fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-21 18:00:13 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1
138393 commits 64 branches 914 tags 1 GiB
Commit graph

125 commits

Author SHA1 Message Date
Kenneth Graunke
572fad1e84 iris: Don't check other batches for our batch BO 
This is an awkward corner case.  We create batches in order, each of
which creates and pins a BO.  The other batches may not be set up yet,
so it may not be safe to ask whether they reference a BO.

Just avoid this for now.  We could avoid it for other context-local BOs
too, but we currently don't have a flag for that (and I'm not certain
whether it's worth it).
 2019-02-21 10:26:11 -08:00
Kenneth Graunke
b92ca4d0f6 iris: don't print the pointer in INTEL_DEBUG=submit 
lots of noise in diff, hope was it would be useful for gdb, but the
the GEM handle is good enough
 2019-02-21 10:26:11 -08:00
Kenneth Graunke
f1a7392be1 iris: Put batches in an array 
We keep re-making this array all over the place
 2019-02-21 10:26:10 -08:00
Kenneth Graunke
db199d9d07 iris: Add wait fences to properly sync between render/compute 
When flushing a batch due to a data dependency, we need to not only
kick off the other batch's work, but stall our execution until it
completes.  Just wait on last_syncpt after flushing it.
 2019-02-21 10:26:10 -08:00
Kenneth Graunke
d69bc4ac12 iris: Hang on to the last batch's sync-point, so we can wait on it 2019-02-21 10:26:10 -08:00
Chris Wilson
fae74234d9 iris: Tag each submitted batch with a syncobj 
(adjusted by Ken to make the signalling sync object immediately on
batch reset, rather than batch finish time.  this will work better
with deferred flushes...)
 2019-02-21 10:26:10 -08:00
Kenneth Graunke
3e332af611 iris: Drop vestiges of throttling code 2019-02-21 10:26:10 -08:00
Chris Wilson
54347c078e iris: Merge two walks of the exec_bos list 2019-02-21 10:26:10 -08:00
Kenneth Graunke
3455f57575 iris: replace vestiges of fence fds with newer exec_fence API 
patch by me and Chris Wilson
 2019-02-21 10:26:10 -08:00
Kenneth Graunke
11da219be9 iris: Avoid synchronizing due to the workaround BO 2019-02-21 10:26:10 -08:00
Kenneth Graunke
30d7bebc8a iris: Avoid cross-batch synchronization on read/reads 
This avoids flushing batches just because e.g. both are reading the same
dynamic state streaming buffer, or shader assembly buffer.
 2019-02-21 10:26:10 -08:00
Kenneth Graunke
b21e916a62 iris: Combine iris_use_pinned_bo and add_exec_bo 2019-02-21 10:26:10 -08:00
Kenneth Graunke
fb4c898842 iris: Use iris_use_pinned_bo rather than add_exec_bo directly 
less special this way
 2019-02-21 10:26:10 -08:00
Kenneth Graunke
25f194d5ac iris: more TODO 2019-02-21 10:26:10 -08:00
Kenneth Graunke
89833eddab iris: leave another TODO 2019-02-21 10:26:10 -08:00
Kenneth Graunke
587e438128 iris: Print the batch name when decoding 2019-02-21 10:26:09 -08:00
Kenneth Graunke
9ec927497e iris: Actually create/destroy HW contexts 
The intention is that render and compute use their own contexts,
and each is PIPELINE_SELECT'd to the right pipeline.  But we hadn't
actually made them, so we got the fd-default context.

Thanks to Chris Wilson for catching this!
 2019-02-21 10:26:09 -08:00
Kenneth Graunke
fbe5d75f11 iris: cross batch flushing 2019-02-21 10:26:09 -08:00
Kenneth Graunke
c3cc525c7a iris: Cross-link iris_batches so they can potentially flush each other 
This makes e.g. the render batch aware of the compute batch, so it can
ask questions like "is this BO referenced by some other batch?" and do
something about that.
 2019-02-21 10:26:09 -08:00
Kenneth Graunke
822f91508e iris: don't mark contains_draw = false when chaining batches 
chaining to a new batch reuses create_batch(), but we don't need to do
the work of pinning BOs we inherit from a previous batch...when that is
actually part of the same execbuf invocation.

