Is it me or is gcc 3.2.2 producing slower code?

[captnjameskirk]

OK, I will admit I have been away from Gentoo for awhile, but I'm back and that's all that matters, right? Well, my newly emerged gcc 3.2.2 system with fairly strong optimization flags (-march=pentium3 -O3 -fomit-frame-pointer -funrool-loops -frerun-loop-opt -frerun-cse-after-loop -falign-functions=4 -fexpensive-optimizations [for a Pentium III/Celeron Copermine]) seems to have certain programs running noticably slower than those compiled with gcc 2.95.3, sometimes 50% slower or more. For example, a game running in X dga-mode now runs at 15 frames per second, whereas before it ran at 35 fps on the same system compiled with gcc 2.95.3 with fewer optimizations. The same goes for Mplayer showing avi files, not quite as drastic but still very noticably slower. And this is in fluxbox with no other programs running, not even gkrellm (the faster version was usually run in kde).

Overall system response is fine, nothing is hanging anywhere, no "out of control" processes, no compile errors. This slowdown is only noticable with very processor-intensive apps, but then again, that's why I switched back to Gentoo, so I could have maxed-out system performance.

Could some of my optimizations actually be causing the number-crunching apps to slow down, or is this a gcc 3.2.2 vs 2.95.3 thing? Or is it just something wacked on my system?
_________________
Are you out of your Vulcan mind?

[TwistedKestrel]

OK, I myself will admit I don't know a whole friggin lot about Linux. BUT the two tasks you mentioned (game in X-dga, playing AVI) sound a lot more video intensive than processor intensive. I myself would suspect that a video driver may have changed in X since you were 'gone' from Gentoo, or just that there might be a stray line in a X configuration file somewhere (or kernel, or something). Now that would be as far as I'd know , but it might help everyone smarter than me (which is everyone) to mention your video card + chipset.

[captnjameskirk]

Actually, the game speed relies more on the cpu, not the graphics card (it's an emulation under xmame). I am using the exact same XF86Config file as with the previous (faster) version, and both were X 4.2, though the new version is compiled with gcc 3.2.2 and the old was compiled with 2.95.3, which is why I'm leaning towards gcc as the culprit.

FWIW, the graphics card is a Trident CybebladeAi1, and the chipset is supported in X and in the XF86Config file.
_________________
Are you out of your Vulcan mind?

[AlterEgo]

If you want to be sure it's your compiler or compiler options:
Freebench is a very suitable tool for benchmarking the performance-effect of compiler and compiler flags.

If that's too much trouble (which it really isn't ) : in my view TwistedKestrel's suggestions make a lot of sense.