RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
ffmpeg/libswresample/arm/audio_convert_neon.S

366 lines
14 KiB
ArmAsm
Raw Normal View History

ARM: libswresample: NEON optimised flat float to s16 conversion Adapted to swr by: Michael Niedermayer <michaelni@gmx.at> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-09-09 18:25:43 +00:00
/*
* Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
*
* This file is part of libswresample.
*
* libswresample is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* libswresample is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with libswresample; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "config.h"
#include "libavutil/arm/asm.S"
function swri_oldapi_conv_flt_to_s16_neon, export=1
swresample/arm: remove unintentional relocation. Branch to global symbol results in reference to PLT, and when compiling for THUMB-2 - in a R_ARM_THM_JUMP19 relocation. Some linkers don't support this relocation (ld.gold), while others can end up truncating the relocation to fit (ld.bfd). Convert this branch through PLT into a direct branch that the assembler can resolve locally. See https://github.com/android-ndk/ndk/issues/337 for background. The current workaround is to disable neon during gstreamer build, which is not optimal and can be reverted after this patch: https://github.com/freedesktop/gstreamer-cerbero/commit/41556c415739fbc3a72c7eaee7e70a565b719b2f Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2018-04-18 23:29:39 +00:00
_swri_oldapi_conv_flt_to_s16_neon:
ARM: libswresample: NEON optimised flat float to s16 conversion Adapted to swr by: Michael Niedermayer <michaelni@gmx.at> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-09-09 18:25:43 +00:00
subs r2, r2, #8
vld1.32 {q0}, [r1,:128]!
vcvt.s32.f32 q8, q0, #31
vld1.32 {q1}, [r1,:128]!
vcvt.s32.f32 q9, q1, #31
beq 3f
bics r12, r2, #15
beq 2f
1: subs r12, r12, #16
vqrshrn.s32 d4, q8, #16
vld1.32 {q0}, [r1,:128]!
vcvt.s32.f32 q0, q0, #31
vqrshrn.s32 d5, q9, #16
vld1.32 {q1}, [r1,:128]!
vcvt.s32.f32 q1, q1, #31
vqrshrn.s32 d6, q0, #16
vst1.16 {q2}, [r0,:128]!
vqrshrn.s32 d7, q1, #16
vld1.32 {q8}, [r1,:128]!
vcvt.s32.f32 q8, q8, #31
vld1.32 {q9}, [r1,:128]!
vcvt.s32.f32 q9, q9, #31
vst1.16 {q3}, [r0,:128]!
bne 1b
ands r2, r2, #15
beq 3f
2: vld1.32 {q0}, [r1,:128]!
vqrshrn.s32 d4, q8, #16
vcvt.s32.f32 q0, q0, #31
vld1.32 {q1}, [r1,:128]!
vqrshrn.s32 d5, q9, #16
vcvt.s32.f32 q1, q1, #31
vqrshrn.s32 d6, q0, #16
vst1.16 {q2}, [r0,:128]!
vqrshrn.s32 d7, q1, #16
vst1.16 {q3}, [r0,:128]!
bx lr
3: vqrshrn.s32 d4, q8, #16
vqrshrn.s32 d5, q9, #16
vst1.16 {q2}, [r0,:128]!
bx lr
endfunc
ARM: libswresample: NEON optimised stereo fltp to s16 conversion Adaption to swr by: Michael Niedermayer <michaelni@gmx.at> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-09-10 14:47:00 +00:00
function swri_oldapi_conv_fltp_to_s16_2ch_neon, export=1
swresample/arm: remove unintentional relocation. Branch to global symbol results in reference to PLT, and when compiling for THUMB-2 - in a R_ARM_THM_JUMP19 relocation. Some linkers don't support this relocation (ld.gold), while others can end up truncating the relocation to fit (ld.bfd). Convert this branch through PLT into a direct branch that the assembler can resolve locally. See https://github.com/android-ndk/ndk/issues/337 for background. The current workaround is to disable neon during gstreamer build, which is not optimal and can be reverted after this patch: https://github.com/freedesktop/gstreamer-cerbero/commit/41556c415739fbc3a72c7eaee7e70a565b719b2f Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2018-04-18 23:29:39 +00:00
_swri_oldapi_conv_fltp_to_s16_2ch_neon:
ARM: libswresample: NEON optimised stereo fltp to s16 conversion Adaption to swr by: Michael Niedermayer <michaelni@gmx.at> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-09-10 14:47:00 +00:00
ldm r1, {r1, r3}
subs r2, r2, #8
vld1.32 {q0}, [r1,:128]!
