griefith/test
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
73,274 Commits 1 Branch 0 Tags
9c91c75ea6514bbc500ae69083f25874b60e1470
Commit Graph

3250 Commits

Author SHA1 Message Date
Sergey Sharybin
5d38c14116 Merge branch 'master' into blender2.8 2018-01-10 13:22:03 +01:00
Sergey Sharybin
5614193745 Task scheduler: Use restrict pointer qualifier 
Those pointers are never to be aliased, so let's be explicit about this and hope
compiler does save some CPU ticks.
 2018-01-10 12:49:51 +01:00
Sergey Sharybin
518c65460e Task scheduler: Use more const qualifiers 2018-01-10 12:27:43 +01:00
Sergey Sharybin
f0610563ee math utils: Add size_t version of min and max functions 2018-01-10 11:06:44 +01:00
Campbell Barton
2b56faebe5 Merge branch 'master' into blender2.8 2018-01-10 19:58:05 +11:00
Campbell Barton
18f53d8822 Cleanup: comment block alignment 2018-01-10 19:57:02 +11:00
Campbell Barton
5ff33ecdf0 Merge branch 'master' into blender2.8 2018-01-10 14:01:09 +11:00
Sergey Sharybin
5fe87a0a8c Task scheduler: Use single thread branch when range fits into single chunk 2018-01-09 18:10:47 +01:00
Sergey Sharybin
4a3b303bb0 Task scheduler: Fix wrong tasks calculation when chunk size is too big 2018-01-09 18:07:34 +01:00
Sergey Sharybin
6deb908a5c Merge branch 'master' into blender2.8 2018-01-09 16:15:33 +01:00
Sergey Sharybin
932d448ae0 Task scheduler: Use const qualifiers in parallel range 2018-01-09 16:09:33 +01:00
Sergey Sharybin
8cffb0a141 Task scheduler: Avoid over-allocation of tasks for parallel ranges 
This seems to only cause extra rthreading overhead on systems with 10s of
threads, without actually solving anything.
 2018-01-09 16:09:33 +01:00
Sergey Sharybin
c4e42d70a4 Task scheduler: Add minimum number of iterations per thread in parallel range 
The idea is to support following: allow doing parallel for on a small range,
each iteration of which takes lots of compute power, but limit such range to
a subset of threads.

For example, on a machine with 44 threads we can occupy 4 threads to handle
range of 64 elements, 16 elements per thread, where each block of 16 elements
is very complex to compute.

The idea should be to use this setting instead of global use_threading flag,
which is only based on size of array. Proper use of the new flag will improve
threadability.

This commit only contains internal task scheduler changes, this setting is not
used yet by any areas.
 2018-01-09 16:09:33 +01:00
Sergey Sharybin
3144f0573a Task scheduler: Simplify parallel range function 
Basically, split it up and avoid extra abstraction level.
 2018-01-09 16:09:33 +01:00
Sergey Sharybin
4c4a7e84c6 Task scheduler: Use single parallel range function with more flexible function 
Now all the fine-tuning is happening using parallel range settings structure,
which avoid passing long lists of arguments, allows extend fine-tuning further,
avoid having lots of various functions which basically does the same thing.
 2018-01-09 16:09:33 +01:00
Sergey Sharybin
d2708b0f73 Task scheduler: Get rid of extended version of parallel range callback 
Wrap all arguments into TLS type of argument. Avoids some branching and also
makes it easier to extend things in the future.
 2018-01-09 16:09:33 +01:00
Campbell Barton
be40389165 Merge branch 'master' into blender2.8 2018-01-03 23:44:47 +11:00
Sergey Sharybin
6efd58dd3e Task scheduler: Clarify why do we need an atomic add of 0 2017-12-22 16:37:25 +01:00
Sergey Sharybin
50f1c9a8af Task scheduler: Start with suspended pool to avoid threading overhead on push 
The idea is to avoid any threading overhead when we start pushing tasks in a
loop. Similarly to how we do it from the new dependency graph. Gives couple of
percent of speedup here, but also improves scalability.
 2017-12-22 12:25:11 +01:00
Sergey Sharybin
a52452c043 Merge branch 'master' into blender2.8 2017-12-20 16:40:49 +01:00
Antonio Vazquez
94a3ee56c7 Fix MSVSC2017 error 
The last compiler version needs this include
 2017-12-20 10:23:12 +01:00
Campbell Barton
cfd1b48492 Merge branch 'master' into blender2.8 2017-12-19 14:15:19 +11:00
Campbell Barton
7a58ff928c Math Lib: signed versions of quaternion angle 
There was no simple way to get the shortest quaternion angle.
 2017-12-19 14:03:29 +11:00
Sergey Sharybin
114013de0d Merge branch 'master' into blender2.8 2017-12-15 17:24:28 +01:00
Sergey Sharybin
5e28b71457 math utils: Add utilities to scan bit and clear it 2017-12-15 16:54:28 +01:00
Sergey Sharybin
de9e5a0926 Math utils: Go away form naive code for highest_order_bit_uint 2017-12-15 16:54:28 +01:00
Sergey Sharybin
412de222f8 Math utils: Add bit scan operations 2017-12-15 16:54:28 +01:00
Sergey Sharybin
2aa0f8a41b Math utils: Cleanup, use _uint suffix for function which operates on uint 2017-12-15 16:54:28 +01:00
Campbell Barton
03a5eccc94 Merge branch 'master' into blender2.8 2017-11-30 18:30:41 +11:00
Campbell Barton
28d2148b09 Haiku OS Support 
D2860 by @miqlas

