blender
/
blender
130
398
Fork 574
Code Issues 6.2k Pull Requests 749 Projects 19 Wiki Activity

Fix for Fix (c): broken atomic lock in own bmesh code. 

That was a nasty one, Debug build would never have any issue (even tried
with 64 threads!), but Release build would deadlock nearly immediately,
even with only 2 threads!

What happened here (I think) is that gcc optimizer would generate a
specific path endlessly looping when initial value of virtual_lock was
FLT_MAX, by-passing re-assignment from v_no[0] and the atomic cas
completely. Which would have been correct, should v_no[0] not have been
shared (and modified) by multiple threads. ;)

Idea of that (broken) for loop was to avoid completely calling the
atomic cas as long as v_no[0] was locked by some other thread, but...
Guess the avoided/missing memory barrier was the root of the issue here.

Lesson of the evening: Remember kids, do not trust your compiler to
understand all possible threading-related side effects, and be explicit
rather than elegant when using atomic ops!

Side-effect lesson: do check both release and debug builds when messing
with said atomic ops...
This commit is contained in:
Bastien Montagne 2017-11-25 23:14:54 +01:00
parent dd6c918b2c
commit 3c1f3c02c6
1 changed files with 7 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
source/blender/bmesh/intern/bmesh_mesh.c
View File

@ -412,14 +412,16 @@ static void mesh_verts_calc_normals_accum_cb(void *userdata, MempoolIterData *mp
* It also assumes that collisions between threads are highly unlikely,
* else performances would be quite bad here. */
float virtual_lock = v_no[0];
for (virtual_lock = v_no[0];
(virtual_lock == FLT_MAX) || (atomic_cas_float(&v_no[0], virtual_lock, FLT_MAX) != virtual_lock);
virtual_lock = v_no[0])
{
while (true) {
/* This loops until following conditions are met:
* - v_no[0] has same value as virtual_lock (i.e. it did not change since last try).
* - v_no_[0] was not FLT_MAX, i.e. it was not locked by another thread.
* - v_no[0] was not FLT_MAX, i.e. it was not locked by another thread.
*/
const float vl = atomic_cas_float(&v_no[0], virtual_lock, FLT_MAX);
if (vl == virtual_lock && vl != FLT_MAX) {
break;
}
virtual_lock = vl;
}
BLI_assert(v_no[0] == FLT_MAX);
/* Now we own that normal value, and can change it.