Open Source LLM

LLVM 11.0-RC3 Released For This Big LLVM/Clang Update

LLVM 11.0 was originally scheduled to be released at the end of August while now it looks like that official milestone is coming in the next few days or week.

Tagged today was LLVM 11.0-RC3 as the belated extra release candidate for this half-year update to the LLVM compiler infrastructure and subprojects like Clang, LLD, FLANG, and libcxx, among others.

LLVM release manager Hans Wennborg wrote in the RC3 announcement that there are no open release blocker bugs remaining. So pending

Open Source LLM

AOMP 11.9 Released For OpenMP Offloading To Radeon GPUs

AOMP 11.9 was released on Friday as AMD's LLVM-based compiler with Clang for C/C++ and Flang for Fortran in offloading capable OpenMP code to Radeon GPUs.

AOMP 11.9 is AMD's latest work on their LLVM 11 derived code-base for OpenMP GPU compute until the necessary patches have worked their way eventually back into the upstream code-base.

This 11.9 release fixes a number of compiler warnings, multiple fixes to the Flang compiler, the

Open Source LLM

Why TensorFlow Lite Has Been Running Slower On Recent Linux Kernels

The Linux 5.0 to 5.9 kernel benchmarking posted this week showed TensorFlow Lite running slower since the Linux 5.5 kernel... On top of looking at the new Linux 5.9 regressions , I also spent some time bisecting and figuring out what happened for TensorFlow Lite last year that has at least for the system under test caused it to run slower for all the kernel releases this year as shown in the aforelinked article.

So with the speedy AMD EPYC box and knowing the regression happened between Linux


Open Source LLM

Clang LTO Patches Updated For The Linux Kernel

Google engineers have posted a new revision of their work from earlier this summer for allowing the Linux kernel to be LTO-optimized using Clang .

Sami Tolvanen of Google sent out the second version of their patches for Clang LTO support with the Linux kernel. These patches allow making use of Clang's link-time optimizations for building Linux x86_64 and AArch64 (ARM64) Linux kernels.

Employing LTO for the Linux kernel has beneficial performance advantages as well as the possibility of smaller binaries

Open Source LLM

Oak Ridge Sponsoring GCC Compiler Improvements For NVIDIA + AMD GPU Offloading

Oak Ridge National Laboratory's Leadership Computing Facility (OLCF) announced on Thursday that they are funding improvements to the GNU Compiler Collection (GCC) around GPU compute offloading.

Oak Ridge's OLCF has entered into contract with Siemens' Mentor Graphics to fund more work on GCC upstream development. The work they are funding is on providing better GPU accelerator offloading support for the upstream, open-source GCC compiler. This will include work on both OpenACC and OpenMP.

Not only are both OpenMP and OpenACC being targeted but this

Open Source LLM

Linux From Scratch 10.0 Released For Rolling Your Own Linux Installation From Source

Just over twenty years after the Linux From Scratch project was started as a guide/book to building all of the software components manually from source, Linux From Scratch 10.0 has been released.

Linux From Scratch continues to teach the user how to roll their own Linux installation from source and while hearing "LFS" has been less common in recent years the effort still continues.

With Linux From Scratch 10.0, the book has gone through a "major reorganization" to cover more cross-compilation techniques

Open Source LLM

Intel mOS, Cachy, Mozilla Layoffs Were The Open-Source Buzz Of August

August was an interesting month in terms of Linux/open-source activity though still another tough month during the coronavirus pandemic.

During the month of August on Phoronix were 273 original news articles and just 10 featured Linux hardware reviews/featured articles. The news count is still close to the average of 10 new/original news articles each and every day of the year while the review/featured count is off the usual track of around one per day... That is due principally due to declining

Open Source LLM

Registration Opens For 2020 Virtual LLVM Developers' Meeting

Like most conferences this year, the annual LLVM Developers' Meeting has become an online-only affair.

The annual LLVM conference normally hosted in Silicon Valley is now taking place entirely online. This virtual event is taking place from 6 to 8 October.

While normally this is an in-person event with an attendee fee, with the event moving to an online-only format, it's now free... If you want it to be. Thanks to sponsors like Apple, Qualcomm, IBM, Google,

Open Source LLM

NVIDIA's DALI 0.25 Deep Learning Library Adds AArch64 SBSA, Performance Improvements

DALI is the project at NVIDIA focused on GPU-accelerated data augmentation and image loading along with other tasks while being optimized for deep learning workflows. DALI 0.25 was released on Friday as the latest step forward for this open-source NVIDIA project.

Yes, DALI (short for the NVIDIA Data Loading Library) is one of NVIDIA's open-source projects . DALI currently describes itself as "a library containing both highly optimized building blocks and an execution engine for data pre-processing in deep learning applications

Open Source LLM

OpenZFS 2.0-RC1 Released With Unified Linux/BSD Support, Zstd Compression & Much More

The first release candidate of the forthcoming OpenZFS 2.0 is now available for testing on both Linux and BSD systems.

OpenZFS 2.0 is a huge feature release for this well maintained, portable open-source ZFS file-system implementation. OpenZFS 2.0 brings unified support for both Linux and now FreeBSD too . FreeBSD just mainlined the OpenZFS code and has been working for many months now on transitioning over to this more maintained and active ZFS file-system code-base.

Making OpenZFS 2.0 even more