Arch Linux outperforms Clear Linux

I have tested clear linux against my good old arch linux.

Here are the specifications:

Kernel

Arch Linux

Linux version 5.8.12-xanmod1-1 (makepkg@archlinux) (gcc (GCC) 10.2.0, GNU ld (GNU Binutils) 2.35) #1 SMP PREEMPT Wed, 30 Sep 2020 18:52:16 +0000

More about my Arch Linux’s Kernel:
Xanmod with Cachy CPU Scheduler, and mq-deadline IO sched)
Compiled with native optimization:

mtune=native, march=native, O3, NUMA disabled

Clear Linux

Linux version 5.7.13-975.native (mockbuild@94d88f9413f140e797a978590df0fcba) (gcc version 10.2.1 20200804 releases/gcc-10.2.0-57-ge5907f3b63 (Clear Linux OS for Intel Architecture), GNU ld (GNU Binutils) 2.35.0.20190203) #1 SMP Wed Aug 5 03:03:05 PDT 2020

Boot Time

Although this is a live boot to clear linux, the systemd-analyze should return the appropriate time after the live root is loaded.

Arch

$ systemd-analyze time
Startup finished in 1.692s (kernel) + 1.000s (userspace) = 2.692s 
graphical.target reached after 987ms in userspace

Clear

$ systemd-analyze time
Startup finished in 4.695s (kernel) + 2.007s (userspace) = 6.702s 
graphical.target reached after 1.894s in userspace

Hardinfo Benchmarks

FPU Raytracing

[Lower is better]
Arch: This Machine Intel Core i3-4150 4x 3500.00 MHz 1.62
Clear: Intel® Core™ i3-4150 CPU @ 3.50GHz 4x 3500.00 MHz 1.66

FPU FFT

[Lower is better]
Arch: This Machine Intel Core i3-4150 4x 3500.00 MHz 1.49
Clear: Intel® Core™ i3-4150 CPU @ 3.50GHz 4x 3500.00 MHz 1.55

CPU N-Queens

[Lower is better]
Arch: This Machine Intel Core i3-4150 4x 3500.00 MHz 3.15
Clear: Intel® Core™ i3-4150 CPU @ 3.50GHz 4x 3500.00 MHz 3.63

CPU Fibonacci

[Lower is better]
Arch: This Machine Intel Core i3-4150 4x 3500.00 MHz 0.37
Clear: Intel® Core™ i3-4150 CPU @ 3.50GHz 4x 3500.00 MHz 0.37

CPU CryptoHash

[Higher is better]
Arch: This Machine Intel Core i3-4150 4x 3500.00 MHz 516.12
Clear: Intel® Core™ i3-4150 CPU @ 3.50GHz 4x 3500.00 MHz 486.19

CPU Blowfish (Multi-core)

[Higher is better]
Arch: This Machine Intel Core i3-4150 4x 3500.00 MHz 21.59
Clear: Intel® Core™ i3-4150 CPU @ 3.50GHz 4x 3500.00 MHz 21.54

CPU Blowfish (Multi-thread)

[Higher is better]
Arch: This Machine Intel Core i3-4150 4x 3500.00 MHz 35.97
Clear: Intel® Core™ i3-4150 CPU @ 3.50GHz 4x 3500.00 MHz 33.43

CPU Blowfish (Single-thread)

[Higher is better]
Arch: This Machine Intel Core i3-4150 4x 3500.00 MHz 10.73
Clear: Intel® Core™ i3-4150 CPU @ 3.50GHz 4x 3500.00 MHz 10.78


Unigine Heaven


My Ruby benchmark

[https://github.com/Souravgoswami/simple-ruby-benchmark/]

Arch

$ ruby benchmark.rb 
:: Details:
Ruby Version: 2.7.1 (x86_64-linux)

