platform/upstream/libdrm.git
14 years agointel: Clear bo->used_as_reloc_target flag on destroy
Chris Wilson [Mon, 30 Nov 2009 23:07:19 +0000 (23:07 +0000)]
intel: Clear bo->used_as_reloc_target flag on destroy

This allows us to keep the assert added in the previous commit that we do
not modify the tree_reloc_size after inserting the buffer into a relocation
tree, which was being hit here:

  #0  0xb78c2424 in __kernel_vsyscall ()
  #1  0xb74f6401 in raise () from /lib/libc.so.6
  #2  0xb74f7b42 in abort () from /lib/libc.so.6
  #3  0xb74ef5a8 in __assert_fail () from /lib/libc.so.6
  #4  0xb737e78b in drm_intel_bo_gem_set_in_aperture_size (bufmgr_gem=<value optimized out>, bo_gem=0x6) at intel_bufmgr_gem.c:373
  #5  0xb737f519 in drm_intel_gem_bo_set_tiling (bo=0xa1030a0, tiling_mode=0xbff6c85c, stride=0) at intel_bufmgr_gem.c:1386
  #6  0xb737f67f in drm_intel_gem_bo_unreference_final (bo=0xa1030a0, time=<value optimized out>) at intel_bufmgr_gem.c:768
  #7  0xb737f5e3 in drm_intel_gem_bo_unreference_locked_timed (bo=0xa1e50d0, time=<value optimized out>) at intel_bufmgr_gem.c:805
  #8  drm_intel_gem_bo_unreference_final (bo=0xa1e50d0, time=<value optimized out>) at intel_bufmgr_gem.c:756
  #9  0xb737fcbb in drm_intel_gem_bo_unreference (bo=0xa1e50d0) at intel_bufmgr_gem.c:821
  #10 0xb737b4e6 in drm_intel_bo_unreference (bo=0x0) at intel_bufmgr.c:80
  #11 0xb7325625 in intel_batch_flush (scrn=0x9d91f78, flush=1) at i830_batchbuffer.c:200

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
14 years agointel: Apply pessimistic alignment to in-aperture buffer size
Chris Wilson [Mon, 30 Nov 2009 22:14:30 +0000 (22:14 +0000)]
intel: Apply pessimistic alignment to in-aperture buffer size

For the older chipsets, i.e. pre-i965, which have severe alignment
restrictions for tiled buffers we need to pessimistically assume that we
will waste the size of buffer to meet those alignment constraints.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
14 years agointel: Only store a buffer in the cache if it is retained.
Chris Wilson [Mon, 30 Nov 2009 20:02:05 +0000 (20:02 +0000)]
intel: Only store a buffer in the cache if it is retained.

If the kernel immediately frees the backing store for a buffer when
marking it purgeable, then there is not point adding to the cache. Free
it immediately, instead.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
14 years agoCorrectly set DRM_MAX_MINOR for all platforms.
Robert Noland [Wed, 25 Nov 2009 21:09:24 +0000 (15:09 -0600)]
Correctly set DRM_MAX_MINOR for all platforms.

DRM_MAJOR is platform specific, but not used outside of xf86drm.c
that I can find.

14 years agoAdd missing DRM_MAX_MINOR define
Alex Deucher [Tue, 24 Nov 2009 20:03:32 +0000 (15:03 -0500)]
Add missing DRM_MAX_MINOR define

lost in 500f5b524000ed5930301f4303744cb4c0a19b75

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
14 years agoMore fixups to allow mesa to build
Robert Noland [Tue, 24 Nov 2009 16:31:41 +0000 (10:31 -0600)]
More fixups to allow mesa to build

14 years agoFinish fixing the build on FreeBSD
Robert Noland [Tue, 24 Nov 2009 15:27:29 +0000 (09:27 -0600)]
Finish fixing the build on FreeBSD

14 years agoAdd drmGetDeviceNameFromFd function
Kristian Høgsberg [Tue, 24 Nov 2009 01:51:34 +0000 (20:51 -0500)]
Add drmGetDeviceNameFromFd function

Determines the /dev filename of the drm fd argument.

