!LQrdVpOJEohPSWMlmf:matrix.org

Smilei users

65 Members
github.com/SmileiPIC/Smilei1 Servers

Load older messages


SenderMessageTime
13 Apr 2021
@davidrblackman:matrix.orgdoubleagentdaveI'll see if I can figure out what is going on16:19:42
@fredpz:matrix.orgfredpzIf you have a somewhat more verbose segfault it could help us figure it out16:55:36
@davidrblackman:matrix.orgdoubleagentdaveThis si the most verbose it gets17:13:26
@davidrblackman:matrix.orgdoubleagentdave

Time-Loop started: number of time-steps n_time = 2

[WARNING] The following `push time` assumes a global number of 256 cores (hyperthreading is unknown)

timestep sim time cpu time [s] ( diff [s] ) push time [ns]
Stack trace (most recent call last):
#5 Object "[0xffffffffffffffff]", at 0xffffffffffffffff, in
#4 Object "/scratch/07303/drblackm/plasma_mirror_reflect/density_test_o/smilei", at 0x4623a8, in
#3 Object "/lib64/libc.so.6", at 0x2b650606a3d4, in __libc_start_main
#2 Object "/scratch/07303/drblackm/plasma_mirror_reflect/density_test_o/smilei", at 0x5d9de7, in main
#1 Object "/scratch/07303/drblackm/plasma_mirror_reflect/density_test_o/smilei", at 0x5e0bf3, in DoubleGrids::syncCurrentsOnRegion(VectorPatch&, Region&, Params&, SmileiMPI*, Timers&, int)
#0 Object "/scratch/07303/drblackm/plasma_mirror_reflect/density_test_o/smilei", at 0x92beb7, in Field3D::add(Field*, Params&, SmileiMPI*, Patch*, Patch*)
Segmentation fault (Address not mapped to object [0x80564c920])

17:13:28
@davidrblackman:matrix.orgdoubleagentdaveignore the warning17:13:38
@davidrblackman:matrix.orgdoubleagentdaveI tried reducing the number of MPi tasks to see if it was a virtual memory problem but still17:14:04
@davidrblackman:matrix.orgdoubleagentdaveit's worth noting that the vector patch statement occurs even if vector is turned off17:15:45
@fredpz:matrix.orgfredpzVectorPatch is unrelated to vectorization17:19:11
@davidrblackman:matrix.orgdoubleagentdaveok, right, i was wondering17:19:26
@davidrblackman:matrix.orgdoubleagentdaveI'm just trying it with more nodes, maybe it's a lack of memory issue still17:21:55
@fredpz:matrix.orgfredpzYou could try to reduce your case. Maybe less particles? There could be a real bug with some features. Try to remove things one by one17:42:52
@davidrblackman:matrix.orgdoubleagentdaveOK, reducing the grid size and particle count (point taken), and there is still the same output17:56:57
@davidrblackman:matrix.orgdoubleagentdavethe ram usage in this case is tiny Particles: Master 0 MB; Max 136 MB; Global 3.54 GB Fields: Master 79 MB; Max 81 MB; Global 5.91 GB SDMD grid: Master 23 MB; Max 23 MB; Global 2.96 GB scalars.txt: Master 0 MB; Max 0 MB; Global 0 GB Fields0.h5: Master 0 MB; Max 0 MB; Global 0 GB 17:57:26
@davidrblackman:matrix.orgdoubleagentdaveso it's not a memory issue17:57:32
@davidrblackman:matrix.orgdoubleagentdavegoign through things one by one hang on17:59:39
@davidrblackman:matrix.orgdoubleagentdaveit's not a problem with any of these things: laser, poisson solver, diagnostic, current filtering, vectoring, custom density profile, 18:04:30
@davidrblackman:matrix.orgdoubleagentdaveGOT IT, not enough ghost cells, need 4 for interpolation=418:05:13
@davidrblackman:matrix.orgdoubleagentdavethanks :D18:05:18
@davidrblackman:matrix.orgdoubleagentdaveMaybe for idiots like me it would be useful for an exception to be raised? The default value maybe should depend on the interpolation rather than be set exactly at 2?18:07:06
@fredpz:matrix.orgfredpzYes good idea18:08:11
@fredpz:matrix.orgfredpzI thought it would be the case18:08:23
@davidrblackman:matrix.orgdoubleagentdaveI missed what the default setting was, sorry if this wasted anyones times18:08:52
@fredpz:matrix.orgfredpzCould you open a short issue for us to remember? 18:08:56
@davidrblackman:matrix.orgdoubleagentdavesure18:09:03
@fredpz:matrix.orgfredpzThank you for spotting the bug 👍18:09:32
@davidrblackman:matrix.orgdoubleagentdaveOne other thing, the pushtime reported is negative, is this normal? it's got a value of -214748364823:04:10
14 Apr 2021
@davidrblackman:matrix.orgdoubleagentdaveI've posted another small weird bug that might be worth looking into.01:47:29
@xxirii:matrix.orgxxiriifredpz: the int conversion when no particle ?05:52:03
@xxirii:matrix.orgxxirii* fredpz: the int conversion when no particle ?05:52:34
@fredpz:matrix.orgfredpzIndeed the negative value is a division by zero06:09:53

There are no newer messages yet.


Back to Room List