Made public this week by CPU security researchers at Graz University of Technology and CISPA Helmholtz Center for Information Security was the research paper published "AMD Prefetch Attacks through Power and Time". The paper points to AMD CPUs suffering from a side-channel leakage vulnerability through timing and power variations of the PREFETCH instruction. The paper argues that AMD CPUs should activate stronger page table isolation by default. AMD has now published their security response where they are not recommending any mitigation changes at this time. But what if Kernel Page Table Isolation (KPTI/PTI) proves necessary for AMD CPUs? Here are some initial benchmarks showing what that performance impact could look like.

Last week was the article on noticing various workloads performing slower on the Linux 5.15 development kernel. There is now a patch pending that in testing so far does indeed correct the performance drop on this forthcoming kernel.

Earlier this month AMD posted their initial public patches for the AMD P-State CPU frequency scaling driver that leverages ACPI CPPC for ultimately aiming to provide better power efficiency and more responsive CPU frequency scaling / performance state decisions on Zen 3 (and Zen 2 eventually) processors. This is part of the effort around AMD and Valve collaborating for better Linux efficiency especially with the AMD-powered Steam Deck.

As usual when the Linux 5.15 merge window began wrapping up, I set out to dive into its performance to see what is in store for this next version of the kernel and whether there was any regressions or other performance changes worth noting. Linux 5.15 overall has been in good shape for the "-rc1" state except noticing that code compilation workloads were taking longer on multiple Linux 5.15-rc1-running systems than Linux 5.14 or prior. Seeing it across multiple systems and a very real-world regression, it was worth bisecting and looking closer so here are the details.

Feature development is over on the Linux 5.15 kernel with Linux 5.15-rc1 being issued. It's now on to testing and bug fixing over the next two months before the kernel is christened as stable. Here is our original Linux 5.15 feature overview about all of the big changes in this next kernel version.

With the in-development Linux 5.15 kernel there is a new option for further protecting the kernel around side channel attacks and information leakage. Enabling the option will ensure that any caller-used register contents are zeroed prior to returning from a function. While the reported performance cost is said to be small, I decided to run some benchmarks when toggling this new Kconfig hardening option.

It's been a while since last running benchmarks evaluating the performance of GCC's profile guided optimizations (PGO) for helping to optimize the performance. But stemming from the discussions around PGO'ing the Linux kernel (though that effort is stalled for now), several Phoronix readers inquired about seeing some fresh PGO figures with GCC 11. So here are such benchmarks of GCC 11 with the upcoming Ubuntu 21.10 running on an AMD Ryzen 9 5950X desktop.

Earlier this month was a look at the LLVM Clang 13 performance on EPYC 7003 showing this forthcoming compiler update to be in good shape for AMD Zen 3, but how is the performance looking on the Intel side? This round of benchmarking is looking at the LLVM Clang 11 / 12 / 13 compiler performance on Intel's flagship Xeon Platinum 8380 "Ice Lake" 2P server configuration.

With LLVM/Clang 13 feature development having ended last week and the 13.0 release candidate being tagged, in starting off the benchmarking cycle first up I was looking at how well this new compiler is performing compared to LLVM Clang 12 stable on an AMD EPYC 7543 (Zen 3) Linux server.

Following the recent benchmarks seeing how AMD's new AOCC 3.1 compiler has brought some performance improvements over the prior AOCC 3.0 release that introduced initial Zen 3 optimizations, here are some benchmarks looking at how that latest AMD Optimizing C/C++ Compiler performance compares to the upstream LLVM Clang 12 compiler for which it is based as well as against GCC 11 as the latest GNU compiler release that remains common to Linux systems.

Last week AMD released their AOCC 3.1 compiler that is their downstream of LLVM Clang/Flang and carrying various yet-to-be-upstreamed patches for benefiting their latest processors. While just a point release, curiosity got the best of me for firing up benchmarks of this latest AMD Optimizing C/C++ Compiler release.

Motivated in part by the recent le9 kernel patches that are already carried by XanMod and not having benchmarked the XanMod or Liquorix Linux kernel downstreams in a while, here are some fresh benchmarks of Liquorix and XanMod against the recent upstream Linux kernel releases.

With the Linux 5.12 kernel bringing support for building the kernel with link-time optimizations (LTO) when using the LLVM Clang compiler, here are some benchmarks looking at that performance impact as well as more generally seeing how the LLVM Clang compiler performance is looking when building the Linux kernel relative to GCC.

With last night's release of Linux 5.14-rc1 the merge window is officially over for this next version of the Linux kernel. With that, here is a look at the highlights for the forthcoming Linux 5.14 kernel based upon our original reporting during the merge window.

For today's benchmarking is a look at how the GNU Compiler Collection has performed over the past few years going from the GCC 8 stable series introduced in 2018 through the recently released GCC 11.1 stable feature release plus also including the current early development snapshot of GCC 12.

Earlier this month were benchmarks looking at GCC 11 performance with varying optimization levels and features like link-time optimizations. Stemming from reader requests, here are now similar reference benchmarks off LLVM Clang 12.0 on the same system with going from -O0 to -Ofast and toggling -march=native and LTO usage.