14 years agoFix build on *BSD
Kristian Høgsberg [Mon, 23 Nov 2009 23:25:08 +0000 (18:25 -0500)]
Fix build on *BSD

This adds a minimal #ifdef clause to drm.h that we'll push upstream.
Once that goes in we can share drm.h between linux, libdrm, and
the bsd's.

14 years agoDon't hardcore 'yes', use in configure.ac output
Kristian Høgsberg [Mon, 23 Nov 2009 17:38:40 +0000 (12:38 -0500)]
Don't hardcore 'yes', use  in configure.ac output

Oops, I assumed intel was always enable, but it just defaults to on.

14 years agoOutput summary of enabled features at the end of configure.ac
Kristian Høgsberg [Mon, 23 Nov 2009 16:41:15 +0000 (11:41 -0500)]
Output summary of enabled features at the end of configure.ac

14 years agoDrop duplicated radeon_*.h files in include/drm
Kristian Høgsberg [Mon, 23 Nov 2009 16:25:46 +0000 (11:25 -0500)]
Drop duplicated radeon_*.h files in include/drm

These files are userspace headers and live in radeon/

14 years agoPut mach64_drm.h back in to avoid breaking mesa build
Kristian Høgsberg [Sat, 21 Nov 2009 01:17:26 +0000 (20:17 -0500)]
Put mach64_drm.h back in to avoid breaking mesa build

We may want to drop mach64 from mesa instead, but that's a different
discussion.

14 years agoInstall kernel headers in ${includedir}/drm
Kristian Høgsberg [Sat, 21 Nov 2009 00:42:59 +0000 (19:42 -0500)]
Install kernel headers in ${includedir}/drm

14 years agonouveau: fix DRM headers
Pekka Paalanen [Fri, 20 Nov 2009 23:00:36 +0000 (01:00 +0200)]
nouveau: fix DRM headers

The nouveau_drmif.h is not the Nouveau DRM kernel ABI file, but purely
user space stuff. Remove it, it does not belong in include/drm/.

Copy the right header from Nouveau kernel v2.6.31-rc9-757-gaca551c.

Signed-off-by: Pekka Paalanen <pq@iki.fi>
14 years agoMerge remote branch 'origin/master' into libdrm
Kristian Høgsberg [Fri, 20 Nov 2009 22:09:03 +0000 (17:09 -0500)]
Merge remote branch 'origin/master' into libdrm

14 years agoUpdate README
Kristian Høgsberg [Fri, 20 Nov 2009 22:08:28 +0000 (17:08 -0500)]
Update README

14 years agolibdrm_radeon: add radeon_bo_is_referenced_by_cs function
Maciej Cencora [Wed, 18 Nov 2009 21:07:13 +0000 (16:07 -0500)]
libdrm_radeon: add radeon_bo_is_referenced_by_cs function

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
14 years agoDrop stale TODO and unused ChangeLog
Kristian Høgsberg [Tue, 17 Nov 2009 16:15:56 +0000 (11:15 -0500)]
Drop stale TODO and unused ChangeLog

14 years agoMove libdrm/ up one level
Kristian Høgsberg [Tue, 17 Nov 2009 16:14:54 +0000 (11:14 -0500)]
Move libdrm/ up one level

14 years agoDrop shared-core, bsd-core, linux-core and scripts subdirs
Kristian Høgsberg [Tue, 17 Nov 2009 14:46:56 +0000 (09:46 -0500)]
Drop shared-core, bsd-core, linux-core and scripts subdirs

14 years agoUse headers copied from kernel instead of shared-core
Kristian Høgsberg [Tue, 17 Nov 2009 14:46:39 +0000 (09:46 -0500)]
Use headers copied from kernel instead of shared-core

14 years agoFix typo in i915 pipe_from_crtc_id ioctl struct name
Kristian Høgsberg [Tue, 17 Nov 2009 14:39:23 +0000 (09:39 -0500)]
Fix typo in i915 pipe_from_crtc_id ioctl struct name

14 years agoCopy headers from kernel v2.6.32-rc6-130-g5b8f0be
Kristian Høgsberg [Tue, 17 Nov 2009 14:23:59 +0000 (09:23 -0500)]
Copy headers from kernel v2.6.32-rc6-130-g5b8f0be

