#2892: SpacetimeX: GPU error in PunctureTracker::PunctureContainer::interpolate
Reporter: Miren Radia
Status: open
Milestone:
Version: development version
Type: bug
Priority: major
Component: CarpetX
Comment (by Miren Radia):
I tried updating the version of AMReX to v25.10 but that did not help. I’m now trying the more drastic step of installing a newer version of CUDA on this system \(specifically 12.6.2\) along with other dependencies \(e.g. UCX, Open MPI, etc.\).
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2892/spacetimex-gpu-er…
#2892: SpacetimeX: GPU error in PunctureTracker::PunctureContainer::interpolate
Reporter: Miren Radia
Status: open
Milestone:
Version: development version
Type: bug
Priority: major
Component: CarpetX
Comment (by Miren Radia):
There are several versions of CarpetX I’ve tried. When I first reported the issue, I was using `8a4c419ce902607673c7ca3b29dd96ac164c0633`. When I rebuilt everything after you told me to try the `puncturetracker` branch for SpacetimeX, I ended up on `8dbd326326f525dd531e6628e8dd8b583030bbd4` \(for CarpetX\).
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2892/spacetimex-gpu-er…
#2892: SpacetimeX: GPU error in PunctureTracker::PunctureContainer::interpolate
Reporter: Miren Radia
Status: open
Milestone:
Version: development version
Type: bug
Priority: major
Component: CarpetX
Comment (by Liwei Ji):
Hi Miren, could you share the git hash of the development branch of CarpetX that you are using?
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2892/spacetimex-gpu-er…
#2892: SpacetimeX: GPU error in PunctureTracker::PunctureContainer::interpolate
Reporter: Miren Radia
Status: open
Milestone:
Version: development version
Type: bug
Priority: major
Component: CarpetX
Comment (by Miren Radia):
I just ran it \(on 2 tursa nodes with 8 Nvidia A100 GPUs\) and ran into the same problem. Since the problem seems to be on the AMReX side, I’ve messaged on their Slack workspace.
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2892/spacetimex-gpu-er…
#2892: SpacetimeX: GPU error in PunctureTracker::PunctureContainer::interpolate
Reporter: Miren Radia
Status: open
Milestone:
Version: development version
Type: bug
Priority: major
Component: CarpetX
Comment (by Miren Radia):
I just tried that branch with the modified qc0 parameter file and now I get the same error as in my original message. When not running through a debugger, the output is just:
```
terminate called after throwing an instance of 'std::runtime_error'
what(): GPU error in file /mnt/lustre/tursafs1/home/dp415/dp415/dc-radi1/ETK/Cactus-CarpetX-dev/configs/sim/scratch/external/AMReX/include/AMReX_Scan.H line 1242 invalid argument
[tu-c0r0n87:342593] *** Process received signal ***
[tu-c0r0n87:342593] Signal: Aborted (6)
[tu-c0r0n87:342593] Signal code: (-6)
[tu-c0r0n87:342593] [ 0] /lib64/libpthread.so.0(+0x12d10)[0x1552d96c0d10]
[tu-c0r0n87:342593] [ 1] /lib64/libc.so.6(gsignal+0x10f)[0x1552d0e8852f]
[tu-c0r0n87:342593] [ 2] /lib64/libc.so.6(abort+0x127)[0x1552d0e5be65]
[tu-c0r0n87:342593] [ 3] /home/y07/shared/utils/core/gcc/12.2.0/lib64/libstdc++.so.6(+0xa9bd9)[0x1552d1720bd9]
[tu-c0r0n87:342593] [ 4] /home/y07/shared/utils/core/gcc/12.2.0/lib64/libstdc++.so.6(+0xb525a)[0x1552d172c25a]
[tu-c0r0n87:342593] [ 5] /home/y07/shared/utils/core/gcc/12.2.0/lib64/libstdc++.so.6(+0xb52c5)[0x1552d172c2c5]
[tu-c0r0n87:342593] [ 6] /home/y07/shared/utils/core/gcc/12.2.0/lib64/libstdc++.so.6(+0xb5517)[0x1552d172c517]
[tu-c0r0n87:342593] [ 7] ./cactus_sim_dev[0x4b305d]
[tu-c0r0n87:342593] [ 8] ./cactus_sim_dev[0x109263a]
[tu-c0r0n87:342593] [ 9] ./cactus_sim_dev[0x109dc22]
[tu-c0r0n87:342593] [10] ./cactus_sim_dev[0x10a9b23]
[tu-c0r0n87:342593] [11] ./cactus_sim_dev[0x1073ed5]
[tu-c0r0n87:342593] [12] ./cactus_sim_dev[0x107084b]
