reactos/lib/3rdparty/libmpg123/synth_x86_64_s32.S

   1 /*
   2         synth_x86_64_s32: SSE optimized synth for x86-64 (s32 output version)
   3
   4         copyright 1995-2009 by the mpg123 project - free software under the terms of the LGPL 2.1
   5         see COPYING and AUTHORS files in distribution or http://mpg123.org
   6         initially written by Taihei Monma
   7 */
   8
   9 #include "mangle.h"
  10
  11 #ifdef _WIN64
  12 /* short *window; */
  13 #define ARG0 %r10
  14 /* short *b0; */
  15 #define ARG1 %rdx
  16 /* short *samples; */
  17 #define ARG2 %r8
  18 /* int bo1; */
  19 #define ARG3 %r9
  20 #else
  21 /* real *window; */
  22 #define ARG0 %rdi
  23 /* real *b0; */
  24 #define ARG1 %rsi
  25 /* real *samples; */
  26 #define ARG2 %rdx
  27 /* int bo1; */
  28 #define ARG3 %rcx
  29 #endif
  30
  31 #define XMMREG_SCALE %xmm15  /* {65536.0, 65536.0, 65536.0, 65536.0} */
  32 #define XMMREG_MAX %xmm14  /* {32767.999, 32767.999, 32767.999, 32767.999} */
  33 #define XMMREG_MIN %xmm13  /* {-32768.0, -32768.0, -32768.0, -32768.0} */
  34 #define XMMREG_CLIP %xmm12
  35
  36 /*
  37         int synth_1to1_s32_x86_64_asm(real *window, real *b0, int32_t *samples, int bo1);
  38         return value: number of clipped samples
  39 */
  40
  41 #ifndef __APPLE__
  42         .section        .rodata
  43 #else
  44         .data
  45 #endif
  46         ALIGN32
  47 ASM_NAME(scale_s32):
  48         .long   1199570944
  49         .long   1199570944
  50         .long   1199570944
  51         .long   1199570944
  52         ALIGN16
  53 ASM_NAME(maxmin_s32):
  54         .long   1191182335
  55         .long   1191182335
  56         .long   1191182335
  57         .long   1191182335
  58         .long   -956301312
  59         .long   -956301312
  60         .long   -956301312
  61         .long   -956301312
  62         .text
  63         ALIGN16
  64 .globl ASM_NAME(synth_1to1_s32_x86_64_asm)
  65 ASM_NAME(synth_1to1_s32_x86_64_asm):
  66 #ifdef _WIN64 /* should save xmm6-15 */
  67         movq            %rcx, ARG0
  68         subq            $168, %rsp /* stack alignment + 10 xmm registers */
  69         movaps          %xmm6, (%rsp)
  70         movaps          %xmm7, 16(%rsp)
  71         movaps          %xmm8, 32(%rsp)
  72         movaps          %xmm9, 48(%rsp)
  73         movaps          %xmm10, 64(%rsp)
  74         movaps          %xmm11, 80(%rsp)
  75         movaps          %xmm12, 96(%rsp)
  76         movaps          %xmm13, 112(%rsp)
  77         movaps          %xmm14, 128(%rsp)
  78         movaps          %xmm15, 144(%rsp)
  79 #endif
  80
  81         leaq            ASM_NAME(scale_s32)(%rip), %rax
  82         movaps          (%rax), XMMREG_SCALE
  83         leaq            ASM_NAME(maxmin_s32)(%rip), %rax
  84         movaps          (%rax), XMMREG_MAX
  85         movaps          16(%rax), XMMREG_MIN
  86
  87         xorps           XMMREG_CLIP, XMMREG_CLIP
  88
  89         andq            $0xf, ARG3
  90         shlq            $2, ARG3
  91         leaq            64(ARG0), ARG0
  92         subq            ARG3, ARG0
  93
  94         movl            $4, %ecx
  95
  96         ALIGN16
  97 Loop_start_1:
  98         movups          (ARG0), %xmm8
  99         movups          16(ARG0), %xmm1
 100         movups          32(ARG0), %xmm2
 101         movups          48(ARG0), %xmm3
 102         movups          128(ARG0), %xmm9
 103         movups          144(ARG0), %xmm5
 104         movups          160(ARG0), %xmm6
 105         movups          176(ARG0), %xmm7
 106         mulps           (ARG1), %xmm8
 107         mulps           16(ARG1), %xmm1
 108         mulps           32(ARG1), %xmm2
 109         mulps           48(ARG1), %xmm3
 110         mulps           64(ARG1), %xmm9
 111         mulps           80(ARG1), %xmm5
 112         mulps           96(ARG1), %xmm6
 113         mulps           112(ARG1), %xmm7
