Language Selection

English French German Italian Portuguese Spanish

Graphics/Benchmarks

Linux Work on Graphics by Intel

Filed under
Graphics/Benchmarks
  • Intel Graphics Compiler Changes For Gen12 - Biggest Changes To The ISA Since i965

    Since June Intel's open-source developers have begun volleying the initial open-source graphics driver code for Tigerlake "Gen12" hardware. To date the Gen12 changes haven't been too invasive even with this being the first generation with the "Xe Graphics" engine branding. But that's now changed with a new patch series showing major changes to the graphics instruction set.

    Now that the Intel open-source developers have begun preparing the changes for their graphics compiler back-end for Xe/Gen12 initially with Tigerlake processors, the major architectural changes are beginning to show compared to Icelake "Gen11" graphics and older. The patches show that the Gen12 graphics ISA is one of the biggest reworks ever to the Intel EU ISA since the original i965 graphics a decade ago.

  • Intel Linux Graphics Stack Gets Another Speed Boost - Helping Civilization VI By ~18%

    Intel's open-source Linux graphics driver crew has been on an exciting spree lately of not only punctually enabling new hardware support but also pushing some big performance improvements for new and existing generations of graphics hardware. Today another performance achievement was unlocked.

    Kenneth Graunke who led the development of their new Gallium3D OpenGL driver discovered an optimization today that not only benefits that "iris" driver but also the i965 classic OpenGL driver and their ANV Vulkan driver. The optimization is to stop redirecting the state cache to the command stream command buffer cache section. By looking through their hardware documentation and comparing the behavior to the Intel Windows driver, this optimization was uncovered.

Firefox 69 / 70 Beta Against Chrome 76 On Ubuntu Linux

Filed under
Graphics/Benchmarks

With Firefox 69 released and Firefox 70 entering beta, here are some fresh web browser benchmarks between Mozilla Firefox and Google Chrome from Ubuntu Linux. On the Firefox size, Firefox 68, 69, and 70 Beta were tested with and without WebRender being enabled and compared to Google's current Chrome 76 stable release.

From an Intel Core i9 7960X workstation with Radeon RX 550 graphics running Ubuntu 19.04, the official Linux x86_64 binaries of these different Chrome/Firefox releases were tested. The Firefox runs with WebRender were forced using the MOZ_WEBRENDER environment variable.

Read more

Phoronix on Linux Graphics and Benchmarks

Filed under
Graphics/Benchmarks
Linux
  • XWayland Sees Updated Protocol Support To Help WLROOTS & KDE

    X.Org Server 1.21's XWayland implementation has added support for the xdg-output-unstable-v1 version 3 protocol to help the likes of KDE and compositors like Sway based on WLROOTS.

    The xdg-output-unstable-v1 version 3 protocol update changes the protocol behavior to ensure a wl_output.done event happens after xdg_output changes. More details on the change in that aforelinked Git commit with KDE and WLROOTS being the mentioned Wayland compositors/users affected by the previous behavior resulting in the XWayland monitor not changing position or the cursor being able to reach all of the surface area.

  • AMD Working On Better Page Fault Handling For Navi / Vega GPUs

    Longtime open-source AMD Linux driver developer Christian König on Wednesday sent out a set of patches providing "graceful" page fault handling support for Navi and Vega graphics processors.

    This graceful page fault handling support is on their trek to providing recoverable page fault handling for newer Radeon GPUs with the AMDGPU Linux kernel driver. This ultimately is for improving the behavior of applications/games to avoid hangs and/or invalid access messages being spewed to their dmesg and the like.

  • Linux Benchmarks Of NVMe SSD Performance With Varying I/O Polling Queues

    A Phoronix reader recently pointed out a little known kernel tunable option to adjust the number of I/O polling queues for NVMe solid-state drives that can potentially help improve the performance/latency. Here are some benchmarks from the NVMe poll_queues option.

mesa 19.2.0-rc2

Filed under
Graphics/Benchmarks

Hi List,

I'd like to announce the availability of mesa-19.2.0-rc2. This is the
culmination of two weeks worth of work. Due to maintenance the Intel CI is not
running, but I've built and tested this locally. I would have preferred to get
more testing, but being two weeks out from -rc1 I wanted to get a release out.

