[x265] [PATCH Review Only] filterHorizontal_p_p_4, 4x4 asm code
Praveen Tiwari
praveen at multicorewareinc.com
Tue Oct 8 08:03:40 CEST 2013
Yes, That is right we don't need to pass width and height but that is
helpfully for now in some cases to know which parameter I need to load
from stack but while finally integrating to Test Bench I will replace with
new one.
Regards,
Praveen
On Tue, Oct 8, 2013 at 11:13 AM, Steve Borho <steve at borho.org> wrote:
>
>
>
> On Tue, Oct 8, 2013 at 12:40 AM, <praveen at multicorewareinc.com> wrote:
>
>> # HG changeset patch
>> # User praveen Tiwari
>> # Date 1381210681 -19800
>> # Node ID d9580d9cad8df6ad00d08ab538cdac27c0eb4e92
>> # Parent d71078917df01e92605158a13b45ab35ee7cfc1c
>> filterHorizontal_p_p_4, 4x4 asm code
>>
>> diff -r d71078917df0 -r d9580d9cad8d source/common/x86/ipfilter8.asm
>> --- a/source/common/x86/ipfilter8.asm Mon Oct 07 12:48:32 2013 +0530
>> +++ b/source/common/x86/ipfilter8.asm Tue Oct 08 11:08:01 2013 +0530
>> @@ -130,3 +130,56 @@
>> RET
>>
>> %endif ; ARCH_X86_64 == 0
>> +
>> +SECTION_RODATA 32
>> +tab_Tm: db 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
>> +
>> +tab_c_512: times 8 dw 512
>> +
>> +SECTION .text
>> +
>> +%macro FILTER_H4_w4 3
>> + movu %1, [srcq - 1]
>> + pshufb %2, %1, Tm0
>> + pmaddubsw %2, coef2
>> + phaddw %2, %1
>> + pmulhrsw %2, %3
>> + packuswb %2, %2
>> +%endmacro
>> +
>> +%macro FILTER_H4_w4_CALL 0
>> + FILTER_H4_w4 x0, x1, x2
>> +
>> + movd [dstq], x1
>> +
>> + add srcq, srcstrideq
>> + add dstq, dststrideq
>> +%endmacro
>> +
>>
>> +;-----------------------------------------------------------------------------
>> +; void filterHorizontal_p_p_4(pixel *src, intptr_t srcStride, pixel
>> *dst, intptr_t dstStride, int width, int height, short const *coeff)
>>
>
> I assume this comment is out of date? you shouldn't need to pass width
> and height
>
>
>>
>> +;-----------------------------------------------------------------------------
>> +INIT_XMM sse4
>> +cglobal filterHorizontal_p_p_4, 4, 5, 5, src, srcstride, dst, dststride
>> +%define coef2 m4
>> +%define Tm0 m3
>> +%define x2 m2
>> +%define x1 m1
>> +%define x0 m0
>> +
>> + mov r4, r6m
>> + movu coef2, [r4]
>> + packsswb coef2, coef2
>> + pshufd coef2, coef2, 0
>> +
>> + mova x2, [tab_c_512]
>> +
>> + mova Tm0, [tab_Tm]
>> +
>> +%rep 3
>> +FILTER_H4_w4_CALL
>> +%endrep
>> +
>> + FILTER_H4_w4 x0, x1, x2
>> + movd [dstq], x1
>> + RET
>> _______________________________________________
>> x265-devel mailing list
>> x265-devel at videolan.org
>> https://mailman.videolan.org/listinfo/x265-devel
>>
>
>
>
> --
> Steve Borho
>
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20131008/620264ff/attachment.html>
More information about the x265-devel
mailing list