Fwd: Questions about Vc

[Daniel Cámpora]

Hi,

[please enable javascript to see the address]>:

> Hi,
>
> just a quick answer to let you know that I have some high priority work to
> do
> the next two weeks...
>
> IIUC your code vectorizes inside each object (vertical vectorization) not
> over
> multiple objects (horizontal vectorization), right. In my experience
> vertical
> vectorization requires shuffles, horizontal vectorization doesn't.
> At this point Vc is much better suited for horizontal vectorization...
>
>
Yes, I fully agree :) Also, gcc / icc will tell you hints towards
horizontal vectorization
always. Vertical is usually more tough (and more fun).


> Once you do vertical vectorization you rely on the number of entries in the
> SIMD register anyway and would be better of with something like
> Vc::simdarray<float, 4> instead of Vc::float_v. With Vc::simdarray a
> shuffle
> API will be much easier to define, because the number of entries is
> portable
> for a given type.
>
>
That's very nice input. I'll have a look into developing something myself
if I find some time.
Was this something you guys intended to support in the short future?

Thanks,

Daniel


> -Matthias
>
> --
> ─────────────────────────────────────────────────────────────
>  Dipl.-Phys. Matthias Kretz
>
>  Web:   http://compeng.uni-frankfurt.de/?mkretz
>
>  SIMD easy and portable: http://compeng.uni-frankfurt.de/?vc
> ─────────────────────────────────────────────────────────────
> _______________________________________________
> Vc mailing list
>[please enable javascript to see the address]
> https://compeng.uni-frankfurt.de/mailman/listinfo/vc
>



-- 
Daniel Hugo Cámpora Pérez
European Organization for Nuclear Research (CERN)
PH LBC, LHCb Online Fellow
[please enable javascript to see the address]
tel. intern 166473
tel. +41 (0) 764876473
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://compeng.uni-frankfurt.de/pipermail/vc/attachments/20140929/9f529c68/attachment.html>