14 years agoAdd makefile rule to copy headers from kernel tree
Kristian Høgsberg [Tue, 17 Nov 2009 14:23:52 +0000 (09:23 -0500)]
Add makefile rule to copy headers from kernel tree

14 years agonouveau: Update some nouveau_class.h definitions from renouveau.xml.
Francisco Jerez [Sun, 15 Nov 2009 03:10:02 +0000 (04:10 +0100)]
nouveau: Update some nouveau_class.h definitions from renouveau.xml.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Pekka Paalanen <pq@iki.fi>
14 years agointel: Export madvise
Chris Wilson [Wed, 11 Nov 2009 13:04:38 +0000 (13:04 +0000)]
intel: Export madvise

Wrap the madvise ioctl for use in APPLE_object_purgeable.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoOnly install libdrm_intel.pc if we build libdrm_intel
Julien Cristau [Sat, 31 Oct 2009 01:25:44 +0000 (02:25 +0100)]
Only install libdrm_intel.pc if we build libdrm_intel

15 years agonouveau: turn another assertion into an error
Ben Skeggs [Wed, 4 Nov 2009 07:31:18 +0000 (17:31 +1000)]
nouveau: turn another assertion into an error

Signed-off-by: Ben Skeggs <skeggsb@beleth.(none)>
15 years agonouveau: modify api slightly to allow caller to handle reloc failures
Ben Skeggs [Wed, 4 Nov 2009 05:23:53 +0000 (15:23 +1000)]
nouveau: modify api slightly to allow caller to handle reloc failures

Signed-off-by: Ben Skeggs <skeggsb@beleth.(none)>
15 years agonouveau: add reloc refcnt to pending bo list
Ben Skeggs [Tue, 27 Oct 2009 04:32:40 +0000 (14:32 +1000)]
nouveau: add reloc refcnt to pending bo list

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
15 years agonouveau: drop rendering on the floor rather than asserting if flush fails
Ben Skeggs [Tue, 3 Nov 2009 03:14:32 +0000 (13:14 +1000)]
nouveau: drop rendering on the floor rather than asserting if flush fails

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
15 years agonouveau: update view of available aperture space after each flush
Ben Skeggs [Mon, 2 Nov 2009 08:49:40 +0000 (18:49 +1000)]
nouveau: update view of available aperture space after each flush

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
15 years agoradeon: fix allocation
Mathias Fröhlich [Tue, 3 Nov 2009 16:41:26 +0000 (11:41 -0500)]
radeon: fix allocation

The old code increments the command stream size by another kbyte, but does
not make sure that the requested packet size fits into the stream. The patch
ensures that the whole next packet fits there and rounds the allocated size to
a power of two.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
15 years agoAdd support for vblank events
Kristian Høgsberg [Fri, 11 Sep 2009 17:27:35 +0000 (13:27 -0400)]
Add support for vblank events

The kernel will now write data to the DRM fd for various event types if
requested.  Currently, the only supported event is a vblank event: it contains
the vblank count for the event as well as a timestamp from when the event
ocurred.  Since the DRM fd is now pollable, it's easy to integrate into
existing event loops.

15 years agointel: Only call clock_gettime once per unreference_final.
Eric Anholt [Tue, 20 Oct 2009 21:19:38 +0000 (14:19 -0700)]
intel: Only call clock_gettime once per unreference_final.

Notably when freeing a batchbuffer, we often end up freeing many of the
buffers it points at as well.  Avoiding repeated calls brings us a 9% CPU
win for cairo-gl.

[ # ]  backend                         test   min(s) median(s) stddev. count
before:
[  0]       gl            firefox-talos-gfx   58.941   58.966   0.75%    3/3
after:
[  0]       gl            firefox-talos-gfx   54.186   54.195   0.49%    3/3

15 years agointel: Improve bo_references performance by skipping the tree walk.
Eric Anholt [Tue, 20 Oct 2009 20:20:55 +0000 (13:20 -0700)]
intel: Improve bo_references performance by skipping the tree walk.

If the target we're asking about hasn't ever been used as a relocation
target, then it obviously hasn't been used as a target by the batch's reloc
tree.  This is the common case for good GL programming where you only map
fresh buffers, and gives us a 5% win in cairo-gl.