[tu-c0r0n87:342593] [13] ./cactus_sim_dev[0x1424f34]
[tu-c0r0n87:342593] [14] ./cactus_sim_dev[0x1426520]
[tu-c0r0n87:342593] [15] ./cactus_sim_dev[0x147e63d]
[tu-c0r0n87:342593] [16] ./cactus_sim_dev[0x135f221]
[tu-c0r0n87:342593] [17] ./cactus_sim_dev[0x147e8b2]
[tu-c0r0n87:342593] [18] ./cactus_sim_dev[0x1485a9c]
[tu-c0r0n87:342593] [19] ./cactus_sim_dev[0x1485bab]
[tu-c0r0n87:342593] [20] ./cactus_sim_dev[0x1480f67]
[tu-c0r0n87:342593] [21] ./cactus_sim_dev[0x1029576]
[tu-c0r0n87:342593] [22] ./cactus_sim_dev[0x147105f]
[tu-c0r0n87:342593] [23] ./cactus_sim_dev[0x13641d4]
[tu-c0r0n87:342593] [24] ./cactus_sim_dev[0x4d4cde]
[tu-c0r0n87:342593] [25] /lib64/libc.so.6(__libc_start_main+0xe5)[0x1552d0e747e5]
[tu-c0r0n87:342593] [26] ./cactus_sim_dev[0x6b10de]
[tu-c0r0n87:342593] *** End of error message ***
```
but I guess we can assume the backtrace is the same as above.
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2892/spacetimex-gpu-er…
#2892: SpacetimeX: GPU error in PunctureTracker::PunctureContainer::interpolate
Reporter: Miren Radia
Status: open
Milestone:
Version: development version
Type: bug
Priority: major
Component: CarpetX
Comment (by Liwei Ji):
Could you try your older setup with branch `puncturetracker` of ETK repo?
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2892/spacetimex-gpu-er…
#2892: SpacetimeX: GPU error in PunctureTracker::PunctureContainer::interpolate
Reporter: Miren Radia
Status: open
Milestone:
Version: development version
Type: bug
Priority: major
Component: CarpetX
Comment (by Liwei Ji):
Thanks Miren. I will take a look now. Stay tuned
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2892/spacetimex-gpu-er…
#2892: SpacetimeX: GPU error in PunctureTracker::PunctureContainer::interpolate
Reporter: Miren Radia
Status: open
Milestone:
Version: development version
Type: bug
Priority: major
Component: CarpetX
Comment (by Miren Radia):
That parameter file with `CarpetX::poison_undefined_value = yes` works fine for me since it doesn’t use the PunctureTracker thorn.
The problems arise when I enable the `PunctureTracker` thorn \(and also `BoxInBox` since it’s required by `PunctureTracker`\). I modified the previous parameter file \(i.e. `qc0.par` with `CarpetX::poison_undefined_value = yes`\) by enabling those two thorns and adding the following parameters:
```
BoxInBox::num_regions = 1
BoxInBox::shape_1 = "cube"
BoxInBox::num_levels_1 = 1
BoxInBox::radius_1 = [-1]
BoxInBox::position_x_1 = 0.0
BoxInBox::position_y_1 = 0.0
BoxInBox::position_z_1 = 0.0
PunctureTracker::verbose = yes
PunctureTracker::track [0] = yes
PunctureTracker::initial_x[0] = +0.5843214365
PunctureTracker::track [1] = yes
PunctureTracker::initial_x[1] = -0.5843214365
```
I then get the following error \(which I mentioned in my original issue description\):
```
ERROR from host tu-c0r0n87 process 0
in thorn CarpetX, file /mnt/lustre/tursafs1/home/dp415/dp415/dc-radi1/ETK/Cactus-CarpetX-main/arrangements/CarpetX/CarpetX/src/valid.cxx:552:
-> CallFunction iteration 0 PunctureTracker_SetupGroup: PunctureTracker::PunctureTracker_Setup checking output: Grid array "BOXINBOX::position_x[0]" has 1 nans on time level 0; expected valid
The interior is valid because: CallFunction iteration 0 PunctureTracker_SetupGroup: PunctureTracker::PunctureTracker_Setup: Mark output variables as valid.
The outer boundary is valid because: CallFunction iteration 0 PunctureTracker_SetupGroup: PunctureTracker::PunctureTracker_Setup: Mark output variables as valid.
The ghost zones are valid because: CallFunction iteration 0 PunctureTracker_SetupGroup: PunctureTracker::PunctureTracker_Setup: Mark output variables as valid.
```
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2892/spacetimex-gpu-er…