| 1 |
/* |
| 2 |
synth_stereo_avx: AVX optimized synth for x86-64 (stereo specific version) |
| 3 |
|
| 4 |
copyright 1995-2026 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 "x86_64_defs.h" |
| 10 |
#include "mangle.h" |
| 11 |
|
| 12 |
#ifdef IS_MSABI |
| 13 |
/* short *window; */ |
| 14 |
#define WINDOW %r10 |
| 15 |
/* short *b0l; */ |
| 16 |
#define B0L %rdx |
| 17 |
/* short *b0r; */ |
| 18 |
#define B0R %r8 |
| 19 |
/* short *samples; */ |
| 20 |
#define SAMPLES %r9 |
| 21 |
#else |
| 22 |
/* short *window; */ |
| 23 |
#define WINDOW %rdi |
| 24 |
/* short *b0l; */ |
| 25 |
#define B0L %rsi |
| 26 |
/* short *b0r; */ |
| 27 |
#define B0R %rdx |
| 28 |
/* short *samples; */ |
| 29 |
#define SAMPLES %r9 |
| 30 |
#endif |
| 31 |
|
| 32 |
/* |
| 33 |
int synth_1to1_s_avx_asm(short *window, short *b0l, short *b0r, short *samples, int bo1); |
| 34 |
return value: number of clipped samples |
| 35 |
*/ |
| 36 |
|
| 37 |
ALIGN16 |
| 38 |
.globl ASM_NAME(INT123_synth_1to1_s_avx_asm) |
| 39 |
ASM_NAME(INT123_synth_1to1_s_avx_asm): |
| 40 |
#ifdef IS_MSABI /* should save xmm6-15 */ |
| 41 |
push %rbp |
| 42 |
mov %rsp, %rbp |
| 43 |
sub $144, %rsp |
| 44 |
movaps %xmm6, (%rsp) |
| 45 |
movaps %xmm7, 16(%rsp) |
| 46 |
movaps %xmm8, 32(%rsp) |
| 47 |
movaps %xmm9, 48(%rsp) |
| 48 |
movaps %xmm10, 64(%rsp) |
| 49 |
movaps %xmm11, 80(%rsp) |
| 50 |
movaps %xmm12, 96(%rsp) |
| 51 |
movaps %xmm13, 112(%rsp) |
| 52 |
movaps %xmm14, 128(%rsp) |
| 53 |
movl 48(%rbp), %eax /* 5th argument; placed after 32-byte shadow space */ |
| 54 |
#endif |
| 55 |
|
| 56 |
#ifdef IS_MSABI |
| 57 |
shl $1, %eax |
| 58 |
mov %rcx, WINDOW |
| 59 |
#else |
| 60 |
mov %r8d, %eax |
| 61 |
shl $1, %eax |
| 62 |
movq %rcx, SAMPLES |
| 63 |
#endif |
| 64 |
add $32, WINDOW |
| 65 |
sub %rax, WINDOW |
| 66 |
|
| 67 |
mov $64, %rax |
| 68 |
movl $4, %ecx |
| 69 |
vpxor %xmm14, %xmm14, %xmm14 |
| 70 |
|
| 71 |
ALIGN16 |
| 72 |
1: |
| 73 |
movups (WINDOW), %xmm8 |
| 74 |
movups 16(WINDOW), %xmm9 |
| 75 |
movups (WINDOW,%rax), %xmm10 |
| 76 |