[ # ]  backend                         test   min(s) median(s) stddev. count
before:
[  0]       gl            firefox-talos-gfx   64.680   64.756   0.06%    3/3
after:
[  0]       gl            firefox-talos-gfx   60.816   60.970   0.29%    3/3

15 years agonouveau: retry if pushbuf ioctl interrupted by signal
Ben Skeggs [Tue, 20 Oct 2009 02:57:46 +0000 (12:57 +1000)]
nouveau: retry if pushbuf ioctl interrupted by signal

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
15 years agonouveau: function to return status of bo being referenced in pushbuf
Ben Skeggs [Sun, 18 Oct 2009 23:04:37 +0000 (09:04 +1000)]
nouveau: function to return status of bo being referenced in pushbuf

Will be used to implement gallium's is_{texture,buffer}_referenced()
callbacks properly.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
15 years agonouveau: nv30: add render target logbase2 width,height
Patrice Mandin [Tue, 13 Oct 2009 20:50:46 +0000 (22:50 +0200)]
nouveau: nv30: add render target logbase2 width,height

15 years agointel: Add the defines for the kernel overlay support landing in 2.6.33.
Eric Anholt [Tue, 13 Oct 2009 18:35:05 +0000 (11:35 -0700)]
intel: Add the defines for the kernel overlay support landing in 2.6.33.

15 years agointel: Fallback to atomic-ops.h [libatomic-ops-dev]
Chris Wilson [Tue, 13 Oct 2009 14:13:00 +0000 (15:13 +0100)]
intel: Fallback to atomic-ops.h [libatomic-ops-dev]

Use the external implementation for atomic operations across a wide
range of architectures.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
15 years agoconfigure: Typo in error message.
Chris Wilson [Tue, 13 Oct 2009 13:25:54 +0000 (14:25 +0100)]
configure: Typo in error message.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
15 years agoconfigure: Conditionally build libdrm_intel
Chris Wilson [Mon, 12 Oct 2009 22:58:47 +0000 (23:58 +0100)]
configure: Conditionally build libdrm_intel

Only build libdrm_intel automatically if we have support for atomic
operations. To force configure to build drm pass --enable-intel, which
will cause the configure to error if no support is found. Or pass
--disable-intel to explicitly prevent libdrm_intel from being built.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
15 years agoBump to 2.4.15 for release. 2.4.15
Eric Anholt [Thu, 8 Oct 2009 23:59:17 +0000 (16:59 -0700)]
Bump to 2.4.15 for release.

15 years agointel: Remove the asserts about the ignored alignment parameter.
Eric Anholt [Thu, 8 Oct 2009 22:39:27 +0000 (15:39 -0700)]
intel: Remove the asserts about the ignored alignment parameter.

I slipped it in with the alloc_tiled changes, since we were explicitly
throwing the parameter away.  It caught some bogus released code, which
we've now fixed, so remove the asserts to keep old drivers working.

15 years agointel: Add a bo_alloc function for tiled BOs.
Jesse Barnes [Tue, 6 Oct 2009 21:34:06 +0000 (14:34 -0700)]
intel: Add a bo_alloc function for tiled BOs.

This simplifies driver code in handling object allocation, and also gives us
an opportunity to possibly cache tiled buffers if it turns out to be a win.

[anholt: This is chopped out of the execbuf2 patch, as it seems to be useful
separately and cleans up the execbuf2 changes to be more obvious]

15 years agointel: Fix up some stale doxygen comments.
Eric Anholt [Tue, 6 Oct 2009 22:25:21 +0000 (15:25 -0700)]
intel: Fix up some stale doxygen comments.

15 years agointel: Reformat to the kernel coding style. Welcome to the 8-space future.
Eric Anholt [Tue, 6 Oct 2009 19:40:42 +0000 (12:40 -0700)]
intel: Reformat to the kernel coding style.  Welcome to the 8-space future.

