There is one important conclusion coming out from these experiments: The fact that a rasterizer is normally stateless makes it very inefficient to modify a single element in a scene.
By stateless I mean they do not keep semantic information about the elements being drawn. For example, lets say in one frame I draw a rectangle, and for the next frame I want to draw the same rectangle somewhere else on the canvas. You already have a batch with all the elements of the scene, happily stored in a vertex buffer object on GPU memory, and the rectangle in question is there somewhere. If you could keep the offset where that rectangle is in the batch, you could modify its attributes without having to drop and re-submit the whole batch.
Each file-system was tested with its stock mount options on the Linux 3.17 Git kernel. No kernel modifications were made to this system under test. The new AMD FX-8370 system was used for the Linux benchmarking system in this article. All of our disk / file-system tests are facilitated by the Phoronix Test Suite.
The CTO of the United States of America is a woman, CEO of HP, Yahoo! and many tech giants are women – leading these companies towards future.
However when it comes to core technologies the number of women participants is quite low and disturbing. I am not aware of any leading open source project which was founded by a woman. So Gnome Foundation started a project called Outreach Program for Women (OPW) to increase the participation of women in free software.
One week after Enlightenment's E19 RC3 release, there's more improvements that have landed that will warrant another new release in the very near future.
Posted to the E19 release manager blog is The Septembering -- going over the latest changes so far this September. There's Covery-detected issues fixed, system tray improvements, and more Wayland improvements. The latest Wayland work includes pointer fixes and improved resolution detection.
E19 features the rewritten Wayland compositor that's working out great natively on Wayland without depending upon X11 or Weston for that matter. The E19 release is due out soon with a ton of notable changes.
Wayland 1.6 is finally close to materializing and should be officially released later this month.
Pekka Paalanen of Collabora has been handling Wayland 1.6 release management in the absence of Wayland founder Kristian Høgsberg. There was a Wayland 1.6 Alpha in late August while out today is Wayland 1.6 RC1 along with a release candidate to the Weston compositor.
If all goes well a second release candidate to Wayland will be out next week and then the final Wayland / Weston 1.6 releases the following week. The Wayland/Weston 1.6 RC1 announcement can be found on the Wayland mailing list.
For now, both Mir and Wayland are under massive development, none of them being used on desktop yet. While Mir is testable via the Ubuntu Touch Next Image, Wayland will be added to the default repositories of Fedora, but will not be used as default.
At first, Canonical intended to use Red Hat’s Wayland on their Ubuntu Touch, but it was difficult for them to submit patches and customizations for the mobile device and so, they decided to do the work themselves and created Mir.
Recently, Canonical has joined the Khronos Group to contribute to the creation of Mir/Wayland drivers.
In adding some extra tests besides what was shared in our large Linux review of the new AMD FX CPUs from earlier in the week, that included a fairly big comparison of Intel and AMD CPUs, here's some more Linux test results for just the FX-8370E, FX-8370, and FX-9590 processors.
This latest article are some more results conducted from the ASRock 990FX Killer AM3+ motherboard for all three eight-core CPUs. The Cooler Master Seidon 120XL water cooler was used for keeping the CPUs running well. Ubuntu 14.04 was still the host on the system while using the Linux 3.17 development kernel and Mesa 10.4-devel via the Oibaf PPA. GCC 4.8.2 was the host compiler.
In anticipation of the LLVM 3.5 release that brings a number of new compiler features -- including possible performance improvements from our benchmarking done earlier today -- here's some benchmarks comparing LLVM Clang 3.5 RC3 to a recent SVN snapshot of the GCC 5.0 compiler that's presently under development.
GCC 5.0 is still under heavy stage one development and isn't anticipated for release likely until H1'2015, so there's still a ways to go with the GNU Compiler Collection, but these brief benchmarks today should provide some nice perspective for how it's shaping up against LLVM Clang 3.5. Of course, as GCC 5.0 nears, we'll have plenty more compiler benchmarks over the months ahead.
For the Linux 3.18 kernel Intel has ready some more DRM graphics driver changes beyond the exciting work already sent into drm-next.
Intel already landed in the drm-next tree improved Cherryview support and many other changes while this week another patch series sent in by Intel OTC's Daniel Vetter just landed into drm-next. The changes outlined by Vetter include:
- basic code for execlist, which is the fancy new cmd submission on gen8. Still disabled by default (Ben, Oscar Mateo, Thomas Daniel et al)
- remove the useless usage of console_lock for I915_FBDEV=n (Chris)
- clean up relations between ctx and ppgtt
- clean up ppgtt lifetime handling (Michel Thierry)
- various cursor code improvements from Ville
- execbuffer code cleanups and secure batch fixes (Chris)
- prep work for dev -> dev_priv transition (Chris)
- some of the prep patches for the seqno -> request object transition (Chris)
- various small improvements all over