Here comes a pretty significant bugfix release for the 2.6 2D series. The goal of this release is to get out the major fixes for GEM and KMS that we think we've pounded on enough to be stable -- certainly more stable than previously. Notable fixes include a significant BO memory usage reduction (which many have suffered from with compositing), textured XV suppor twith KMS, and rotation support with KMS. Some infrequent failure to render/xv with GEM on 965 (dmesg warnings about being unable to bind objects) should also be fixed. But perhaps the exciting thing for most people will be the dynamic front buffer allocation. We nearly slipped this into 2.6.0, but decided that it was just a little too new. Well, turns out it was actually in good shape, and it's time to get it out there. You'll need UXA to do this. Bill Nottingham (1): Quirk MSI IM-945GSE-A LVDS, TV outputs. Eric Anholt (23): Remove logical context setup. Fix compile failure after 45f45c73469f1bd46a1b6fb206f2e9e5e4fd66b3 warning fix. Move 965 video setup to a separate function so we can move it around. Do check_aperture_space and batch_start_atomic for i965 video. Move i965 render state bo setup back to prepare_composite. Use intel_emit_reloc from video to prettify 965 render bind_bo setup. Move filter computation from emit_batch_header to prepare_composite. i965: Pull check_aperture out to a separate function and make it dtrt. Move i965 render transform setup from emit_composite_state to prepare_composite. Move i965 render vb setup to use time, and decouple state emit from it. Use drm_intel_bo_subdata to put render vb data in. Fix build with server 1.4. Remove 965 render wm scratch space, which was just unused. Move 965 render unit state to BOs. Move i965 render kernels to BOs. Move i965 render sampler state to BOs. Don't forget the new state bos in check_aperture. dri2: Use modesetting's master fd instead of opening our own non-master. uxa: Ask for BOs ready for rendering for pixmaps. Don't do AdjustFrame in KMS mode. Fix distcheck from drmmode_display.h deletion. Bump version to 2.6.2 for release. Helge Bahmann (1): Move disable_render_standby to EnterVT instead of startup. Jesse Barnes (2): Fixup bogus VBT modes when detected Support tiled back/depth on 915-class hardware with DRI2. Keith Packard (3): Make i830_allocate_memory take tiling parameters. Resize framebuffer on screen size change (requires UXA and DRI2) Assign rotation memory dri_bo to rotation pixmap. Kristian H?gsberg (12): Implement front buffer resize for KMS. Fix last-minute "cleanup" that broke the patch. Un-revert the I915_SETPARAM_NUM_USED_FENCES commit reverted by accident. Fix front buffer memset() for non-KMS case. KMS: Hook up crtc::gamma_set. Make Xv used a buffer object instead of i830_memory. Dont allocate overlay registers in KMS mode. Fix i915 textured video to work with the i830_memory -> bo change. Access the Xv buffer through the GTT for the non-KMS case. KMS: Hook up rotated shadow buffers. KMS: Fix bug that prevented EDID data from getting propagated. Update kms to work with drmModeModeInfo API update. Ma Ling (1): Don't disable vga centering bit. Vincent Mussard (1): quirk for AOpen MP45 Zhenyu Wang (1): TV quirk for HP Compaq nx6310 git tag: 2.6.2 http://xorg.freedesktop.org/archive/individual/driver/xf86-video-intel-2.6.2.tar.bz2 MD5: c171bb4175cfcea2662079ef20ee2020 xf86-video-intel-2.6.2.tar.bz2 SHA1: 8da19bf6c66fa66d42eedae322ee98397ee58397 xf86-video-intel-2.6.2.tar.bz2 http://xorg.freedesktop.org/archive/individual/driver/xf86-video-intel-2.6.2.tar.gz MD5: 9db24b22c07d0a356388414ae7c3e7f1 xf86-video-intel-2.6.2.tar.gz SHA1: 403da992f9352f4f2ee8ad13b392c47cc897bb91 xf86-video-intel-2.6.2.tar.gz -- Eric Anholt eric at anholt.net eric.anholt at intel.com -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 197 bytes Desc: This is a digitally signed message part Url : http://lists.freedesktop.org/archives/xorg-announce/attachments/20090224/cb7967e5/attachment.pgp