This is done with:
Lindent *.[ch]
perl -pi -e 's|drm_intel_bo \* |drm_intel_bo *|g' *.[ch]
perl -pi -e 's|drm_intel_bufmgr \* |drm_intel_bufmgr *|g' *.[ch]
perl -pi -e 's|drm_intel_bo_gem \* |drm_intel_bo_gem *|g' *.[ch]
perl -pi -e 's|drm_intel_bufmgr_gem \* |drm_intel_bufmgr_gem *|g' *.[ch]
perl -pi -e 's|_fake \* |_fake *|g' *.[ch]
hand-editing to whack indented comments into line and other touchups.

15 years agointel: Don't allocate more relocation entries than the BO could support.
Eric Anholt [Mon, 5 Oct 2009 23:35:32 +0000 (16:35 -0700)]
intel: Don't allocate more relocation entries than the BO could support.

This saves 32k of relocation entry storage for many 965 state buffers.  No
noticeable impact on performance for cairo-gl firefox.

15 years agotests: Disable intel-specific tests with --disable-intel
Chris Wilson [Sat, 3 Oct 2009 22:17:57 +0000 (23:17 +0100)]
tests: Disable intel-specific tests with --disable-intel

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
15 years agointel: Add a configure option to *disable* building libdrm-intel
Chris Wilson [Sat, 3 Oct 2009 21:56:04 +0000 (22:56 +0100)]
intel: Add a configure option to *disable* building libdrm-intel

In conjunction with the atomic operation patch, it may be more
convenient for some people to disable building libdrm-intel and its
dependencies upon the atomic intrinsics then it is for them to use a
supported compiler.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
15 years agointel: report errno
Chris Wilson [Sun, 20 Sep 2009 02:11:59 +0000 (03:11 +0100)]
intel: report errno

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
15 years agolibdrm_radeon: Update RADEON_TILING_* flags to what's in current kernels.
Michel Dänzer [Sat, 3 Oct 2009 15:37:07 +0000 (17:37 +0200)]
libdrm_radeon: Update RADEON_TILING_* flags to what's in current kernels.

15 years agolibdrm_radeon: Zero-initialize structures to silence valgrind warnings
Nicolai Hähnle [Sat, 3 Oct 2009 11:43:42 +0000 (13:43 +0200)]
libdrm_radeon: Zero-initialize structures to silence valgrind warnings

Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
15 years agointel: Use atomic refcounters
Chris Wilson [Fri, 2 Oct 2009 03:39:22 +0000 (04:39 +0100)]
intel: Use atomic refcounters

As the target architecture for Intel GPUs is the x86, we can presume to
have reasonable compiler support for Intel atomic intrinsics, i.e. gcc,
and so use those in preference to pulling in a complicated mess of
fragile assembly.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[anholt: hand-resolved against my previous commit.  This brings cairo-gl
firefox-talos-gfx time from 65 seconds back down to 62 seconds.]
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agointel: Mark cached bo as purgeable
Chris Wilson [Fri, 2 Oct 2009 03:31:34 +0000 (04:31 +0100)]
intel: Mark cached bo as purgeable

Set the DONTNEED flag on cached buffers so that the kernel is free to
discard those when under memory pressure.

[anholt: This takes firefox-talos-gfx time from ~62 seconds to ~65 seconds
on my GM965, but it seems like a hit worth taking for the improved
functionality from saving memory]

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agointel: Don't free the reloc list when putting a freed BO in the cache.
Eric Anholt [Fri, 2 Oct 2009 18:11:31 +0000 (11:11 -0700)]
intel: Don't free the reloc list when putting a freed BO in the cache.

This takes firefox-talos-gfx from 74 seconds to 70 seconds on my GM965.

15 years agointel: Add a new function to check if a BO's reloc tree references some BO.
Eric Anholt [Fri, 2 Oct 2009 02:09:26 +0000 (19:09 -0700)]
intel: Add a new function to check if a BO's reloc tree references some BO.

There are a bunch of places in GL where if we can't do this we have to
flush the batchbuffer, and the cost of lookups here is outweighed by flush
savings.

15 years agoBump to 2.4.14 for release. libdrm-2.4.14
Eric Anholt [Mon, 21 Sep 2009 22:29:58 +0000 (15:29 -0700)]
Bump to 2.4.14 for release.

15 years agointel: Remove the max_entries stuff that complicated bo reuse.
Eric Anholt [Mon, 7 Sep 2009 06:17:14 +0000 (23:17 -0700)]
intel: Remove the max_entries stuff that complicated bo reuse.

