public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Jakub Jelinek <jakub@redhat.com>
To: Hongyu Wang <wwwhhhyyy333@gmail.com>
Cc: Hongyu Wang <hongyu.wang@intel.com>, gcc-patches@gcc.gnu.org
Subject: Re: [PATCH] libgomp: Fix default value of GOMP_SPINCOUNT [PR 109062]
Date: Wed, 8 Mar 2023 09:35:36 +0100	[thread overview]
Message-ID: <ZAhI2G6icli9FYp4@tucnak> (raw)
In-Reply-To: <CA+OydW=RyumvMfggDeYJh3PXUPkmeBk4XtU02bGpbnk7brwfLA@mail.gmail.com>

On Wed, Mar 08, 2023 at 04:21:46PM +0800, Hongyu Wang wrote:
> Hongyu Wang <wwwhhhyyy333@gmail.com> 于2023年3月8日周三 16:07写道:
> >
> > > I think the right spot to fix this would be instead in initialize_icvs,
> > > change the
> > >   icvs->wait_policy = 0;
> > > in there to
> > >   icvs->wait_policy = -1;
> > > That way it will be the default for all the devices, not just the
> > > initial one.
> >
> > It doesn't work, for the code that determines value of wait_policy:
> >
> > if (none != NULL && gomp_get_icv_flag (none->flags, GOMP_ICV_WAIT_POLICY))
> >   wait_policy = none->icvs.wait_policy;
> > else if (all != NULL && gomp_get_icv_flag (all->flags, GOMP_ICV_WAIT_POLICY))
> >   wait_policy = all->icvs.wait_policy;
> >
> > gomp_get_icv_flag (none->flags, GOMP_ICV_WAIT_POLICY) returns true only when
> > OMP_WAIT_POLICY is explicitly set, so the initial icvs->wait_policy
> > could not affect the global wait_policy that used to set
> > GOMP_SPINCOUNT.
> 
> Also the global variable wait_policy here is only used for setting
> spin_count related values that do not
> belong to any ICV, so there is no need to set icvs->wait_policy since
> for OMP_WAIT_POLICY_(DEV|ALL)
> itself only has value 0 for passive and value 1 for active.

Seems for many ICVs the default values are done through
gomp_default_icv_values, but that doesn't cover wait_policy.
For other vars, the defaults are provided through just initializers of
those vars on the var definitions, e.g.:
char *gomp_affinity_format_var = "level %L thread %i affinity %A";
So, I'd do the initialize_icvs change and change
static int wait_policy;
to
static int wait_policy = -1;

	Jakub


  reply	other threads:[~2023-03-08  8:35 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-03-08  6:31 Hongyu Wang
2023-03-08  7:35 ` Jakub Jelinek
2023-03-08  8:07   ` Hongyu Wang
2023-03-08  8:21     ` Hongyu Wang
2023-03-08  8:35       ` Jakub Jelinek [this message]
2023-03-08  8:54         ` Hongyu Wang
2023-03-08  9:03           ` Jakub Jelinek

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ZAhI2G6icli9FYp4@tucnak \
    --to=jakub@redhat.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=hongyu.wang@intel.com \
    --cc=wwwhhhyyy333@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).