CC: gcc
CFLAGS: -march=x86-64 -mtune=generic -O2 -pipe -fno-plt -fPIC
--------------------------------------------------------------------------------
:: Please stop all your apps, perhaps reboot your system, and run the benchmark
:: Don't even move your mouse during the benchmark for consistent result!
| Ready? (0)...................................................................
CPU Blowfish Test
:: CPU Blowfish Iteration 1: 0.119s
:: CPU Blowfish Iteration 2: 0.068s
:: CPU Blowfish Iteration 3: 0.068s
:: CPU Blowfish Iteration 4: 0.068s
:: CPU Blowfish Iteration 5: 0.068s
:: CPU Blowfish Iteration 6: 0.068s
:: CPU Blowfish Iteration 7: 0.067s
:: CPU Blowfish Iteration 8: 0.067s
:: CPU Blowfish Iteration 9: 0.067s
:: CPU Blowfish Iteration 10: 0.071s
Total time taken: 0.731s
--------------------------------------------------------------------------------
FPU Test
:: FPU Math Iteration 1: 0.377s
:: FPU Math Iteration 2: 0.379s
:: FPU Math Iteration 3: 0.377s
:: FPU Math Iteration 4: 0.377s
:: FPU Math Iteration 5: 0.377s
:: FPU Math Iteration 6: 0.378s
:: FPU Math Iteration 7: 0.377s
:: FPU Math Iteration 8: 0.377s
:: FPU Math Iteration 9: 0.377s
:: FPU Math Iteration 10: 0.377s
Total time taken: 3.773s
--------------------------------------------------------------------------------
CPU Fibonacci Test
:: CPU Fibonacci Iteration 1: 0.476s
:: CPU Fibonacci Iteration 2: 0.438s
:: CPU Fibonacci Iteration 3: 0.412s
:: CPU Fibonacci Iteration 4: 0.408s
:: CPU Fibonacci Iteration 5: 0.396s
:: CPU Fibonacci Iteration 6: 0.401s
:: CPU Fibonacci Iteration 7: 0.403s
:: CPU Fibonacci Iteration 8: 0.402s
:: CPU Fibonacci Iteration 9: 0.407s
:: CPU Fibonacci Iteration 10: 0.419s
Total time taken: 4.162s
--------------------------------------------------------------------------------
CPU Anagram Hunt
:: CPU Anagram Iteration 1: 0.304s
:: CPU Anagram Iteration 2: 0.305s
:: CPU Anagram Iteration 3: 0.306s
:: CPU Anagram Iteration 4: 0.301s
:: CPU Anagram Iteration 5: 0.299s
:: CPU Anagram Iteration 6: 0.302s
:: CPU Anagram Iteration 7: 0.299s
:: CPU Anagram Iteration 8: 0.299s
:: CPU Anagram Iteration 9: 0.300s
:: CPU Anagram Iteration 10: 0.299s
Total time taken: 3.014s
--------------------------------------------------------------------------------
CPU 2 Million Prime Numbers
:: Prime Numbers Iteration 1: 0.326s
:: Prime Numbers Iteration 2: 0.327s
:: Prime Numbers Iteration 3: 0.325s
:: Prime Numbers Iteration 4: 0.325s
:: Prime Numbers Iteration 5: 0.328s
:: Prime Numbers Iteration 6: 0.325s
:: Prime Numbers Iteration 7: 0.330s
:: Prime Numbers Iteration 8: 0.326s
:: Prime Numbers Iteration 9: 0.325s
:: Prime Numbers Iteration 10: 0.325s
Total time taken: 3.262s
--------------------------------------------------------------------------------
CPU 2k Pi Digits
:: 2K Pi Digits Iteration 1: 0.362s
:: 2K Pi Digits Iteration 2: 0.391s
:: 2K Pi Digits Iteration 3: 0.392s
:: 2K Pi Digits Iteration 4: 0.394s
:: 2K Pi Digits Iteration 5: 0.335s
:: 2K Pi Digits Iteration 6: 0.390s
:: 2K Pi Digits Iteration 7: 0.392s
:: 2K Pi Digits Iteration 8: 0.392s
:: 2K Pi Digits Iteration 9: 0.336s
:: 2K Pi Digits Iteration 10: 0.389s
Total time taken: 3.773s
--------------------------------------------------------------------------------
Tests: 6 | Total time: 18.72s | Avg Test Time: 3.12s

Clear

$ ruby benchmark.rb 
:: Details:
Ruby Version: 2.7.1 (x86_64-linux-gnu)