I thought I was going to do all sorts of crazy experiments with it.  I never
did, and it turned out the free-after-a-few-seconds plan is working out fine.

15 years agointel: Remove the old swrast flag for reducing cache flushing.
Eric Anholt [Mon, 7 Sep 2009 06:02:21 +0000 (23:02 -0700)]
intel: Remove the old swrast flag for reducing cache flushing.

It hasn't been doing anything effective since
52e5d24fae4af6f2f4a5304a516c8c5ab347a11b, and we pretty much don't bo_map
pinned buffers any more anyway.

15 years agodrmModeFreeConnector: free encoders and properties
Keith Packard [Fri, 18 Sep 2009 00:28:08 +0000 (17:28 -0700)]
drmModeFreeConnector: free encoders and properties

These were leaking.

Signed-off-by: Keith Packard <keithp@keithp.com>
15 years agoradeon: fix 32/64 bit issue with sign extension
Dave Airlie [Mon, 14 Sep 2009 21:29:02 +0000 (07:29 +1000)]
radeon: fix 32/64 bit issue with sign extension

Not sure what intptr_t was up to here.

Reported and tested by: Kevin DeKorte
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agolibdrm_intel: include B43 chipset check
Zhenyu Wang [Mon, 7 Sep 2009 08:17:04 +0000 (16:17 +0800)]
libdrm_intel: include B43 chipset check

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
15 years agolibdrm_intel: add new pci ids
Zhenyu Wang [Fri, 4 Sep 2009 01:24:23 +0000 (09:24 +0800)]
libdrm_intel: add new pci ids

New ids for G41, Clarkdale and Arrandale.
Make sure we don't need to count fence also on new chips.

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
15 years agoBump to version 2.4.13 for release. libdrm-2.4.13
Eric Anholt [Fri, 28 Aug 2009 22:20:22 +0000 (15:20 -0700)]
Bump to version 2.4.13 for release.

15 years agoAdd drm_intel_bo_busy to query whether mapping a BO would block.
Eric Anholt [Fri, 28 Aug 2009 01:32:07 +0000 (18:32 -0700)]
Add drm_intel_bo_busy to query whether mapping a BO would block.

15 years agoRevert "libdrm_radeon: Always wait for BO idle in bo_map."
Michel Dänzer [Fri, 28 Aug 2009 09:42:07 +0000 (11:42 +0200)]
Revert "libdrm_radeon: Always wait for BO idle in bo_map."

This reverts commit 0a732983f059c353b267b6bf877e1f0eea4e033f.

Paul Nieminen and Dave Airlie pointed out on IRC that this shouldn't be
necessary. I was seeing visual corruption in X before I made this change, but
I can't reproduce that anymore so it was probably an unrelated issue.

15 years agolibdrm_radeon: Always wait for BO idle in bo_map.
Michel Dänzer [Thu, 27 Aug 2009 06:36:58 +0000 (08:36 +0200)]
libdrm_radeon: Always wait for BO idle in bo_map.

This allows users to eliminate explicit bo_wait calls before bo_map calls.

15 years agoKill last remnant of replacefb ioctl
Jakob Bornecrantz [Fri, 21 Aug 2009 12:06:51 +0000 (14:06 +0200)]
Kill last remnant of replacefb ioctl

Kenrels doesn't expose this ioctl

15 years agoradeon: pull in z pipe changes from kernel
Alex Deucher [Mon, 24 Aug 2009 22:15:03 +0000 (18:15 -0400)]
radeon: pull in z pipe changes from kernel

15 years agoradeon: add support for busy/domain check interface.
Pauli Nieminen [Sat, 22 Aug 2009 03:16:18 +0000 (13:16 +1000)]
radeon: add support for busy/domain check interface.

airlied: modified the interface to drop busy return value, just return
it normally, also fixed int->uint32_t for domain

Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
15 years agonouveau: fix a thinko in copyless pushbuf ioctl
Ben Skeggs [Wed, 19 Aug 2009 05:55:05 +0000 (15:55 +1000)]
nouveau: fix a thinko in copyless pushbuf ioctl

No idea why G80 doesn't hit this, but, this fixes at least one NV40 card.

