This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: TSX lock elision for glibc v14
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: Andi Kleen <andi at firstfloor dot org>
- Cc: libc-alpha at sourceware dot org, Roland McGrath <roland at hack dot frob dot com>, Torvald Riegel <triegel at redhat dot com>, Rich Felker <dalias at aerifal dot cx>
- Date: Fri, 28 Jun 2013 13:31:22 -0400
- Subject: Re: TSX lock elision for glibc v14
- References: <1372398717-16530-1-git-send-email-andi at firstfloor dot org> <51CD4878 dot 8010903 at redhat dot com>
On 06/28/2013 04:25 AM, Carlos O'Donell wrote:
> On 06/28/2013 01:51 AM, Andi Kleen wrote:
>> All open issues addressed. Should be (hopefully) ready for merging now.
>
> I've reviewed patches 1 through 9.
>
> Patches 1 through 9 are almost 100% non-ABI / non-API,
> and with minor work we can make them clean.
>
> I would like Torvald and Rich to comment on patch #5
> to confirm that my interpretation is a defensible
> position for rwlock.
Patch #5 is still outsanding and needs consensus.
I would seem that Andi and I agree that POSIX is sufficiently
flexible to allow an elided rwlock.
I'm looking for feedback on my comments.
> I would like to see a v2 of patch 6 with tests for
> the new interfaces moved to the patch that adds the
> interface (patch #11).
Andi is reworking patch 6 to include a minimal set of
changes to ensure the existing testsuite runs cleanly.
> I would like to see patch 8 reworked slightly again
> to move the new interface documentation into the patch
> that adds the new interface (patch #11).
I've provided Andi with an incremental patch to cleanup
the documentation and I've indicated my intent to help
check this in if he's out of time.
I really want to drive these last 3 issues to completion
so we can get a baseline elision framework in place.
A baseline framework in 2.18 is important in that:
* It will reduce the work needed for a distro to enable
elision usefully for their experimental users.
* It will prevent fragmented and varied patches at the
distro level. We've already found several defects
with the elision support that will propagate to the
distros if we have baseline support.
* It will simplify the review of future patches that
depend upon the baseline support.
Cheers,
Carlos.