Language Selection

English French German Italian Portuguese Spanish

Graphics/Benchmarks

Graphics: RADV, Wayland's Weston 8.0 Reaches Alpha, Mesa 20.0

Filed under
Graphics/Benchmarks
  • RADV's ACO Compiler Back-End Now Supported For Older AMD "Sea Islands" GPUs

    The Valve-backed "ACO" compiler back-end for the open-source Radeon "RADV" Vulkan driver has added support now for AMD GCN 1.1 "Sea Islands" graphics cards.

    Sea Islands includes the Radeon Rx 200 series with the R7 260/260X/290/295 series and these 2nd Gen GCN GPUs also ended up in the Radeon HD 7790, Radeon HD 8770, Radeon R7 360, Radeon R9 390/390X, and Radeon RX 455. Up to now the ACO compiler back-end has only supported GCN1.2/GFX8 and newer but with the latest Mesa 20.0-devel code as of today in Mesa Git there is now ACO support for GCN 1.1 Sea Islands.

  • Wayland's Weston 8.0 Reaches Alpha With EGL Partial Updates, Headless OpenGL

    Weston 8.0 is another significant update for this Wayland reference compositor in it offers EGL_KHR_partial_update support to reduce GPU vRAM usage on supported drivers/GPUs thanks to handling partial screen updates, support for building the DRM back-end without Mesa's GBM, greater hardware planes usage, the Weston headless back-end now supports OpenGL, a direct display extension, HDCP support in the DRM back-end, and various other improvements / features.

  • weston 7.0.91
    This is the alpha release for weston 8.0. Here is a highlight of the
    main new features:
    
    - DRM hardware planes should be used more effectively
    - Headless backend now supports OpenGL
    - DRM backend can now be built without GBM
    - EGL_KHR_partial_update is now used when available, reducing memory
      bandwidth usage on some GPUs
    - Logging framework improvements
    - Documentation for weston can now be built
    
    A lot of fixes have been merged too. Thanks to all contributors!
    
    Full commit history below.
    
    Adam Jackson (5):
          simple-dmabuf-egl: Allow QueryDmaBufModifiers to report no modifiers
          gl-renderer: Fix possible memory leak when no dmabuf modifers are supported
          libweston: Fix integer underflow in weston_layer_mask_is_infinite
          image-loader: Fix undefined left shift in premultiply_data
          tests: Fix undefined left shift in internal-screenshot-test
    
    Ankit Nautiyal (6):
          backend-drm: Add support for content-protection
          libweston: Add functions to modify disable_planes counter for an output
          libweston: Add function to schedule idle task for updating surface protection
          libweston: Notify the client, when output recording is started/stopped
          man: Declare drm-backend support for HDCP
          backend-drm: Check for HDCP Content Type property before setting
    
    Daniel Stone (8):
          renderer-gl: Assert function presence matches extensions
          remoting: Use DRM FourCC formats instead of GBM formats
          Revert "backend-drm: Teach drm_property_info_populate() to retrieve range values"
          config-parser: Export get_full_path and destroy
          backend-drm: Use aspect-ratio bit definitions from libdrm
          config-parser: Make get_bool be bool
          tests/config-parser: Remove useless duplicate test
          option-parser: Make bools boolean
    
    Drew DeVault (1):
          simple-dmabuf-egl: update to xdg-shell stable
    
    Eero Tamminen (1):
          Add include for missing symbols
    
    Emmanuel Gil Peyrot (1):
          shared: Use memfd_create() when available
    
    Harish Krupo (3):
          gl-renderer: Censor protected views when output is recorded
          clients/window: Add viewport destination support
          desktop-shell: Set 1x1 buffers for solid-color backgrounds
    
    Jeffy Chen (2):
          clients: Drop corresponding touch points when destroying windows
          clients: Add more sanity checks to catch destroyed window
    
