Intel News Archives

With the Intel VT-d 4.0 specification there is performance monitoring "PerfMon" infrastructure introduced. A new patch series from Intel is preparing for IOMMU performance monitoring with the Linux kernel code.

Back in July Intel engineers published the initial open-source driver code around the new Versatile Processing Unit "VPU" coming with Meteor Lake. This VPU block with 14th Gen Core CPUs is intended for AI inference acceleration for deep learning software.

For years Intel has been developing HAXM as a hardware-accelerated execution manager with a focus on using it for the Android Emulator and QEMU in conjunction with Intel VT enabled processors. HAXM works not only on Linux but Windows, macOS, and some BSDs. Unfortunately, Intel has decided to discontinue development of HAXM.

Merged today for the LLVM 16 compiler stack is support for Intel's next-generation Xeon Scalable "Emerald Rapids" processors with -march=emeraldrapids now being supported.

Since GCC 11 there has been support for AMX and the upcoming Sapphire Rapids CPU features, which has been further improved in the open-source compiler over the past two years. GCC 13 meanwhile as the next GNU Compiler Collection release is bringing Meteor Lake and Sierra Forest, Grand Ridge, and Granite Rapids. Basic enablement of Intel's Emerald Rapids meanwhile was merged yesterday for GCC 13 too.

Intel engineers had submitted support for Linear Address Masking (LAM) with the recently-closed Linux 6.2 merge window but it was rejected by Linus Torvalds. In working toward re-submitted it for the v6.3 cycle or later, an updated Linux LAM patch series was posted today.

If you are running the newest Intel Raptor Lake processors with integrated graphics and the latest Intel Arc Graphics discrete graphics cards under Linux, you are currently relying on the Intel "i915" DRM kernel graphics driver... As implied by the name, it's been used with Intel graphics going back to the old 915G chipset days nearly twenty years ago. But Intel has been working on a new "Xe" kernel graphics driver they have initially announced today and aim to make it production-ready in 2023 for supporting their modern Xe Graphics hardware.

Intel's OpenVINO toolkit for deep learning is out with a major release ahead of the holidays and now has full support for Xeon Scalable "Sapphire Rapids" as well as full support now for their discrete GPUs.

Intel last year published documentation concerning a feature for future CPUs that they dubbed FRED, the Flexible Return and Event Delivery. FRED has the capability of helping system performance and response time while now initial patches for the Linux kernel have been published for supporting FRED.

Intel overnight released oneDNN 3.0 as the newest major release to this open-source project for assisting in building deep learning applications. This oneAPI software component can already be used by PyTorch, ONNX, MATLAB, and other prominent software while the v3.0 release prepares it for future Intel hardware.

The "char/misc" changes have been merged for the in-development Linux 6.2 as the random catch-all area of the kernel for drivers not fitting well in other subsystem areas. Notably with this update for Linux 6.2 is continued work on enabling the Intel-owned Habana Labs Gaudi2 AI accelerator.

Linus Torvalds can be known for his hardware commentary at times like hoping AVX-512 "dies a painful death", Intel's "bad policies" around ECC memory, and giving NVIDIA the finger. The latest colorful commentary by the Linux creator is around Intel's new Linear Address Masking (LAM) feature that aimed to land in Linux 6.2 but is now delayed until the code can be reworked.

After recently getting H.264 and H.265 video decode working for the Mesa RADV Vulkan driver with the current Vulkan Video extensions, David Airlie of Red Hat resumed his prior work on enabling the Vulkan Video extensions for the open-source Intel "ANV" driver too.

Ahead of 4th Gen Intel Xeon Scalable, Xeon CPU Max, and Intel Data Center GPUs shipping, Intel today announced the oneAPI 2023 tools release.

The x86/microcode changes that were merged this week into the Linux 6.2 kernel address prior shortcomings with the Intel In-Field Scan (IFS) driver so it's now deemed ready to help in spotting out faulty silicon across a fleet of systems in production or prior to commissioning new hardware.

There's a fair amount going on in the power management space for both Intel x86_64 and Arm hardware with the Linux 6.2 kernel.

In addition to the in-development Linux 6.2 bringing TDX guest attestation support for use with new processors, another new hardware security feature being enabled with this next kernel release is Asynchronous Exit Notification for Software Guard Extensions (SGX).

The work by Intel engineers the past few months on Call Depth Tracking as a less costly mitigation for Retbleed on Skylake-era processors is now set to be merged for the Linux 6.2 kernel.

