Run time differences in Linux gcc versions

Hi all,
In my home computer I've got Slackware 12 with gcc 4.1.2. I wrote a C++ program (could attach the code if you want) and I got a run time of around 2.2s. I transferred the same C++ program to my work computer that has Fedora Core 4 with gcc 4.0.0 and I found that it takes 16s to run. The work computer has more RAM (1 GB) than my home computer (512 MB) and both are AMD 64 bit proc.

I compiled and installed gcc 4.1.2 in my work computer and now find the run times have reduced to 2.2s. Is there a difference in the gcc compiler versions?

Thanks.

Ehm, quite obviously the answer to that is "Yes".

What the difference is, and how why that makes so much difference in this case is something we could only [possibly] say by looking at the code [and more likely by investigating the difference in the produced assembler code].

--
Mats

gcc can change very dramatically from version to version. The difference is not surprising.

Thanks for the responses.

I am attaching the program.

Especially if the program does something that is very well optimized by a new optimizer in the new version.

Note also that 4.1 is a big improvement over 4.0. 4.0 introduced a new optimization framework, but only 4.1 really makes use of it. 4.0 is also said to be quite riddled with bugs. You shouldn't use it.

If you didn't try various optimization levels (-ON for various values of N, 0 is the default) with both versions, it's possible that 4.1.2 was able to optimize at default level something that 4.0.0 would need a higher level to be able to find.

Yes, I would say that optimization levels may affect the code efficiency, and the x.0 versions of gcc often have limitations in the efficency of new features, whilst the later update versions are much better at using these new features.

--
Mats

Using the -ON optimization options reduces run times to 9.6s. I used all various levels but couldn't get run times lesser than that. Guess I'll use gcc 4.1.2 on all my computers.

Thanks.