 114
 115         addps           %xmm1, %xmm8
 116         addps           %xmm2, %xmm3
 117         addps           %xmm5, %xmm9
 118         addps           %xmm7, %xmm6
 119         addps           %xmm3, %xmm8
 120         addps           %xmm6, %xmm9
 121         leaq            256(ARG0), ARG0
 122         leaq            128(ARG1), ARG1
 123
 124         movups          (ARG0), %xmm10
 125         movups          16(ARG0), %xmm1
 126         movups          32(ARG0), %xmm2
 127         movups          48(ARG0), %xmm3
 128         movups          128(ARG0), %xmm11
 129         movups          144(ARG0), %xmm5
 130         movups          160(ARG0), %xmm6
 131         movups          176(ARG0), %xmm7
 132         mulps           (ARG1), %xmm10
 133         mulps           16(ARG1), %xmm1
 134         mulps           32(ARG1), %xmm2
 135         mulps           48(ARG1), %xmm3
 136         mulps           64(ARG1), %xmm11
 137         mulps           80(ARG1), %xmm5
 138         mulps           96(ARG1), %xmm6
 139         mulps           112(ARG1), %xmm7
 140
 141         addps           %xmm1, %xmm10
 142         addps           %xmm2, %xmm3
 143         addps           %xmm5, %xmm11
 144         addps           %xmm7, %xmm6
 145         addps           %xmm3, %xmm10
 146         addps           %xmm6, %xmm11
 147         leaq            256(ARG0), ARG0
 148         leaq            128(ARG1), ARG1
 149
 150         movaps          %xmm8, %xmm0
 151         movaps          %xmm10, %xmm1
 152         unpcklps        %xmm9, %xmm8
 153         unpcklps        %xmm11, %xmm10
 154         unpckhps        %xmm9, %xmm0
 155         unpckhps        %xmm11, %xmm1
 156         movaps          %xmm8, %xmm2
 157         movaps          %xmm0, %xmm3
 158         movlhps         %xmm10, %xmm8
 159         movhlps         %xmm2, %xmm10
 160         movlhps         %xmm1, %xmm0
 161         movhlps         %xmm3, %xmm1
 162         subps           %xmm10, %xmm8
 163         subps           %xmm1, %xmm0
 164         addps           %xmm8, %xmm0
 165
 166         movups          (ARG2), %xmm1
 167         movups          16(ARG2), %xmm2
 168         movaps          %xmm0, %xmm3
 169         movaps          %xmm0, %xmm4
 170         mulps           XMMREG_SCALE, %xmm0
 171         cmpnleps        XMMREG_MAX, %xmm3
 172         cmpltps         XMMREG_MIN, %xmm4
 173         cvtps2dq        %xmm0, %xmm0
 174         xorps           %xmm3, %xmm0
 175         shufps          $0xdd, %xmm2, %xmm1
 176         movaps          %xmm0, %xmm2
 177         unpcklps        %xmm1, %xmm0
 178         unpckhps        %xmm1, %xmm2
 179         movups          %xmm0, (ARG2)
 180         movups          %xmm2, 16(ARG2)
 181
 182         psrld           $31, %xmm3
 183         psrld           $31, %xmm4
 184         paddd           %xmm4, %xmm3
 185         paddd           %xmm3, XMMREG_CLIP
 186
 187         leaq            32(ARG2), ARG2
 188         decl            %ecx
 189         jnz                     Loop_start_1
 190
 191         movl            $4, %ecx
 192
 193         ALIGN16
 194 Loop_start_2:
 195         movups          (ARG0), %xmm8
 196         movups          16(ARG0), %xmm1
 197         movups          32(ARG0), %xmm2
 198         movups          48(ARG0), %xmm3
 199         movups          128(ARG0), %xmm9
 200         movups          144(ARG0), %xmm5
 201         movups          160(ARG0), %xmm6
 202         movups          176(ARG0), %xmm7
 203         mulps           (ARG1), %xmm8
 204         mulps           16(ARG1), %xmm1
 205         mulps           32(ARG1), %xmm2
 206         mulps           48(ARG1), %xmm3
 207         mulps           -64(ARG1), %xmm9
 208         mulps           -48(ARG1), %xmm5
 209         mulps           -32(ARG1), %xmm6
 210         mulps           -16(ARG1), %xmm7
 211
 212         addps           %xmm1, %xmm8