What first entered the kernel as the "Software Defined Silicon" and now set to be marketed as Intel On Demand is ready to go with Linux 6.2 for this CPU license activation model appearing with upcoming Intel Xeon server processors.

The Intel TDX guest support has been merged into the in-development Linux 6.2 kernel.

Going back to the summer Intel posted their initial open-source Linux driver for their Versatile Processing Unit "VPU" debuting with Meteor Lake. Since then they have continued refining this open-source VPU Linux driver and with the latest patch series have adapted it to make use of the new accelerator framework/subsystem premiering in Linux 6.2.

While Intel's GPU compute stack for Linux is fully open-source, one area where it still has room for improvement is getting it packaged up on more Linux distributions. The reference binaries published by Intel for their Compute-Runtime and Level Zero components are just Debian/Ubuntu packages but with time -- and as Arc Graphics and other hardware becomes available -- we are seeing more distributions taking a stab at offering up their own package builds.

Released this morning is Intel IGC 1.0.12504.5 for Linux and Windows systems as the newest version of the open-source Intel Graphics Compiler. As the first tagged update in nearly two months, IGC 1.0.12504.5 is a big one.

Intel's oneAPI Deep Neural Network Library "oneDNN" is preparing to embark on its v3.0 release that improves performance not only for current and upcoming Intel hardware but also furthers along the NVIDIA and AMD support too.

With the new Mesa 22.3 release one of the changes for vintage hardware users is the introduction of "HasVK" as a Vulkan driver forked from Intel's ANV codebase.

Intel engineers have published their "2022Q3" patch queue for FFmpeg along with a "2022Q4 RC1" set too for representing the latest yet-to-be-merged patches for improving FFmpeg video acceleration with Intel graphics.

CVE-2022-4139 was made public today as an i915 kernel graphics driver security issue affecting all Gen12 graphics -- from integrated Tigerlake graphics up through the latest Raptor Lake graphics as well as the in-development Meteor Lake code plus the discrete GPUs of DG2/Alchemist and Arctic Sound.

Beyond Intel's well regarded open-source software work deep down the stack on everything from ensuring timely new hardware support to optimizing the Linux kernel with various performance optimizations and investing heavily in the upstream GNU and LLVM toolchains (and much more!), they also make terrific software ecosystem advancements higher up the stack. One of their most promising areas of open-source software work in recent years has been around oneAPI. Next week Intel will be hosting a virtual oneAPI DevSummit for AI and HPC where they will be showing off their latest software advancements.

Back in September was a big patch set working out classes of tasks for hybrid CPUs and more properly implementing Intel Thread Director for Linux. This work to better the performance/efficiency of modern Intel Core CPUs with a mix of P and E cores has now advanced past the "request for comments" stage with a new patch series sent out on Monday.

While these days the Intel-owned Habana Labs Linux software stack is a shining example of an open-source AI accelerator solution with mainline kernel driver support and also helping bring together the new compute accelerator subsystem, it wasn't always so blessed. Initially there was the closed-source user-space bits that fortunately last year was opened up with SynapseAI Core.

While for years Intel has been very well regarded -- and rightfully so -- for their open-source Linux hardware support, occasionally there are exceptions. One such exception currently is Intel's IPU6 drivers for their MIPI cameras found on many newer Alder lake laptops and presumably upcoming Raptor Lake laptops too. The IPU6 drivers remain outside of the Linux kernel and will still likely be that way for sometime.

A patch was merged today to Mesa 23.0 as part of the effort for building the Intel OpenGL and Vulkan Linux drivers for non-x86/x86_64 architectures. This is part of the ongoing effort to enable Intel discrete GPUs to eventually work on the likes of AArch64, POWER, and RISC-V systems.

Back in August an Intel engineer posted a patch series to help Linux on Intel hybrid CPUs with Hyper Threading. That work benefiting newer Alder Lake and Raptor Lake processors is around avoiding unnecessary migrations within SMT domains. Finally an updated version of that patch series has now been posted as this Intel hybrid improvement works its way toward the kernel.

Over the past year since being the first to report on Software Defined Silicon (SDSi) for license-activated hardware features with future CPUs, we've been left to wonder all what will be incorporated into this controversial Software Defined Silicon. Intel recently shared that SDSi will be marketed as Intel On Demand and now they have published a web page outlining more about this after-purchase upgrades for activating extra processor features.