instead, just flag it when setting primary_batch_size = 0, in
iris_batch_reset
 2019-02-21 10:26:09 -08:00
Kenneth Graunke
eff081cdd9 iris: Support multiple binder BOs, update Surface State Base Address 2019-02-21 10:26:08 -08:00
Kenneth Graunke
888efcd192 iris: Allow inlining of require/get_command_space 
eliminates so many callqs for ptr++
 2019-02-21 10:26:08 -08:00
Kenneth Graunke
42dccb1233 iris: use consistent copyright formatting 
some of them had typos, didn't say 'authors or copyright holders',
or other mistakes.  This is now https://opensource.org/licenses/MIT
text, formatted consistently.
 2019-02-21 10:26:08 -08:00
Kenneth Graunke
2d578e71d5 iris: get angry about execbuf failures 
want this to be easy to detect for now
 2019-02-21 10:26:08 -08:00
Kenneth Graunke
a378ee3607 iris: simplify batch len qword alignment 
Split from a patch by Chris Wilson so I can test it independently
 2019-02-21 10:26:08 -08:00
Kenneth Graunke
621cb43f41 iris: rename ring to engine 
makes more sense these days.  split from a patch by Chris Wilson
 2019-02-21 10:26:08 -08:00
Kenneth Graunke
dfe1ee4f6f iris: comment everything 
1. Write the code
2. Add comments
3. PROFIT (or just avoid cost of explaining or relearning things...)
 2019-02-21 10:26:08 -08:00
Kenneth Graunke
a3a998f19a iris: iris - fix QWord aligned endings after batch chaining rework 
I need to save the primary batch size after expanding it to include
MI_BATCH_BUFFER_END and the QWord padding NOP
 2019-02-21 10:26:07 -08:00
Kenneth Graunke
aacbcbbf47 iris: colorize batchbuffer failures to make them stand out 2019-02-21 10:26:07 -08:00
Kenneth Graunke
8e2b71b190 iris: bad inherited comments 2019-02-21 10:26:07 -08:00
Kenneth Graunke
8c54433275 iris: Handle batch submission failure "better" 
We used to not reset the batch, and just keep appending to it, so you'd
get the same invalid contents over and over.

I'd also really like to know about this, so aborting seems wise for now,
if not for the long term
 2019-02-21 10:26:07 -08:00
Kenneth Graunke
9226ebfa85 iris: print second batch size separately 2019-02-21 10:26:07 -08:00
Kenneth Graunke
0a43c9defa iris: print refcounts in INTEL_DEBUG=submit 2019-02-21 10:26:07 -08:00
Kenneth Graunke
432790bacd iris: print binder utilization in INTEL_DEBUG=submit 2019-02-21 10:26:07 -08:00
Kenneth Graunke
c763ecaa65 iris: plug leaks 2019-02-21 10:26:07 -08:00
Kenneth Graunke
3f863cf680 iris: fix the validation list on new batches 2019-02-21 10:26:06 -08:00
Kenneth Graunke
9ea05ccf1f iris: completely rewrite binder 
now we get a new one per batch, and flush if it fills up
 2019-02-21 10:26:06 -08:00
Kenneth Graunke
4d24874236 iris: Lower the max number of decoded VBO lines 
saint foo, vbo lines!
 2019-02-21 10:26:06 -08:00
Kenneth Graunke
48ddd7212d iris: fix decoding and undo testing code 2019-02-21 10:26:06 -08:00
Kenneth Graunke
f31eea1f00 iris: fix batch chaining... 
don't chain a batch just for the end
 2019-02-21 10:26:06 -08:00
Kenneth Graunke
604a1a1614 iris: chaining not growing 2019-02-21 10:26:06 -08:00
Kenneth Graunke
053fb51125 iris: just turn batch reset_and_clear_caches into reset 2019-02-21 10:26:06 -08:00
Kenneth Graunke
ca735c5e0c iris: delete growing code and just die for now 
we need proper batch chaining.  without relocations, we can't grow,
since we've only allocated so much VMA for the batch, and the mechanism
only works if we can pin it at the old address
 2019-02-21 10:26:06 -08:00
Kenneth Graunke
75a1639262 iris: drop the 48b printout, we never use anything else 2019-02-21 10:26:06 -08:00
Kenneth Graunke
60d708bb80 iris: copy over i965's cache tracking 
needed to split out vtbl so I can pipe control without ice
 2019-02-21 10:26:06 -08:00
Kenneth Graunke
c75a1254a4 iris: Move get_command_space to iris_batch.c 
for reuse in blorp.  it's a better interface anyway.
 2019-02-21 10:26:06 -08:00
Kenneth Graunke
0148bd6839 iris: decoder fixes 2019-02-21 10:26:05 -08:00
Kenneth Graunke
d890aee15d iris: SBA once at context creation, not per batch 
hooray!
 2019-02-21 10:26:05 -08:00
Kenneth Graunke
bf90d8a125 iris: delete more trash 2019-02-21 10:26:05 -08:00
Kenneth Graunke
b69a85bc4d iris: validation dumping improvements 
backported from i965.  don't bother with (pinned) because everything is.
 2019-02-21 10:26:05 -08:00