    Leandro Ribeiro (11):
          build: bump libdrm requirement to newer version (2.4.83)
          backend-drm: remove unecessary ifdef checks
          backend-drm: remove unnecessary ifdefs
          move frame_signal emission to weston_output_repaint()
          screenshooter: stop using frame_signal void *data parameter to get weston_output
          tests: stop using frame_signal 'void *data' argument to get weston_output
          renderer: change frame_signal emission to pass previous_damage as data argument
          screenshooter: get previous_damage from data argument instead of weston_output
          screen-share: get previous_damage from data argument instead of weston_output
          Revert "move frame_signal emission to weston_output_repaint()"
          libweston: remove previous_damage from struct weston_output
    
    Link Mauve (1):
          xwayland: Remove unused variable
    
    Loïc Yhuel (1):
          libweston: fix possible crash after a view is removed the layer
    
    Marius Vlad (53):
          weston-log: s/scope/sub, leftover from the logging framework
          libweston: Fix rename of weston_compositor_destroy() reference
          weston-log: 'new_subscriber' is actually 'new_subscription'
          weston-log: Add 'destroy_subscription' callback for the subscription
          weston-log-internal: Allow to hang-off data over the subscription
          weston-log: Add a subscription iterator
          libweston: Clean-up timeline to make room for a new approach
          libweston: Introduce timeline subscription and timeline subscription object
          libweston: Create the 'timeline' scope
          libweston: Convert timeline points to use the timeline scope
          libweston: Notify timeline of object modification
          libweston: Remove timeline-object and clean-up
          doc/sphinx: Add some documentation about timeline points
          compositor: Allow protocol to be displayed when asked for, even if we're not supplying debug argument
          libweston: Init weston_output's 'destroy_signal' before timeline has a chance to emit a
          compositor: Pass the entire string in one-shot when writting logger data
          weston-log: Avoid prefix-matching the scope name when checking for a
          backend-drm: Teach drm_property_info_populate() to retrieve range values
          backend-drm: Teach drm_property_info_populate() to retrieve range values
          backend-drm: Add zpos DRM-property
          backend-drm: Add a helper to display plane type as a 'string'
          backend-drm: Hard-code zpos values if HW doesn't exposes them
          libweston: Add a new helper weston_view_has_valid_buffer
          libweston: Add a new helper to check if the view spawns the entire
          backend-drm: Construct a zpos candidate list of planes
          backend-drm: Place pixel format checks for the overlay plane in its own
          backend-drm: Place pixel format checks for the cursor plane in its own
          backend-drm: Check pixel format before constructing the zpos candidate list
          backend-drm: Allow for views to reach overlays/underlays planes
          backend-drm: Pass the plane to prepare_overlay_view
          backend-drm: Pass the drm_fb to each prepare_overlay/scanout_view functions
          backend-drm: Move plane's availability in drm_output_try_view_on_plane()
          backend-drm: Print whenever a view will reach the renderer region
          backend-drm: Print whenever a view could not placed on the primary due to
          compositor: Fix some warning when passing debugoptimized to meson
          protocol: Add weston-direct-display extension
          libweston: Add weston-direct-display server side implementation
          libweston: Add the ability to determine if a dmabuf is scanout-capable
          backend-drm: Add dmabuf scan-out check for DRM-backend
          renderer-gl: Avoid retrieving the EGL image it direct_display flag was set
          renderer-gl: Display a solid shader color when direct-display is in use
          clients/simple-dmabuf-egl: Make use of direct-display
          clients/simple-dmabuf-drm: Make use of direct-display
          backend-drm: Assign the primary plane the lowest zpos value
          backend-drm: Skip testing plane state if plane is not enabled
          backend-drm: Turn zpos duplicate check into an hard assert
          backend-drm: Further checks to skip plane assignment to HW planes
          weston-log-flight-rec: Add a global variable to access the ring buffer
          weston-log-flight-rec: Don't allow more than one flight recorder to be
          weston-log-flight-rec: Fix useless comparison when displaying the
          doc/scripts/gdb: Added gdb script to dump contents of the flight recorder
          clients/fullscreen: Refuse to resize the surface size when fullscreen'ed
          gitlab-ci: Update ci-templates to latest SHA commit
    
