mlp_iirorder_1;
  42 extern void ff_mlp_iirorder_0;
  43
  44 static const void *firtable[9] = { &ff_mlp_firorder_0, &ff_mlp_firorder_1,
  45                                    &ff_mlp_firorder_2, &ff_mlp_firorder_3,
  46                                    &ff_mlp_firorder_4, &ff_mlp_firorder_5,
  47                                    &ff_mlp_firorder_6, &ff_mlp_firorder_7,
  48                                    &ff_mlp_firorder_8 };
  49 static const void *iirtable[5] = { &ff_mlp_iirorder_0, &ff_mlp_iirorder_1,
  50                                    &ff_mlp_iirorder_2, &ff_mlp_iirorder_3,
  51                                    &ff_mlp_iirorder_4 };
  52
  53 #if ARCH_X86_64
  54
  55 #define MLPMUL(label, offset, offs, offc)   \
  56     LABEL_MANGLE(label)":             \n\t" \
  57     "movslq "offset"+"offs"(%0), %%rax\n\t" \
  58     "movslq "offset"+"offc"(%1), %%rdx\n\t" \
  59     "imul                 %%rdx, %%rax\n\t" \
  60     "add                  %%rax, %%rsi\n\t"
  61
  62 #define FIRMULREG(label, offset, firc)\
  63     LABEL_MANGLE(label)":       \n\t" \
  64     "movslq "#offset"(%0), %%rax\n\t" \
  65     "imul        %"#firc", %%rax\n\t" \
  66     "add            %%rax, %%rsi\n\t"
  67
  68 #define CLEAR_ACCUM                   \
  69     "xor            %%rsi, %%rsi\n\t"
  70
  71 #define SHIFT_ACCUM                   \
  72     "shr     %%cl,         %%rsi\n\t"
  73
  74 #define ACCUM    "%%rdx"
  75 #define RESULT   "%%rsi"
  76 #define RESULT32 "%%esi"
  77
  78 #else /* if ARCH_X86_32 */
  79
  80 #define MLPMUL(label, offset, offs, offc)  \
  81     LABEL_MANGLE(label)":            \n\t" \
  82     "mov   "offset"+"offs"(%0), %%eax\n\t" \
  83     "imull "offset"+"offc"(%1)       \n\t" \
  84     "add                %%eax , %%esi\n\t" \
  85     "adc                %%edx , %%ecx\n\t"
  86
  87 #define FIRMULREG(label, offset, firc)  \
  88     MLPMUL(label, #offset, "0", "0")
  89
  90 #define CLEAR_ACCUM                  \
  91     "xor           %%esi, %%esi\n\t" \
  92     "xor           %%ecx, %%ecx\n\t"
  93
  94 #define SHIFT_ACCUM                  \
  95     "mov           %%ecx, %%edx\n\t" \
  96     "mov           %%esi, %%eax\n\t" \
  97     "movzbl        %7   , %%ecx\n\t" \
  98     "shrd    %%cl, %%edx, %%eax\n\t" \
  99
 100 #define ACCUM    "%%edx"
 101 #define RESULT   "%%eax"
 102 #define RESULT32 "%%eax"
 103
 104 #endif /* !ARCH_X86_64 */
 105
 106 #define BINC  AV_STRINGIFY(4* MAX_CHANNELS)
 107 #define IOFFS AV_STRINGIFY(4*(MAX_FIR_ORDER + MAX_BLOCKSIZE))
 108 #define IOFFC AV_STRINGIFY(4* MAX_FIR_ORDER)
 109
 110 #define FIRMUL(label, offset) MLPMUL(label, #offset,   "0",   "0")
 111 #define IIRMUL(label, offset) MLPMUL(label, #offset, IOFFS, IOFFC)
 112
 113 static void mlp_filter_channel_x86(int32_t *state, const int32_t *coeff,
 114                                    int firorder, int iirorder,
 115                                    unsigned int filter_shift, int32_t mask,
 116                                    int blocksize, int32_t *sample_buffer)
 117 {
 118     const void *firjump = firtable[firorder];
 119     const void *iirjump = iirtable[iirorder];
 120
 121     blocksize = -blocksize;
 122
 123     __asm__ volatile(
 124         "1:                           \n\t"
 125         CLEAR_ACCUM
 126         "jmp  *%5                     \n\t"
 127         FIRMUL   (ff_mlp_firorder_8, 0x1c   )
 128         FIRMUL   (ff_mlp_firorder_7, 0x18   )
 129         FIRMUL   (ff_mlp_firorder_6, 0x14   )
 130         FIRMUL   (ff_mlp_firorder_5, 0x10   )
 131         FIRMUL   (ff_mlp_firorder_4, 0x0c   )
 132         FIRMULREG(ff_mlp_firorder_3, 0x08,10)
 133         FIRMULREG(ff_mlp_firorder_2, 0x04, 9)
 134         FIRMULREG(ff_mlp_firorder_1, 0x00, 8)
 135         LABEL_MANGLE(ff_mlp_firorder_0)":\n\t"
 136         "jmp  *%6                     \n\t"
 137         IIRMUL   (ff_mlp_iirorder_4, 0x0c   )
 138         IIRMUL   (ff_mlp_iirorder_3, 0x08   )
 139         IIRMUL   (ff_mlp_iirorder_2, 0x04   )
 140         IIRMUL   (ff_mlp_iirorder_1, 0x00   )
 141         LABEL_MANGLE(ff_mlp_iirorder_0)":\n\t"
 142         SHIFT_ACCUM
 143         "mov  "RESULT"  ,"ACCUM"      \n\t"
 144         "add  (%2)      ,"RESULT"     \n\t"
 145         "and   %4       ,"RESULT"     \n\t"
 146         "sub   $4       ,  %0         \n\t"
 147         "mov  "RESULT32", (%0)        \n\t"
 148         "mov  "RESULT32", (%2)        \n\t"
 149         "add $"BINC"    ,  %2         \n\t"
 150         "sub  "ACCUM"   ,"RESULT"     \n\t"
 151         "mov  "RESULT32","IOFFS"(%0)  \n\t"
 152         "incl              %3         \n\t"
 153         "js 1b                        \n\t"
 154         : /* 0*/"+r"(state),
 155           /* 1*/"+r"(coeff),
 156           /* 2*/"+r"(sample_buffer),
 157 #if ARCH_X86_64
 158           /* 3*/"+r"(blocksize)
 159         : /* 4*/"r"((x86_reg)mask), /* 5*/"r"(firjump),
 160           /* 6*/"r"(iirjump)      , /* 7*/"c"(filter_shift)
 161         , /* 8*/"r"((int64_t)coeff[0])
 162         , /* 9*/"r"((int64_t)coeff[1])
 163         , /*10*/"r"((int64_t)coeff[2])
 164         : "rax", "rdx", "rsi"
 165 #else /* ARCH_X86_32 */
 166           /* 3*/"+m"(blocksize)
 167         : /* 4*/"m"(         mask), /* 5*/"m"(firjump),
 168           /* 6*/"m"(iirjump)      , /* 7*/"m"(filter_shift)
 169         : "eax", "edx", "esi", "ecx"
 170 #endif /* !ARCH_X86_64 */
 171     );
 172 }
 173
 174 #endif /* HAVE_7REGS && HAVE_TEN_OPERANDS */
 175
 176 void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx)
 177 {
 178 #if HAVE_7REGS && HAVE_TEN_OPERANDS
 179     c->mlp_filter_channel = mlp_filter_channel_x86;
 180 #endif
 181 }