ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

AOMP 22.0-1 Brings Many Improvements For AMD's Fortran Compiler GPU Offloading

([AMD] 16 October 09:42 AM EDT AOMP 22.0-1)

AOMP 22.0-1 was released on Wednesday as the newest routine update to this downstream of LLVM/Clang/Flang maintained by AMD that continues to carry their latest modifications for enhancing the C/C++/Fortran compiler offloading support to AMD Radeon/Instinct hardware using the likes of OpenMP and OpenACC.


LLVM/Clang 22 Merges Support For Intel Nova Lake "-march=novalake"

([Intel] 16 October 09:25 AM EDT LLVM Clang 22)

Merged today to LLVM Git for next spring's LLVM 22.0 release is support for the Intel Nova Lake ISA targeting with the "-march=novalake" option.


Mesa NVK Lands Support For VK_NVX_image_view_handle - Needed For NVIDIA DLSS

([Nouveau] 16 October 08:24 AM EDT VK_NVX_image_view_handle)

Just days ago Valve developer Autumn Ashton announced initial NVIDIA DLSS upscaling for the open-source Mesa NVK driver. One of those needed Vulkan extensions, VK_NVX_image_view_handle, is already merged to Mesa Git.


Latest Linux Patches For Homa Posted: TCP Alternative With 10~100x Lower Tail Latency

([Linux Networking] 16 October 06:43 AM EDT Homa)

Posted to the Linux networking mailing list on Wednesday were the latest patches for enabling the Homa transport protocol. Homa is the clean-sheet design aiming to become an alternative to TCP within data centers and capable of offering 10~100x reductions in tail latency for short messages.


Proposed Patches Make Upstream Linux Kernel Usable For The OpenWrt One Router

([Linux Networking] 16 October 06:30 AM EDT OpenWrt One)

Announced nearly one year ago was the OpenWrt One as a router/AP that is "hacker-friendly" and open-source. The OpenWrt One is powered by a Filogic 820 SoC and features WiFi 6. This official OpenWrt device is manufactured by Banana Pi. While there is downstream open-source code available for customizing the OpenWrt One to your heart's content, the upstream Linux kernel support to date hasn't been full-featured.


Intel Proposes "SYCLBIN" As New Format For SYCL Device Code

([LLVM] 16 October 06:15 AM EDT SYCLBIN)

SYCLBIN has been proposed by a longtime Intel compiler expert as a new way for storing SYCL device code for use as part of their GPU/XPU programming ambitions.


Mesa 26.0 Enters Feature Development With Mesa 25.3 Branched

([Mesa] 16 October 05:56 AM EDT Mesa 26.0)

The latest Mesa Git code is now under version Mesa 26.0-devel with the Mesa 25.3 code being branched overnight for what will become this quarter's stable feature release.


Mesa Gallium3D Driver Merged For Arm Ethos NPUs

([AI] 15 October 08:50 PM EDT Arm Ethos)

Arm has been working on an open-source Linux kernel accelerator "accel" driver for their Ethos NPUs. That kernel driver continues being revised and under review for inclusion into a future mainline Linux kernel release. Already though a Gallium3D driver for Mesa has been merged for leveraging the Ethos NPU.


ollama Rolls Out Experimental Vulkan Support For Expanded AMD & Intel GPU Coverage

([AI] 15 October 08:24 PM EDT ollama + Vulkan)

The ollama 0.12.6-rc0 software released this evening and with it comes experimental Vulkan API support.


Valve Developer Contributes Open-Source Driver Fixes For 12 Year Old Hawaii GPUs

([Radeon] 15 October 04:40 PM EDT Old GPU Fixes)

Valve's open-source Linux graphics team continues carrying out great feats from getting NVIDIA DLSS working on the open-source NVK driver to enhancing Linux GPU driver support for hardware long forgotten about by the hardware vendors. The latest improvement from Valve's stellar group of open-source contributors are some fixes that benefit old Radeon Rx 200 series graphics cards that are more than one decade old.


PyTorch 2.9 Released With Easier Install Support For AMD ROCm & Intel XPUs

([AI] 15 October 03:26 PM EDT PyTorch 2.9)

PyTorch 2.9 is out today ahead of the PyTorch Conference happening next week in San Francisco. Notable with PyTorch 2.9 is better AMD ROCm and Intel XPU installation support via expanded wheel variant support.