    Miguel A. Vico (2):
          desktop-shell: Avoid NULL output dereference when getting surface label
          compositor: Do not trigger invalid destructors when hotunplugging
    
    Nicholas Niro (2):
          backend-drm: Fix for gbm modifiers when they are not available.
          backend-drm: Added support for legacy fd_import
    
    Olivier Fourdan (1):
          xwm: Use Xwayland allow commits for repaint
    
    Pekka Paalanen (73):
          backend-headless: fix comment on use_pixman
          backend-headless: refactor into headless_output_enable_pixman
          backend-headless: refactor into headless_output_disable_pixman
          backend-headless: make renderer type an enum
          clients: fix len-string formatting
          gl_renderer: remove unused NO_EGL_PLATFORM
          gl-renderer: fix typo native_window to native_display
          gl-renderer: remove platform_attribs
          gl-renderer: remove gl_renderer_display
          gl-renderer: remove gl_renderer_output_surface
          gl-renderer: remove print_egl_error_state
          backend-drm: use format db for fallback too
          gl-renderer: move into egl-glue.c
          gl_renderer: introduce gl_renderer_get_egl_config()
          gl-renderer: use gl_renderer_get_egl_config() for display_create
          gl-renderer: do not even pick a config with configless_context
          pixel-formats: add RGBA bits and type fields
          gl-renderer: use pixel_format_info internally for EGL
          gl-renderer: fuzzy EGLConfig matching for non-GBM
          backend-wayland: use DRM formats for EGLConfig
          backend-x11: use DRM formats for EGLConfig
          gl-renderer: remove EGLConfig attributes from API
          gl-renderer: configs for pbuffers too
          gl-renderer: pbuffer config for non-surfaceless
          gl-renderer: prefer the base EGLConfig
          gl-renderer: improve get_egl_config errors
          gl-renderer: print detailed EGLConfig list
          gl-renderer: use EGLConfig printer for window outputs
          build: shells do not need matrix.c
          build: use dependency for matrix.c
          xwm: dnd does not need cairo-util.h
          Unify the include style of shared/ headers
          build: simplify include_directories
          xwm: no need for compositor/weston.h
          gl-renderer: display_create needs surface type
          gl-renderer: document display_create
          gl-renderer: document output_window_create
          gl-renderer: add EGL surfaceless platform support
          noop-renderer: zero-initialize struct
          headless, gl-renderer: support pbuffer outputs
          compositor: add use-gl option to headless
          gitlab-ci: fix pages
          build: separate deps for int and ext libweston users
          build: link libm explicitly
          build: link libdl explicitly to DRM backend
          backend-x11: need libdrm headers in build
          build: reduce sub-dependencies of libweston
          compositor: turn weston main() into a lib
          cms-colord: work around unresolved symbols
          backend-rdp: work around unresolved symbols
          Link Weston plugins to libexec-weston.so
          tests: surface-screenshot needs libshared
          build: do not allow unresolved symbols
          libweston: drop a misleading dmabuf comment
          tests: remove static data from viewporter
          tests: remove static data from ivi-layout-test-plugin
          tests: remove static data from ivi-shell-app-test
          tests: remove static data from ivi-layout-test-client
          tests: remove static data from presentation
          tests: fix test-shell init error path
          ivi-shell: fix init error path
          colord: remove destroy listener on clean-up
          Use weston_compositor_add_destroy_listener_once() in plugins
          libweston: allow double-loading modules
          compositor: allow double-loading modules
          tests: write image to current directory by default
          tests/subsurface-shot: hardcode reference image names
          tests: replace fprintf() with testlog()
          tests/xwayland: do not call exit(SUCCESS)
          tests: rename struct weston_test to weston_test_entry
          tests/ivi: rename test_section
          tests: drop FAIL_TEST
          libweston: do not include weston.h
    
