Talos Regression: Tp4 (Memset) increase 11.00% on XP Firefox
Nobody
Regression: Tp4 (Memset) increase 11.00% on XP Firefox
Previous results: 94025000.0 from build 20100315035441 of revision 17a3e86d5ec5 at 2010-03-15 05:57:00 on talos-r3-xp-019 run # 0
New results: 104365000.0 from build 20100315042737 of revision 8ff6056308bd at 2010-03-15 06:24:00 on talos-r3-xp-027 run # 0
Suspected checkin range: from revision 17a3e86d5ec5 to revision 8ff6056308bd
Nobody
Regression: Tp4 (Private Bytes) increase 11.03% on XP Firefox
Previous results: 91539700.0 from build 20100315035441 of revision 17a3e86d5ec5 at 2010-03-15 05:57:00 on talos-r3-xp-019 run # 0
New results: 101639000.0 from build 20100315042737 of revision 8ff6056308bd at 2010-03-15 06:24:00 on talos-r3-xp-027 run # 0
Nobody
Improvement: Tp4 decrease 5.96% on XP Firefox
Previous results: 276.737 from build 20100315035441 of revision 17a3e86d5ec5 at 2010-03-15 05:57:00 on talos-r3-xp-019 run # 0
New results: 260.232 from build 20100315042737 of revision 8ff6056308bd at 2010-03-15 06:24:00 on talos-r3-xp-027 run # 0
Daniel Holbert
to http://hg.mozilla.org/mozilla-central/rev/8ff6056308bd
which makes sense as a possible regressor (" bug 502906 - part 3 -
factor out Uniscribe and GDI shapers from Windows GDI font code.")
So I think this is a real regression... Commenting on bug.
On 03/16/2010 04:25 AM, Nobody wrote:
> Regression: Tp4 (Memset) increase 7.09%
> <http://graphs.mozilla.org/graph.html#tests=[{"machine":351,"test":45,"branch":1},{"machine":352,"test":45,"branch":1},{"machine":353,"test":45,"branch":1},{"machine":354,"test":45,"branch":1},{"machine":355,"test":45,"branch":1},{"machine":356,"test":45,"branch":1},{"machine":357,"test":45,"branch":1},{"machine":358,"test":45,"branch":1},{"machine":359,"test":45,"branch":1},{"machine":360,"test":45,"branch":1},{"machine":361,"test":45,"branch":1},{"machine":362,"test":45,"branch":1},{"machine":363,"test":45,"branch":1},{"machine":364,"test":45,"branch":1},{"machine":365,"test":45,"branch":1},{"machine":366,"test":45,"branch":1},{"machine":367,"test":45,"branch":1},{"machine":368,"test":45,"branch":1},{"machine":369,"test":45,"branch":1},{"machine":370,"test":45,"branch":1}]&sel=1268569440,1268742240>
> on Win7 Firefox
>
> Previous results: 87278200.0 from build 20100315035441 of revision
> 17a3e86d5ec5 at 2010-03-15 04:57:00 on talos-r3-w7-019 run # 0
>
> New results: 93468000.0 from build 20100315042737 of revision
> 8ff6056308bd at 2010-03-15 05:24:00 on talos-r3-w7-013 run # 0
>
> Suspected checkin range: from revision 17a3e86d5ec5 to revision
> 8ff6056308bd <
> http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=17a3e86d5ec5&tochange=8ff6056308bd>
>
>
>
> _______________________________________________
> dev-tree-management mailing list
> dev-tree-...@lists.mozilla.org
> https://lists.mozilla.org/listinfo/dev-tree-management
Benjamin Smedberg
> These Windows Tp4 (Memset) & Tp4 (Private Bytes) regressions all point
> to http://hg.mozilla.org/mozilla-central/rev/8ff6056308bd
> which makes sense as a possible regressor (" bug 502906 - part 3 -
> factor out Uniscribe and GDI shapers from Windows GDI font code.")
>
> So I think this is a real regression... Commenting on bug.
Yeah, it looks like we're trading memory for speed here. It may be a valid
tradeoff, but it doesn't look like that was discussed/expected.
--BDS