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>
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>
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>
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.
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>
Robert Noland [Tue, 24 Nov 2009 16:31:41 +0000 (10:31 -0600)]
More fixups to allow mesa to build
Robert Noland [Tue, 24 Nov 2009 15:27:29 +0000 (09:27 -0600)]
Finish fixing the build on FreeBSD
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.
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.
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.
Kristian Høgsberg [Mon, 23 Nov 2009 16:41:15 +0000 (11:41 -0500)]
Output summary of enabled features at the end of configure.ac
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/
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.
Kristian Høgsberg [Sat, 21 Nov 2009 00:42:59 +0000 (19:42 -0500)]
Install kernel headers in ${includedir}/drm
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>
Kristian Høgsberg [Fri, 20 Nov 2009 22:09:03 +0000 (17:09 -0500)]
Merge remote branch 'origin/master' into libdrm
Kristian Høgsberg [Fri, 20 Nov 2009 22:08:28 +0000 (17:08 -0500)]
Update README
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>
Kristian Høgsberg [Tue, 17 Nov 2009 16:15:56 +0000 (11:15 -0500)]
Drop stale TODO and unused ChangeLog
Kristian Høgsberg [Tue, 17 Nov 2009 16:14:54 +0000 (11:14 -0500)]
Move libdrm/ up one level
Kristian Høgsberg [Tue, 17 Nov 2009 14:46:56 +0000 (09:46 -0500)]
Drop shared-core, bsd-core, linux-core and scripts subdirs
Kristian Høgsberg [Tue, 17 Nov 2009 14:46:39 +0000 (09:46 -0500)]
Use headers copied from kernel instead of shared-core
Kristian Høgsberg [Tue, 17 Nov 2009 14:39:23 +0000 (09:39 -0500)]
Fix typo in i915 pipe_from_crtc_id ioctl struct name
Kristian Høgsberg [Tue, 17 Nov 2009 14:23:59 +0000 (09:23 -0500)]
Copy headers from kernel v2.6.32-rc6-130-g5b8f0be
Kristian Høgsberg [Tue, 17 Nov 2009 14:23:52 +0000 (09:23 -0500)]
Add makefile rule to copy headers from kernel tree
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>
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>
Julien Cristau [Sat, 31 Oct 2009 01:25:44 +0000 (02:25 +0100)]
Only install libdrm_intel.pc if we build libdrm_intel
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)>
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)>
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>
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>
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>
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>
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.
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
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
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>
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>
Patrice Mandin [Tue, 13 Oct 2009 20:50:46 +0000 (22:50 +0200)]
nouveau: nv30: add render target logbase2 width,height
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.
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>
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>
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>
Eric Anholt [Thu, 8 Oct 2009 23:59:17 +0000 (16:59 -0700)]
Bump to 2.4.15 for release.
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.
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]
Eric Anholt [Tue, 6 Oct 2009 22:25:21 +0000 (15:25 -0700)]
intel: Fix up some stale doxygen comments.
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.
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.
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>
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>
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>
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.
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>
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>
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>
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.
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.
Eric Anholt [Mon, 21 Sep 2009 22:29:58 +0000 (15:29 -0700)]
Bump to 2.4.14 for release.
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.
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.
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>
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>
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>
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>
Eric Anholt [Fri, 28 Aug 2009 22:20:22 +0000 (15:20 -0700)]
Bump to version 2.4.13 for release.
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.
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.
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.
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
Alex Deucher [Mon, 24 Aug 2009 22:15:03 +0000 (18:15 -0400)]
radeon: pull in z pipe changes from kernel
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>
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.
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>
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>
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>
Ben Skeggs [Wed, 12 Aug 2009 04:21:00 +0000 (14:21 +1000)]
nouveau: support for copy-less pushbuf ioctl
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
Dave Airlie [Mon, 17 Aug 2009 11:21:02 +0000 (21:21 +1000)]
radeon: fix bo wait at map time.
Christoph Brill [Sun, 16 Aug 2009 06:26:25 +0000 (08:26 +0200)]
Filter radeon pkgconfig file as do intel and nouveau
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>
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>
Dave Airlie [Sat, 1 Aug 2009 07:19:43 +0000 (17:19 +1000)]
libdrm_radeon: add tiling support
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
Ben Skeggs [Mon, 27 Jul 2009 22:12:21 +0000 (08:12 +1000)]
nouveau: don't overwrite user-specified bo size needlessly
Ben Skeggs [Mon, 27 Jul 2009 21:46:20 +0000 (07:46 +1000)]
nouveau: user buffers need to be mappable
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
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.
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
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
Eric Anholt [Wed, 15 Jul 2009 17:04:44 +0000 (10:04 -0700)]
Bump to version 2.4.12 for release.
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.
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
Ben Skeggs [Thu, 9 Jul 2009 01:41:47 +0000 (11:41 +1000)]
nouveau: fix pin for buffers created with bo_wrap
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.
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>
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>
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.
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