    Sebastian Wick (7):
          shared: add read-only anonymous file abstraction
          CI: build wayland from source
          input: bump wl_seat version to 6
          clients/window: bump wl_seat version to 6
          input: bump wl_seat version to 7
          clients/window: bump wl_seat version to 7
          input: use ro_anonymous_file to minimize duplication of keymap files
    
    Simon Ser (4):
          build: reopen master for regular development
          clients: drop simple-dmabuf-drm
          clients: remove leftover from simple-dmabuf-drm
          build: bump to version 7.0.91 for the alpha release
    
    Stefan Agner (10):
          backend-rdp: don't use shadow buffer for the RDP backend
          backend-headless: fix build issue without gl-renderer
          clients: avoid build error without gl-renderer
          gitlab-ci: add build configuration without gl-renderer
          backend-drm: use DRM_ constants everywhere
          remoting: make sure GL renderer is enabled
          backend-drm: separate out DRM virtual support
          backend-drm: make GBM optional
          weston-launch: show when a signal is sent to a child
          weston-launch: use exec to ensure signal delivery
    
    Veeresh Kadasani (1):
          simple-dmabuf-egl: make application generic
    
    Vivek Kasireddy (2):
          gl-renderer: Replace EGL_*_WL macros with locally defined enums
          gl-renderer: Add support for XYUV format (v2)
    
    sichem (1):
          libweston: Bring back 'weston_output_move'
    
    git tag: 7.0.91
    
    
  • Mesa 20.0 Now Includes Intel's Gallium3D Driver To Build By Default

    As part of the ongoing effort for Intel's plans to use their new Gallium3D OpenGL Linux driver by default on next quarter's Mesa 20.0 for Broadwell "Gen8" graphics and newer, another step in that direction was achieved on Friday.

    Intel's "Iris" Gallium3D driver is still making good progress in its goal for Mesa 20.0 to switch the default "i965" classic driver to Intel Gallium3D for Broadwell and newer hardware. Earlier this week was adding a build-time option to change the Intel OpenGL driver default so those building from source or distribution vendors can change the default on their own with ease.

CentOS 6 Through CentOS 8 Benchmarks On Intel Xeon

Filed under
Graphics/Benchmarks

Complementing the CentOS 8 benchmarks I did following the release of that Red Hat Enterprise Linux 8 rebuild in late September, here are tests going back further for showing the performance of CentOS 6, CentOS 7, and CentOS 8 all benchmarked from the same Intel Xeon Scalable server. These tests were done about a month ago albeit with all the hardware launches, new child, and other factors, only now getting to posting the data.

These benchmarks are of CentOS 6, CentOS 7, and CentOS 8 with all available stable release updates for each as of early November (prior to TAA, JCC Erratum, and other more recent disclosures). This was done to look at how the performance of these CentOS releases compare that track RHEL6, RHEL7, and RHEL8 respectively. Additionally, for each operating system was also a secondary run when booted with mitigations disabled to also provide a look at the CentOS Linux performance with the various CPU security mitigations disabled.

Read more

RadeonSI NIR Benchmarks Show Great Progress With Mesa 20.0

Filed under
Graphics/Benchmarks

With AMD last week having enabled OpenGL 4.6 for their RadeonSI OpenGL Linux driver when enabling the NIR intermediate representation support, you may be wondering how using NIR is stacking up these days compared to the default TGSI route. Here are some benchmarks on Polaris, Vega, and Navi for comparing this driver option that ultimately allows OpenGL 4.6 to be flipped on.

