The 2014Q2 highlights are: Broadwell support, Baytrail improvements and many improvements and optimizations on power saving features.
The support for graphics and media acceleration is now fully functional and was validated on 5th Generation Intel® Core™ processors with Intel® HD Graphics (Broadwell).
There are fixes and new features on all components of the stack as detailed below.
Highlighted new features or fixed bugs
Baytrail improvements and fixes.
5.4 GHz DisplayPort support (needed for 4k Displays).
Large cursor support (useful for high-dpi screens).
Improvements on fastboot support.
Fine-grained display power domain and lot of runtime power management infrastructure work.
Manage display power wells.
Runtime D3 support for Haswell.
GT power well force-wake improvements.
Watermark computation fixes.
Framebuffer Compression (FBC) fixes.
UMS support deprecated.
Kernel option to disable the legacy fbdev support.
Complete and detailed list can be found at Maintainer's (Daniel Vetter) announcements:
Neat drm/i915 stuff for 3.14 and Neat drm/i915 stuff for 3.15.
Allow Xserver work without privileges under the supervision of systemd/logind by adding:
Support tu server fds: where driver receives which fd to use.
Backlight helper to continue providing a RandR backlight property.
Many bug fixes including:
Flush when changing blend modes on Ironlake, or else single glyphs are sometimes rendered incorrectly.
Fix pixmap offsets for pixman fallbacks onto Composite redirected.
Fix blending onto 8-bit destinations.
Handle failure to create DRI bo more carefully.
Actually turn off displays with DPMS off for UXA.
Complete and detailed list can be found at Maintainer's (Chris Wilson) announcements.
Many bug fixes:
DOTA lighting effects on Haswell.
Batch overwritten with garbage on Ivybridge and Haswell.
Piglit spec/NV_conditional_render_blitframebuffer fails.
glClear on a multisample texture doesn't work.
GL_LINE_STIPPLE displays incorrect colors.
GLX_INTEL_swap_event crashes driver when swapping window buffers.
Complete and detailed list can be found at all Mesa's 10.2 release notes.
Media – Libva/Intel-vaapi-driver
API: fix VA_ENC_PACKED_HEADER_RAW_DATA value.
API: add STE filter to the VPP pipeline.
Update doxygen tags in libva.
A workaround for https://bugs.freedesktop.org/show_bug.cgi?id=76755 in case user still uses older version xorg-server.
Add support for STE on Broadwell.
Add support for YV16.
Add support for user specified tiling and stride.
Fix VP8 decoding on Broadwell.
Fix the wrong alpha when convert NV12 into RGBA.
Export JPEG format by vaDeriveImage().
Add support for MADI on SNB.
H.264: fix the support for grayscale format (Y800).
Fix vaGetConfigAttributes() to validate the profile/entrypoint pair.
Fix vaCreateConfig() to not override user chroma format.
Fix the scaling issue on IVB/HSW/BDW.
Quality improvement for H.264 encoding on BDW.
The main issue on this stack release is related to Broadwell power savings features. RC6, Package C States and FBC might not work properly.
On BDW, RC6 and PC7-PC10 are not ready. Turbo functional and FBC power consumption testing fail on BDW.
bug 61485 libva [SNB/IVB/HSW] SSIM value is low with att_mismatch_att.bits (MPEG2) medium nor
bug 61826 libva [SNB/IVB/HSW] Some VC1 files failed by low ssim value medium nor
bug 62849 libva [HSW] I/IPB MPEG2 encoding: the gap is more than 3DB compare with windows data medium nor
bug 69252 libva [IVB/HSW/BYT]Low Y/U/V PSNR when NV12 to RGBX medium nor
bug 69253 libva [HSW/BDW]ProcAmp failed with blurred screen medium nor
bug 70522 libva Mplayer interrupted while playing two mpeg2 files medium nor
bug 75479 libva [BDW] AVC_cqp_encoding had lower pass rate comparing with IVB medium maj
bug 730911 gstreamer-vaapi gstreamer-vaapi maintainer(s) random fail when doing transcoding test(MPEG2->H264)
bug 76247 DRI [BDW]igt/pm_rps/min-max-config-loaded fails medium nor
bug 76877 DRI [BYT]Sysfs RC6 residency counter is inaccurate medium nor
bug 78687 DRI [All]Subcases dpms-lpsp and dpms-non-lpsp of igt/pm_pc8 fail medium nor
bug 78948 DRI [ HSW/BYT/BDW]igt/pm_rpm/system-suspend fails medium nor
bug 80078 DRI [BDW]GFX device consume the same Watt as Idle when it's active medium nor
This release was validated on the following hardware, and part of the test results are published here.
5th Generation Intel® Core™ processors with Intel® HD Graphics 6000
5th Generation Intel® Core™ processors with Intel® HD Graphics 5500
4th Generation Intel® Core™ processors with Intel® Iris™ Pro Graphics 5200
4th Generation Intel® Core™ processors with Intel® HD Graphics 5000
4th Generation Intel® Core™ processors with Intel® HD Graphics 4600
3rd Generation Intel® Core™ processors with Intel® HD Graphics 4000
3rd Generation Intel® Core™ processors with Intel® HD Graphics 2500
2nd Generation Intel® Core™ processors with Intel® HD Graphics 3000
2nd Generation Intel® Core™ processors with Intel® HD Graphics 2000
Intel® Core™ Processors with Intel® HD Graphics
Intel® Celeron® N2820 processor with Intel® HD Graphics (Bay Trail)
Intel® Atom™ N450 with Intel® Graphics Media Accelerator 3150 (Pine Trail)
Mobile Intel® GM45 Express Chipset