15 years agolibdrm_radeon: Optimize copy of table to cs buffer with specialized call.
Pauli Nieminen [Tue, 18 Aug 2009 15:51:38 +0000 (18:51 +0300)]
libdrm_radeon: Optimize copy of table to cs buffer with specialized call.

Using this call in OUT_BATCH_TABLE reduces radeonEmitState cpu usage from
9% to 5% and emit_vpu goes from 7% to 1.5%. I did use calgrind to profile
gears for cpu hotspots with r500 card.

Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
15 years agolibdrm_radeon: Fix loops so that compiler can optimize them.
Pauli Nieminen [Tue, 18 Aug 2009 15:51:37 +0000 (18:51 +0300)]
libdrm_radeon: Fix loops so that compiler can optimize them.

GCC did war about optimization not possible because possible forever loop.

Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
15 years agolibdrm/radeon: Update head of linked list not to point freed memory.
Pauli Nieminen [Fri, 7 Aug 2009 17:03:26 +0000 (20:03 +0300)]
libdrm/radeon: Update head of linked list not to point freed memory.

Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
15 years agonouveau: support for copy-less pushbuf ioctl
Ben Skeggs [Wed, 12 Aug 2009 04:21:00 +0000 (14:21 +1000)]
nouveau: support for copy-less pushbuf ioctl

15 years agonouveau: for the moment, assert if we exceed some reloc limits
Ben Skeggs [Tue, 18 Aug 2009 04:25:50 +0000 (14:25 +1000)]
nouveau: for the moment, assert if we exceed some reloc limits

Nasty, but nicer than silently not writing into the pushbuf

15 years agoradeon: fix bo wait at map time.
Dave Airlie [Mon, 17 Aug 2009 11:21:02 +0000 (21:21 +1000)]
radeon: fix bo wait at map time.

15 years agoFilter radeon pkgconfig file as do intel and nouveau
Christoph Brill [Sun, 16 Aug 2009 06:26:25 +0000 (08:26 +0200)]
Filter radeon pkgconfig file as do intel and nouveau

15 years agoradeon: fix GTT writing space check
Dave Airlie [Sat, 15 Aug 2009 11:32:35 +0000 (21:32 +1000)]
radeon: fix GTT writing space check

Noticed by vehemens on irc.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agolink libdrm_intel with -lrt for new use of clock_gettime().
Anssi Hannula [Fri, 24 Jul 2009 18:58:43 +0000 (21:58 +0300)]
link libdrm_intel with -lrt for new use of clock_gettime().

Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agolibdrm_radeon: add tiling support
Dave Airlie [Sat, 1 Aug 2009 07:19:43 +0000 (17:19 +1000)]
libdrm_radeon: add tiling support

15 years agoradeon: don't map/unmap explicitly.
Dave Airlie [Sat, 1 Aug 2009 06:47:24 +0000 (16:47 +1000)]
radeon: don't map/unmap explicitly.

This caches the mapping and just use mapping as a sync point

15 years agonouveau: don't overwrite user-specified bo size needlessly
Ben Skeggs [Mon, 27 Jul 2009 22:12:21 +0000 (08:12 +1000)]
nouveau: don't overwrite user-specified bo size needlessly

15 years agonouveau: user buffers need to be mappable
Ben Skeggs [Mon, 27 Jul 2009 21:46:20 +0000 (07:46 +1000)]
nouveau: user buffers need to be mappable

15 years agonouveau: drm api 0.0.15, update object header, remove fake bo support
Ben Skeggs [Sun, 26 Jul 2009 21:23:09 +0000 (07:23 +1000)]
nouveau: drm api 0.0.15, update object header, remove fake bo support

15 years agolibdrm/nouveau: unmap before ufree means nvbo->sysmem != NULL, which inbalances cpu_p...
Maarten Maathuis [Mon, 27 Jul 2009 17:24:13 +0000 (19:24 +0200)]
libdrm/nouveau: unmap before ufree means nvbo->sysmem != NULL, which inbalances cpu_prep/cpu_finish

- The bo was mapped with sysmem == NULL, so this means cpu prep is called.
- The bo was unmapped with sysmem != NULL, so this means cpu finish is not called.
- This can lead to a non-zero "cpu writers" count in ttm_bo.

