As requested by spyderous, I'm reposting the bug I reported at https://bugs.gentoo.org/show_bug.cgi?id=107751 . For those who would rather not click, I'm basically proposing that the patches Ubuntu add to glxgears which prevent the FPS rate being shown unless the user specifies "-iacknowledgethatthistoolisnotabenchmark" be added to the main Mesa tree. Now, this is not extremely important. But it's quite annoying seeing countless glxgears comparisons all over the place. They can just be ignored, but I think an acknowledgement like this would be a good idea as it seems many users believe it is a valid benchmark. Just an idea I thought I'd throw around, critisisms/flames welcome :)
I was about to file a bug for this, then thought I'd better check, and noticed there was one already. I heartily endorse this product and/or opinion. All over the world, people post glxgears results, only to be told by X / Mesa developers and their pupils that glxgears isn't a benchmark. Since glxgears isn't a benchmark, it shouldn't act like one. It shouldn't spit out FPS numbers to the console by default.
I concur, but please use a sane option name like "-show-fps" with a suitable explanation in the help text -- "-iacknowledgethatthistoolisnotabenchmark" is just childish.
Actually gears *is* a benchmark -- primarily a benchmark of swapbuffers. The fact that swapbuffers has gotten a lot slower recently isn't a justification for turning off the output. I can't really think of any reason to run gears except to get the fps number -- what other purpose does that application have?
bugzilla-daemon@freedesktop.org writes: > --- Comment #3 from Keith Whitwell <keith@tungstengraphics.com> 2009-06-18 0 > 2:47:22 PST --- > > I can't really think of any reason to run gears except to get the fps number > what other purpose does that application have? We sometimes tell our users to run gears to verify they're getting hardware acceleration. Despite the `gears is not a benchmark' arguments, the difference between a swrast-backed gears and a <insert gpu here>-backed gears is significant enough that one doesn't actually need to look at any numbers.
(In reply to comment #4) > > We sometimes tell our users to run gears to verify they're getting > hardware acceleration. Why not just have them look at the glxinfo output?
bugzilla-daemon@freedesktop.org writes: > --- Comment #5 from Dan Nicholson <dbn.lists@gmail.com> 2009-06-18 14:47:04 > PST --- > (In reply to comment #4) > > > > We sometimes tell our users to run gears to verify they're getting > > hardware acceleration. > > Why not just have them look at the glxinfo output? We do, when we ask them to send us information. Or sometimes we know the user, and know that they'd be capable of reading glxinfo output and piecing together whether or not they're getting accelerated rendering. glxgears is easier though, as it's more obvious to a user that everything is setup correctly. If there was a single line we could grep out, we might choose to ask them to run glxinfo in more cases. As of now, we'd have to explain why `direct rendering' is a Good Thing, and there's too many cases for "OpenGL vendor string"/glGetString(GL_VENDOR) to be worth explaining.
(In reply to comment #4) > We sometimes tell our users to run gears to verify they're getting > hardware acceleration. Despite the `gears is not a benchmark' > arguments, the difference between a swrast-backed gears and a <insert > gpu here>-backed gears is significant enough that one doesn't actually > need to look at any numbers. I'm not sure I understand -- glxgears looks smooth to me, regardless of whether it's software rastered or hardware accelerated. I need to see the FPS number (200~300 fps vs. > 1000fps) to see whether I get the one or the other.
bugzilla-daemon@freedesktop.org writes: > --- Comment #7 from Nils Philippsen <nphilipp@redhat.com> 2009-06-19 02:20:5 > 9 PST --- > (In reply to comment #4) > > We sometimes tell our users to run gears to verify they're getting > > hardware acceleration. Despite the `gears is not a benchmark' > > arguments, the difference between a swrast-backed gears and a <insert > > gpu here>-backed gears is significant enough that one doesn't actually > > need to look at any numbers. > > I'm not sure I understand -- glxgears looks smooth to me, regardless > of wheth er it's software rastered or hardware accelerated. I need to > see the FPS number (200~300 fps vs. > 1000fps) to see whether I get > the one or the other. When I have an indirect rendering context, gears seems to stutter at times. Perhaps stutter is a bad word; the speed changes noticeably. On my beefy new workstation though, yeah, it's no longer clear which gears is the hw-backed one and which gears is swrast-backed. That wasn't the case last time I tested, though in retrospect I guess it's been a while since I've done so.
(In reply to comment #3) > Actually gears *is* a benchmark -- primarily a benchmark of swapbuffers. The > fact that swapbuffers has gotten a lot slower recently isn't a justification > for turning off the output. > > I can't really think of any reason to run gears except to get the fps number -- > what other purpose does that application have? I agree. Anyway, no matter if we print fps or not, some people will always consider it a benchmark of something else than swapbuffers. That's their problem, not ours. Closing.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.