Intel has published a new version of its oneVPL library as its open-source oneAPI Video Processing Library for video encoding/decoding and media processing across CPUs, GPUs, and other accelerators.

Intel open-source engineers continue working on getting their Trust Domain Extensions (TDX) support squared away for the mainline Linux kernel. With the upcoming Linux 6.2 cycle, the TDX guest driver is now ready.

In addition to Linux 6.2 promoting [DG2] Arc Graphics to stable, this next kernel version will no longer deem the Intel In-Field Scan (IFS) driver as "broken" now that it's API/ABI is in good shape.

A final batch of drm-intel-next feature patches were submitted on Friday to complement the drm-intel-gt-next patches. Most exciting with this last minute PR for Linux 6.2 is the DG2/Alchemist discrete GPU support no longer being treated as experimental.

For going along with the i915 DRM kernel driver support to premiere in Linux 6.2, the Mesa 23.0 development code for Intel's Vulkan driver is exposing performance metrics / hardware counters for DG2 "Alchemist" Arc Graphics hardware.

Going back to late 2020 Intel's open-source/Linux engineers have been working on Linear Address Masking "LAM" enablement for that feature coming with future processors. With the upcoming Linux 6.2, the kernel-side enablement for Intel LAM appears to be finally wrapped up.

Intel today published their "20221108" CPU microcode collection alongside announcing various security disclosures for the quarter. Fortunately on the CPU microcode side, the changes are all focused on functional issues.

Over the past three years one of Intel's many promising open-source software projects has been the Rust-written Cloud Hypervisor. Cloud Hypervisor started as just a modern, security-focused, cloud-centric Rust VMM hypervisor for modern hardware/software. It began as just one of many open-source software projects at Intel but last year was folded into the Linux Foundation umbrella while Intel continues to be a major contributor to the project. Coming as a bit of a surprise today is AMD announcing they have joined the Cloud Hypervisor project.

With next-generation Meteor Lake CPUs the integrated graphics are set to have native HDMI 2.1 display capabilities. Intel's open-source Linux kernel driver has begun those HDMI 2.1 preparations and sent out today were early patches for enabling HDMI 2.1 Fixed Rate Link (FRL).

Intel compiler engineers continue being very busy working to land as much of the new CPU feature support as they can into GCC 13 for what is the next annual compiler release that will debut as GCC 13.1 in the early months of 2023.

Future Intel CPUs and some existing processors via a microcode update will support a new feature called the Asynchronous EXit (AEX) notification mechanism to help with Software Guard Extensions (SGX) enclave security. Patches for the Linux kernel are pending for implementing this Intel AEX Notify support with capable processors.

On the same day as new GCC compiler patches for next-generation processors coming out for those wares that are more than one year out, the GNU Compiler Collection dropped a remnant from Intel's past: the Many Integrated Core "MIC" architecture support with Xeon Phi for offloading.

A patch on its way to the mainline Linux kernel tunes Alder Lake N and Raptor Lake P mobile processors to have slightly lower power consumption out-of-the-box to help with battery life and thermal characteristics.

Stemming from the recent discussion of Intel's open-source Linux driver for Arc Graphics not yet running on POWER, another rather interesting support caveat was also raised. It turns out updating the GSC firmware for Arc Graphics hardware currently requires the Intel Management Engine (ME) functionality, which basically limits the graphics card firmware updating in turn to systems with Intel CPUs. (Update: See end of article.)

While Xeon Scalable "Sapphire Rapids" will finally see its formal launch in January as recently revealed by Intel, it will then be succeeded down the road by Emerald Rapids. Succeeding Emerald Rapids will then be Granite Rapids to which there is now an initial GCC compiler enablement patch posted. Granite Rapids won't be out until at least well into 2024 while fortunately they have already begun their compiler enablement work to ensure that new CPU instructions and other capabilities are in place well ahead of launch.

Besides open-source drivers being loved by Linux enthusiasts for the greater technical clarity/insight, better security with the ability to verify the driver's behavior, and better durability of the driver over the longer-term, another common open-source driver benefit is being able to get the drivers working on other CPU architectures not otherwise a focus by the upstream hardware vendor. With Intel's open-source graphics driver stack for Arc Graphics and also in the data center with the Data Center GPU Flex Series and forthcoming Ponte Vecchio, it's drawn interest from ARM, RISC-V, and POWER folks. Unfortunately at least in the case of the POWER9 hardware, the current Intel Linux graphics driver isn't yet building properly there.