CC: gcc
CFLAGS: -O2 -g -feliminate-unused-debug-types -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=32 -Wformat -Wformat-security -Wl,-z,now,-z,relro,-z,max-page-size=0x1000,-z,separate-code -Wno-error -ftree-vectorize -ftree-slp-vectorize -Wl,--enable-new-dtags -fcf-protection=return -Wl,--build-id=sha1 -Wa,-mbranches-within-32B-boundaries -march=westmere -mtune=haswell -fasynchronous-unwind-tables -fno-omit-frame-pointer -Wp,-D_REENTRANT -fno-lto -fstack-protector-strong -mzero-caller-saved-regs=used  -fPIC -fPIC
--------------------------------------------------------------------------------
:: Please stop all your apps, perhaps reboot your system, and run the benchmark
:: Don't even move your mouse during the benchmark for consistent result!
| Ready? (0)...................................................................
:: Downloading the words...
CPU Blowfish Test
:: CPU Blowfish Iteration 1: 0.067s
:: CPU Blowfish Iteration 2: 0.067s
:: CPU Blowfish Iteration 3: 0.067s
:: CPU Blowfish Iteration 4: 0.068s
:: CPU Blowfish Iteration 5: 0.067s
:: CPU Blowfish Iteration 6: 0.067s
:: CPU Blowfish Iteration 7: 0.067s
:: CPU Blowfish Iteration 8: 0.068s
:: CPU Blowfish Iteration 9: 0.067s
:: CPU Blowfish Iteration 10: 0.067s
Total time taken: 0.672s
--------------------------------------------------------------------------------
FPU Test
:: FPU Math Iteration 1: 0.410s
:: FPU Math Iteration 2: 0.411s
:: FPU Math Iteration 3: 0.410s
:: FPU Math Iteration 4: 0.411s
:: FPU Math Iteration 5: 0.410s
:: FPU Math Iteration 6: 0.410s
:: FPU Math Iteration 7: 0.412s
:: FPU Math Iteration 8: 0.410s
:: FPU Math Iteration 9: 0.410s
:: FPU Math Iteration 10: 0.411s
Total time taken: 4.105s
--------------------------------------------------------------------------------
CPU Fibonacci Test
:: CPU Fibonacci Iteration 1: 0.547s
:: CPU Fibonacci Iteration 2: 0.471s
:: CPU Fibonacci Iteration 3: 0.442s
:: CPU Fibonacci Iteration 4: 0.444s
:: CPU Fibonacci Iteration 5: 0.435s
:: CPU Fibonacci Iteration 6: 0.423s
:: CPU Fibonacci Iteration 7: 0.421s
:: CPU Fibonacci Iteration 8: 0.426s
:: CPU Fibonacci Iteration 9: 0.416s
:: CPU Fibonacci Iteration 10: 0.420s
Total time taken: 4.445s
--------------------------------------------------------------------------------
CPU Anagram Hunt
:: CPU Anagram Iteration 1: 0.302s
:: CPU Anagram Iteration 2: 0.302s
:: CPU Anagram Iteration 3: 0.303s
:: CPU Anagram Iteration 4: 0.301s
:: CPU Anagram Iteration 5: 0.302s
:: CPU Anagram Iteration 6: 0.301s
:: CPU Anagram Iteration 7: 0.301s
:: CPU Anagram Iteration 8: 0.304s
:: CPU Anagram Iteration 9: 0.302s
:: CPU Anagram Iteration 10: 0.302s
Total time taken: 3.020s
--------------------------------------------------------------------------------
CPU 2 Million Prime Numbers
:: Prime Numbers Iteration 1: 0.342s
:: Prime Numbers Iteration 2: 0.342s
:: Prime Numbers Iteration 3: 0.342s
:: Prime Numbers Iteration 4: 0.345s
:: Prime Numbers Iteration 5: 0.344s
:: Prime Numbers Iteration 6: 0.345s
:: Prime Numbers Iteration 7: 0.360s
:: Prime Numbers Iteration 8: 0.350s
:: Prime Numbers Iteration 9: 0.349s
:: Prime Numbers Iteration 10: 0.349s
Total time taken: 3.468s
--------------------------------------------------------------------------------
CPU 2k Pi Digits
:: 2K Pi Digits Iteration 1: 0.371s
:: 2K Pi Digits Iteration 2: 0.354s
:: 2K Pi Digits Iteration 3: 0.425s
:: 2K Pi Digits Iteration 4: 0.420s
:: 2K Pi Digits Iteration 5: 0.424s
:: 2K Pi Digits Iteration 6: 0.424s
:: 2K Pi Digits Iteration 7: 0.423s
:: 2K Pi Digits Iteration 8: 0.345s
:: 2K Pi Digits Iteration 9: 0.420s
:: 2K Pi Digits Iteration 10: 0.426s
Total time taken: 4.032s
--------------------------------------------------------------------------------
Tests: 6 | Total time: 19.74s | Avg Test Time: 3.29s