Dylan


Shortlog:
--------

Alex Smith (1):
      radv: Change memory type order for GPUs without dedicated VRAM

Alyssa Rosenzweig (1):
      pan/midgard: Fix writeout combining

Andres Rodriguez (1):
      radv: additional query fixes

Bas Nieuwenhuizen (3):
      radv: Use correct vgpr_comp_cnt for VS if both prim_id and instance_id are needed.
      radv: Emit VGT_GS_ONCHIP_CNTL for tess on GFX10.
      radv: Disable NGG for geometry shaders.

Danylo Piliaiev (1):
      nir/loop_unroll: Prepare loop for unrolling in wrapper_unroll

Dave Airlie (2):
      virgl: fix format conversion for recent gallium changes.
      gallivm: fix atomic compare-and-swap

Dylan Baker (1):
      bump version to 19.2-rc2

Ian Romanick (7):
      nir/algrbraic: Don't optimize open-coded bitfield reverse when lowering is enabled
      intel/compiler: Request bitfield_reverse lowering on pre-Gen7 hardware
      nir/algebraic: Mark some value range analysis-based optimizations imprecise
      nir/range-analysis: Adjust result range of exp2 to account for flush-to-zero
      nir/range-analysis: Adjust result range of multiplication to account for flush-to-zero
      nir/range-analysis: Fix incorrect fadd range result for (ne_zero, ne_zero)
      nir/range-analysis: Handle constants in nir_op_mov just like nir_op_bcsel

Ilia Mirkin (1):
      gallium/vl: use compute preference for all multimedia, not just blit

Jose Maria Casanova Crespo (1):
      mesa: recover target_check before get_current_tex_objects

Kenneth Graunke (15):
      gallium/ddebug: Wrap resource_get_param if available
      gallium/trace: Wrap resource_get_param if available
      gallium/rbug: Wrap resource_get_param if available
      gallium/noop: Implement resource_get_param
      iris: Replace devinfo->gen with GEN_GEN
      iris: Fix broken aux.possible/sampler_usages bitmask handling
      iris: Update fast clear colors on Gen9 with direct immediate writes.
      iris: Drop copy format hacks from copy region based transfer path.
      iris: Avoid unnecessary resolves on transfer maps
      iris: Fix large timeout handling in rel2abs()
      isl: Drop UnormPathInColorPipe for buffer surfaces.
      isl: Don't set UnormPathInColorPipe for integer surfaces.
      util: Add a _mesa_i64roundevenf() helper.
      mesa: Fix _mesa_float_to_unorm() on 32-bit systems.
      iris: Fix partial fast clear checks to account for miplevel.

Lionel Landwerlin (2):
      util/timespec: use unsigned 64 bit integers for nsec values
      util: fix compilation on macos

Marek Olšák (18):
      radeonsi/gfx10: fix the legacy pipeline by storing as_ngg in the shader cache
      radeonsi: move some global shader cache flags to per-binary flags
      radeonsi/gfx10: fix tessellation for the legacy pipeline
      radeonsi/gfx10: fix the PRIMITIVES_GENERATED query if using legacy streamout
      radeonsi/gfx10: create the GS copy shader if using legacy streamout
      radeonsi/gfx10: add as_ngg variant for VS as ES to select Wave32/64
      radeonsi/gfx10: fix InstanceID for legacy VS+GS
      radeonsi/gfx10: don't initialize VGT_INSTANCE_STEP_RATE_0
      radeonsi/gfx10: always use the legacy pipeline for streamout
      radeonsi/gfx10: finish up Navi14, add PCI ID
      radeonsi/gfx10: add AMD_DEBUG=nongg
      winsys/amdgpu+radeon: process AMD_DEBUG in addition to R600_DEBUG
      radeonsi: add PKT3_CONTEXT_REG_RMW
      radeonsi/gfx10: remove incorrect ngg/pos_writes_edgeflag variables
      radeonsi/gfx10: set PA_CL_VS_OUT_CNTL with CONTEXT_REG_RMW to fix edge flags
      radeonsi: consolidate determining VGPR_COMP_CNT for API VS
      radeonsi: unbind blend/DSA/rasterizer state correctly in delete functions
      radeonsi: fix scratch buffer WAVESIZE setting leading to corruption

Paulo Zanoni (1):
      intel/fs: grab fail_msg from v32 instead of v16 when v32->run_cs fails

Pierre-Eric Pelloux-Prayer (1):
      glsl: replace 'x + (-x)' with constant 0

Rafael Antognolli (1):
      anv: Only re-emit non-dynamic state that has changed.