movups 16(WINDOW,%rax), %xmm11 |
| 77 |
vpmaddwd (B0L), %xmm8, %xmm0 |
| 78 |
vpmaddwd 16(B0L), %xmm9, %xmm1 |
| 79 |
vpmaddwd (B0R), %xmm8, %xmm2 |
| 80 |
vpmaddwd 16(B0R), %xmm9, %xmm3 |
| 81 |
vpmaddwd 32(B0L), %xmm10, %xmm4 |
| 82 |
vpmaddwd 48(B0L), %xmm11, %xmm5 |
| 83 |
vpmaddwd 32(B0R), %xmm10, %xmm6 |
| 84 |
vpmaddwd 48(B0R), %xmm11, %xmm7 |
| 85 |
vpaddd %xmm1, %xmm0, %xmm8 |
| 86 |
vpaddd %xmm3, %xmm2, %xmm0 |
| 87 |
vpaddd %xmm5, %xmm4, %xmm9 |
| 88 |
vpaddd %xmm7, %xmm6, %xmm1 |
| 89 |
lea (WINDOW,%rax,2), WINDOW |
| 90 |
add %rax, B0L |
| 91 |
add %rax, B0R |
| 92 |
|
| 93 |
movups (WINDOW), %xmm10 |
| 94 |
movups 16(WINDOW), %xmm11 |
| 95 |
movups (WINDOW,%rax), %xmm12 |
| 96 |
movups 16(WINDOW,%rax), %xmm13 |
| 97 |
vpmaddwd (B0L), %xmm10, %xmm2 |
| 98 |
vpmaddwd 16(B0L), %xmm11, %xmm3 |
| 99 |
vpmaddwd (B0R), %xmm10, %xmm4 |
| 100 |
vpmaddwd 16(B0R), %xmm11, %xmm5 |
| 101 |
vpmaddwd 32(B0L), %xmm12, %xmm6 |
| 102 |
vpmaddwd 48(B0L), %xmm13, %xmm10 |
| 103 |
vpmaddwd 32(B0R), %xmm12, %xmm7 |
| 104 |
vpmaddwd 48(B0R), %xmm13, %xmm11 |
| 105 |
vpaddd %xmm3, %xmm2, %xmm2 |
| 106 |
vpaddd %xmm5, %xmm4, %xmm3 |
| 107 |
vpaddd %xmm6, %xmm10, %xmm4 |
| 108 |
vpaddd %xmm7, %xmm11, %xmm5 |
| 109 |
lea (WINDOW,%rax,2), WINDOW |
| 110 |
add %rax, B0L |
| 111 |
add %rax, B0R |
| 112 |
|
| 113 |
vpunpckldq %xmm0, %xmm8, %xmm6 |
| 114 |
vpunpckhdq %xmm0, %xmm8, %xmm0 |
| 115 |
vpunpckldq %xmm1, %xmm9, %xmm7 |
| 116 |
vpunpckhdq %xmm1, %xmm9, %xmm1 |
| 117 |
vpaddd %xmm6, %xmm0, %xmm0 |
| 118 |
vpaddd %xmm7, %xmm1, %xmm1 |
| 119 |
vpunpckldq %xmm3, %xmm2, %xmm6 |
| 120 |
vpunpckhdq %xmm3, %xmm2, %xmm2 |
| 121 |
vpunpckldq %xmm5, %xmm4, %xmm7 |
| 122 |
vpunpckhdq %xmm5, %xmm4, %xmm3 |
| 123 |
vpaddd %xmm6, %xmm2, %xmm2 |
| 124 |
vpaddd %xmm7, %xmm3, %xmm3 |
| 125 |
|
| 126 |
vpunpcklqdq %xmm1, %xmm0, %xmm4 |
| 127 |
vpunpckhqdq %xmm1, %xmm0, %xmm0 |
| 128 |
vpunpcklqdq %xmm3, %xmm2, %xmm5 |
| 129 |
vpunpckhqdq %xmm3, %xmm2, %xmm1 |
| 130 |
vpaddd %xmm0, %xmm4, %xmm0 |
| 131 |
vpaddd %xmm1, %xmm5, %xmm1 |
| 132 |
vpsrad $13, %xmm0, %xmm0 |