vcvt.s32.f32 q8, q0, #31
vld1.32 {q1}, [r1,:128]!
vcvt.s32.f32 q9, q1, #31
vld1.32 {q10}, [r3,:128]!
vcvt.s32.f32 q10, q10, #31
vld1.32 {q11}, [r3,:128]!
vcvt.s32.f32 q11, q11, #31
beq 3f
bics r12, r2, #15
beq 2f
1: subs r12, r12, #16
vld1.32 {q0}, [r1,:128]!
vcvt.s32.f32 q0, q0, #31
vsri.32 q10, q8, #16
vld1.32 {q1}, [r1,:128]!
vcvt.s32.f32 q1, q1, #31
vld1.32 {q12}, [r3,:128]!
vcvt.s32.f32 q12, q12, #31
vld1.32 {q13}, [r3,:128]!
vsri.32 q11, q9, #16
vst1.16 {q10}, [r0,:128]!
vcvt.s32.f32 q13, q13, #31
vst1.16 {q11}, [r0,:128]!
vsri.32 q12, q0, #16
vld1.32 {q8}, [r1,:128]!
vsri.32 q13, q1, #16
vst1.16 {q12}, [r0,:128]!
vcvt.s32.f32 q8, q8, #31
vld1.32 {q9}, [r1,:128]!
vcvt.s32.f32 q9, q9, #31
vld1.32 {q10}, [r3,:128]!
vcvt.s32.f32 q10, q10, #31
vld1.32 {q11}, [r3,:128]!
vcvt.s32.f32 q11, q11, #31
vst1.16 {q13}, [r0,:128]!
bne 1b
ands r2, r2, #15
beq 3f
2: vsri.32 q10, q8, #16
vld1.32 {q0}, [r1,:128]!
vcvt.s32.f32 q0, q0, #31
vld1.32 {q1}, [r1,:128]!
vcvt.s32.f32 q1, q1, #31
vld1.32 {q12}, [r3,:128]!
vcvt.s32.f32 q12, q12, #31
vsri.32 q11, q9, #16
vld1.32 {q13}, [r3,:128]!
vcvt.s32.f32 q13, q13, #31
vst1.16 {q10}, [r0,:128]!
vsri.32 q12, q0, #16
vst1.16 {q11}, [r0,:128]!
vsri.32 q13, q1, #16
vst1.16 {q12-q13},[r0,:128]!
bx lr
3: vsri.32 q10, q8, #16
vsri.32 q11, q9, #16
vst1.16 {q10-q11},[r0,:128]!
bx lr
endfunc
ARM: libswresample: NEON optimised generic fltp to s16 conversion Adapted to swr by: Michael Niedermayer <michaelni@gmx.at> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-09-10 14:47:00 +00:00
function swri_oldapi_conv_fltp_to_s16_nch_neon, export=1
cmp r3, #2
itt lt
ldrlt r1, [r1]
swresample/arm: remove unintentional relocation. Branch to global symbol results in reference to PLT, and when compiling for THUMB-2 - in a R_ARM_THM_JUMP19 relocation. Some linkers don't support this relocation (ld.gold), while others can end up truncating the relocation to fit (ld.bfd). Convert this branch through PLT into a direct branch that the assembler can resolve locally. See https://github.com/android-ndk/ndk/issues/337 for background. The current workaround is to disable neon during gstreamer build, which is not optimal and can be reverted after this patch: https://github.com/freedesktop/gstreamer-cerbero/commit/41556c415739fbc3a72c7eaee7e70a565b719b2f Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2018-04-18 23:29:39 +00:00
blt _swri_oldapi_conv_flt_to_s16_neon
beq _swri_oldapi_conv_fltp_to_s16_2ch_neon
ARM: libswresample: NEON optimised generic fltp to s16 conversion Adapted to swr by: Michael Niedermayer <michaelni@gmx.at> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-09-10 14:47:00 +00:00
push {r4-r8, lr}
cmp r3, #4
lsl r12, r3, #1
blt 4f
@ 4 channels
5: ldm r1!, {r4-r7}
mov lr, r2
mov r8, r0
vld1.32 {q8}, [r4,:128]!
vcvt.s32.f32 q8, q8, #31
vld1.32 {q9}, [r5,:128]!
vcvt.s32.f32 q9, q9, #31
vld1.32 {q10}, [r6,:128]!
vcvt.s32.f32 q10, q10, #31
vld1.32 {q11}, [r7,:128]!
vcvt.s32.f32 q11, q11, #31
6: subs lr, lr, #8
vld1.32 {q0}, [r4,:128]!
