Opening The Gates To Our Daily Open-Source Linux Benchmark Results

[phoronix]

Phoronix: Opening The Gates To Our Daily Open-Source Linux Benchmark Results

For a few months now I've been talking about the LinuxBenchmarking.com initiative to provide daily benchmark results of the latest development Git/SVN code for various open-source projects in a fully-automated manner... Among the projects being tracked have been the Linux kernel, GCC, LLVM Clang, etc. There's dozens of systems at Phoronix Media in "the basement server room" doing nothing but running these upstream benchmarks day in and day out. The data flow is now open at LinuxBenchmarking.com.

http://www.phoronix.com/vr.php?view=21708

[darkbasic]

Wonderful, good job Michael!
I see lots of regressions in r600g...

[Michael]

Wonderful, good job Michael!
I see lots of regressions in r600g...

Thanks.

On what systems/days? Not sure if anything relevant to any system changes, but there's obviously highlighted any system hardware/software differences worth reporting -- when it came to the graphics tracker, one OS had to be reloaded in the past few months and there were a few systems acting up early on but now they all seem to be running fairly stable and well.

As a reminder to anyone, patches always welcome to extend it -- like one of the features I haven't yet done is the system log viewing via LinuxBenchmarking.com but anyone interested could look at my Git commits today to see how I implemented the basic exporter and viewer of the exported data (the viewer is less than 200 lines of PHP and HTML), etc. I did do basic mobile support for the viewer but any further improvements there would be another nice way to get started, etc. Or of course anyone not wanting to offer patches or even just suggestions and feedback but are able to help out financially can offer a tip/donation. It's been a long road getting the server room all setup and this viewer all setup but the PTS/Phoromatic infrastructure continues chugging along smoothly.

[Creak]

Good job Michael, this is an instant bookmark for me

I don't understand why there are some gaps in the open-source drivers graph lines (http://linuxbenchmarking.com/?latest...linux-graphics), and also why some configurations are there on one benchmark, but not the other (e.g. Core i7 3960X).

It would also be great to have a more relevant name for the configs. For instance, using this "radeon (HD 6770)", instead of "Core i7 3960X".

Anyway, I suppose all the glitches will go away as the whole structure stabilizes
Thanks for all that, it's impressive!

[Michael]

Good job Michael, this is an instant bookmark for me

I don't understand why there are some gaps in the open-source drivers graph lines (http://linuxbenchmarking.com/?latest...linux-graphics), and also why some configurations are there on one benchmark, but not the other (e.g. Core i7 3960X).

It would also be great to have a more relevant name for the configs. For instance, using this "radeon (HD 6770)", instead of "Core i7 3960X".

Anyway, I suppose all the glitches will go away as the whole structure stabilizes
Thanks for all that, it's impressive!

Some of the gaps are due to either time when moving systems from spare bedroom down into the basement server room, when doing upgrades on room, when electricity goes down, or just other such scenarios. Or when a kernel would hang and Phoromatic would know it but I wouldn't get around to manually rebooting system until next day, etc.

The i7 3960X system in particularly is a bit finicky with some of the tests for either compatibility issues with R600g or something, haven't had too much time to explore.

Any naming improvements also welcome... For the Core i7 3960X system, originally I had it part of a different tracker but when having to replace the hard drive in it, ended up re-trasking it to the graphics tracker and just didn't rename it. I'll rename it now and with my next export/import of data to the LinuxBenchmarking.com server, it will update all of the graphics results for it to more accurate naming.

Thanks.

[utack]

May the results be with you!

[Creak]

Some of the gaps are due to either time when moving systems from spare bedroom down into the basement server room, when doing upgrades on room, when electricity goes down, or just other such scenarios. Or when a kernel would hang and Phoromatic would know it but I wouldn't get around to manually rebooting system until next day, etc.

The i7 3960X system in particularly is a bit finicky with some of the tests for either compatibility issues with R600g or something, haven't had too much time to explore.

Any naming improvements also welcome... For the Core i7 3960X system, originally I had it part of a different tracker but when having to replace the hard drive in it, ended up re-trasking it to the graphics tracker and just didn't rename it. I'll rename it now and with my next export/import of data to the LinuxBenchmarking.com server, it will update all of the graphics results for it to more accurate naming.

Thanks.

Thanks a lot!

[Michael]

Thanks a lot!

Cheers! Let me know if any other feedback arises.

[darkbasic]

Michael I would expect to see fps when I hover with the mouse on the graphs! Of course it should hook on the nearest result, it shouldn't report the fps corresponding to the mouse's absolute position.

[curaga]

Style quip: the orange is really unreadable against the white bg, and especially against the option grey. http://imgur.com/6KVSXuC