linux-bench
linux-bench copied to clipboard
Firefox Compile Benchmark
I think this is a good one to add and Ian @ Anandtech requested it also.
Basic idea: Download FF source to RAM disk Compile FF.
Maybe we use FF 40.0:
dependency
apt-get install zip
working on FF compile
wget https://archive.mozilla.org/pub/firefox/releases/40.0/source/firefox-40.0.source.tar.bz2 tar -xjvf firefox-40.0.source.tar.bz2 cd mozilla-release ./mach build
This is failing but should be close
How complete a build are we testing? We can probably build a basic FF without GTK dependencies, etc.. or if we go full steam, lots of dependencies.
I think it is more about being able to be consistent across architectures less about exactly how much we can compile. On Aug 27, 2015 7:43 AM, "Charles Nguyen" [email protected] wrote:
How complete a build are we testing? We can probably build a basic FF without GTK dependencies, etc.. or if we go full steam, lots of dependencies.
— Reply to this email directly or view it on GitHub https://github.com/STH-Dev/linux-bench/issues/69#issuecomment-135456105.
I have a working build, tested in a clean docker. Some things to think about:
- Do the options look ok?
- Compile speed can vary due to number of cores in the make step.
- Need to point to /dev/shm though the shm in my Docker is 64M and the build directory is 5.9GB. This may be a problem for the Docker version, but better for bare-metal version. https://github.com/docker/docker/pull/4981 I may try to integrate in later, but documented for now.
apt-get install python gtk+-2.0 dbus yasm libasound2-dev libxt-dev zip export SHELL=/bin/sh make
wget https://archive.mozilla.org/pub/firefox/releases/40.0/source/firefox-40.0.source.tar.bz2 tar xvfj firefox-40.0.source.tar.bz2 mkdir -p ffbuild; cd ffbuild ../mozilla-release/configure --disable-dbus --disable-pulseaudio --disable-gstreamer --disable-necko-wifi make -j 10
What about making a tmpfs and using that?
I think that is a standard workaround, passing /dev/shm as a volume into the docker container. Have to enable privileged mode and whatnot, but we can do that. Makes running it more complicated for users. I think we can do some testing on actual performance difference from tmpfs and disk.
Please excuse autocorrect typos, sent from a phone. On Aug 28, 2015 8:14 PM, "Sean Kelly" [email protected] wrote:
What about making a tmpfs and using that?
— Reply to this email directly or view it on GitHub https://github.com/STH-Dev/linux-bench/issues/69#issuecomment-135924531.