Zhigang Gong on the behalf of the Intel China team released Glamor 0.4 as the latest work on the unique Glamor acceleration architecture
for the Intel X.Org driver.
Zhigang Gong released Glamor 0.4 on Saturday morning via this mailing list message
. It's a nice in-depth release announcement covering the changes, build process, and future plans.
Glamor 0.4 now properly works with DRI2 and GLX_EXT_texture_from_pixmap support, fully supports GLX including AIGLX, optimizations in the fall-back paths, fully support all color formats for the OpenGL ES 2.0 (GLES2) port, many bug-fixes, and a FBO/texture cache pool mechanism to reduce the overhead of texture/FBO creation and destruction. This cache pool mechanism is said to improve the performance by 15~20% or for a PowerVR 545 platform about a 10x performance improvement.
Enabling Intel Glamor acceleration requires rebuilding Mesa with special parameters, building xf86-video-intel DDX with the --enable-glamor
option, and building the actual Glamor source-code. A special xorg.conf
file for Glamor is also needed.
The Glamor source itself can be looked at via FreeDesktop.org CGit
For Glamor 0.5, Intel's plans include full gradient optimizations including linear and radial gradients, large pixmap support, full trapezoid optimization, and fine tuning the FBO caching mechanism.
Those not familiar with the Intel Glamor acceleration architecture can learn more via this Wiki page
. Coming up soon will be Intel Ivy Bridge Linux benchmarks running Git and comparing the stock xf86-video-intel DDX UXA acceleration, SNA acceleration, and Glamor acceleration.