[FFmpeg-devel] [PATCH 5/9] x86: simple_idct10_template: fix overflow in pass

[Christophe Gisquet]

2015-10-13 15:43 GMT+02:00 Michael Niedermayer <michael at niedermayer.cc>:
> On Tue, Oct 13, 2015 at 01:33:07PM +0200, Christophe Gisquet wrote:
>> Hi,
>>
>> 2015-10-13 13:10 GMT+02:00 Michael Niedermayer <michael at niedermayer.cc>:
>> > hmm, iam a bit concerned that adding the rounder (which effectively is
>> > 0.5) causes a overflow, that would if iam not mistaken imlpy that
>> > things are very close to overflowing already without it
>>
>> It's true, but the immediate cause here is the input coeffs
>> overflowing (ie, once the rounder is added, a positive value is seen
>> as a negative one to pmaddwd).
>
> have you tried using saturating additions for the rounder ?

This would prevent an absolutely wrong result due to overflow, but
would not be sufficient for bitexactness.

The cases where the bitexactness issue arise may be pathological, not
appearing in fate, though.

But I'll check.

-- 
Christophe