Samuel Pitoiset (7):
      radv: allow to enable VK_AMD_shader_ballot only on GFX8+
      radv: add a new debug option called RADV_DEBUG=noshaderballot
      radv: force enable VK_AMD_shader_ballot for Wolfenstein Youngblood
      ac: fix exclusive scans on GFX8-GFX9
      radv/gfx10: don't initialize VGT_INSTANCE_STEP_RATE_0
      radv/gfx10: do not use NGG with NAVI14
      radv: fix getting the index type size for uint8_t

Tapani Pälli (3):
      util: fix os_create_anonymous_file on android
      iris/android: fix build and link with libmesa_intel_perf
      egl: reset blob cache set/get functions on terminate

Thong Thai (1):
      Revert "radeonsi: don't emit PKT3_CONTEXT_CONTROL on amdgpu"

Read more

Also: Mesa 19.2-RC2 Released Following Delay - Many Iris, RADV & RadeonSI Fixes

Running Intel's Clear Linux On AMD EPYC Rome? Still Significant Performance Uplift Over Ubuntu

Filed under
Graphics/Benchmarks

The current AMD EPYC 7742 2P benchmarking that is happening at Phoronix is an interesting Linux/BSD operating system performance comparison. That's in the works while so far are some Ubuntu and Clear Linux numbers. Yes, Intel's open-source Clear Linux platform does run fine generally on AMD hardware -- including the new AMD "Rome" processors -- and generally does still run damn fast. Here is a look at Clear Linux on this 128 core / 256 thread server with Clear Linux against Ubuntu 19.04 as well as the upcoming Ubuntu 19.10.

To the surprise of many in the community, Intel's Clear Linux operating system generally works fine on AMD hardware while retaining many of the performance benefits compared to general purpose Linux distributions. About the extent of Clear Linux issues on AMD hardware I have personally encountered are sometimes NVMe SSDs not being detected, mostly due to power management quirks between the chipset and various power management tuning done by Intel's open-source team, but with time they have rectified those issues. One issue currently encountered is Clear Linux not detecting the PCI Express 4.0 Corsair Force MP600 SSD on AMD Ryzen 3000 + X570 setups, but I'll have some similar Ryzen tests with a PCIe 3 SSD there in the days ahead.

Read more

Graphics: NVIDIA, X.Org and Mesa

Filed under
Graphics/Benchmarks
  • NVIDIA Lands Another New OpenGL Extension In 2019 Around Multi-GPU/SLI

    While most games/engines and software in general are moving from OpenGL to Vulkan, NVIDIA is still investing in their OpenGL driver stack and even adding new multi-GPU/SLI functionality to their driver and as part of that introducing new extensions.

    Back in July I wrote about new OpenGL extensions for multi-GPU rendering that at the time were about exposing explicit controls over multi-cast rendering, the progress fence extension for coordinating operations between multiple GPU command streams, and a WGL-catered extension for creating contexts in a multi-GPU environment.

  • X.Org's Modesetting Driver Flips Off Atomic By Default

    While atomic mode-setting has been around for several years now and to provide a modern mode-setting interface that can test modes prior to the actual operation and reduce possible flickering during mode-setting events and also being faster, the common xf86-video-modesetting driver has at least temporarily disabled the support by default.

    Within the latest X.Org Server 1.21 Git and requested for back-porting to X.Org Server 1.20 is disabling atomic support by default. This stems from a black screen when doing rotation on a different CRTC and other problems along the atomic code-path for this in-tree driver. So by default the generic modesetting driver will use the legacy code-path by default to avoid messing up existing users.

