Tag Archives: Linux Hardware Reviews

OpenMandriva Lx 4.1 Aiming To PGO More Packages, Use IWD For WiFi Connections


OPERATING SYSTEMS --

While OpenMandriva Lx 4.0 was just released last month, we are already looking forward to OpenMandriva 4.1 for a number of improvements and some new features.

OpenMandriva’s developer board provides an interesting look at what’s ahead for OpenMandriva Lx 4.1. Already completed for this next milestone include migrating to LLVM Clang 9, and using LD.lld and BFD as the default linkers.

Meanwhile they are currently tackling using Profile Guided Optimizations (PGO) for more packages to improve the performance of their default binaries. Using PGO should help the likes of Python, Firefox, OpenSSL, LZ4, MPFR, Ogg, Vorbis, and many other packages they are evaluating for PGO’ing.

Also notable is switching to Intel IWD as an alternative to WPA_Supplicant for dealing with WiFi connections. They are also eyeing a replacement for Firewalld, other LLVM toolchain changes, moving to a merged /usr layout, updating their Java stack, and other changes.

Those curious what else is coming for OpenMandriva Lx 4.1 can learn more via GitHub.


FreeBSD 12 Runs Refreshingly Easy On AMD Ryzen 9 3900X – Benchmarks Against Ubuntu 18.04 LTS


While newer Linux distributions have run into problems on the new AMD Zen 2 desktop CPUs (fixed by a systemd patch or fundamentally by a BIOS update) and DragonFlyBSD needed a separate boot fix, FreeBSD 12.0 installed out-of-the-box fine on the AMD Ryzen 9 3900X test system with ASUS ROG CROSSHAIR VIII HERO WiFi motherboard.

I was curious about the FreeBSD support for AMD Zen 2 CPUs and new X570 motherboards, so this weekend I tried out FreeBSD 12.0. Fortunately, the experience was great! This current FreeBSD 12.0 AMD64 image installed effortlessly — no boot problems, networking did work out-of-the-box with this ASUS X570 motherboard, and there were no other issues at least as core functionality is concerned. So in no time I was off to the races in running FreeBSD 12.0 benchmarks on the Ryzen 9 3900X 12-core / 24-thread CPU.

I also attempted to try DragonFlyBSD with its latest daily ISO/IMG following the Zen 2 fix this week by Matthew Dillon. Unfortunately, even with the latest daily ISO I ran into a panic at boot time. So as a result, today are just some FreeBSD 12.0 vs. Ubuntu 18.04 benchmarks for reference. Matthew Dillon did have some interesting comments in our forums about his (great) experiences with these new CPUs, some limitations, and about the original DragonFlyBSD issue.

This system test configuration was the Ryzen 9 3900X at stock speeds, 2 x 8GB DDR4-3600 memory, ASUS ROG CROSSHAIR VIII HERO motherboard, and 2TB Corsair Force MP600 PCIe 4.0 NVMe SSD. Ubuntu 18.04 LTS was benchmarked against FreeBSD 12.0 with its default LLVM Clang 6.0 compiler and then again when switching to the GCC 8.3 compiler.

Ubuntu 18.04.2 LTS wins most of the benchmarks, but FreeBSD 12.0 was able to hold its ground fairly well in many of the benchmarks. Switching over to the GCC compiler did help address the difference in some of these benchmarks. All of these tests were carried out via the Phoronix Test Suite on Linux and BSD. Let’s check out some of those interesting numbers.




NVIDIA’s Graphics Driver Will Run Into Problems With Linux 5.3 On IBM POWER


NVIDIA --

For those using the NVIDIA proprietary graphics driver on an IBM POWER system, it could be a while before seeing Linux 5.3+ kernel support. Upstream has removed code depended upon by the NVIDIA binary driver for supporting the POWER architecture and as is the case they don’t care that it will break NVIDIA driver support since it’s binary/out-of-tree.

The POWER changes for Linux 5.3 remove NPU DMA code. In the pull request they do acknowledge this DMA code is “used by the out-of-tree Nvidia driver, as well as some other functions only used by drivers that haven’t (yet?) made it upstream.”

The patch removing the NPU DMA code by Linux kernel veteran Christoph Hellwig does acknowledge this basically reverts the POWER support for NVIDIA NVLink GPUs. The code is being dropped since it’s no longer being used by the in-tree kernel code and thus a burden when it comes to maintaining the upstream DMA code.

IBM developer Alexey Kardashevskiy did warn that this particular code is “heavily” used by NVIDIA’s graphics driver. Hellwig responded though that “Not by the [driver / code] that actually exists in the kernel tree, so it simply doesn’t matter.

This isn’t just a function or two being removed but amounts to 1,280 lines of code now stripped out of the kernel that was used by the NVIDIA binary driver on POWER. The NVIDIA POWER support will now break on Linux 5.3 but hopefully NVIDIA will be able to come up with a timely solution to fix their driver on 5.3 and newer.


XFS Gets Cleaned Up In Linux 5.3 Kernel Development Activity


LINUX STORAGE --

While not too eventful on the end-user feature front, the XFS file-system has seen another round of clean-ups with the ongoing Linux 5.3 merge window.

XFS maintainer Darrick Wong characterized the feature work for XFS in Linux 5.3 as “significant amounts of consolidations and cleanups in the log code; restructuring of the log to issue struct bios directly; new bulkstat ioctls to return v5 fs inode information (and fix all the padding problems of the old ioctl); the beginnings of multithreaded inode walks (e.g. quotacheck); and a reduction in memory usage in the online scrub code leading to reduced runtimes.

It’s not immediately exciting for end-users but hopefully will cleanup some open issues and the start of multi-threaded inode walks could help with performance along with the memory reduction work.

The complete list of Linux 5.3 XFS work can be found via this pull request.


Wayland’s Weston Gets Option To Enable HDCP Support Per-Output


WAYLAND --

An Intel open-source developer contributed support to Wayland’s reference Weston compositor for enabling HDCP support on a per-output basis using a new allow_hdcp option.

From the weston.ini configuration file, High-bandwidth Digital Content Protection can be enabled per-output via the “allow_hdcp” option within each output section. HDCP otherwise is always enabled by default for the display outputs.

The support that made it into Weston’s development code this week is for easily being able to toggle it via the Weston configuration file. This, of course, depends upon the lower-level HDCP driver work that has been ongoing for a while now particularly on the Intel Linux graphics side where the likes of Google for Chrome OS has been interested in this functionality.

For the conventional Linux desktop, we haven’t seen much (any?) software at this stage that actually makes use of the HDCP interfaces that have come with time to the Intel Direct Rendering Manager driver but the support is there though doesn’t impose any restrictions otherwise on the users with the implementation.