Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-23 Thread Ehsan Akhgari
On 2013-01-23 3:38 PM, Joe Drew wrote: On 2013-01-22 5:27 PM, Mike Hommey wrote: FWIW, IIRC my experiments last time we had this problem, LTCG alone accounts for less than a third of the performance boost we get from PGO on Talos. Did you happen to measure how big the linker got in memory doin

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-23 Thread Mike Hommey
On Wed, Jan 23, 2013 at 03:38:05PM -0500, Joe Drew wrote: > On 2013-01-22 5:27 PM, Mike Hommey wrote: > >FWIW, IIRC my experiments last time we had this problem, LTCG alone > >accounts for less than a third of the performance boost we get from > >PGO on Talos. > > Did you happen to measure how big

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-23 Thread Joe Drew
On 2013-01-22 5:27 PM, Mike Hommey wrote: FWIW, IIRC my experiments last time we had this problem, LTCG alone accounts for less than a third of the performance boost we get from PGO on Talos. Did you happen to measure how big the linker got in memory doing only LTCG? joe __

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Mike Hommey
On Tue, Jan 22, 2013 at 05:09:10PM -0500, Ehsan Akhgari wrote: > On 2013-01-22 4:40 PM, Robert O'Callahan wrote: > >On Wed, Jan 23, 2013 at 4:31 AM, Ehsan Akhgari >> wrote: > > > >But note that unless a given code path is examined throughout the > >profiling

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Ehsan Akhgari
On 2013-01-22 4:40 PM, Robert O'Callahan wrote: On Wed, Jan 23, 2013 at 4:31 AM, Ehsan Akhgari mailto:ehsan.akhg...@gmail.com>> wrote: But note that unless a given code path is examined throughout the profiling phase of a PGO build, PGO will probably have negligible effect on it, if

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Robert O'Callahan
On Wed, Jan 23, 2013 at 4:31 AM, Ehsan Akhgari wrote: > But note that unless a given code path is examined throughout the > profiling phase of a PGO build, PGO will probably have negligible effect on > it, if any. The PGO compiler looks for hot code paths and tries to > optimize those, so for exa

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Ehsan Akhgari
OK, everyone. Both mozilla-central and mozilla-inbound are *temporarily* reopened now. Please be gentle. Cheers, -- Ehsan On Tue, Jan 22, 2013 at 9:06 AM, Ehsan Akhgari wrote: > Status update #3: > > It seems like with PGO disabled for all of the above modules, we'

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Ehsan Akhgari
On 2013-01-22 9:45 AM, Marco Bonardo wrote: On 22/01/2013 15:36, Benjamin Smedberg wrote: On 1/22/2013 9:28 AM, Axel Hecht wrote: How are the perf numbers looking? One of the reasons for asking is that I expect RDF to be part of the startup and window-open codepaths, at least. I would not exp

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Marco Bonardo
On 22/01/2013 15:36, Benjamin Smedberg wrote: On 1/22/2013 9:28 AM, Axel Hecht wrote: How are the perf numbers looking? One of the reasons for asking is that I expect RDF to be part of the startup and window-open codepaths, at least. I would not expect PGO to optimize any of the RDF code for s

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Benjamin Smedberg
On 1/22/2013 9:28 AM, Axel Hecht wrote: How are the perf numbers looking? One of the reasons for asking is that I expect RDF to be part of the startup and window-open codepaths, at least. I would not expect PGO to optimize any of the RDF code for speed, even if they were in the startup codepat

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Axel Hecht
How are the perf numbers looking? One of the reasons for asking is that I expect RDF to be part of the startup and window-open codepaths, at least. I'm not overly concerned, but wanted to make sure we look. Axel On 22.01.13 15:06, Ehsan Akhgari wrote: Status update #3: It seems like with P