Ruby on Arch with GCC -mtune=native -march=native -O3:

$ ruby benchmark.rb 
:: Details:
Ruby Version: 2.7.1 (x86_64-linux)

CC: gcc
CFLAGS: -O3 -pipe -fno-plt -march=native -mtune=native -fPIC
--------------------------------------------------------------------------------
:: Please stop all your apps, perhaps reboot your system, and run the benchmark
:: Don't even move your mouse during the benchmark for consistent result!
| Ready? (0)...................................................................
CPU Blowfish Test
:: CPU Blowfish Iteration 1: 0.092s
:: CPU Blowfish Iteration 2: 0.068s
:: CPU Blowfish Iteration 3: 0.068s
:: CPU Blowfish Iteration 4: 0.068s
:: CPU Blowfish Iteration 5: 0.067s
:: CPU Blowfish Iteration 6: 0.069s
:: CPU Blowfish Iteration 7: 0.067s
:: CPU Blowfish Iteration 8: 0.067s
:: CPU Blowfish Iteration 9: 0.067s
:: CPU Blowfish Iteration 10: 0.070s
Total time taken: 0.703s
--------------------------------------------------------------------------------
FPU Test
:: FPU Math Iteration 1: 0.386s
:: FPU Math Iteration 2: 0.386s
:: FPU Math Iteration 3: 0.386s
:: FPU Math Iteration 4: 0.386s
:: FPU Math Iteration 5: 0.387s
:: FPU Math Iteration 6: 0.386s
:: FPU Math Iteration 7: 0.387s
:: FPU Math Iteration 8: 0.386s
:: FPU Math Iteration 9: 0.386s
:: FPU Math Iteration 10: 0.386s
Total time taken: 3.862s
--------------------------------------------------------------------------------
CPU Fibonacci Test
:: CPU Fibonacci Iteration 1: 0.441s
:: CPU Fibonacci Iteration 2: 0.409s
:: CPU Fibonacci Iteration 3: 0.410s
:: CPU Fibonacci Iteration 4: 0.410s
:: CPU Fibonacci Iteration 5: 0.402s
:: CPU Fibonacci Iteration 6: 0.401s
:: CPU Fibonacci Iteration 7: 0.398s
:: CPU Fibonacci Iteration 8: 0.401s
:: CPU Fibonacci Iteration 9: 0.395s
:: CPU Fibonacci Iteration 10: 0.404s
Total time taken: 4.071s
--------------------------------------------------------------------------------
CPU Anagram Hunt
:: CPU Anagram Iteration 1: 0.309s
:: CPU Anagram Iteration 2: 0.310s
:: CPU Anagram Iteration 3: 0.308s
:: CPU Anagram Iteration 4: 0.303s
:: CPU Anagram Iteration 5: 0.303s
:: CPU Anagram Iteration 6: 0.302s
:: CPU Anagram Iteration 7: 0.301s
:: CPU Anagram Iteration 8: 0.302s
:: CPU Anagram Iteration 9: 0.302s
:: CPU Anagram Iteration 10: 0.302s
Total time taken: 3.042s
--------------------------------------------------------------------------------
CPU 2 Million Prime Numbers
:: Prime Numbers Iteration 1: 0.290s
:: Prime Numbers Iteration 2: 0.288s
:: Prime Numbers Iteration 3: 0.291s
:: Prime Numbers Iteration 4: 0.291s
:: Prime Numbers Iteration 5: 0.291s
:: Prime Numbers Iteration 6: 0.294s
:: Prime Numbers Iteration 7: 0.294s
:: Prime Numbers Iteration 8: 0.296s
:: Prime Numbers Iteration 9: 0.294s
:: Prime Numbers Iteration 10: 0.294s
Total time taken: 2.923s
--------------------------------------------------------------------------------
CPU 2k Pi Digits
:: 2K Pi Digits Iteration 1: 0.356s
:: 2K Pi Digits Iteration 2: 0.404s
:: 2K Pi Digits Iteration 3: 0.405s
:: 2K Pi Digits Iteration 4: 0.405s
:: 2K Pi Digits Iteration 5: 0.338s
:: 2K Pi Digits Iteration 6: 0.403s
:: 2K Pi Digits Iteration 7: 0.406s
:: 2K Pi Digits Iteration 8: 0.405s
:: 2K Pi Digits Iteration 9: 0.338s
:: 2K Pi Digits Iteration 10: 0.403s
Total time taken: 3.863s
--------------------------------------------------------------------------------
Tests: 6 | Total time: 18.46s | Avg Test Time: 3.08s

So the conclusion is Clear Linux is a little bit slower than Arch. Why?