 213         addps           %xmm2, %xmm3
 214         addps           %xmm5, %xmm9
 215         addps           %xmm7, %xmm6
 216         addps           %xmm3, %xmm8
 217         addps           %xmm6, %xmm9
 218         leaq            256(ARG0), ARG0
 219         leaq            -128(ARG1), ARG1
 220
 221         movups          (ARG0), %xmm10
 222         movups          16(ARG0), %xmm1
 223         movups          32(ARG0), %xmm2
 224         movups          48(ARG0), %xmm3
 225         movups          128(ARG0), %xmm11
 226         movups          144(ARG0), %xmm5
 227         movups          160(ARG0), %xmm6
 228         movups          176(ARG0), %xmm7
 229         mulps           (ARG1), %xmm10
 230         mulps           16(ARG1), %xmm1
 231         mulps           32(ARG1), %xmm2
 232         mulps           48(ARG1), %xmm3
 233         mulps           -64(ARG1), %xmm11
 234         mulps           -48(ARG1), %xmm5
 235         mulps           -32(ARG1), %xmm6
 236         mulps           -16(ARG1), %xmm7
 237
 238         addps           %xmm1, %xmm10
 239         addps           %xmm2, %xmm3
 240         addps           %xmm5, %xmm11
 241         addps           %xmm7, %xmm6
 242         addps           %xmm3, %xmm10
 243         addps           %xmm6, %xmm11
 244         leaq            256(ARG0), ARG0
 245         leaq            -128(ARG1), ARG1
 246
 247         movaps          %xmm8, %xmm0
 248         movaps          %xmm10, %xmm1
 249         unpcklps        %xmm9, %xmm8
 250         unpcklps        %xmm11, %xmm10
 251         unpckhps        %xmm9, %xmm0
 252         unpckhps        %xmm11, %xmm1
 253         movaps          %xmm8, %xmm2
 254         movaps          %xmm0, %xmm3
 255         movlhps         %xmm10, %xmm8
 256         movhlps         %xmm2, %xmm10
 257         movlhps         %xmm1, %xmm0
 258         movhlps         %xmm3, %xmm1
 259         addps           %xmm10, %xmm8
 260         addps           %xmm1, %xmm0
 261         addps           %xmm8, %xmm0
 262
 263         movups          (ARG2), %xmm1
 264         movups          16(ARG2), %xmm2
 265         movaps          %xmm0, %xmm3
 266         movaps          %xmm0, %xmm4
 267         mulps           XMMREG_SCALE, %xmm0
 268         cmpnleps        XMMREG_MAX, %xmm3
 269         cmpltps         XMMREG_MIN, %xmm4
 270         cvtps2dq        %xmm0, %xmm0
 271         xorps           %xmm3, %xmm0
 272         shufps          $0xdd, %xmm2, %xmm1
 273         movaps          %xmm0, %xmm2
 274         unpcklps        %xmm1, %xmm0
 275         unpckhps        %xmm1, %xmm2
 276         movups          %xmm0, (ARG2)
 277         movups          %xmm2, 16(ARG2)
 278
 279         psrld           $31, %xmm3
 280         psrld           $31, %xmm4
 281         paddd           %xmm4, %xmm3
 282         paddd           %xmm3, XMMREG_CLIP
 283
 284         leaq            32(ARG2), ARG2
 285         decl            %ecx
 286         jnz                     Loop_start_2
 287
 288         pshuflw         $0xee, XMMREG_CLIP, %xmm0
 289         movhlps         XMMREG_CLIP, %xmm1
 290         pshuflw         $0xee, %xmm1, %xmm2
 291         paddd           %xmm0, XMMREG_CLIP
 292         paddd           %xmm1, XMMREG_CLIP
 293         paddd           %xmm2, XMMREG_CLIP
 294
 295         movd            XMMREG_CLIP, %eax
 296
 297 #ifdef _WIN64
 298         movaps          (%rsp), %xmm6
 299         movaps          16(%rsp), %xmm7
 300         movaps          32(%rsp), %xmm8
 301         movaps          48(%rsp), %xmm9
 302         movaps          64(%rsp), %xmm10
 303         movaps          80(%rsp), %xmm11
 304         movaps          96(%rsp), %xmm12
 305         movaps          112(%rsp), %xmm13
 306         movaps          128(%rsp), %xmm14
 307         movaps          144(%rsp), %xmm15
 308         addq            $168, %rsp
 309 #endif
 310         ret
 311
 312 NONEXEC_STACK