NIR is the modern intermediate representation used by a majority of Mesa drivers now in some capacity as an alternative to the likes of TGSI as what had been the default IR for Gallium3D drivers. With RadeonSI they have been transitioning to NIR since that has been the growing trend of these open-source drivers for sharing IR optimizations and the like. As well, NIR is being wired up in order to re-use some code-paths used currently by the "RADV" Radeon Vulkan driver to share some of the SPIR-V work that was needed in order for RadeonSI to have OpenGL 4.6 support. Like on the Intel side when they crossed the OpenGL 4.6 milestone recently, the big blocker to GL 4.6 on these drivers was handling SPIR-V ingestion with GL_ARB_gl_spirv / GL_ARB_spirv_extensions.

Read more

Nvidia Is Preparing An Unexpected Surprise For Linux Users In 2020

Filed under
Graphics/Benchmarks
Linux

Each year Nvidia hosts the GPU Technology Conference, a global gathering of AI developers, data scientists, graphic artists, and pretty much anyone in the technology industry working with GPUs in their chosen fields. The event packs in keynotes with roadmaps and reveals, face-time with Nvidia engineers, and hundreds of sessions to participate in. GTC 2020, though, looks to include a special surprise for Linux users and open source enthusiasts.

Supporting Nouveau eh? That’s the open source Linux driver used to drive Nvidia graphics cards (Nvidia also supplies a proprietary driver for Linux), and Nvidia’s historical lack of contributions is what led Linus Torvalds to famously flip Nvidia the bird and utter words I can’t print here. (I can link to them though. . .)

The community of developers working on the Nouveau driver have experienced several roadblocks throughout the years. Paramount among them is the inability to achieve normal GPU clock speeds due to Nvidia’s locked down firmware on many models of graphics cards. This leads to undesirable performance and a multitude of potential video display issues across many Linux distributions.

Read more

An Extensive Look At The AMD Naples vs. Rome Power Efficiency / Performance-Per-Watt

Filed under
Graphics/Benchmarks

Since the AMD EPYC 7002 "Rome" series launch in August we have continue to be captivated by the raw performance of AMD's Zen 2 server processors across many different workloads as covered now in countless articles. The performance-per-dollar / TCO is also extremely competitive against Intel's Xeon Scalable line-up, but how is the power efficiency of these 7nm EPYC processors? We waited to deliver those numbers until having a retail Rome board for carrying out those tests and now after that and then several weeks of benchmarking, here is an extensive exploration of the AMD EPYC 7002 series power efficiency as well as a look at the peak clock frequencies being achieved in various workloads to also provide some performance-per-clock metrics compared to Naples.

Read more

Graphics: Mir, X.Org, Gallium3D, GPUOpen, Mesa, Lima and Libinput

Filed under
Graphics/Benchmarks
Linux
  • Mir 1.6 Released With New Wayland, DispmanX Platform Support

    Mir 1.6 is out today with the latest batch of features for this Ubuntu-focused display server that offers Wayland client compatibility.

    The two big additions to Mir 1.6 are on the graphics platform front. First, there is now a "Wayland platform" for running Mir as a nested compositor on top of a Wayland compositor. Secondly, the rpi-dispmanx platform is for Broadcom's DispmanX API.

  • Before Ending 2019, Vintage SiS X.Org Driver Sees A New Release

    xf86-video-sis 0.12.0 is available this week as a new version of the SiS display driver for X.Org systems in supporting Silicon Integrated Systems' display hardware.

    This X.Org user-space mode-setting driver has seen its first update in four months but prior to that it hadn't seen any update to the open-source code in three years.

  • RadeonSI Lands SDMA Copy Support For Vega/GFX9

    The RadeonSI Gallium3D driver has finally landed SDMA copy support for Vega/GFX9 graphics hardware, which should principally benefit compute shaders and other cases.

  • AMD's GPUOpen Releases Vulkan Memory Allocator 2.3

    AMD's GPUOpen team has released their first official update to the open-source Vulkan Memory Allocator project in nearly one year.

    Vulkan Memory Allocator is an easy-to-use Vulkan memory allocation library that in the two and a half years since being open-sourced has been picked up for use by multiple games/engineers, Vulkan code samples, and other projects.

