| Age | Commit message (Collapse) | Author |
|---|
|
Move the tracking of the last emitted instructions into the core
batchbuffer routines and take advantage of the shadow batch copy to
avoid extra memory allocations and copies.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
It's faster. Not only is the memcpy more efficiently performed in the
kernel (making up for the system call overhead), but by not using mmap
we remove the greater overhead of tracking the vma of every batch.
And it means we can read back from the batch buffer without incurring
the cost of a uncached read through the GTT.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
|
|
No measurable performance difference on cairo-perf-trace, but
simplifies the code and should have cache benefit in general.
|
|
|
|
The slightly less mechanical change of converting the emit_reloc calls
will follow.
|
|
This saves 6.6KB on the 965 driver, and appears to speed
firefox-talos-gfx up by 1-2%. Unlike many other asserts in the
driver, when we make a mistake that would trigger one of these it
generally shows up all the time for developers, so turning it off for
release seems fine.
|
|
This manages to cut down another 3800 bytes.
|
|
Cuts another 1800 bytes from the driver.
|
|
Shaves 800 bytes off the driver.
|
|
This improves tiled texture performance of OA on my 945 from 25.3fps
to 29.0fps, whereas untiled is 28.2fps, by avoiding stalls for fence
register changes.
|
|
|
|
|
|
This should do all the things that MI_FLUSH did, but it can be pipelined
so that further rendering isn't blocked on the flush completion unless
necessary.
|
|
|
|
This should catch the common programming error where we attempt to
emit a relocation to beyond the end of the target buffer.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
|
|
This avoids issues with dereferencing stale cliprects around intel_draw_buffer
time. Additionally, take advantage of cliprects staying constant for FBOs and
DRI2, and emit cliprects in the batchbuffer instead of having to flush batch
each time they change.
|
|
Makefile.template
|
|
|
|
|
|
This reverts commit 7c81124d7c4a4d1da9f48cbf7e82ab1a3a970a7a.
|
|
This reverts commit 53675e5c05c0598b7ea206d5c27dbcae786a2c03.
Conflicts:
src/mesa/drivers/dri/i965/brw_wm_surface_state.c
|
|
Conflicts:
src/mesa/drivers/dri/common/dri_bufmgr.c
src/mesa/drivers/dri/i965/brw_wm_surface_state.c
|
|
pointed out and debugged by stringfellow on #dri-devel
|
|
Most of these were to ensure that caches got synchronized between 2d (or meta)
rendering and later use of the target as a source, such as for texture
miptree setup. Those are replaced with intel_batchbuffer_emit_mi_flush(),
which just drops an MI_FLUSH. Most of the remainder were to ensure that
REFERENCES_CLIPRECTS batchbuffers got flushed before the lock was dropped.
Those are now replaced by automatically flushing those when dropping the lock.
|
|
Track DRM GEM name changes.
Add driver hooks for bo_subdata and bo_get_subdata so that GEM can use pread
and pwrite.
|
|
Make sure 'used' tracks the right value through the whole function.
Also, use GLint for intel_batchbuffer_space in case we do bad things
in the future.
|
|
The GEM flags are much more descriptive for what we need. Since this makes
bufmgr_fake rather device-specific, move it to the intel common directory.
We've wanted to do device-specific stuff to it before.
|
|
Fencing was used in two places: ensuring that we didn't get too many frames
ahead of ourselves, and glFinish. glFinish will be satisfied by waiting on
buffers like we would do for CPU access on them. The "don't get too far ahead"
is now the responsibility of the execution manager (kernel).
|
|
The fix for pageflipping with cliprects ended up causing a batch flush at
an inopportune time, which is fixed by moving it up.
Additionally, the recovery code for handling batch wraps at bad times is
replaced by just checking for the space up front, and using a no_batch_wrap
assert like on 965 to make sure that we weren't wrong about how much space that
was.
|
|
The previous code would reference freed memory on window moves.
|
|
The previous change gave us only two modes, one which looped over the batch
per cliprect (3d drawing) and one that didn't (state updeast).
However, we really want 4:
- Batch doesn't care about cliprects (state updates)
- Batch needs DRAWING_RECTANGLE looping per cliprect (3d drawing)
- Batch needs to be executed just once (region fills, copies, etc.)
- Batch already includes cliprect handling, and must be flushed by unlock time
(copybuffers, clears).
All callers should now be fixed to use one of these states for any batchbuffer
emits. Thanks to Keith Whitwell for pointing out the failure.
|
|
In particular, batch buffers are no longer flushed when switching from
CLIPRECTS to NO_CLIPRECTS or vice versa, and 965 just uses DRM cliprect
handling for primitives instead of trying to sneak in its own to avoid the
DRM stuff. The disadvantage is that we will re-execute state updates per
cliprect, but the advantage is that we will be able to accumulate larger
batch buffers, which were proving to be a major overhead.
|
|
This is currently believed to work but be a significant performance loss.
Performance recovery should be soon to follow.
The dri_bo_fake_disable_backing_store() call was added to allow backing store
disable like bufmgr_fake.c did, which is a significant performance win (though
it's missing the no-fence-subdata part).
This commit is a squash merge of the 965-ttm branch, which had some history
I wanted to avoid pulling due to noisiness and brokenness at many points
for git-bisecting.
|
|
This could lead to incorrect rendering or even lockups.
|
|
|
|
This reverts commit b2f1aa2389473ed09170713301b042661d70a48e.
Somehow I ended up with my branch's save-this-while-I-work-on-master commit
actually on master.
|
|
|
