This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Performance regression testing?

On Nov 28, 2005, at 8:41 PM, Hans-Peter Nilsson wrote:
runtime,-O1,zlib-1.1.4:minigzip,previous 0.32

Ah, ok, good. I'd eject the ,previous to the filename, and reorder slightly, but, certainly that is trivial to do.

Can't be compared with each other

I suspect we're in agreement, though I missed the detail of what the two items were to know exactly what you meant.

...and emitting PASS/FAIL after _comparing_ two or more arbitrary
runs according to some criteria?  Like above?  ;-)

Yes. :-)

You have to elaborate here.  How does "biasing" the number of
cycles to make it 0 help?

No, I don't mean bias. I only meant to imply direction, bigger == worse, and that the numbers are non-negative, that is all. As for normalization, yeah, well, that's a very interesting topic in its own right, I won't dwell on it here. Let's just say that it winds up being a vote of confidence/relevance of each testcase.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]