  • Chromium's Ozone Wayland Back-End Is Now Considered Beta, Aiming To Ship Next Year

    For years there has been work on a Wayland back-end to Ozone, the Google component for abstracting user-interface elements and input/window handling among other tasks across platforms. It looks like in 2020 the Ozone Wayland support will be in good standing and promoted out of beta.

    We were tipped off to a recent presentation by Igalia's Alexander Dunaev on their work contributing to the Ozone Wayland code. From consulting firm Igalia's perspective, they have been focused on bringing up Ozone Wayland support in the embedded Linux context considering the number of consumer devices now shipping that use Wayland and Chromium or CEF. But all their embedded Linux work for Ozone Wayland also benefits the Linux desktop.

  • Mesa Developers Weigh Renaming Gallium "State Tracker" To "API"

    Gallium3D state tracker terminology has been around a decade now in referring to the portions of the architecture that are ultimately implementing various graphics / compute / video APIs. Marek proposed keeping the Mesa OpenGL state tracker term but in renaming the other state trackers to being "API implementations" as that terminology is technically more accurate for the likes of Clover OpenCL, VA-API, VDPAU, and the other state trackers / APIs implemented.

  • Lima Gets Tiling While Vulkan Turnip Lands SSBO + Compute Shaders

    The Lima Gallium3D driver that supports older Mali 400/450 hardware with an open-source OpenGL driver hasn't been seeing too many improvements these days compared to the likes of the Panfrost Gallium3D driver for the newer Arm Mali Bifrost/Midgard architectures. But hitting Mesa 20.0-devel yesterday was tiling support for Lima. This should improve the performance for this open-source Mali driver and also end up working around the driver's broken mipmapping support for linear textures.

  • Libinput 1.15 Is On Approach With Various Improvements/Fixes For Linux Input Handling

    Peter Hutterer has been preparing libinput 1.15 as the next update to this open-source input handling library used by Linux systems both on X.Org and Wayland.

    Compared to past releases that have seen exciting changes on supporting new input devices like the Dell Totem, scrolling enhancements, and other major additions, there isn't too much of that with libinput 1.15.

Latest on Mesa Graphics

Filed under
Graphics/Benchmarks
  • mesa 19.3.0-rc6
    Hi list,
    
    Available today is mesa 19.3.0-rc6. Things are starting to slow down, and there
    are now only two open issues in the 19.3 milestone, so I'm hopeful that next
    week will bring 19.3.0 final, and not an rc7, but I can always be surprised.
    
    By far radv + aco dominate the list of changes, but there's also changes to anv,
    panfrost, core gallium, fixes for OpenBSD, iris, and nir.
    
    Dylan
    
    Shortlog
    ========
    
    Bas Nieuwenhuizen (2):
          radv: Unify max_descriptor_set_size.
          radv: Fix timeline semaphore refcounting.
    
    Boris Brezillon (2):
          gallium: Fix the ->set_damage_region() implementation
          panfrost: Make sure we reset the damage region of RTs at flush time
    
    Christian Gmeiner (1):
          etnaviv: remove dead code
    
    Daniel Schürmann (2):
          aco: don't split live-ranges of linear VGPRs
          aco: fix a couple of value numbering issues
    
    Dylan Baker (1):
          VERSION: bump version for 19.3-rc6
    
    Jason Ekstrand (1):
          anv: Set up SBE_SWIZ properly for gl_Viewport
    
    Jonathan Gray (2):
          winsys/amdgpu: avoid double simple_mtx_unlock()
          i965: update Makefile.sources for perf changes
    
    Jordan Justen (1):
          iris: Allow max dynamic pool size of 2GB for gen12
    
    Kenneth Graunke (2):
          driconf, glsl: Add a vs_position_always_invariant option
          drirc: Set vs_position_always_invariant for Shadow of Mordor on Intel
    