Mesa 25.2.5 Released With Very Important Intel Driver Fix

([Mesa] 15 October 01:40 PM EDT Mesa 25.2.5)

Mesa 25.2.5 is out today as the newest bi-weekly point release for these open-source OpenGL and Vulkan drivers. Particularly if you are on Intel graphics of Battlemage or Lunar Lake and potentially older, Mesa 25.2.5 contains a very important bug fix for various rendering issues and potential game hangs/crashes.


Open 3D Engine O3DE 25.10 Brings Build Improvements, Vulkan & Linux Fixes

([Linux Gaming] 15 October 01:30 PM EDT Open 3D Engine)

It's been four years now since the Open 3D Engine was born out of Amazon's Lumberyard project and hosted by the Linux Foundation. Today marks the release of the Open 3D Engine "O3DE" 25.10 release with the newest features and fixes for this cross-platform game/graphics engine.


Intel ISH Firmware Upstreamed Ahead Of Intel Panther Lake Laptops

([Intel] 15 October 01:03 PM EDT Intel ISH + Panther Lake)

In preparation for Panther Lake laptops shipping in early 2026, after Intel appears to have largely wrapped up work on the Linux driver enablement they are now ensuring the necessary firmware bits are out in time.


AMD EPYC 9005 Brings Incredible Performance To The Cloud With Amazon M8a Benchmarks

([Processors] 15 October 10:30 AM EDT 6 Comments)

Last week Amazon/AWS announced the new EC2 M8a instances as their latest-generation, general-purpose compute instances now powered by AMD EPYC 9005 "Turin" processors. Amazon announced the M8a as having up to 30% higher performance and up to 19% better price performance over M7a. With my testing of both at 32 vCPUs, the new AMD EPYC Turin instance provided 1.59x the performance over the prior-generation EPYC Genoa instance!


Apple Announces M5 With Much Faster GPU For AI

([Apple] 15 October 09:25 AM EDT Apple M5)

Apple today announced the M5 SoC as the newest in the Apple Silicon family and with claims of 4x the peak GPU compute performance for AI of the M4.


Tinygrad Gains A Mesa NIR Backend - Initially Supporting NVK/NAK & LLVMpipe Execution

([AI] 15 October 09:09 AM EDT Tinygrad + Mesa NIR)

Merged today to the Tinygrad deep learning framework is a Mesa NIR back-end to allow targeting that common intermediate representation used by these open-source Linux GPU drivers. Initially supported with this Tinygrad NIR back-end is the open-source NVIDIA Vulkan driver "NVK" with its Rust-based NAK compiler as well as the CPU-based LLVMpipe driver.


AMD HIP-RT Is Stable For Blender 5.0 But Will Be Off By Default Until Blender 5.1

([Radeon] 15 October 08:24 AM EDT AMD HIP-RT)

AMD's HIP-RT is used by the Blender 3D modeling software for GPU-accelerated ray-tracing on Radeon GPUs. For Blender 5.0 the AMD HIP-RT support is expected to be declared "stable" but will not be enabled by default until Blender 5.1.


Intel Wildcat Lake "-march=wildcatlake" Added To GCC & LLVM Clang Compilers

([Intel] 15 October 06:38 AM EDT Wildcat Lake)

While the open-source GCC and LLVM/Clang compilers saw Panther Lake support added in early 2024, only overnight was support upstreamed to GCC and Clang for the similar Wildcat Lake target.


Mesa's Zink OpenGL-On-Vulkan Driver Lands Mesh Shader Support

([Mesa] 15 October 06:02 AM EDT GL_EXT_mesh_shader)

In development for around the past year was the cross-vendor mesh shading extension for OpenGL. Last week GL_EXT_mesh_shader was merged to the OpenGL Registry for this mesh shader support and first new GL extension in a while.


More

Technicality, n.:
In an English court a man named Home was tried for slander in having
accused a neighbor of murder. His exact words were: "Sir Thomas Holt
hath taken a cleaver and stricken his cook upon the head, so that one
side of his head fell on one shoulder and the other side upon the
other shoulder." The defendant was acquitted by instruction of the
court, the learned judges holding that the words did not charge murder,
for they did not affirm the death of the cook, that being only an
inference.
-- Ambrose Bierce, "The Devil's Dictionary"