At least you need to have same kernel version for both distro, for the benchmark.

The latest CL has 5.8.12 kernel, so you should upgrade your CL machine and re-run the benchmark.

2 Likes

You cannot compare a live installed system to an ISO booted image…

Ok I am installing the OS on my spare HDD. The problem is Arch runs on SSD, so I will be doing CPU only tests when the installation is done and I have fully upgraded the system like my Arch installation.

I might also try benchmarking the Xanmod kernel (which is obviously faster) - still I think Clear Linux is performance oriented, and that should be faster with the default ones.

Clear Linux to me, is just like Arch, with a bloated GNOME rather than XFCE on my Arch. I don’t want to invest a lot of time in Clear for now, but I don’t think even if I install it on my SSD it can’t beat the Arch, or at least will do similar result. But I am skeptical with the CPU benchmarks.

Don’t think so, the boot time might differ a bit, sure. But clear should win. The reason is it’s running off of my RAM, which is even 25x faster than my SATA SSD is.

But if you run systemd-analyze blame, you will know why it’s slow. On my Arch, I have journaling storage set to none, which means no journaling, no extra writing, no journal flush. I have masked lvm2-monitor as well. The GRUB parameter on my Arch is:

GRUB_CMDLINE_LINUX_DEFAULT="fsck.mode=skip quiet loglevel=0 rd.systemd.show_status=false nowatchdog mitigations=off libahci.ignore_sss=1

Yes, I don’t mind security much for my own system, and it never got hacked with this mitigations=off. I don’t mind using fsck.mode=skip because I am using XFS file system.

On clear, I agree those stuff are not in use, but still… I don’t think I need to tweak it a lot. I don’t really like how a distro that is emerged for performance, doesn’t use ZRAM, instead suggest using swap. I don’t like how I am forced to use EXT4 (235) (Otherwise system is not installable). XFS performs better than EXT4 and it’s also more robust. Same goes for F2FS.

This is incorrect. The swapfile does exist, but is barely minimal and hardly used.

After upgrading, all the benchmarks are lost to the older Clear Linux I benchmarked.

FPU Raytracing (lower is better): Now 1.67 | Prev 1.66
FPU FFT (lower is better): Now 1.56 | Prev 1.55
CPU ZLib (higher is better): Now 0.12 | Prev 0.13
CPU-NQueens (lower is better): Now 3.58 | Prev 3.63
CPU-Fibonacci (lower is better): Now 0.38 | Prev 0.37
CPU Cryptohash (higher is better): Now 473.90 | Prev 486.18
CPU Blowfish (higher is better):
Multi-core: Now 21.47 | Prev 21.54
Multi-thread: Now 33.14 | Prev 33.43
Single-thread: Now 10.77 | Prev 10.78

So it didn’t got any better than Arch.

the next is to set the same cmdline parameters you have set for Arch.

I just added mitigations=off (and confirmed by cat /proc/cmdline), because watchdog is not running on my system, on Arch I used to blacklist iTCO_wdt, no need to do that here.

Ok so the results is a bit better than this one, but worse than the liveroot I ran. The CPU N-Queens is now 3.56 (lower is better). But not close to my Arch system.

I am willing to try out the xanmod kernel, but I compiled it from AUR, the PKGBUILD automatically fetches the patches, makes initcpio preset, copies the kernel to boot, all automatically, so I used to compile the package from AUR.

But in clear I don’t think that the Arch package is installable. Surely I can unpack the compiled binaries and copy them manually and set permission, but that’s too much work…

It really depends on what and how you are testing. The PTS has not been a great measure of a distributions performance (which I consider the packages that are shipped in the repo) and in fact is absolutely wrong in many circumstances. To give an obvious example, Michael reports a geometric mean of F33 being 13% higher than F32…but clearly F32 is faster…(which was generated from the same openbenchmarking data)

If you want to tinker or rebuild packages from source (and remove all the security features), you will easily get faster results than CL. There’s not enough performance to recommend its use as a home host OS imo.