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Ehsan Akhgari
On Tue, Jan 22, 2013 at 7:35 AM, Mike Hommey wrote: > On Tue, Jan 22, 2013 at 01:30:50PM +0100, Mike Hommey wrote: > > On Mon, Jan 21, 2013 at 11:32:34PM -0500, Ehsan Akhgari wrote: > > > Second status update: > > > > > > The numbers from disabling PGO on image, accessible and webrtc are in, > an

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Ehsan Akhgari
Status update #3: It seems like with PGO disabled for all of the above modules, we've now decreased the linker max vmem size by about 500MB, which is nice. There is one PGO build bustage < https://tbpl.mozilla.org/php/getParsedLog.php?id=19006659&tree=Mozilla-Inbound> which has been re-triggered,

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Mike Hommey
On Tue, Jan 22, 2013 at 01:30:50PM +0100, Mike Hommey wrote: > On Mon, Jan 21, 2013 at 11:32:34PM -0500, Ehsan Akhgari wrote: > > Second status update: > > > > The numbers from disabling PGO on image, accessible and webrtc are in, and > > the linker max vmem size is down by only ~200MB, which is q

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-22 Thread Mike Hommey
On Mon, Jan 21, 2013 at 11:32:34PM -0500, Ehsan Akhgari wrote: > Second status update: > > The numbers from disabling PGO on image, accessible and webrtc are in, and > the linker max vmem size is down by only ~200MB, which is quite > disappointing, especially since according to Randell, putting we

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-21 Thread Ehsan Akhgari
Second status update: The numbers from disabling PGO on image, accessible and webrtc are in, and the linker max vmem size is down by only ~200MB, which is quite disappointing, especially since according to Randell, putting webrtc outside of libxul should buy us something around 600MB... So, as de

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-21 Thread Ehsan Akhgari
Status update: we have landed three patches on mozilla-inbound which disable PGO on the following directories (rdf/, image/ and accessible/) and I have triggered PGO builds on top of them to see how much they can shave off of the linker's vmem usage. Randel is also working on taking some webrt

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-21 Thread Mike Hommey
On Mon, Jan 21, 2013 at 10:56:28PM +0100, Mike Hommey wrote: > On Tue, Jan 22, 2013 at 10:52:39AM +1300, Robert O'Callahan wrote: > > Exactly what control do we have over what gets PGOed? In particular: > > > > 1) Are we able to exclude particular object files or libraries from PGO > > when we lin

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-21 Thread Kyle Huey
On Mon, Jan 21, 2013 at 1:52 PM, Robert O'Callahan wrote: > Exactly what control do we have over what gets PGOed? In particular: > > 1) Are we able to exclude particular object files or libraries from PGO > when we link libxul, reducing PGO memory usage for the final link? I think > you said "yes"

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-21 Thread Mike Hommey
On Tue, Jan 22, 2013 at 10:52:39AM +1300, Robert O'Callahan wrote: > Exactly what control do we have over what gets PGOed? In particular: > > 1) Are we able to exclude particular object files or libraries from PGO > when we link libxul, reducing PGO memory usage for the final link? I think > you s

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-21 Thread Robert O'Callahan
Exactly what control do we have over what gets PGOed? In particular: 1) Are we able to exclude particular object files or libraries from PGO when we link libxul, reducing PGO memory usage for the final link? I think you said "yes" on IRC. 2) Are we able to collect a set of object files, link them

Re: mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-21 Thread Marco Bonardo
On 21/01/2013 21:47, Ehsan Akhgari wrote: 1. mozilla-central/inbound will remain closed to all check-ins except the ones that do not add any code which gets linked into libxul on Windows. Both inbound and central are APPROVAL REQUIRED as of now, any patches that don't touch libxul can land wit

mozilla-central/inbound closed -- we hit the Windows PGO memory limit

2013-01-21 Thread Ehsan Akhgari
I'm sorry that I have to be the bearer of bad news about our trees, but we have finally started to hit the linker maximum memory size when linking libxul as part of our Windows PGO builds, and as a result, mozilla-central and inbound are CLOSED for now. I propose the following: 1. mozilla-cen