[x265] [PATCH 2 of 3] asm: routines for chroma hps filter functions for 8xN block size

[Steve Borho]

On Tue, Nov 12, 2013 at 9:35 AM, Nabajit Deka
<nabajit at multicorewareinc.com>wrote:

> Function prototype is wrong here, dst should be int16_t *dst.
>

waiting for a resend of this series



>
> On Tue, Nov 12, 2013 at 8:36 PM, <nabajit at multicorewareinc.com> wrote:
>
>> # HG changeset patch
>> # User Nabajit Deka
>> # Date 1384268587 -19800
>> #      Tue Nov 12 20:33:07 2013 +0530
>> # Node ID 676529ca9c45a14e54a1ad0976252b03fdcd9cd2
>> # Parent  c9851effbce88c9a70f712fbfaf7e83616c5615f
>> asm: routines for chroma hps filter functions for 8xN block size.
>>
>> diff -r c9851effbce8 -r 676529ca9c45 source/common/x86/ipfilter8.asm
>> --- a/source/common/x86/ipfilter8.asm   Tue Nov 12 20:24:34 2013 +0530
>> +++ b/source/common/x86/ipfilter8.asm   Tue Nov 12 20:33:07 2013 +0530
>> @@ -3600,3 +3600,62 @@
>>
>>  FILTER_HORIZ_CHROMA 6, 8
>>  FILTER_HORIZ_CHROMA 12, 16
>> +
>> +%macro PROCESS_CHROMA_W8 3
>> +    movu        %1, [srcq - 1]
>> +    pshufb      %2, %1, Tm0
>> +    pmaddubsw   %2, coef2
>> +    pshufb      %1, %1, Tm1
>> +    pmaddubsw   %1, coef2
>> +    phaddw      %2, %1
>> +    psubw       %2, %3
>> +    movu        [dstq],      %2
>> +%endmacro
>> +
>>
>> +;-----------------------------------------------------------------------------
>> +; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel
>> *dst, intptr_t dstStride, int coeffIdx)
>>
>> +;-----------------------------------------------------------------------------
>> +%macro FILTER_HORIZ_CHROMA_8xN 2
>> +INIT_XMM sse4
>> +cglobal interp_4tap_horiz_ps_%1x%2, 4, 6, 6, src, srcstride, dst,
>> dststride
>> +%define coef2       m5
>> +%define Tm0         m4
>> +%define Tm1         m3
>> +%define t2          m2
>> +%define t1          m1
>> +%define t0          m0
>> +
>> +mov         r4d,        r4m
>> +add         dststrided, dststrided
>> +
>> +%ifdef PIC
>> +lea         r5,          [tab_ChromaCoeff]
>> +movd        coef2,       [r5 + r4 * 4]
>> +%else
>> +movd        coef2,       [tab_ChromaCoeff + r4 * 4]
>> +%endif
>> +
>> +mov         r5d,         %2
>> +
>> +pshufd      coef2,       coef2,      0
>> +mova        t2,          [tab_c_8192]
>> +mova        Tm0,         [tab_Tm]
>> +mova        Tm1,         [tab_Tm + 16]
>> +
>> +.loop
>> +PROCESS_CHROMA_W8   t0, t1, t2
>> +add         srcq,        srcstrideq
>> +add         dstq,        dststrideq
>> +
>> +dec         r5d
>> +jnz         .loop
>> +
>> +RET
>> +%endmacro
>> +
>> +FILTER_HORIZ_CHROMA_8xN 8, 2
>> +FILTER_HORIZ_CHROMA_8xN 8, 4
>> +FILTER_HORIZ_CHROMA_8xN 8, 6
>> +FILTER_HORIZ_CHROMA_8xN 8, 8
>> +FILTER_HORIZ_CHROMA_8xN 8, 16
>> +FILTER_HORIZ_CHROMA_8xN 8, 32
>>
>
>
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel
>
>


-- 
Steve Borho
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20131112/3cb0ff6a/attachment-0001.html>