vcvt.s32.f32 q0, q0, #31
vsri.32 q9, q8, #16
vld1.32 {q1}, [r5,:128]!
vcvt.s32.f32 q1, q1, #31
vsri.32 q11, q10, #16
vld1.32 {q2}, [r6,:128]!
vcvt.s32.f32 q2, q2, #31
vzip.32 d18, d22
vld1.32 {q3}, [r7,:128]!
vcvt.s32.f32 q3, q3, #31
vzip.32 d19, d23
vst1.16 {d18}, [r8], r12
vsri.32 q1, q0, #16
vst1.16 {d22}, [r8], r12
vsri.32 q3, q2, #16
vst1.16 {d19}, [r8], r12
vzip.32 d2, d6
vst1.16 {d23}, [r8], r12
vzip.32 d3, d7
beq 7f
vld1.32 {q8}, [r4,:128]!
vcvt.s32.f32 q8, q8, #31
vst1.16 {d2}, [r8], r12
vld1.32 {q9}, [r5,:128]!
vcvt.s32.f32 q9, q9, #31
vst1.16 {d6}, [r8], r12
vld1.32 {q10}, [r6,:128]!
vcvt.s32.f32 q10, q10, #31
vst1.16 {d3}, [r8], r12
vld1.32 {q11}, [r7,:128]!
vcvt.s32.f32 q11, q11, #31
vst1.16 {d7}, [r8], r12
b 6b
7: vst1.16 {d2}, [r8], r12
vst1.16 {d6}, [r8], r12
vst1.16 {d3}, [r8], r12
vst1.16 {d7}, [r8], r12
subs r3, r3, #4
it eq
popeq {r4-r8, pc}
cmp r3, #4
add r0, r0, #8
bge 5b
@ 2 channels
4: cmp r3, #2
blt 4f
ldm r1!, {r4-r5}
mov lr, r2
mov r8, r0
tst lr, #8
vld1.32 {q8}, [r4,:128]!
vcvt.s32.f32 q8, q8, #31
vld1.32 {q9}, [r5,:128]!
vcvt.s32.f32 q9, q9, #31
vld1.32 {q10}, [r4,:128]!
vcvt.s32.f32 q10, q10, #31
vld1.32 {q11}, [r5,:128]!
vcvt.s32.f32 q11, q11, #31
beq 6f
subs lr, lr, #8
beq 7f
vsri.32 d18, d16, #16
vsri.32 d19, d17, #16
vld1.32 {q8}, [r4,:128]!
vcvt.s32.f32 q8, q8, #31
vst1.32 {d18[0]}, [r8], r12
vsri.32 d22, d20, #16
vst1.32 {d18[1]}, [r8], r12
vsri.32 d23, d21, #16
vst1.32 {d19[0]}, [r8], r12
vst1.32 {d19[1]}, [r8], r12
vld1.32 {q9}, [r5,:128]!
vcvt.s32.f32 q9, q9, #31
vst1.32 {d22[0]}, [r8], r12
vst1.32 {d22[1]}, [r8], r12
vld1.32 {q10}, [r4,:128]!
vcvt.s32.f32 q10, q10, #31
vst1.32 {d23[0]}, [r8], r12
vst1.32 {d23[1]}, [r8], r12
vld1.32 {q11}, [r5,:128]!
vcvt.s32.f32 q11, q11, #31
6: subs lr, lr, #16
vld1.32 {q0}, [r4,:128]!
vcvt.s32.f32 q0, q0, #31
vsri.32 d18, d16, #16
vld1.32 {q1}, [r5,:128]!
vcvt.s32.f32 q1, q1, #31
vsri.32 d19, d17, #16
vld1.32 {q2}, [r4,:128]!
vcvt.s32.f32 q2, q2, #31
vld1.32 {q3}, [r5,:128]!
vcvt.s32.f32 q3, q3, #31
vst1.32 {d18[0]}, [r8], r12
vsri.32 d22, d20, #16
vst1.32 {d18[1]}, [r8], r12
vsri.32 d23, d21, #16
vst1.32 {d19[0]}, [r8], r12
vsri.32 d2, d0, #16
vst1.32 {d19[1]}, [r8], r12
vsri.32 d3, d1, #16
vst1.32 {d22[0]}, [r8], r12
vsri.32 d6, d4, #16
vst1.32 {d22[1]}, [r8], r12
vsri.32 d7, d5, #16
vst1.32 {d23[0]}, [r8], r12
vst1.32 {d23[1]}, [r8], r12
beq 6f
vld1.32 {q8}, [r4,:128]!
