GCC 14 vs. LLVM Clang 18 Compiler Performance On Fedora 40

[swastika]

I have filed performance bug reports a long time ago (for both gcc and clang) and they were more-or-less ignored. Kindly let me know if this time it'd be different ...

If they say it is going to be different, how can you possibly rely on that? If you filed it, reference them so we know what you are talking about but there should be no expectation of a commitment unless you are willing to pay for these services. Free software ! = free labor

[avis]

Quake and Doom are not horribly coded nor are they entirely GPU bound. It is valid to use them as a test case.

Are you talking about Quake 3 and Doom 2, because AFAIK there's nothing more recent which is open source.

Doom 2 runs at ~3000fps and Quake 3 runs at ~1500fps, so what are you trying to find out and how this information is applicable to the average user?

[avis]

I have filed performance bug reports a long time ago (for both gcc and clang) and they were more-or-less ignored. Kindly let me know if this time it'd be different ...

99% of bugs that I've filed against GCC received proper attention. No idea why you get this treatment. Maybe you could provide the links to the appropriate bug reports.

[swastika]

Are you talking about Quake 3 and Doom 2, because AFAIK there's nothing more recent which is open source.

Doom 2 runs at ~3000fps and Quake 3 runs at ~1500fps, so what are you trying to find out and how this information is applicable to the average user?

FYI, Doom 3 is open source but more importantly, you don't need game engines to be open source to benchmark them. It is industry practice to provide these engines to the press for benchmarks. Average user does not care about benchmarks really. It's mostly tech enthusiasts who obsess over micro benchmarks.

[phoronix38]

I would like to see Eigen and/or xsimd benchmark. I have seen quite big wins for Clang with these libraries.

[vient]

Doom 2 runs at ~3000fps and Quake 3 runs at ~1500fps, so what are you trying to find out and how this information is applicable to the average user?

I would say this information is a lot more meaningful for average user than some SMHasher or JohnTheRipper result. A game is a complex application with lots of different activities inside so it's more representative of an "average" application (or at least "average" game which is also valuable info for average user).

If average users read Phoronix, is another question.

[vladpetric]

99% of bugs that I've filed against GCC received proper attention. No idea why you get this treatment. Maybe you could provide the links to the appropriate bug reports.

For ICEs and bad output, I agree. Not so for performance bugs in my experience.

[carewolf]

Not a lot of sense in comparing video codecs because they include hand written assembly for most important/heavy parts of processing, so the compiler's role is quite minimal.

SMHasher SHA3-256 could be sped up on GCC a lot: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113235

No fix is proposed just yet.

--param max-completely-peel-times=30

[carewolf]

For ICEs and bad output, I agree. Not so for performance bugs in my experience.

The performce bugs I have filed has been adressed, especially if you debug it down to general cases and assembler. Though I also fixed a few of them myself. The only one I can remember that wasn't fixed, was because it was caused by a specific micro-ops limit in Sandybridge processors, so slightly more agressive optimization caused a regression, but they didn't want to hardcore such specific limits for all processors. clang also doesnt have that.

[avis]

--param max-completely-peel-times=30

This is a workaround/hack, not a fix. I can read as well, thank you very much. Michael will not change PTS and apply this flag to this particular benchmark because we're testing standard compilers, not a load of hacks to speed up certain tests arbitrarily.