You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by Weldon Washburn <we...@gmail.com> on 2007/03/13 19:12:31 UTC
[drlvm][threading] JIRA H3065/H3231 -- how about a discussion on the relationship between these two patches?
H3231 mentions that it may no longer be possible to use the "initial-exec"
compile option on Linux if the TLS inlining from H3065 is committed. Is
anyone looking at this issue? I would hate to commit a TLS inlining mod
that subsequently needs to be reversed to accomodate "initial-exec".
I suggest all parties working on H3065 and H3231 to discuss the issues on
the dev list before adding any more patches to JIRA.
--
Weldon Washburn
Intel Enterprise Solutions Software Division
Re: [drlvm][threading] JIRA H3065/H3231 -- how about a discussion on the relationship between these two patches?
Posted by Weldon Washburn <we...@gmail.com>.
On 3/14/07, Rustem Rafikov <r....@gmail.com> wrote:
>
> I mistook thinking that there may be a conflict. I looked more carefully
> at
> 3065 patch: tm_self_tls var is not moved from the hythread library. The
> launcher depends on hythr.so (doesn't use dlopen for loading it). So,
> hythr's TLS block will be placed in a static TLS for a thread when loading
> the module. And gcc produces correct code for inlined methods accessing
> the
> tm_self_tls var. It should work and moreover I tried with both patches and
> it works.
Thanks. This helps. Please keep watching 3065. 3065 needs to be reworked
to make it committable. It would be good for you to tell us if future
proposed 3065 patches are OK with H3231.
Thanks, Rustem
>
> On 3/13/07, Weldon Washburn <we...@gmail.com> wrote:
> >
> > H3231 mentions that it may no longer be possible to use the
> "initial-exec"
> > compile option on Linux if the TLS inlining from H3065 is committed. Is
> > anyone looking at this issue? I would hate to commit a TLS inlining mod
> > that subsequently needs to be reversed to accomodate "initial-exec".
> > I suggest all parties working on H3065 and H3231 to discuss the issues
> on
> > the dev list before adding any more patches to JIRA.
> >
> > --
> > Weldon Washburn
> > Intel Enterprise Solutions Software Division
> >
>
--
Weldon Washburn
Intel Enterprise Solutions Software Division
Re: [drlvm][threading] JIRA H3065/H3231 -- how about a discussion on the relationship between these two patches?
Posted by Rustem Rafikov <r....@gmail.com>.
I mistook thinking that there may be a conflict. I looked more carefully at
3065 patch: tm_self_tls var is not moved from the hythread library. The
launcher depends on hythr.so (doesn't use dlopen for loading it). So,
hythr's TLS block will be placed in a static TLS for a thread when loading
the module. And gcc produces correct code for inlined methods accessing the
tm_self_tls var. It should work and moreover I tried with both patches and
it works.
Thanks, Rustem
On 3/13/07, Weldon Washburn <we...@gmail.com> wrote:
>
> H3231 mentions that it may no longer be possible to use the "initial-exec"
> compile option on Linux if the TLS inlining from H3065 is committed. Is
> anyone looking at this issue? I would hate to commit a TLS inlining mod
> that subsequently needs to be reversed to accomodate "initial-exec".
> I suggest all parties working on H3065 and H3231 to discuss the issues on
> the dev list before adding any more patches to JIRA.
>
> --
> Weldon Washburn
> Intel Enterprise Solutions Software Division
>