1/*** 2 This file is part of PulseAudio. 3 4 Copyright 2004-2006 Lennart Poettering 5 Copyright 2009 Wim Taymans <wim.taymans@collabora.co.uk> 6 7 PulseAudio is free software; you can redistribute it and/or modify 8 it under the terms of the GNU Lesser General Public License as published 9 by the Free Software Foundation; either version 2.1 of the License, 10 or (at your option) any later version. 11 12 PulseAudio is distributed in the hope that it will be useful, but 13 WITHOUT ANY WARRANTY; without even the implied warranty of 14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 15 General Public License for more details. 16 17 You should have received a copy of the GNU Lesser General Public License 18 along with PulseAudio; if not, see <http://www.gnu.org/licenses/>. 19***/ 20 21#ifdef HAVE_CONFIG_H 22#include <config.h> 23#endif 24 25#include <pulse/rtclock.h> 26 27#include <pulsecore/random.h> 28#include <pulsecore/macro.h> 29#include <pulsecore/endianmacros.h> 30 31#include "cpu-x86.h" 32 33#include "sample-util.h" 34 35#if (!defined(__APPLE__) && !defined(__FreeBSD__) && !defined(__FreeBSD_kernel__) && defined (__i386__)) || defined (__amd64__) 36/* in s: 2 int16_t samples 37 * in v: 2 int32_t volumes, fixed point 16:16 38 * out s: contains scaled and clamped int16_t samples. 39 * 40 * We calculate the high 32 bits of a 32x16 multiply which we then 41 * clamp to 16 bits. The calculation is: 42 * 43 * vl = (v & 0xffff) 44 * vh = (v >> 16) 45 * s = ((s * vl) >> 16) + (s * vh); 46 * 47 * For the first multiply we have to do a sign correction as we need to 48 * multiply a signed int with an unsigned int. Hacker's delight 8-3 gives a 49 * simple formula to correct the sign of the high word after the signed 50 * multiply. 51 */ 52#define VOLUME_32x16(s,v) /* .. | vh | vl | */ \ 53 " pxor %%mm4, %%mm4 \n\t" /* .. | 0 | 0 | */ \ 54 " punpcklwd %%mm4, "#s" \n\t" /* .. | 0 | p0 | */ \ 55 " pcmpgtw "#v", %%mm4 \n\t" /* .. | 0 | s(vl) | */ \ 56 " pand "#s", %%mm4 \n\t" /* .. | 0 | (p0) | (vl >> 15) & p */ \ 57 " movq "#s", %%mm5 \n\t" \ 58 " pmulhw "#v", "#s" \n\t" /* .. | 0 | vl*p0 | */ \ 59 " paddw %%mm4, "#s" \n\t" /* .. | 0 | vl*p0 | + sign correct */ \ 60 " pslld $16, "#s" \n\t" /* .. | vl*p0 | 0 | */ \ 61 " psrld $16, "#v" \n\t" /* .. | 0 | vh | */ \ 62 " psrad $16, "#s" \n\t" /* .. | vl*p0 | sign extend */ \ 63 " pmaddwd %%mm5, "#v" \n\t" /* .. | p0 * vh | */ \ 64 " paddd "#s", "#v" \n\t" /* .. | p0 * v0 | */ \ 65 " packssdw "#v", "#v" \n\t" /* .. | p1*v1 | p0*v0 | */ 66 67/* approximately advances %3 = (%3 + a) % b. This function requires that 68 * a <= b. */ 69#define MOD_ADD(a,b) \ 70 " add "#a", %3 \n\t" \ 71 " mov %3, %4 \n\t" \ 72 " sub "#b", %4 \n\t" \ 73 " cmovae %4, %3 \n\t" 74 75/* swap 16 bits */ 76#define SWAP_16(s) \ 77 " movq "#s", %%mm4 \n\t" /* .. | h l | */ \ 78 " psrlw $8, %%mm4 \n\t" /* .. | 0 h | */ \ 79 " psllw $8, "#s" \n\t" /* .. | l 0 | */ \ 80 " por %%mm4, "#s" \n\t" /* .. | l h | */ 81 82/* swap 2 registers 16 bits