| 133 |
vpsrad $13, %xmm1, %xmm1 |
| 134 |
vpackssdw %xmm1, %xmm0, %xmm2 |
| 135 |
vpcmpeqd %xmm3, %xmm3, %xmm3 |
| 136 |
vpslld $16, %xmm0, %xmm0 |
| 137 |
vpslld $16, %xmm1, %xmm1 |
| 138 |
vpsrld $16, %xmm0, %xmm0 |
| 139 |
vpsrld $16, %xmm1, %xmm1 |
| 140 |
vpackusdw %xmm1, %xmm0, %xmm0 |
| 141 |
vpcmpeqw %xmm2, %xmm0, %xmm0 |
| 142 |
vpxor %xmm3, %xmm0, %xmm0 |
| 143 |
vpaddw %xmm0, %xmm14, %xmm14 |
| 144 |
|
| 145 |
movups %xmm2, (SAMPLES) |
| 146 |
add $16, SAMPLES |
| 147 |
dec %ecx |
| 148 |
jnz 1b |
| 149 |
|
| 150 |
movl $4, %ecx |
| 151 |
|
| 152 |
ALIGN16 |
| 153 |
1: |
| 154 |
movups (WINDOW), %xmm8 |
| 155 |
movups 16(WINDOW), %xmm9 |
| 156 |
movups (WINDOW,%rax), %xmm10 |
| 157 |
movups 16(WINDOW,%rax), %xmm11 |
| 158 |
vpmaddwd (B0L), %xmm8, %xmm0 |
| 159 |
vpmaddwd 16(B0L), %xmm9, %xmm1 |
| 160 |
vpmaddwd (B0R), %xmm8, %xmm2 |
| 161 |
vpmaddwd 16(B0R), %xmm9, %xmm3 |
| 162 |
vpmaddwd -32(B0L), %xmm10, %xmm4 |
| 163 |
vpmaddwd -16(B0L), %xmm11, %xmm5 |
| 164 |
vpmaddwd -32(B0R), %xmm10, %xmm6 |
| 165 |
vpmaddwd -16(B0R), %xmm11, %xmm7 |
| 166 |
vpaddd %xmm1, %xmm0, %xmm8 |
| 167 |
vpaddd %xmm3, %xmm2, %xmm0 |
| 168 |
vpaddd %xmm5, %xmm4, %xmm9 |
| 169 |
vpaddd %xmm7, %xmm6, %xmm1 |
| 170 |
lea (WINDOW,%rax,2), WINDOW |
| 171 |
sub %rax, B0L |
| 172 |
sub %rax, B0R |
| 173 |
|
| 174 |
movups (WINDOW), %xmm10 |
| 175 |
movups 16(WINDOW), %xmm11 |
| 176 |
movups (WINDOW,%rax), %xmm12 |
| 177 |
movups 16(WINDOW,%rax), %xmm13 |
| 178 |
vpmaddwd (B0L), %xmm10, %xmm2 |
| 179 |
vpmaddwd 16(B0L), %xmm11, %xmm3 |
| 180 |
vpmaddwd (B0R), %xmm10, %xmm4 |
| 181 |
vpmaddwd 16(B0R), %xmm11, %xmm5 |
| 182 |
vpmaddwd -32(B0L), %xmm12, %xmm6 |
| 183 |
vpmaddwd -16(B0L), %xmm13, %xmm10 |
| 184 |
vpmaddwd -32(B0R), %xmm12, %xmm7 |
| 185 |
vpmaddwd -16(B0R), %xmm13, %xmm11 |
| 186 |
vpaddd %xmm3, %xmm2, %xmm2 |
| 187 |
vpaddd %xmm5, %xmm4, %xmm3 |
| 188 |
vpaddd %xmm6, %xmm10, %xmm4 |
| 189 |
vpaddd %xmm7, %xmm11, %xmm5 |
| 190 |
lea (WINDOW,%rax,2), WINDOW |
| 191 |
sub %rax, B0L |
| 192 |
sub %rax, B0R |
