Accéder au contenu.
Menu Sympa

starpu-devel - Re: [starpu-devel] Corrupted large ooc files with unistd backend

Objet : Developers list for StarPU

Archives de la liste

Re: [starpu-devel] Corrupted large ooc files with unistd backend


Chronologique Discussions 
  • From: Samuel Thibault <samuel.thibault@inria.fr>
  • To: barbier@imacs.polytechnique.fr
  • Cc: starpu-devel@inria.fr
  • Subject: Re: [starpu-devel] Corrupted large ooc files with unistd backend
  • Date: Mon, 5 Dec 2022 18:31:18 +0100
  • Authentication-results: mail3-relais-sop.national.inria.fr; dkim=none (message not signed) header.i=none; spf=SoftFail smtp.mailfrom=samuel.thibault@inria.fr; dmarc=fail (p=none dis=none) d=inria.fr
  • Organization: I am not organized

barbier@imacs.polytechnique.fr, le lun. 05 déc. 2022 09:45:16 +0100, a ecrit:
> write.2 manpage constains this note on Linux:
>
> On Linux, write() (and similar system calls) will transfer at most
> 0x7ffff000 (2,147,479,552) bytes, returning the number of bytes actually
> transferred. (This is true on both 32-bit and 64-bit systems.)

Oh :/

> As a result, files contain only zeros after offset 0x7ffff000.
> This is pretty easy to fix in
> starpu_unistd_global_read/write with a loop, but I do not see how to deal
> with asynchronous IO; I discarded AIO when size is too large in attached
> patch.

Ok, I have pushed that for a start. I have noted in the code
that we could make starpu_unistd_global_test_request and
starpu_unistd_global_test_request resubmit an updated request.

Thanks!
Samuel



Archives gérées par MHonArc 2.6.19+.

Haut de le page