Even though Haiku is a niche OS, only minor changes are needed.
 2017-11-30 18:05:21 +11:00
Brecht Van Lommel
84d39ab97b Merge branch 'master' into blender2.8 2017-11-29 18:13:06 +01:00
Brecht Van Lommel
4f7280da40 Fix paint float color blending bugs with alpha 1.0 and vivid light. 
For some blend modes there would be no effect with factor 1.0, even if factor
0.999 would give a very different image. Now the result should have no
discontinuity.

Differential Revision: https://developer.blender.org/D2925
 2017-11-29 18:01:36 +01:00
Campbell Barton
f3b384862d Merge branch 'master' into blender2.8 2017-11-24 14:30:36 +11:00
Bastien Montagne
b63442e0b6 Minor cleanup for own recent commits. 2017-11-23 22:43:11 +01:00
Bastien Montagne
a786baa193 Merge branch 'master' into blender2.8 2017-11-23 21:36:27 +01:00
Bastien Montagne
efb86b712d Add a new parallel looper for MemPool items to BLI_task. 
It merely uses the new thread-safe iterators system of mempool, quite
straight forward.

Note that to avoid possible confusion with two void pointers as
parameters of the callback, a dummy opaque struct pointer is used
instead for the second parameter (pointer generated by iteration over
mempool), callback functions must explicitely convert it to expected
real type.

Also added a basic gtest for this new feature.
 2017-11-23 21:14:43 +01:00
Bastien Montagne
b84e6dfee4 Add ability to use more than one mempool iterator simultaneously. 
This will allow threaded tasks to 'consume' all mempool items in
parallel tasks, each one working on a whole chunk at once (to reduce
concurrency managing overhead).
 2017-11-23 21:12:00 +01:00
Bastien Montagne
5330e7fd29 Merge branch 'master' into blender2.8 2017-11-23 16:29:00 +01:00
Bastien Montagne
497e2b3dfa Cleanup: use signed atomic ops when needed. 2017-11-23 16:24:34 +01:00
Sergey Sharybin
5d2b3a966e Merge branch 'master' into blender2.8 2017-11-22 12:50:17 +01:00
Brecht Van Lommel
96415cb52a Code cleanup: fix harmless compiler warning. 2017-11-20 23:32:06 +01:00
Julian Eisel
7f96323cd0 Merge branch 'master' into blender2.8 2017-11-19 13:16:14 +01:00
Campbell Barton
83e4e65074 Cleanup: outdated comment 2017-11-19 03:25:52 +11:00
Campbell Barton
40c8a18229 Cleanup: style 2017-11-18 17:22:54 +11:00
Bastien Montagne
562da211c2 Merge branch 'master' into blender2.8 2017-11-16 12:50:53 +01:00
Bastien Montagne
d697e3d46e BLI listbase: add bytes finding helpers. 
Quite similar to string ones actually, except more generic. Used in
id_override_static branch currently.
 2017-11-16 12:49:31 +01:00
Campbell Barton
ae8130e287 Merge branch 'master' into blender2.8 2017-11-15 23:17:25 +11:00
Campbell Barton
474362b363 Cleanup: use 'uint' 2017-11-15 22:58:39 +11:00
Campbell Barton
9cbf374814 GSet: utils to access data stored outside the set 2017-11-15 22:45:37 +11:00
Bastien Montagne
2a7f8d3cb3 Merge branch 'master' into blender2.8 
Conflicts:
	source/blender/blenlib/BLI_math_matrix.h
	source/blender/blenlib/intern/math_matrix.c
	source/blender/blenlib/intern/rand.c
	source/blender/editors/animation/anim_channels_edit.c
	source/blender/makesrna/intern/rna_mask.c
 2017-11-15 10:46:19 +01:00