Quantcast

AVX-512 development status and Valgrind multi-processing prospects

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

AVX-512 development status and Valgrind multi-processing prospects

Mineeva, Tatyana A

Hello Julian,

 

We are getting close to running a small real AVX-512 application (NAS IS benchmark) under Nulgrind, but probably it would take a few more weeks. We have figured how to implement masked instructions and AVX-512 instructions which do not have AVX-2 equivalents, we just need to fix some bugs and add more instructions now.

 

One question, mostly out of curiosity:

KNL architecture is better suited for parallel applications, and Valgrind’s serialization could cause a significant overhead on KNLs. I know there were some attempts to make Valgrind multi-threaded. Could you please point out the main difficulties of parallelizing Valgrind?

 

Thank you,

Tanya


--------------------------------------------------------------------
Joint Stock Company Intel A/O
Registered legal address: Krylatsky Hills Business Park,
17 Krylatskaya Str., Bldg 4, Moscow 121614,
Russian Federation

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Valgrind-developers mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/valgrind-developers
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: AVX-512 development status and Valgrind multi-processing prospects

Philippe Waroquiers
On Fri, 2017-04-28 at 17:46 +0000, Mineeva, Tatyana A wrote:
> One question, mostly out of curiosity:
> KNL architecture is better suited for parallel applications, and
> Valgrind’s serialization could cause a significant overhead on KNLs. I
> know there were some attempts to make Valgrind multi-threaded. Could
> you please point out the main difficulties of parallelizing Valgrind?

A presentation was done at FOSDEM about a (very partial) prototype :
See
https://archive.fosdem.org/2015/schedule/event/valgrind_multi_prototype/


In the SVN MTV branch, the file docs/internals/mtV.txt gives some
more information.
This proto has shown that there is a lot of work (still) to do but
IMO, the biggest challenge was to have a memcheck VA bit data
structure that is both fast, thread safe, and not causing false positive
or false negative.

Philippe




------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Valgrind-developers mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/valgrind-developers
Loading...