| 193 |
|
| 194 |
vpunpckldq %xmm0, %xmm8, %xmm6 |
| 195 |
vpunpckhdq %xmm0, %xmm8, %xmm0 |
| 196 |
vpunpckldq %xmm1, %xmm9, %xmm7 |
| 197 |
vpunpckhdq %xmm1, %xmm9, %xmm1 |
| 198 |
vpaddd %xmm6, %xmm0, %xmm0 |
| 199 |
vpaddd %xmm7, %xmm1, %xmm1 |
| 200 |
vpunpckldq %xmm3, %xmm2, %xmm6 |
| 201 |
vpunpckhdq %xmm3, %xmm2, %xmm2 |
| 202 |
vpunpckldq %xmm5, %xmm4, %xmm7 |
| 203 |
vpunpckhdq %xmm5, %xmm4, %xmm3 |
| 204 |
vpaddd %xmm6, %xmm2, %xmm2 |
| 205 |
vpaddd %xmm7, %xmm3, %xmm3 |
| 206 |
|
| 207 |
vpunpcklqdq %xmm1, %xmm0, %xmm4 |
| 208 |
vpunpckhqdq %xmm1, %xmm0, %xmm0 |
| 209 |
vpunpcklqdq %xmm3, %xmm2, %xmm5 |
| 210 |
vpunpckhqdq %xmm3, %xmm2, %xmm1 |
| 211 |
vpaddd %xmm0, %xmm4, %xmm0 |
| 212 |
vpaddd %xmm1, %xmm5, %xmm1 |
| 213 |
vpsrad $13, %xmm0, %xmm0 |
| 214 |
vpsrad $13, %xmm1, %xmm1 |
| 215 |
vpackssdw %xmm1, %xmm0, %xmm2 |
| 216 |
vpcmpeqd %xmm3, %xmm3, %xmm3 |
| 217 |
vpslld $16, %xmm0, %xmm0 |
| 218 |
vpslld $16, %xmm1, %xmm1 |
| 219 |
vpsrld $16, %xmm0, %xmm0 |
| 220 |
vpsrld $16, %xmm1, %xmm1 |
| 221 |
vpackusdw %xmm1, %xmm0, %xmm0 |
| 222 |
vpcmpeqw %xmm2, %xmm0, %xmm0 |
| 223 |
vpxor %xmm3, %xmm0, %xmm0 |
| 224 |
vpaddw %xmm0, %xmm14, %xmm14 |
| 225 |
|
| 226 |
movups %xmm2, (SAMPLES) |
| 227 |
add $16, SAMPLES |
| 228 |
dec %ecx |
| 229 |
jnz 1b |
| 230 |
|
| 231 |
pxor %xmm1, %xmm1 |
| 232 |
psubw %xmm14, %xmm1 |
| 233 |
pshufd $0x4e, %xmm1, %xmm0 |
| 234 |
paddw %xmm1, %xmm0 |
| 235 |
pshuflw $0x4e, %xmm0, %xmm1 |
| 236 |
paddw %xmm1, %xmm0 |
| 237 |
pshuflw $0x11, %xmm0, %xmm1 |
| 238 |
paddw %xmm1, %xmm0 |
| 239 |
movd %xmm0, %eax |
| 240 |
and $0x7f, %eax |
| 241 |
|
| 242 |
#ifdef IS_MSABI |
| 243 |
movaps (%rsp), %xmm6 |
| 244 |
movaps 16(%rsp), %xmm7 |
| 245 |
movaps 32(%rsp), %xmm8 |
| 246 |
movaps 48(%rsp), %xmm9 |
| 247 |
movaps 64(%rsp), %xmm10 |
| 248 |
movaps 80(%rsp), %xmm11 |
| 249 |
movaps 96(%rsp), %xmm12 |
| 250 |
movaps 112(%rsp), %xmm13 |
| 251 |
movaps 128(%rsp), %xmm14 |
| 252 |
mov %rbp, %rsp |
| 253 |
pop %rbp |
| 254 |
#endif |
| 255 |
ret |
| 256 |
|
| 257 |
NONEXEC_STACK |