On Wed, 2009-02-25 at 14:10 +0100, Jacek Luczak wrote:> Eric Anholt pisze: > > Here comes a pretty significant bugfix release for the 2.6 2D series. > > The goal of this release is to get out the major fixes for GEM and KMS > > that we think we've pounded on enough to be stable -- certainly more > > stable than previously. Notable fixes include a significant BO memory > > usage reduction (which many have suffered from with compositing), > > textured XV suppor twith KMS, and rotation support with KMS. Some > > infrequent failure to render/xv with GEM on 965 (dmesg warnings about > > being unable to bind objects) should also be fixed. > > > > But perhaps the exciting thing for most people will be the dynamic front > > buffer allocation. We nearly slipped this into 2.6.0, but decided that > > it was just a little too new. Well, turns out it was actually in good > > shape, and it's time to get it out there. You'll need UXA to do this. > > Hi Eric, > > this release is totally unusable while running in UXA. System eats lot of > memory, including swapping. Is this that, reported earlier, ,,memory leak''? As > a result X are really slow and lot of lockups occur (everything freeze for a few > seconds). At the end I can't even switch to text console, but system reacts on > power button and it goes down successfully. With EXA there's no such issue: > total system memory usage around 40%, no lockups. > > If it's not know issue I will try to bisect it down. > > -Jacek > > ------------------ > Details: > 1. X.Org X Server 1.5.99.903 (1.6.0 RC 3) > 2. Linux Kernel 2.6.29-rc6Can you give me exact steps to reproduce this leak? -- Eric Anholt eric at anholt.net eric.anholt at intel.com -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 197 bytes Desc: This is a digitally signed message part Url : http://lists.freedesktop.org/archives/xorg-announce/attachments/20090225/ea825708/attachment.pgp
On Wed, 2009-02-25 at 19:13 +0100, Brice Goglin wrote:> Eric Anholt wrote: > > On Wed, 2009-02-25 at 14:10 +0100, Jacek Luczak wrote: > >> > >> this release is totally unusable while running in UXA. System eats lot of > >> memory, including swapping. Is this that, reported earlier, ,,memory leak''? As > >> a result X are really slow and lot of lockups occur (everything freeze for a few > >> seconds). At the end I can't even switch to text console, but system reacts on > >> power button and it goes down successfully. With EXA there's no such issue: > >> total system memory usage around 40%, no lockups. > >> > >> If it's not know issue I will try to bisect it down. > >> > >> -Jacek > >> > >> ------------------ > >> Details: > >> 1. X.Org X Server 1.5.99.903 (1.6.0 RC 3) > >> 2. Linux Kernel 2.6.29-rc6 > > > > Can you give me exact steps to reproduce this leak? > > Same problem here with libdrm 2.4.5, intel 2.6.2, Xserver 1.6-rc2, > Mesa 7.3 on > Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated > Graphics Controller [8086:27a2] (rev 03) > > Nothing specific needed to reproduce the problem, just use X :) > As soon as I start a big application such as firefox ou thunderbird, > my 1GB RAM is entirely used after a couple seconds and the system > becomes verrrry slow. > > Only downgrade intel to 2.6.1 => 200MB used, no problem.I think the problem here was the DRI2 tiling fix, which was great for the 915-class 3D performance regression but bad for 915-class 2D. I've pushed a fix to master that should help. If it does, I'll try to get a 2.6.3 out soon. commit 5bfd73cd31ba197a62f549cdbad1a1270b571027 Author: Eric Anholt <eric at anholt.net> Date: Fri Feb 27 19:09:49 2009 -0800 Only allocate pixmaps aligned for tiling when requested by DRI2 GetBuffers. This saves massive quantities of memory on pre-965 since the DRI2 tiling enable caused the minimum size of any pixmap to be 1MB. -- Eric Anholt eric at anholt.net eric.anholt at intel.com -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 197 bytes Desc: This is a digitally signed message part Url : http://lists.freedesktop.org/archives/xorg-announce/attachments/20090227/d2a57986/attachment.pgp
Possibly Parallel Threads
- [ANNOUNCE] xf86-video-intel 2.13.902
- Non-Gnome desktop
- "enable dri3 support without glamor" causes gnome-shell regression on nv4x
- [Mesa-dev] [PATCH 2/2] docs: mark ARB_texture_gather as done on nvc0
- "enable dri3 support without glamor" causes gnome-shell regression on nv4x