Most of the tests of mine don’t agree with Phoronix.

I have previously benchmarked the stock kernel with zen kernel, zen wins. Zen vs. Liquorix, Zen wins in most cases. Zen vs. Xanmod, Xanmod wins. The tests result is consistent after each reboot. The tests results are consistent up to 2 decimal places.

But in Clear Linux, the tests is really consistent, but not after I added mitigations=off, which always change deviates the benchmark a tiny little bit.

In my Arch however, when I do benchmark, it sits absolutely idle, there’s no software manager on my Arch, like Clear. I use to run pacman to update my system.

Phoronix on the other hand has a behemoth CPU, and an expensive cooler. When a big CPU heats up, it will always add some deviation to the results. On my system however, the system is consistent, not that much powerful, I use i3 4th gen (haswell) CPU.

In my ruby benchmarks however, the results are consistent up to 1 decimal place.

The real challenges is to make a system more responsive, there’s no real way to benchmark that AFAIK. So to test, put the system under 100% load and test, xanmod with cachy really wins in that, but there’s no way to benchmark that thing. There are commands like cyclictest and perf, but that’s not realistic.

IMO, A well tuned Arch and a well custom tuned Clear will provide same performance. I just posted here only to show this. But it’s good to see there’s a linux providing bleeding elde packages, and faster than any other distros Debian, Ubuntu, Linux Mint, OpenSUSE, Fedora, CentOS (slowest linux), etc. My choice will go like this: Arch, Manjaro, Gentoo, Clear…

Too compare the performance, theoretically it’s necessary to include a wide variety of hardwares.

So the question then is, is xanmod consistent faster for your hardware, or for all hardwares.

Also, I fully disagree with the practice of using software optimisation to improve performance on old hardwares. The practical solution is always to replace them with newer ones.

I’m more interested in how xanmod kernel performs with “a behemoth CPU, and an expensive cooler”.

That I can’t really provide. 4 of my neighbours went straight to Arch from windows. They have no idea what a command is, but I have installed Arch with every GUI alternative and with nice themes. They say Linux is better than Windows XD!

I have benchmarked the xanmod kernel on my neighbours laptops and desktops, 1 have Ryzen GPU in laptop, and NVIDIA GPU on PC.

There are a lot of different hardware. But I have given Xanmod on those PCs. I saw that the benchmark is a bit better now, the games (xonotic git) can be played at higher quality now and still maintains the same FPS.


Anyways a little bit off-topic topic:
Openbenchmarking.org says intel’s ICC compiler is wayyy faster than GCC:

That looks like 1.5x faster.
So I compiled one ray tracer written in C++, and benchmarked. Yes it’s faster than GCC and clang:

But in most cases:

  1. ICC produces slower binary than clang when -O3 is used in most cases. Like in a benchmark I did with memory test in Linux, ICC was slowest.
  2. ICC breaks games like SuperTuxKart where lightning in the sky doesn’t go away in a map.
  3. ICC Can’t compile Ruby or Linux kernel.

So it doesn’t sound a lot better, and it’s not really better in most of the programs, except the ray tracer only shined.


That is to say, Phoronix benchmarks aren’t always the same in most cases. I tried ICC on my laptop and desktop though, pretty limited number of devices.


BTW, I encourage benchmarking the Xanmod kernel with the Cachy scheduler. It’s makes the system really faster that people start noticing the performance they are getting.

Oh and one step I forgot to do in Clear is that setting the governor to performance, which is default for Xanmod…

1 Like

Well, 1.5x faster itself looks fishy.

Unlike the standardized test programs used in benchmarking tools which are not arbitrarily chosen, the performance gain on example programs nearly always disappear in reality…

You will find people claiming clang is better than gcc, or golden linker is better than stock linker, etc. Though they are true for some cases, it’s unpredictable.

1 Like

In many cases clang is indeed faster.
But in other cases GCC is faster.

It’s really hard to know which one is faster unless it’s already thoroughly benchmarked… And again, that can be faster for that system, but slower on others… Which is confusing and there’s nothing to do about it unless people start writing Linux is assembly :joy::laughing: Just kidding…

1 Like