ac3: armv6 optimised bit_alloc_calc_bap

Signed-off-by: Mans Rullgard <mans@mansr.com>
This commit is contained in:
Mans Rullgard 2011-03-11 02:50:57 +00:00
parent 5085a1a068
commit 182826c884
3 changed files with 94 additions and 0 deletions

View File

@ -1,6 +1,8 @@
OBJS-$(CONFIG_AC3DSP) += arm/ac3dsp_init_arm.o
OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_init_arm.o \
ARMV6-OBJS-$(CONFIG_AC3DSP) += arm/ac3dsp_armv6.o
OBJS-$(CONFIG_VP5_DECODER) += arm/vp56dsp_init_arm.o
OBJS-$(CONFIG_VP6_DECODER) += arm/vp56dsp_init_arm.o
OBJS-$(CONFIG_VP8_DECODER) += arm/vp8dsp_init_arm.o

View File

@ -0,0 +1,83 @@
/*
* Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
*
* This file is part of Libav.
*
* Libav 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.
*
* Libav 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 Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "asm.S"
function ff_ac3_bit_alloc_calc_bap_armv6, export=1
ldr r12, [sp]
cmp r12, #-960
beq 4f
push {r4-r11,lr}
add r5, sp, #40
movrel r4, X(ff_ac3_bin_to_band_tab)
movrel lr, X(ff_ac3_band_start_tab)
ldm r5, {r5-r7}
ldrb r4, [r4, r2]
add r1, r1, r2, lsl #1 @ psd + start
add r0, r0, r4, lsl #1 @ mask + band
add r4, lr, r4
add r7, r7, r2 @ bap + start
ldrb r10, [r4], #1
1:
ldrsh r9, [r0], #2 @ mask[band]
movw r8, #0x1fe0
sub r9, r9, r12 @ - snr_offset
mov r11, r10
ldrb r10, [r4], #1 @ band_start_tab[band++]
subs r9, r9, r5 @ - floor
movlt r9, #0
cmp r10, r3 @ - end
and r9, r9, r8 @ & 0x1fe0
subgt r8, r3, r11
suble r8, r10, r11
add r9, r9, r5 @ + floor => m
tst r8, #1
add r2, r7, r8
bne 3f
b 5f
2:
ldrsh r8, [r1], #2
ldrsh lr, [r1], #2
sub r8, r8, r9
sub lr, lr, r9
usat r8, #6, r8, asr #5 @ address
usat lr, #6, lr, asr #5
ldrb r8, [r6, r8] @ bap_tab[address]
ldrb lr, [r6, lr]
strb r8, [r7], #1 @ bap[bin]
strb lr, [r7], #1
5: cmp r7, r2
blo 2b
cmp r3, r11
bgt 1b
pop {r4-r11,pc}
3:
ldrsh r8, [r1], #2 @ psd[bin]
sub r8, r8, r9 @ - m
usat r8, #6, r8, asr #5 @ address
ldrb r8, [r6, r8] @ bap_tab[address]
strb r8, [r7], #1 @ bap[bin]
b 5b
4:
ldr r0, [sp, #12]
mov r1, #0
mov r2, #256
b memset
endfunc

View File

@ -29,8 +29,17 @@ void ff_ac3_lshift_int16_neon(int16_t *src, unsigned len, unsigned shift);
void ff_ac3_rshift_int32_neon(int32_t *src, unsigned len, unsigned shift);
void ff_float_to_fixed24_neon(int32_t *dst, const float *src, unsigned int len);
void ff_ac3_bit_alloc_calc_bap_armv6(int16_t *mask, int16_t *psd,
int start, int end,
int snr_offset, int floor,
const uint8_t *bap_tab, uint8_t *bap);
av_cold void ff_ac3dsp_init_arm(AC3DSPContext *c, int bit_exact)
{
if (HAVE_ARMV6) {
c->bit_alloc_calc_bap = ff_ac3_bit_alloc_calc_bap_armv6;
}
if (HAVE_NEON) {
c->ac3_exponent_min = ff_ac3_exponent_min_neon;
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_neon;