| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | * SIMD-optimized LPC functions | ||
| 3 | * Copyright (c) 2007 Loren Merritt | ||
| 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/attributes.h" | ||
| 23 | #include "libavutil/x86/asm.h" | ||
| 24 | #include "libavutil/x86/cpu.h" | ||
| 25 | #include "libavcodec/lpc.h" | ||
| 26 | |||
| 27 | void ff_lpc_apply_welch_window_sse2(const int32_t *data, ptrdiff_t len, | ||
| 28 | double *w_data); | ||
| 29 | void ff_lpc_apply_welch_window_avx2(const int32_t *data, ptrdiff_t len, | ||
| 30 | double *w_data); | ||
| 31 | |||
| 32 | DECLARE_ASM_CONST(16, double, pd_1)[2] = { 1.0, 1.0 }; | ||
| 33 | |||
| 34 | #if HAVE_SSE2_INLINE | ||
| 35 | |||
| 36 | 5200 | static void lpc_compute_autocorr_sse2(const double *data, ptrdiff_t len, int lag, | |
| 37 | double *autoc) | ||
| 38 | { | ||
| 39 | int j; | ||
| 40 | |||
| 41 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 5200 times.
|
5200 | if((x86_reg)data & 15) |
| 42 | ✗ | data++; | |
| 43 | |||
| 44 |
2/2✓ Branch 0 taken 20800 times.
✓ Branch 1 taken 5200 times.
|
26000 | for(j=0; j<lag; j+=2){ |
| 45 | 20800 | x86_reg i = -len*sizeof(double); | |
| 46 |
2/2✓ Branch 0 taken 5200 times.
✓ Branch 1 taken 15600 times.
|
20800 | if(j == lag-2) { |
| 47 | 5200 | __asm__ volatile( | |
| 48 | "movsd "MANGLE(pd_1)", %%xmm0 \n\t" | ||
| 49 | "movsd "MANGLE(pd_1)", %%xmm1 \n\t" | ||
| 50 | "movsd "MANGLE(pd_1)", %%xmm2 \n\t" | ||
| 51 | "1: \n\t" | ||
| 52 | "movapd (%2,%0), %%xmm3 \n\t" | ||
| 53 | "movupd -8(%3,%0), %%xmm4 \n\t" | ||
| 54 | "movapd (%3,%0), %%xmm5 \n\t" | ||
| 55 | "mulpd %%xmm3, %%xmm4 \n\t" | ||
| 56 | "mulpd %%xmm3, %%xmm5 \n\t" | ||
| 57 | "mulpd -16(%3,%0), %%xmm3 \n\t" | ||
| 58 | "addpd %%xmm4, %%xmm1 \n\t" | ||
| 59 | "addpd %%xmm5, %%xmm0 \n\t" | ||
| 60 | "addpd %%xmm3, %%xmm2 \n\t" | ||
| 61 | "add $16, %0 \n\t" | ||
| 62 | "jl 1b \n\t" | ||
| 63 | "movhlps %%xmm0, %%xmm3 \n\t" | ||
| 64 | "movhlps %%xmm1, %%xmm4 \n\t" | ||
| 65 | "movhlps %%xmm2, %%xmm5 \n\t" | ||
| 66 | "addsd %%xmm3, %%xmm0 \n\t" | ||
| 67 | "addsd %%xmm4, %%xmm1 \n\t" | ||
| 68 | "addsd %%xmm5, %%xmm2 \n\t" | ||
| 69 | "movsd %%xmm0, (%1) \n\t" | ||
| 70 | "movsd %%xmm1, 8(%1) \n\t" | ||
| 71 | "movsd %%xmm2, 16(%1) \n\t" | ||
| 72 | :"+&r"(i) | ||
| 73 | 5200 | :"r"(autoc+j), "r"(data+len), "r"(data+len-j) | |
| 74 | NAMED_CONSTRAINTS_ARRAY_ADD(pd_1) | ||
| 75 | :"memory" | ||
| 76 | ); | ||
| 77 | } else { | ||
| 78 | 15600 | __asm__ volatile( | |
| 79 | "movsd "MANGLE(pd_1)", %%xmm0 \n\t" | ||
| 80 | "movsd "MANGLE(pd_1)", %%xmm1 \n\t" | ||
| 81 | "1: \n\t" | ||
| 82 | "movapd (%3,%0), %%xmm3 \n\t" | ||
| 83 | "movupd -8(%4,%0), %%xmm4 \n\t" | ||
| 84 | "mulpd %%xmm3, %%xmm4 \n\t" | ||
| 85 | "mulpd (%4,%0), %%xmm3 \n\t" | ||
| 86 | "addpd %%xmm4, %%xmm1 \n\t" | ||
| 87 | "addpd %%xmm3, %%xmm0 \n\t" | ||
| 88 | "add $16, %0 \n\t" | ||
| 89 | "jl 1b \n\t" | ||
| 90 | "movhlps %%xmm0, %%xmm3 \n\t" | ||
| 91 | "movhlps %%xmm1, %%xmm4 \n\t" | ||
| 92 | "addsd %%xmm3, %%xmm0 \n\t" | ||
| 93 | "addsd %%xmm4, %%xmm1 \n\t" | ||
| 94 | "movsd %%xmm0, %1 \n\t" | ||
| 95 | "movsd %%xmm1, %2 \n\t" | ||
| 96 | 15600 | :"+&r"(i), "=m"(autoc[j]), "=m"(autoc[j+1]) | |
| 97 | 15600 | :"r"(data+len), "r"(data+len-j) | |
| 98 | NAMED_CONSTRAINTS_ARRAY_ADD(pd_1) | ||
| 99 | ); | ||
| 100 | } | ||
| 101 | } | ||
| 102 | 5200 | } | |
| 103 | |||
| 104 | #endif /* HAVE_SSE2_INLINE */ | ||
| 105 | |||
| 106 | 155 | av_cold void ff_lpc_init_x86(LPCContext *c) | |
| 107 | { | ||
| 108 | 155 | int cpu_flags = av_get_cpu_flags(); | |
| 109 | |||
| 110 | #if HAVE_SSE2_INLINE | ||
| 111 |
2/2✓ Branch 0 taken 25 times.
✓ Branch 1 taken 130 times.
|
155 | if (INLINE_SSE2_SLOW(cpu_flags)) |
| 112 | 25 | c->lpc_compute_autocorr = lpc_compute_autocorr_sse2; | |
| 113 | #endif | ||
| 114 | |||
| 115 |
2/2✓ Branch 0 taken 25 times.
✓ Branch 1 taken 130 times.
|
155 | if (EXTERNAL_SSE2(cpu_flags)) |
| 116 | 25 | c->lpc_apply_welch_window = ff_lpc_apply_welch_window_sse2; | |
| 117 | |||
| 118 |
2/2✓ Branch 0 taken 17 times.
✓ Branch 1 taken 138 times.
|
155 | if (EXTERNAL_AVX2(cpu_flags)) |
| 119 | 17 | c->lpc_apply_welch_window = ff_lpc_apply_welch_window_avx2; | |
| 120 | 155 | } | |
| 121 |