    Rhys Perry (5):
          aco: propagate p_wqm on an image_sample's coordinate p_create_vector
          aco: fix i2i64
          aco: add v_nop inbetween exec write and VMEM/DS/FLAT
          radv: set writes_memory for global memory stores/atomics
          nir/lower_io_to_vector: don't create arrays when not needed
    
    Samuel Pitoiset (2):
          radv: fix enabling sample shading with SampleID/SamplePosition
          radv/gfx10: fix implementation of exclusive scans
    
    
    git tag: mesa-19.3.0-rc6
    
  • Mesa 19.3 Might Release Next Week But For Now There's RC6 With Several ACO+RADV Fixes

    Mesa 19.3 continues running behind schedule but stands chances for releasing next week if the lingering blocker bugs are closed.

    Mesa 19.3-RC6 was released today as the newest weekly release candidate and it brought with it several random RADV fixes, a number of ACO compiler back-end fixes that also benefit RADV, a few Gallium3D fixes, an Intel Iris Gen12 fix, and a workaround for Shadow of Mordor on Intel graphics.

  • Mesa Devs Discuss Potentially Dropping Non-Gallium Drivers Or Forking Code For Gallium

    Longtime open-source AMD graphics driver developer Marek Olšák has kicked off a discussion over the possibility in the not too distant future of either dropping non-Gallium3D drivers from Mesa (and moving them off to a maintenance branch or the like) or forking some of Mesa's existing code to allow it to be better optimized for Gallium3D use-cases. Due to raised concerns, other possibilities are also being expressed like simply moving ahead with optimizing the Mesa code-base for Gallium3D at a cost of potentially hitting dead code more often with the classic drivers.

    As it stands now, the only relevant non-Gallium3D driver in the Mesa code-base is Intel i965. While that's currently the default Intel driver, for Broadwell "Gen8" graphics and newer they will be transitioning to their new Iris Gallium3D driver by default expected to happen for Mesa 20.0. The i965 driver will still be around for Haswell and older generations to come -- either within mainline Mesa or some maintenance branch. As part of this new Mesa discussion was a hypothetical comment about creating a new Intel Gallium3D driver for Haswell and older, but that's extremely unlikely to happen and was just brought up as a matter of being thorough. There aren't the extra resources available to create an Intel Gallium3D driver for aging Haswell and older hardware plus that it would likely take around a year to develop and even longer before reaching performance parity to i965.

  • Remove classic drivers or fork src/mesa for gallium?
    Hi,
    
    Here are 2 proposals to simplify and better optimize the GL->Gallium
    translation.
    
    1) Move classic drivers to a fork of Mesa, and remove them from master.
    Classic drivers won't share any code with master. glvnd will load them, but
    glvnd is not ready for this yet.
    
    2) Keep classic drivers. Fork src/mesa for Gallium. I think only mesa/main,
    mesa/vbo, mesa/program, and drivers/dri/common need to be forked and
    mesa/state_tracker moved. src/gallium/state-trackers/gl/ can be the target
    location.
    
    Option 2 is more acceptable to people who want to keep classic drivers in
    the tree and it can be done right now.
    
    Opinions?
    
    Thanks,
    Marek
    

NVIDIA GeForce GTX 1650 SUPER Linux Performance

Filed under
Graphics/Benchmarks

For those looking to spend less than $200 USD on a graphics card, the recently launched NVIDIA GeForce GTX 1650 SUPER offers great value starting at $159 USD and working well with the NVIDIA Linux driver for providing decent 1080p Linux gaming performance as well as OpenCL / CUDA support. Here are benchmarks of the GTX 1650 SUPER alongside a total of 18 lower-end/mid-range AMD Radeon and NVIDIA GeForce graphics cards on Ubuntu Linux.

The NVIDIA GeForce GTX 1650 SUPER features 1280 CUDA cores, a reference 1530MHz base clock, 1725MHz boost clock, 4GB of GDDR6 video memory on a 128-bit bus, and other common NVIDIA Turing GPU features sans this being a GTX part and not RTX thus no RT cores.

