OpenMandriva Lx 4.2 Released For This Linux Distro Built By The Clang Compiler

[phoronix]

Phoronix: OpenMandriva Lx 4.2 Released For This Linux Distro Built By The Clang Compiler

One year after OpenMandriva Lx 4.1 released, OpenMandriva Lx 4.2 is now available. Continuing to make OpenMandriva Lx rather unique among Linux distributions is its use of the LLVM Clang compiler by default rather than GCC. Another unique "selling point" of OpenMandriva is its AMD Zen optimized version where the entire package set is rebuilt with Zen optimizations...

http://www.phoronix.com/scan.php?page=news_item&px=OpenMandriva-Lx-4.2-Released

[Charlie68]

Sorry for the ignorance, but what are the advantages of using LLVM Clang instead of GCC?

[tildearrow]

KDE Plasma 5.20.25

This version of Plasma does not exist...

[ms178]

Sorry for the ignorance, but what are the advantages of using LLVM Clang instead of GCC?

Some people simply dislike GNU tools. And while there are plenty of differences under the hood between GCC and Clang, you would need to benchmark every program to see which compiler leads to the best performing binaries for a particular application. Phoronix testing showed that past Mandriva versions were slower than other GCC-based distributions. Maybe the developers invested more time in compiler tuning this time, I hope we will see fresh numbers soon of such a distribution comparison.

[jabl]

Sorry for the ignorance, but what are the advantages of using LLVM Clang instead of GCC?

For the end user, very little difference. clang has roughly caught up with gcc.

For compiler developers, llvm/clang has a more modern and modular structure, making it easier to hack on, so it has sort of become the standard vehicle for academic compiler research. Also a lot of corporations that are, for better or worse, suspicious of GPLv3, prefer to contribute to LLVM. So it wouldn't surprise me if LLVM where to eventually overtake GCC as the "standard" open source compiler.

[xpris]

Phoronix testing showed that past Mandriva versions were slower than other GCC-based distributions. Maybe the developers invested more time in compiler tuning this time, I hope we will see fresh numbers soon of such a distribution comparison.

Previous release of OpenMandriva perform not that bad. Look at this:

https://www.phoronix.com/scan.php?page=article&item=openmandriva-41-zen&num=2

Some tests win, in some is in the middle and some loss. Not bad but in every benchmarks we need to the cpugovernor. For example, Clear uses performance by default which gives a performance boost, but causes your computer to use a lot more electricity and heats - which is especially important in laptops. In OpenMandriva we use ondenamed - which many other systems use. Therefore, do not be surprised if the system performs similarly to others, but loses to one of the cpugovernor in the performance.

[ms178]

Previous release of OpenMandriva perform not that bad. Look at this:

https://www.phoronix.com/scan.php?page=article&item=openmandriva-41-zen&num=2

Some tests win, in some is in the middle and some loss. Not bad but in every benchmarks we need to the cpugovernor. For example, Clear uses performance by default which gives a performance boost, but causes your computer to use a lot more electricity and heats - which is especially important in laptops. In OpenMandriva we use ondenamed - which many other systems use. Therefore, do not be surprised if the system performs similarly to others, but loses to one of the cpugovernor in the performance.

Granted, there are other differences such as the used CPU governor which have an impact, not just the compiler. And yes, the newer tests were showing some improvements relative to other distributions. I was referring to these tests from 2019: https://www.phoronix.com/scan.php?pa...r1-linux&num=8

[xpris]

Sorry for the ignorance, but what are the advantages of using LLVM Clang instead of GCC?

Bero (leader of OpenMandriva) talked about advantages of using Clang instead of GCC at one of the conferences. You can hear about it in the video below:

https://www.youtube.com/watch?v=QinoajSKQ1k

Quick:
- better compile times,
- less pain with crosscompilers
- sanitizers
- more readable code
and more.

[xpris]

Granted, there are other differences such as the used CPU governor which have an impact, not just the compiler. And yes, the newer tests were showing some improvements relative to other distributions. I was referring to these tests from 2019: https://www.phoronix.com/scan.php?pa...r1-linux&num=8

As you will see on page 5. In the configuration table you will see that OpenMandriva back then used cpugovernor on acpi-cpufreq conservative. When other systems are either ondenamed or performance. It had its advantages and disadvantages. That is why ondenamed is already used in new editions.
When we combine some distribution operating with the same cpugovernor the differences in compilers seem to be minimal.

[ms178]

As you will see on page 5. In the configuration table you will see that OpenMandriva back then used cpugovernor on acpi-cpufreq conservative. When other systems are either ondenamed or performance. It had its advantages and disadvantages. That is why ondenamed is already used in new editions.
When we combine some distribution operating with the same cpugovernor the differences in compilers seem to be minimal.

From my own testing between GCC 10.2.1 vs. Clang 11.0.1 on Mesa, I found Clang to be better performing (68 vs 72 fps in Company of Heroes 2) for my setup (with all other settings being equal). But I doubt that benchmarking performance-sensitive packages and chosing the better performing compiler would be practical for smaller distributions (AFAIK Clear Linux does this).