Kostya <[email protected]> writes:

> On Thu, Mar 24, 2011 at 03:34:32PM +0000, Mans Rullgard wrote:
>> ---
>>  libavcodec/ac3dsp.c              |    2 +
>>  libavcodec/ac3dsp.h              |    1 +
>>  libavcodec/arm/Makefile          |    3 ++
>>  libavcodec/arm/ac3dsp_init_arm.c |   35 ++++++++++++++++++++++
>>  libavcodec/arm/ac3dsp_neon.S     |   59 
>> ++++++++++++++++++++++++++++++++++++++
>>  5 files changed, 100 insertions(+), 0 deletions(-)
>>  create mode 100644 libavcodec/arm/ac3dsp_init_arm.c
>>  create mode 100644 libavcodec/arm/ac3dsp_neon.S
>> 
> [...]
>> diff --git a/libavcodec/arm/ac3dsp_neon.S b/libavcodec/arm/ac3dsp_neon.S
>> new file mode 100644
>> index 0000000..e69a3c2
>> --- /dev/null
>> +++ b/libavcodec/arm/ac3dsp_neon.S
>> @@ -0,0 +1,59 @@
>> +/*
>> + * Copyright (c) 2011 Mans Rullgard <[email protected]>
>> + *
>> + * 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_max_msb_abs_int16_neon, export=1
>> +        vmov.i16        q0,  #0
>> +        vmov.i16        q2,  #0
>> +1:      vld1.16         {q1},     [r0,:128]!
>> +        vabs.s16        q1,  q1
>> +        vld1.16         {q3},     [r0,:128]!
>> +        vabs.s16        q3,  q3
>> +        vorr            q0,  q0,  q1
>> +        vorr            q2,  q2,  q3
>> +        subs            r1,  r1,  #16
>> +        bgt             1b
>> +        vorr            q0,  q0,  q2
>> +        vorr            d0,  d0,  d1
>> +        vpmax.u16       d0,  d0,  d0
>> +        vpmax.u16       d0,  d0,  d0
>> +        vmov.u16        r0,  d0[0]
>> +        bx              lr
>> +endfunc
>
> this seems to be slightly incorrect (say, we have 2 and 1 as input,
> max will be 3) does that affect anything?

The comment for the function in ac3dsp.h says only the MSB matters, so
the bitwise or of all the values works fine.

>> +function ff_ac3_exponent_min_neon, export=1
>> +        cmp             r1,  #0
>> +        bxeq            lr
>> +        push            {lr}
>> +        mov             r12, #256
>> +1:
>> +        vld1.8          {q0},     [r0,:128]
>> +        mov             lr,  r1
>> +        add             r3,  r0,  #256
>> +2:      vld1.8          {q1},     [r3,:128], r12
>> +        subs            lr,  lr,  #1
>> +        vmin.u8         q0,  q0,  q1
>> +        bgt             2b
>> +        subs            r2,  r2,  #16
>> +        vst1.8          {q0},     [r0,:128]!
>> +        bgt             1b
>> +        pop             {pc}
>> +endfunc
>
> nice hack but otherwise looks ok

Hack?  I see no hack.

-- 
Måns Rullgård
[email protected]
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to