for better pairing */ 83#define SWAP_16_2(s1,s2) \ 84 " movq "#s1", %%mm4 \n\t" /* .. | h l | */ \ 85 " movq "#s2", %%mm5 \n\t" \ 86 " psrlw $8, %%mm4 \n\t" /* .. | 0 h | */ \ 87 " psrlw $8, %%mm5 \n\t" \ 88 " psllw $8, "#s1" \n\t" /* .. | l 0 | */ \ 89 " psllw $8, "#s2" \n\t" \ 90 " por %%mm4, "#s1" \n\t" /* .. | l h | */ \ 91 " por %%mm5, "#s2" \n\t" 92 93static void pa_volume_s16ne_mmx(int16_t *samples, const int32_t *volumes, unsigned channels, unsigned length) { 94 pa_reg_x86 channel, temp; 95 96 /* Channels must be at least 4, and always a multiple of the original number. 97 * This is also the max amount we overread the volume array, which should 98 * have enough padding. */ 99 channels = channels == 3 ? 6 : PA_MAX (4U, channels); 100 101 __asm__ __volatile__ ( 102 " xor %3, %3 \n\t" 103 " sar $1, %2 \n\t" /* length /= sizeof (int16_t) */ 104 105 " test $1, %2 \n\t" /* check for odd samples */ 106 " je 2f \n\t" 107 108 " movd (%q1, %3, 4), %%mm0 \n\t" /* | v0h | v0l | */ 109 " movw (%0), %w4 \n\t" /* .. | p0 | */ 110 " movd %4, %%mm1 \n\t" 111 VOLUME_32x16 (%%mm1, %%mm0) 112 " movd %%mm0, %4 \n\t" /* .. | p0*v0 | */ 113 " movw %w4, (%0) \n\t" 114 " add $2, %0 \n\t" 115 MOD_ADD ($1, %5) 116 117 "2: \n\t" 118 " sar $1, %2 \n\t" /* prepare for processing 2 samples at a time */ 119 " test $1, %2 \n\t" /* check for odd samples */ 120 " je 4f \n\t" 121 122 "3: \n\t" /* do samples in groups of 2 */ 123 " movq (%q1, %3, 4), %%mm0 \n\t" /* | v1h | v1l | v0h | v0l | */ 124 " movd (%0), %%mm1 \n\t" /* .. | p1 | p0 | */ 125 VOLUME_32x16 (%%mm1, %%mm0) 126 " movd %%mm0, (%0) \n\t" /* .. | p1*v1 | p0*v0 | */ 127 " add $4, %0 \n\t" 128 MOD_ADD ($2, %5) 129 130 "4: \n\t" 131 " sar $1, %2 \n\t" /* prepare for processing 4 samples at a time */ 132 " cmp $0, %2 \n\t" 133 " je 6f \n\t" 134 135 "5: \n\t" /* do samples in groups of 4 */ 136 " movq (%q1, %3, 4), %%mm0 \n\t" /* | v1h | v1l | v0h | v0l | */ 137 " movq 8(%q1, %3, 4), %%mm2 \n\t" /* | v3h | v3l | v2h | v2l | */ 138 " movd (%0), %%mm1 \n\t" /* .. | p1 | p0 | */ 139 " movd 4(%0), %%mm3 \n\t" /* .. | p3 | p2 | */ 140 VOLUME_32x16 (%%mm1, %%mm0) 141 VOLUME_32x16 (%%mm3, %%mm2) 142 " movd %%mm0, (%0) \n\t" /* .. | p1*v1 | p0*v0 | */ 143 " movd %%mm2, 4(%0) \n\t" /* .. | p3*v3 | p2*v2 | */ 144 " add $8, %0 \n\t" 145 MOD_ADD ($4, %5) 146 " dec %2 \n\t" 147 " jne 5b \n\t" 148 149 "6: \n\t" 150 " emms \n\t" 151 152 : "+r" (samples), "+r" (volumes), "+r" (length), "=&D" (channel), "=&r" (temp) 153#if defined (__i386__) 154 : "m" (channels) 155#else 156 : "r" ((pa_reg_x86)channels) 157#endif 158 : "cc" 159 ); 160} 161 162static void pa_volume_s16re_mmx(int16_t *samples, const int32_t *volumes, unsigned channels, unsigned length) { 163 pa_reg_x86 channel, temp; 164 165 /* Channels must be at least 4, and always a multiple of the original number. 166 * This is also the max amount we overread the volume array, which should 167 * have enough padding. */ 168 channels = channels == 3 ? 