  • Mesa's Gallium3D LLVMpipe Driver Adds Compute Shader Support

    Red Hat's David Airlie has been refocusing efforts recently on improving the state of the LLVMpipe driver that implements OpenGL / OpenGL ES on top of CPUs using LLVM. In the past few weeks he's been wiring up more GL4 / GLES 3.1 extensions and this morning the latest achievement is supporting OpenGL compute shaders!

    Following a long series of patches, Airlie has OpenGL compute shader support working for LLVMpipe in tandem with various LLVM/Gallivm changes.

AMD Ryzen 9 3900X Linux CPU Frequency Scaling Governor Benchmarks

Filed under
Graphics/Benchmarks

Given the recent talk about the Schedutil CPU frequency scaling governor and its future along with CPU frequency scaling behavior in general on AMD Zen 2 processors, here are some benchmarks of the Ryzen 9 3900X when tested with the different Linux "CPUFreq" governor options.

This round of testing is just for reference purposes in looking at how these different CPU frequency scaling governor tunables compare. The power consumption between the modes weren't compared due to the WattsUp Pro meters being busy on other systems and this testing was done prior to the recent Windows/Linux power discoveries; in fact, this testing was from back in July albeit slipped under the radar for publishing these figures.

From the Linux 5.3 Git kernel as of the end of July while running Ubuntu 18.04 LTS, the CPUFreq governor options of ondemand, performance, powersave, and schedutil were tested. For most Linux distributions, "ondemand" is the default governor while "performance" is known for some workloads to deliver better performance when the CPUFreq driver doesn't respond fast enough for ramping up the CPU clock frequencies. Schedutil meanwhile is the newest governor and makes use of the kernel's scheduler utilization data to try to make more informed decisions about what to do with the CPU frequencies. Schedutil has an interesting future ahead, but as we've shown several times, there generally are shortcomings still compared to the likes of Schedutil. The powersave governor, lastly, keeps the clocks at a minimum and really isn't useful for a desktop unless really trying to conserve power. But for this article we're just looking at the performance difference on the Ryzen 9 3900X for easily changing this Linux kernel tunable.

Read more

NAS Parallel Benchmarks: EPYC 7601 vs. EPYC 7742 vs. Xeon Platinum 8280

Filed under
Graphics/Benchmarks

Not included as part of our original EPYC 7742 / EPYC 7002 "Rome" Linux benchmarks was the NAS Parallel Benchmarks (NPB) developed by NASA. While an MPI testing favorite, there were build issues with the older version of NPB packaged by the Phoronix Test Suite. But with recently having updated that test profile against the latest NPB upstream, here are some results for the EPYC 7742 2P, EPYC 7601 2P, and dual Xeon Platinum 8280 benchmark results. Separately, there's also results now for NeatBench 5 with this video editing plug-in test case now part of the Phoronix Test Suite.

Read more

Benchmarks: Phoronix Test Suite 9.0 M3 and More AMD Tests

Filed under
Graphics/Benchmarks
  • Phoronix Test Suite 9.0 M3 Brings Improvements Around Offline/Private Testing

    The third and likely final development milestone release ahead of this month's Phoronix Test Suite 9.0-Asker release is now available for cross-platform, fully-automated benchmarking.

    Phoronix Test Suite 9.0 Milestone 3 most notably brings improvements for running the Phoronix Test Suite in an offline environment particularly when there is no OpenBenchmarking.org test/suite data present. Phoronix Test Suite 9.0 is now shipping a static cache of all the tests/suites as of release time to help offline users. Those running with Internet connectivity will continue to query OpenBenchmarking.org for new/updated profiles. This should help those that initially run Phoronix Test Suite in an offline environment.

  • AMD Firmware Update To Bring Boost Performance Optimizations

    There has been a lot of talk recently of AMD Ryzen 3000 series processors reportedly not hitting their boost clock frequencies, whether stock coolers are adequate for hitting the boost frequencies, and other concerns around the boost behavior on these new Zen 2 processors. AMD issued a statement today they are rolling out a new BIOS/firmware update to help with boost clock frequency optimizations.