Read more

Linux Graphics: RISC-V/Think Silicon, ELC Europe and Mesa

Filed under
Graphics/Benchmarks
  • Think Silicon® demonstrates early preview of Industry’s first RISC-V ISA based 3D GPU at the RISC-V Summit

    Think Silicon, recognized for the successful ultra-low power NEMA® GPU-Series for MCU driven SoCs, announced the demonstration of the industry’s first RISC-V ISA based 3D GPU -- the NEOX|V™. Attendees at the RISC-V Summit, in San Jose, California, will have the first opportunity to witness this new GPU innovation designed for the rapid deployment of Computer Graphics, Machine Learning and open GPGPU compute framework applications.

    Offering a myriad of flexible possibilities, NEOX|V ™ IP is designed to be easily configured for applications such as computer graphics, machine learning, vision/video processing and general-purpose compute. The new offering provides a platform for implementation in multiple embedded and external devices across many consumer and industrial vertical markets including Graphics, Compute, and AI for IoT/Edge/Compute.

  • NEOX V Announced By Think Silicon As First RISC-V 3D GPU

    While there has been the Libre RISC-V community-driven effort to create a RISC-V graphics processor that basically amounts to a RISC-V core with vector extensions/improvements and running a Vulkan software implementation (though they are now reportedly eyeing POWER instead of RISC-V), Think Silicon has announced the first actual RISC-V ISA based 3D graphics processor.

  • ELCE Lyon: Everything Great About Upstream Graphics

    At ELC Europe in Lyon I held a nice little presentation about the state of upstream graphics drivers, and how absolutely awesome it all is. Of course with a big focus on SoC and embedded drivers. Slides and the video recording

  • Mesa Adds Option For Changing Intel's OpenGL Driver Default

    While originally Intel planned to transition their OpenGL driver default to the modern "Iris" Gallium3D driver rather than the longstanding "i965" DRI driver for Mesa 19.3, that was pushed back to Mesa 20.0 for introduction in Q1'2020. In aiming to make that revised milestone a reality, a new option has been added to Mesa 20.0 with the Meson build system for being able to indicate the Intel OpenGL driver preference.

    The plan is for Mesa 20.0 to default to their new Gallium3D driver with Broadwell "Gen8" graphics and newer, including Icelake "Gen11". It's with Tiger Lake "Gen12" graphics where there is only support being implemented anyhow on this Gallium3D driver and not the older i965 OpenGL driver. As it stands right now when building Mesa, the i965 driver is used by default and then an environment variable allows overriding the driver to load in order to use Iris Gallium3D.

FreeBSD 12.1 Runs Refreshingly Well With AMD Ryzen Threadripper 3970X - Benchmarks Against Windows + Linux

Filed under
Graphics/Benchmarks

For those of you interested in AMD's new Ryzen Threadripper 3960X/3970X processors with TRX40 motherboards for running FreeBSD, the experience in our initial testing has been surprisingly pleasant. In fact, it works out-of-the-box which one could argue is better than the current Linux support that needs the MCE workaround for booting. Here are some benchmarks of FreeBSD 12.1 on the Threadripper 3970X compared to Linux and Windows for this new HEDT platform.

It was refreshing to see FreeBSD 12.1 booting and running just fine with the Ryzen Threadripper 3970X 32-core/64-thread processor from the ASUS ROG ZENITH II EXTREME motherboard and all core functionality working including the PCIe 4.0 NVMe SSD storage, onboard networking, etc. The system was running with 4 x 16GB DDR4-3600 memory, 1TB Corsair Force MP600 NVMe SSD, and Radeon RX 580 graphics. It was refreshing to see FreeBSD 12.1 running well with this high-end AMD Threadripper system considering Linux even needed a boot workaround.

Read more

Syndicate content