62 Benchmarks, 12 Systems, 4 Compilers: Our Most Extensive Benchmarks Yet Of GCC vs. Clang Performance

[phoronix]

Phoronix: 62 Benchmarks, 12 Systems, 4 Compilers: Our Most Extensive Benchmarks Yet Of GCC vs. Clang Performance

After nearly two weeks of benchmarking, here is a look at our most extensive Linux x86_64 compiler comparison yet between the latest stable and development releases of the GCC and LLVM Clang C/C++ compilers. Tested with GCC 8, GCC 9.0.1 development, LLVM Clang 7.0.1, and LLVM Clang 8.0 SVN were tests on 12 distinct 64-bit systems and a total of 62 benchmarks run on each system with each of the four compilers... Here's a look at this massive data set for seeing the current GCC vs. Clang performance.

http://www.phoronix.com/vr.php?view=27499

[tildearrow]

Just for reference, the Summer Nature video lasts 3603 frames. As an example, the 2990WX scores 169.9 FPS (3603/time).

[tildearrow]

Typo:

Prmeium gets you ad-free access to the site

[Michael]

You didn't accept my dav1d patches yet?

Queued for the next time there is a dav1d update / version bump.

[HoermI]

Wow, I really like the summary of all benchmarks!
Somehow though, it would be a bit more motivating to read the whole article, if the summary graph was at the end.
Another little suggestion would be graphs that switch between absolute values and percentages when clicking on them.
Still, thank you so much for these thorough benchmarks!

[linner]

I'm glad someone does this testing. Kind of works out the way I expected. Honestly not that much difference with gcc very slightly better.

Now lets see how they compare to ICC. ie. we know Clear Linux is fast. Is it just the compiler?

[Michael]

I'm glad someone does this testing. Kind of works out the way I expected. Honestly not that much difference with gcc very slightly better.

Now lets see how they compare to ICC. ie. we know Clear Linux is fast. Is it just the compiler?

Clear Linux does NOT use ICC, but GCC and Clang.

[linner]

Clear Linux does NOT use ICC, but GCC and Clang.

Really? I read they use ICC for a good portion of stuff. I guess you can tell I've never actually tried Clear Linux.

[boxie]

That is a massive series of tests Michael - well done! Good to see that on average that the compilers are so close

For shits and giggles - do you have an ARM SBC that you can throw into that mix? I think that is where the next big battleground in compiler performance is

[Compartmentalisation]

62 benchmarks, 12 systems, 4 compilers to tell me that it's all about the same for me as a consumer of them! It's interesting that sometimes very specific processors can find more significant differences in some benchmarks. It's also interesting that I'm not the only one getting slow encodes with VP9.