  • PowerTop, AMD CPUFreq CPPC & Other Power Tests From The Ryzen 9 3900X On Linux

    Idle load was about three Watts less for the overall AC system power consumption, the average power draw under idle and load came out to 2~3 Watts, and the peak AC power draw was just one Watt lower. That's the out-of-the-box state compared to setting all the PowerTop tunables to their "good" state. So it saved 1~3 Watts, but still off the 10~30 Watt differences seen on the same hardware between Ubuntu 19.04 (plus Linux 5.3) and Windows 10 1903.

DAV1D Experimenting With Vulkan & OpenGL ES GPU Offloading

Filed under
Graphics/Benchmarks
Movies
OSS

There isn't any AV1 video decode/encode built into the video engines of today's GPUs, but the DAV1D project CPU-based AV1 decoder is experimenting with offloading some aspects of the process to current generation hardware with OpenGL ES and Vulkan.

There are experimental branches of the dav1d decoder that exploit OpenGL ES and Vulkan for offloading some elements of the decode process to the graphics hardware. At the moment Self-Guided Restoration (SGR) and Constrained Directional. Enhancement Filter (CDEF) is among the functionality implemented for OpenGL ES / Vulkan.

Read more

Syndicate content

More in Tux Machines

Android Leftovers

Canonical Outs New Linux Kernel Security Update for Ubuntu 18.04 and 16.04 LTS

Affecting both the Linux 4.15 kernel used in Ubuntu 18.04 LTS (Bionic Beaver) and Ubuntu 16.04.6 LTS (Xenial Xerus) systems, the new security patch fixed an improperly implemented Spectre mitigation in the ptrace susbsystem (CVE-2019-15902), which could allow a local attacker to expose sensitive information. It also addresses a buffer overread (CVE-2019-15918) discovered that the SMB networking file system implementation, which could allow an attacker to expose sensitive information (kernel memory), two flaws (CVE-2019-15117 and CVE-2019-15118) discovered in the USB audio driver that may allow a physically proximate attacker to crash the system, and a flaw (CVE-2019-14821) in the KVM hypervisor implementation that let a local attacker to crash the system. Read more

Leftovers: MX-19, Versalogic and Security

  • MX-19 “patito feo” released!

    We are pleased to offer MX-19 for your use. As usual, this iso includes the latest updates from debian 10.1 (buster), antiX and MX repos.

  • Compact Apollo Lake SBC aims sky high

    Versalogic’s Linux-ready, sandwich-style “Harrier” SBC has an Apollo Lake processor and a compact 95 x 55mm footprint, ECC RAM support, and ruggedization features designed for high altitude UAVs. Versalogic announced a Harrier SBC due in Q1 2020 that revises the compact, COM-and-carrier design of its three-year-old, Intel Bay Trail based Osprey, but advances to the newer Intel Apollo Lake. The Osprey is similarly bereft of real-world ports to enable easier real-world deployments in constrained environments.

  • Security updates for Tuesday

    Security updates have been issued by CentOS (jss and kernel), Debian (libpcap, openjdk-8, and tcpdump), Fedora (java-11-openjdk), openSUSE (libreoffice), Oracle (java-1.7.0-openjdk), Red Hat (java-1.7.0-openjdk, python, and wget), Scientific Linux (java-1.7.0-openjdk), SUSE (ceph, ceph-iscsi, ses-manual_en, dhcp, openconnect, and procps), and Ubuntu (exiv2, linux, linux-aws, linux-azure, linux-gcp, linux-kvm, linux-raspi2, linux, linux-aws, linux-kvm, linux-raspi2, linux-snapdragon, linux-hwe, linux-azure, linux-gcp, linux-gke-5.0, linux-snapdragon, and uw-imap).

  • Password lessons: Longer is better, so is salt

    Infosec pros who had no idea of how easily a stolen list of hashed passwords could be cracked got a sobering lesson at this month’s SecTor security conference in Toronto. There, Will Hunt, co-founder of the U.K. based In.security consulting firm, casually talked of systems that can be built around a common (about $1,500) Nvidea GTX 2080 graphics card that could make 100 billion guesses a second in a brute force attack.

Unix Celebrates 50 Years

Today and tomorrow Nokia Bell Labs is hosting a two-day event celebrating 50 years of the Unix operating system, reflecting on Unix’s past and exploring the future of computing. Speakers and panelists include many of the original team that built Unix and designed the C programming language. Read more