Phoronix Test Suite 4.6.1 Drives OS X Benchmarking

[phoronix]

Phoronix: Phoronix Test Suite 4.6.1 Drives OS X Benchmarking

The first point release to Phoronix Test Suite "Utsira" has been released with some marked improvements for Apple OS X and other minor enhancements...

http://www.phoronix.com/vr.php?view=MTM4NTc

[pandev92]

how can I install the suite on osx?

[mendieta]

GUI?

Anyone knows if the web browser based GUI is already functional? Thanks!

[mendieta]

Oh, and one suggestion, Michael, if you are reading. Can you please output results in a way that "more is better" always? This way, once can run a suite of tests and get a global score, say through the geometric mean. It would also make the results easier to digest when browsing the plots

Thanks!

[Michael]

how can I install the suite on osx?

Just download the package, extract it, and run ./phoronix-test-suite benchmark whatever

[Michael]

Anyone knows if the web browser based GUI is already functional? Thanks!

The HTML5 GUI is still held up by not having any compelling artwork yet, so I guess my current aim is for it to be mastered for PTS 5.0.

Oh, and one suggestion, Michael, if you are reading. Can you please output results in a way that "more is better" always? This way, once can run a suite of tests and get a global score, say through the geometric mean. It would also make the results easier to digest when browsing the plots

Thanks!

Geometric mean should already accomodate for that and divide LIB results by 1 / result. The overview graphs and other select areas also do that to make a HIB result. However, in terms of actually showing all graphs that way, I don't have any meaningful way to illustate it while still giving an actually meaningful number.

[mendieta]

Geometric mean should already accomodate for that and divide LIB results by 1 / result. The overview graphs and other select areas also do that to make a HIB result. However, in terms of actually showing all graphs that way, I don't have any meaningful way to illustate it while still giving an actually meaningful number.

Oh, I believe there is a bug then:
http://openbenchmarking.org/result/1...Phenom+X3+8650

But it would be great if it worked the way you described it. Also, note that you _can_ return something meaningful. For instance, for a timed compilation, ou can return number of compilations per hour. That is a meaningful quantity.

[mendieta]

Oh, I believe there is a bug then:
http://openbenchmarking.org/result/1...Phenom+X3+8650

But it would be great if it worked the way you described it. Also, note that you _can_ return something meaningful. For instance, for a timed compilation, ou can return number of compilations per hour. That is a meaningful quantity.

Sorry, I should have been clearer, here in these tests, there are 3 processors, with very different capabilities. One of them is always about twice as fast as the slowest, and the other three times as fast, but the geometric mean is about the same for all of them ... by the way, is there a bugtracker for the for PTS? I couldn't find one ...