vcvt.s32.f32 q8, q8, #31
vst1.32 {d2[0]}, [r8], r12
vst1.32 {d2[1]}, [r8], r12
vld1.32 {q9}, [r5,:128]!
vcvt.s32.f32 q9, q9, #31
vst1.32 {d3[0]}, [r8], r12
vst1.32 {d3[1]}, [r8], r12
vld1.32 {q10}, [r4,:128]!
vcvt.s32.f32 q10, q10, #31
vst1.32 {d6[0]}, [r8], r12
vst1.32 {d6[1]}, [r8], r12
vld1.32 {q11}, [r5,:128]!
vcvt.s32.f32 q11, q11, #31
vst1.32 {d7[0]}, [r8], r12
vst1.32 {d7[1]}, [r8], r12
bgt 6b
6: vst1.32 {d2[0]}, [r8], r12
vst1.32 {d2[1]}, [r8], r12
vst1.32 {d3[0]}, [r8], r12
vst1.32 {d3[1]}, [r8], r12
vst1.32 {d6[0]}, [r8], r12
vst1.32 {d6[1]}, [r8], r12
vst1.32 {d7[0]}, [r8], r12
vst1.32 {d7[1]}, [r8], r12
b 8f
7: vsri.32 d18, d16, #16
vsri.32 d19, d17, #16
vst1.32 {d18[0]}, [r8], r12
vsri.32 d22, d20, #16
vst1.32 {d18[1]}, [r8], r12
vsri.32 d23, d21, #16
vst1.32 {d19[0]}, [r8], r12
vst1.32 {d19[1]}, [r8], r12
vst1.32 {d22[0]}, [r8], r12
vst1.32 {d22[1]}, [r8], r12
vst1.32 {d23[0]}, [r8], r12
vst1.32 {d23[1]}, [r8], r12
8: subs r3, r3, #2
add r0, r0, #4
it eq
popeq {r4-r8, pc}
@ 1 channel
4: ldr r4, [r1]
tst r2, #8
mov lr, r2
mov r5, r0
vld1.32 {q0}, [r4,:128]!
vcvt.s32.f32 q0, q0, #31
vld1.32 {q1}, [r4,:128]!
vcvt.s32.f32 q1, q1, #31
bne 8f
6: subs lr, lr, #16
vld1.32 {q2}, [r4,:128]!
vcvt.s32.f32 q2, q2, #31
vld1.32 {q3}, [r4,:128]!
vcvt.s32.f32 q3, q3, #31
vst1.16 {d0[1]}, [r5,:16], r12
vst1.16 {d0[3]}, [r5,:16], r12
vst1.16 {d1[1]}, [r5,:16], r12
vst1.16 {d1[3]}, [r5,:16], r12
vst1.16 {d2[1]}, [r5,:16], r12
vst1.16 {d2[3]}, [r5,:16], r12
vst1.16 {d3[1]}, [r5,:16], r12
vst1.16 {d3[3]}, [r5,:16], r12
beq 7f
vld1.32 {q0}, [r4,:128]!
vcvt.s32.f32 q0, q0, #31
vld1.32 {q1}, [r4,:128]!
vcvt.s32.f32 q1, q1, #31
7: vst1.16 {d4[1]}, [r5,:16], r12
vst1.16 {d4[3]}, [r5,:16], r12
vst1.16 {d5[1]}, [r5,:16], r12
vst1.16 {d5[3]}, [r5,:16], r12
vst1.16 {d6[1]}, [r5,:16], r12
vst1.16 {d6[3]}, [r5,:16], r12
vst1.16 {d7[1]}, [r5,:16], r12
vst1.16 {d7[3]}, [r5,:16], r12
bgt 6b
pop {r4-r8, pc}
8: subs lr, lr, #8
vst1.16 {d0[1]}, [r5,:16], r12
vst1.16 {d0[3]}, [r5,:16], r12
vst1.16 {d1[1]}, [r5,:16], r12
vst1.16 {d1[3]}, [r5,:16], r12
vst1.16 {d2[1]}, [r5,:16], r12
vst1.16 {d2[3]}, [r5,:16], r12
vst1.16 {d3[1]}, [r5,:16], r12
vst1.16 {d3[3]}, [r5,:16], r12
it eq
popeq {r4-r8, pc}
vld1.32 {q0}, [r4,:128]!
vcvt.s32.f32 q0, q0, #31
vld1.32 {q1}, [r4,:128]!
vcvt.s32.f32 q1, q1, #31
b 6b
endfunc