On 7/2/14, 2:52 PM, Kim Gräsman wrote:
On Wed, Jul 2, 2014 at 9:45 PM, Kim Gräsman <kim.gras...@gmail.com> wrote:
Ah, I thought it was just made non-default. Thanks, waiting for that
full build to complete!
OK, so I now have preliminary perf data.
- Before is ``mach build -p content`` with the plain pymake in mozilla-central
- After is ``mach build -p content`` with pymake from GitHub with an
additional patch to remove the iter- prefix on all dict enumeration. I
also had to revert this [1] commit to be able to build without
failing.
Before After
34,701 36,98
35,097 36,196
34,993 35,697
35,054 36,741
35,244 34,736
34,68 36,555
35,494 35,043
35,437 36,282
Average 35,0875 36,02875
So it's almost a full second slower on average for the null build of content.
The only perf-sensitive change I've done so far is the dict iteration
stuff, but I can run a test without it tomorrow, to make sure it's not
something else.
How do you feel about this?
That performance hit is negligible and probably within expected
deviation due to system load, page cache state, etc. Let's nix the iter
calls.
Thank you for measuring this.
_______________________________________________
dev-builds mailing list
dev-builds@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-builds
