| 1 | ;****************************************************************************** |
| 2 | ;* FLAC DSP functions |
| 3 | ;* |
| 4 | ;* Copyright (c) 2014 James Darnley <james.darnley@gmail.com> |
| 5 | ;* |
| 6 | ;* This file is part of FFmpeg. |
| 7 | ;* |
| 8 | ;* FFmpeg is free software; you can redistribute it and/or modify |
| 9 | ;* it under the terms of the GNU General Public License as published by |
| 10 | ;* the Free Software Foundation; either version 2 of the License, or |
| 11 | ;* (at your option) any later version. |
| 12 | ;* |
| 13 | ;* FFmpeg is distributed in the hope that it will be useful, |
| 14 | ;* but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | ;* GNU General Public License for more details. |
| 17 | ;* |
| 18 | ;* You should have received a copy of the GNU General Public License along |
| 19 | ;* with FFmpeg; if not, write to the Free Software Foundation, Inc., |
| 20 | ;* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. |
| 21 | ;****************************************************************************** |
| 22 | |
| 23 | %include "libavutil/x86/x86util.asm" |
| 24 | |
| 25 | SECTION_TEXT |
| 26 | |
| 27 | INIT_XMM sse4 |
| 28 | %if ARCH_X86_64 |
| 29 | cglobal flac_enc_lpc_16, 5, 7, 8, 0, res, smp, len, order, coefs |
| 30 | DECLARE_REG_TMP 5, 6 |
| 31 | %define length r2d |
| 32 | |
| 33 | movsxd orderq, orderd |
| 34 | %else |
| 35 | cglobal flac_enc_lpc_16, 5, 6, 8, 0, res, smp, len, order, coefs |
| 36 | DECLARE_REG_TMP 2, 5 |
| 37 | %define length r2mp |
| 38 | %endif |
| 39 | |
| 40 | ; Here we assume that the maximum order value is 32. This means that we only |
| 41 | ; need to copy a maximum of 32 samples. Therefore we let the preprocessor |
| 42 | ; unroll this loop and copy all 32. |
| 43 | %assign iter 0 |
| 44 | %rep 32/(mmsize/4) |
| 45 | movu m0, [smpq+iter] |
| 46 | movu [resq+iter], m0 |
| 47 | %assign iter iter+mmsize |
| 48 | %endrep |
| 49 | |
| 50 | lea resq, [resq+orderq*4] |
| 51 | lea smpq, [smpq+orderq*4] |
| 52 | lea coefsq, [coefsq+orderq*4] |
| 53 | sub length, orderd |
| 54 | movd m3, r5m |
| 55 | neg orderq |
| 56 | |
| 57 | %define posj t0q |
| 58 | %define negj t1q |
| 59 | |
| 60 | .looplen: |
| 61 | pxor m0, m0 |
| 62 | pxor m4, m4 |
| 63 | pxor m6, m6 |
| 64 | mov posj, orderq |
| 65 | xor negj, negj |
| 66 | |
| 67 | .looporder: |
| 68 | movd m2, [coefsq+posj*4] ; c = coefs[j] |
| 69 | SPLATD m2 |
| 70 | movu m1, [smpq+negj*4-4] ; s = smp[i-j-1] |
| 71 | movu m5, [smpq+negj*4-4+mmsize] |
| 72 | movu m7, [smpq+negj*4-4+mmsize*2] |
| 73 | pmulld m1, m2 |
| 74 | pmulld m5, m2 |
| 75 | pmulld m7, m2 |
| 76 | paddd m0, m1 ; p += c * s |
| 77 | paddd m4, m5 |
| 78 | paddd m6, m7 |
| 79 | |
| 80 | dec negj |
| 81 | inc posj |
| 82 | jnz .looporder |
| 83 | |
| 84 | psrad m0, m3 ; p >>= shift |
| 85 | psrad m4, m3 |
| 86 | psrad m6, m3 |
| 87 | movu m1, [smpq] |
| 88 | movu m5, [smpq+mmsize] |
| 89 | movu m7, [smpq+mmsize*2] |
| 90 | psubd m1, m0 ; smp[i] - p |
| 91 | psubd m5, m4 |
| 92 | psubd m7, m6 |
| 93 | movu [resq], m1 ; res[i] = smp[i] - (p >> shift) |
| 94 | movu [resq+mmsize], m5 |
| 95 | movu [resq+mmsize*2], m7 |
| 96 | |
| 97 | add resq, 3*mmsize |
| 98 | add smpq, 3*mmsize |
| 99 | sub length, (3*mmsize)/4 |
| 100 | jg .looplen |
| 101 | RET |