15 years agoAdd RADEON_INIT_R600_CP too
Alex Deucher [Wed, 15 Jul 2009 18:51:38 +0000 (14:51 -0400)]
Add RADEON_INIT_R600_CP too

this rounds out the r6xx/r7xx stuff in radeon_drm.h

15 years agoAdd R600_SCRATCH_REG_OFFSET to radeon_drm.h
Alex Deucher [Wed, 15 Jul 2009 18:40:41 +0000 (14:40 -0400)]
Add R600_SCRATCH_REG_OFFSET to radeon_drm.h

This lets us libdrm from master for r6xx/r7xx 3d

15 years agoBump to version 2.4.12 for release. libdrm-2.4.12
Eric Anholt [Wed, 15 Jul 2009 17:04:44 +0000 (10:04 -0700)]
Bump to version 2.4.12 for release.

15 years agointel: Free buffers in the BO cache that haven't been reused in a while.
Eric Anholt [Fri, 10 Jul 2009 00:49:46 +0000 (17:49 -0700)]
intel: Free buffers in the BO cache that haven't been reused in a while.

The goal of the BO cache is to keep buffers on hand for fast continuous use,
as in every frame of a game or every batchbuffer of the X Server.  Keeping
older buffers on hand not only doesn't serve this purpose, it may hurt
performance by resulting in disk cache getting kicked out, or even driving
the system to swap.

Bug #20766.

15 years agoconfigure: move AC_USE_SYSTEM_EXTENSIONS up higher
Dave Airlie [Thu, 9 Jul 2009 16:38:13 +0000 (02:38 +1000)]
configure: move AC_USE_SYSTEM_EXTENSIONS up higher

this was giving me a warning on automake 1.11

15 years agonouveau: fix pin for buffers created with bo_wrap
Ben Skeggs [Thu, 9 Jul 2009 01:41:47 +0000 (11:41 +1000)]
nouveau: fix pin for buffers created with bo_wrap

15 years agos/AC_USE_SYSTEM_MACROS/AC_USE_SYSTEM_EXTENSIONS/
Ian Romanick [Mon, 6 Jul 2009 22:36:20 +0000 (15:36 -0700)]
s/AC_USE_SYSTEM_MACROS/AC_USE_SYSTEM_EXTENSIONS/

Not sure what I was thinking.

15 years agolibdrm: Make chown check for return value
Pauli Nieminen [Mon, 6 Jul 2009 20:37:20 +0000 (23:37 +0300)]
libdrm: Make chown check for return value

If call was interrupted by signal we have to make call again.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
15 years agolibdrm: Set _XOPEN_SOURCE and _GNU_SOURCE
Ian Romanick [Mon, 6 Jul 2009 20:23:46 +0000 (13:23 -0700)]
libdrm: Set _XOPEN_SOURCE and _GNU_SOURCE

Several POSIX extensions are used in the libdrm code (e.g., mknod and ffs).
Set _XOPEN_SOURCE and _GNU_SOURCE to something reasonable to ensure that
prototypes for these functions are available.  This is done in configure.ac
using AC_USE_SYSTEM_MACROS.  This requires autoconf 2.60 or later.  Eventually
the code should check for the existance of these defines and do something
reasonable if they are not available.

Inspired by a patch by Pauli Nieminen and suggestions from Julien Cristau.
Thanks.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
15 years agointel: Fix up math errors when allocating very large BOs.
Eric Anholt [Mon, 6 Jul 2009 18:55:28 +0000 (11:55 -0700)]
intel: Fix up math errors when allocating very large BOs.

The logbase2 would overflow and wrap the size around to 0, making the code
allocate a 4kb object instead.  By simplifying the code to just walk the
14-entry bucket array comparing sizes instead of indexing on
ffs(1 << logbase2(size)), we avoid silly math errors and have code of
approximately the same speed.

Many thanks to Simon Farnsworth for debugging and providing a working patch.
Bug #27365.

15 years agointel: fix compiler warning about returning a void value.
Edward O'Callaghan [Thu, 2 Jul 2009 18:03:48 +0000 (11:03 -0700)]
intel: fix compiler warning about returning a void value.

bug #21999