some tests with high poly mesh data in hashes.
- empty buckets before 4-5%, after 1-2%
- speedup for hash lookups, in my tests lookups take approx ~60% of the time they did before.
In previous optimization in outliner I assumed that order in treehash was not important.
But testing outliner in datablocks mode revealed a problem: when user expands multiple recursive levels and then closes any element, it always closed the top level of recursion.
Now it should work fine with recursive trees.
Now treehash contains groups of elements indexed by (id,nr,type). Adding an element with the same (id,nr,type) results in appending it to existing group. No duplicates are possible in treehash.
This commit should also make lookups a little bit faster, because searching in small arrays by "used" is faster than searching in hashtable with duplicates by "id,nr,type,used".
- plane_point_side_v3(), a bit like line_point_side_v2()
- isect_point_planes_v3(), moved from paint_hide.c
functions to convert between point/normal pairs.
- plane_from_point_normal_v3()
- plane_to_point_normal_v3()
The reason of this is because PATH_MAX is not guaranteed
to be defined on all platforms and Hurd doesn't define it.
So either we need to support arbitrary long file path or
we need to define own maximum path length.
The rule here would be:
- If it's not big trouble to support arbitrary long paths
(i.e. in ghost by using std::string instead of char*)
then arbitrary long path shall be implemented.
- For other cases to use PATH_MAX please include BLI_fileops.h
which takes care of making sure PATH_MAX is defined.
Additional change: get rid of own changes made yesterday
which were supposed to make storage.c work fine in cases
PATH_MAX is not define, but on the second though it lead
to unneeded complication of the code.
Thanks Campbell for review!
- Re-arrange locks, so no actual memory allocation
(which is relatively slow) happens from inside
the lock. operation system will take care of locks
which might be needed there on it's own.
- Use spin lock instead of mutex, since it's just
list operations happens from inside lock, no need
in mutex here.
- Use atomic operations for memory in use and total
used blocks counters.
This makes guarded allocator almost the same speed
as non-guarded one in files from Tube project.
There're still MemHead/MemTail overhead which might
be bad for CPU cache utilization
Particle system code used global variable to sort hair by orig index,
which is not safe for threading at all.
Replaced this with usage of reentrant version of qsort, which is
now implemented in BLI. It was moved from recast navigation code
to BLI, so more areas could use it (if needed).
--
svn merge -r59086:59087 ^/branches/soc-2013-depsgraph_mt
When blender is built in debug mode, BLI_mempool_free will
ensure address passed to the function actually belongs to
this pool.
--
svn merge -r58710:58711 ^/branches/soc-2013-depsgraph_mt
This new macros could be used to benchmark overall
execution time of some chunk of code, running in cycle.
The usage is:
void foo(void) {
TIMEIT_BLOCK_INIT(overall_bar);
for (...) {
...
TIMEIT_BLOCK_BEGIN(over_bar);
bar();
TIMEIT_BLOCK_END(oberall_bar);
...
}
TIMEIT_BLOCK_STATS(overall_bar)
}
This would print total time which was spent on
running function bar().
--
svn merge -r58281:58283 ^/branches/soc-2013-depsgraph_mt
