[FFmpeg-devel] [PATCH 04/11] x86: dcadsp: implement SSE lfe_dir

[Christophe Gisquet]

Hi,

2014-02-07 Loren Merritt <lorenm at u.washington.edu>:
> On Thu, 6 Feb 2014, Christophe Gisquet wrote:
>
>> diff --git a/libavcodec/x86/dcadsp.asm b/libavcodec/x86/dcadsp.asm
>> index 03593ce..4a682be 100644
>> --- a/libavcodec/x86/dcadsp.asm
>> +++ b/libavcodec/x86/dcadsp.asm
>> @@ -88,3 +88,108 @@ INT8X8_FMUL_INT32  3
>>
>>  INIT_XMM sse4
>>  INT8X8_FMUL_INT32  3
>> +
>> +; %1=v0/v1  %2=in1  %3=in2
>> +%macro FIR_LOOP 2-3
>> +.loop%1:
>> +%define va          m1
>> +%define vb          m2
>> +%if %1
>> +%define OFFSET      0
>> +%else
>> +%define OFFSET      NUM_COEF*count
>> +%endif
>> +; for v0, incrementint and for v1, decrementing
>> +    mova        va, [cf0q + OFFSET]
>> +    mova        vb, [cf0q + OFFSET + 4*NUM_COEF]
>> +%if %0 == 3
>> +    mova        m4, [cf0q + OFFSET + mmsize]
>> +    mova     SCALE, [cf0q + OFFSET + 4*NUM_COEF + mmsize]
>> +%endif
>> +    mulps       va, %2
>> +    mulps       vb, %2
>> +%if %0 == 3
>> +    mulps       m4, %3
>> +    mulps    SCALE, %3
>> +    addps       va, m4
>> +    addps       vb, SCALE
>> +%endif
>> +    ; va = va1 va2 va3 va4
>> +    ; vb = vb1 vb2 vb3 vb4
>> +%if %1
>> +%define   O1    vb
>> +%define   O2    va
>> +%else
>> +%define   O1    va
>> +%define   O2    vb
>> +%endif
>
> Can this be simplified with
> %if %1
> SWAP va, vb
> %endif
> and no O1, O2 variables?
>
>> +    mova        m4, O1
>> +    unpcklps    O1, O2 ; va3 vb3 va4 vb4
>> +    unpckhps    m4, O2 ; va1 vb1 va2 vb2
>> +    addps       m4, O1 ; va1+3 vb1+3 va2+4 vb2+4
>> +    movhlps     O2, m4 ; va1+3  vb1+3
>> +    addps       O2, m4 ; va0..4 vb0..4
>> +%if %1
>> +    movh    [outq + count], O2
>> +    sub       cf0q, 8*NUM_COEF
>> +%else
>> +    movh    [outq + count], O2
>
> factor out of the %if

All was ok, so here's a new patch.

-- 
Christophe
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0004-x86-dcadsp-implement-SSE-lfe_dir.patch
Type: text/x-patch
Size: 3966 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20140207/55a9e29d/attachment.bin>