Given the recent forum discussion stemming from the -O3 optimization level still too unsafe for the Linux kernel (in part due to older, buggy compilers) and some users wondering about the current -O2 versus -O3 compiler optimization level impact, here is a fresh round of reference benchmarks using GCC 11.1 on Fedora Workstation 33 looking at various optimization levels and optimizations tested on dozens of different application benchmarks to see the overall impact on performance.

As part of the curiosity-driven benchmarks and areas of technical interest now that we've gotten some of our initial Intel Xeon Platinum 8380 "Ice Lake" benchmarks out of the way has been looking into the performance of Linux's P-State CPU frequency scaling driver on the 3rd Gen Xeon Scalable server. Benchmarked for the latesting testing was the power/efficiency out-of-the-box with P-State powersave as used by default with many Linux distributions against the P-State "performance" mode as well as putting P-State into passive mode to be able to via intel_cpufreq to try the Schedutil governor that relies on the kernel's scheduler utilization data for making frequency scaling decisions. Here is a number of power/performance governor benchmarks with the dual Xeon Platinum 8380 server in these varying kernel configurations.

Recently we have been running a number of compiler benchmarks looking at the recently released LLVM Clang 12 and GCC 11 open-source code compilers. There is as healthy and competitive competition as ever between GCC and Clang with the mainline Linux kernel these days working well under Clang, more software projects shifting to Clang by default, and the performance being as tight as ever between GCC and Clang for compiled C/C++ code on x86_64 and AArch64. In today's article are benchmarks of Clang 12 vs. GCC 11 on the dual Intel Xeon Platinum 8380 Ice Lake server.

Earlier this week I posted some benchmarks looking at the compiler performance of GCC 11 vs. LLVM Clang 12 on the Intel Core i9 11900K "Rocket Lake" processor while in this article the same tests and same software are being carried out on an AMD Ryzen 9 5950X "Zen 3" desktop. With these AMD Linux tests the Clang 12 compiler not only yielded the fastest binaries at -O2 but carried through in the more optimized configurations as well.

For those wondering how GCC and LLVM Clang are competing when running on Intel's latest Rocket Lake processors, here are some GCC 11 vs. LLVM Clang 12 compiler benchmarks with the Core i9 11900K running from the newly-released Fedora Workstation 34 featuring these very latest compilers. The compiler benchmarks were carried out at multiple optimization levels on each compiler.

Following the two week merge window, feature development on the Linux 5.13 kernel is slated to end today with the release of Linux 5.13-rc1. Here is a look at some of the most interesting new features and improvements for this kernel that in turn should debut as stable around the end of June.

For those wondering how the recent releases of the GCC 11 and LLVM 12 (Clang 12) open-source compilers are competing on AMD Zen 3, here are some recently conducted benchmarks looking at that showdown on an AMD EPYC 7763 1P server.

Released last week was the LuxCoreRender 2.5 open-source physically based renderer. Significant with this v2.5 update is OptiX/RTX acceleration support in addition to its existing CUDA, OpenCL, and CPU render paths. Given that, here are some fresh benchmarks of LuxCoreRender 2.5 across an assortment of NVIDIA graphics cards.

Here is a look at the AVX / AVX2 / AVX-512 performance on the Intel Core i9 11900K "Rocket Lake" when building a set of relevant open-source benchmarks limited to AVX, AVX2, and AVX-512 caps each time while also monitoring the CPU package power consumption during the tests for looking at the performance-per-Watt in providing some fresh reference metrics over AVX-512 on Linux with the latest Intel "Rocket Lake" processors.

This past week AMD published a security analysis of AMD Zen 3's new Predictive Store Forwarding (PSF) functionality. In there they did acknowledge there is the possibility where bad PSF functionality could lead to a side-channel attack albeit the real-world exposure would be quite low. In any case they are allowing interested users to disable the Predictive Store Forwarding functionality, but what they didn't comment on in that paper was what performance overhead to expect if disabling PSF. So my Easter weekend turned into AMD Zen 3 PSF benchmarking.

For those wondering about the state of speculative execution vulnerabilities and what software-based mitigations are required for Intel's new Rocket Lake processors, here is the rundown along with benchmarks when disabling those present Linux kernel mitigations.

Launched last week with the AMD EPYC 7003 "Milan" processors was the AOCC 3.0 code compiler as AMD's downstream of LLVM Clang with various patches now catering to optimized for Zen 3. Last week some preliminary benchmarks of AOCC 3.0 on the Ryzen 9 5950X were carried out to good results. Since then I have begun putting AOCC 3.0 through its paces on a AMD EPYC 7003 series server to overall great results.

This week alongside the EPYC 7003 series launch was the introduction of AOCC 3.0 as AMD's Zen-optimized LLVM/Clang downstream. We have started putting this updated compiler through its paces to see what it means for AMD Zen 3 performance.

With GCC 11 stable likely to be released next month, here is the latest in our compiler testing against the current GCC 10 stable release. This round of tests was carried out on a System76 Thelio Major with Ryzen Threadripper 3990X HEDT processor.