6 : PA_MAX (4U, channels); 169 170 __asm__ __volatile__ ( 171 " xor %3, %3 \n\t" 172 " sar $1, %2 \n\t" /* length /= sizeof (int16_t) */ 173 " pcmpeqw %%mm6, %%mm6 \n\t" /* .. | ffff | ffff | */ 174 " pcmpeqw %%mm7, %%mm7 \n\t" /* .. | ffff | ffff | */ 175 " pslld $16, %%mm6 \n\t" /* .. | ffff | 0 | */ 176 " psrld $31, %%mm7 \n\t" /* .. | 0 | 1 | */ 177 178 " test $1, %2 \n\t" /* check for odd samples */ 179 " je 2f \n\t" 180 181 " movd (%q1, %3, 4), %%mm0 \n\t" /* | v0h | v0l | */ 182 " movw (%0), %w4 \n\t" /* .. | p0 | */ 183 " rorw $8, %w4 \n\t" 184 " movd %4, %%mm1 \n\t" 185 VOLUME_32x16 (%%mm1, %%mm0) 186 " movd %%mm0, %4 \n\t" /* .. | p0*v0 | */ 187 " rorw $8, %w4 \n\t" 188 " movw %w4, (%0) \n\t" 189 " add $2, %0 \n\t" 190 MOD_ADD ($1, %5) 191 192 "2: \n\t" 193 " sar $1, %2 \n\t" /* prepare for processing 2 samples at a time */ 194 " test $1, %2 \n\t" /* check for odd samples */ 195 " je 4f \n\t" 196 197 "3: \n\t" /* do samples in groups of 2 */ 198 " movq (%q1, %3, 4), %%mm0 \n\t" /* | v1h | v1l | v0h | v0l | */ 199 " movd (%0), %%mm1 \n\t" /* .. | p1 | p0 | */ 200 SWAP_16 (%%mm1) 201 VOLUME_32x16 (%%mm1, %%mm0) 202 SWAP_16 (%%mm0) 203 " movd %%mm0, (%0) \n\t" /* .. | p1*v1 | p0*v0 | */ 204 " add $4, %0 \n\t" 205 MOD_ADD ($2, %5) 206 207 "4: \n\t" 208 " sar $1, %2 \n\t" /* prepare for processing 4 samples at a time */ 209 " cmp $0, %2 \n\t" 210 " je 6f \n\t" 211 212 "5: \n\t" /* do samples in groups of 4 */ 213 " movq (%q1, %3, 4), %%mm0 \n\t" /* | v1h | v1l | v0h | v0l | */ 214 " movq 8(%q1, %3, 4), %%mm2 \n\t" /* | v3h | v3l | v2h | v2l | */ 215 " movd (%0), %%mm1 \n\t" /* .. | p1 | p0 | */ 216 " movd 4(%0), %%mm3 \n\t" /* .. | p3 | p2 | */ 217 SWAP_16_2 (%%mm1, %%mm3) 218 VOLUME_32x16 (%%mm1, %%mm0) 219 VOLUME_32x16 (%%mm3, %%mm2) 220 SWAP_16_2 (%%mm0, %%mm2) 221 " movd %%mm0, (%0) \n\t" /* .. | p1*v1 | p0*v0 | */ 222 " movd %%mm2, 4(%0) \n\t" /* .. | p3*v3 | p2*v2 | */ 223 " add $8, %0 \n\t" 224 MOD_ADD ($4, %5) 225 " dec %2 \n\t" 226 " jne 5b \n\t" 227 228 "6: \n\t" 229 " emms \n\t" 230 231 : "+r" (samples), "+r" (volumes), "+r" (length), "=&D" (channel), "=&r" (temp) 232#if defined (__i386__) 233 : "m" (channels) 234#else 235 : "r" ((pa_reg_x86)channels) 236#endif 237 : "cc" 238 ); 239} 240 241#endif /* (!defined(__APPLE__) && !defined(__FreeBSD__) && !defined(__FreeBSD_kernel__) && defined (__i386__)) || defined (__amd64__) */ 242 243void pa_volume_func_init_mmx(pa_cpu_x86_flag_t flags) { 244#if (!defined(__APPLE__) && !defined(__FreeBSD__) && !defined(__FreeBSD_kernel__) && defined (__i386__)) || defined (__amd64__) 245 if ((flags & PA_CPU_X86_MMX) && (flags & PA_CPU_X86_CMOV)) { 246 pa_log_info("Initialising MMX optimized volume functions."); 247 248 pa_set_volume_func(PA_SAMPLE_S16NE, (pa_do_volume_func_t) pa_volume_s16ne_mmx); 249 pa_set_volume_func(PA_SAMPLE_S16RE, (pa_do_volume_func_t) pa_volume_s16re_mmx); 250 } 251#endif /* (!defined(__APPLE__) && !defined(__FreeBSD__) && !defined(__FreeBSD_kernel__) && defined (__i386__)) || defined (__amd64__) */ 252} 253