Linux 4.20 kernel slower than its previous stable releases, Spectre flaw to be blamed, according to Phoronix

On the 4th of November, Linux 4.20 rc-1 was released with a host of notable changes right from AMD Vega 20 support getting squared away, AMD Picasso APU support, Intel 2.5G Ethernet support, the removal of Speck, and other new hardware support additions and software features. The release that was supposed to upgrade the kernel’s performance, did not succeed in doing so. On the contrary, the kernel is much slower as compared to previous Linux kernel stable releases.

In a blog released by Phoronix, Michael Larabel,e lead developer of the Phoronix Test Suite, Phoromatic, and OpenBenchmarking.org, discussed the results of some tests conducted on the kernel. He bisected the 4.20 kernel merge window to explore the reasons for the significant slowdowns in the kernel for many real-world workloads.

The article attributes this degrade in performance to the Spectre Flaws in the processor. In order to mitigate against the Spectre flaw, an intentional kernel change was made.The change is termed as "STIBP" for cross-hyperthread Spectre mitigation on Intel processors. Single Thread Indirect Branch Predictors (STIBP) prevents cross-hyperthread control of decisions that are made by indirect branch predictors. The STIBP addition in Linux 4.20 will affect systems that have up-to-date/available microcode with this support and where a user’s CPU has Hyper-Threading enabled/present.

Performance issues in Linux 4.20

Michael has done a detailed analysis of the kernel performance and here are some of his findings.

Many synthetic and real-world tests showed that the Intel Core i9 performance was not upto the mark.

The Rodinia scientific OpenMP tests took 30% longer, Java-based DaCapo tests taking up to ~50% more time to complete, the code compilation tests also extended in length.

There was lower PostgreSQL database server performance and longer Blender3D rendering times. All this was noticed in Core i9 7960X and Core i9 7980XE test systems while the AMD Threadripper 2990WX performance was unaffected by the Linux 4.20 upgrade.

The latest Linux kernel Git benchmarks also saw a significant pullback in performance from the early days of the Linux 4.20 merge window up through the very latest kernel code as of today. Those affected systems included a low-end Core i3 7100 as well as a Xeon E5 v3 and Core i7 systems.

The tests conducted found the Smallpt renderer to slow down significantly

PHP performance took a major dive, HMMer also faced a major setback compared to the current Linux 4.19 stable series.

What is surprising is that there are mitigations against Spectre, Meltdown, Foreshadow, etc in Linux 4.19 as well. But 4.20 shows an additional performance drop on top of all the previously outlined performance hits this year. In the entire testing phase, the AMD systems didn’t appear to be impacted. This would mean if a user disables Spectre V2 mitigations to account for better performance- the system’s security could be compromised.

You can head over to Phoronix for a complete analysis of the test outputs and more information on this news.

Soon, RHEL (Red Hat Enterprise Linux) won’t support KDE
Red Hat releases Red Hat Enterprise Linux 8 beta; deprecates Btrfs filesystem
The Ceph Foundation has been launched by the Linux Foundation to support the open source storage project