Issue with free list used for memory allocation when using geoclaw for fine resolution inundation over multiple AMR refinement regions

[naveenragur]

While running a model for high resolution inundation modelling I ran into a simulation crash. Maybe due to how memory allocation is handled for extensive AMR systems.

Simulation log ended with following message:
AMRCLAW: level 4 CFL = .680E+00 dt = 0.6073E-01 final t = 0.372212E+01
AMRCLAW: level 4 CFL = .680E+00 dt = 0.6073E-01 final t = 0.378284E+01
AMRCLAW: level 4 CFL = .680E+00 dt = 0.6073E-01 final t = 0.384357E+01
free list full with 5000 items
Note: The following floating-point exceptions are signalling: IEEE_UNDERFLOW_FLAG IEEE_DENORMAL ==> runclaw: Done executing /home/nrr/projects/sim_tests/noto/tsunami/geoclaw/xgeoclaw via clawutil.runclaw.py ==> runclaw: Output is in /home/nrr/projects/sim_tests/noto/tsunami/geoclaw/_output come from

Solution tried:
Increased the lfdim parameter from 5000 default to 50000 in line 236 of clawpack-v5.13.0/amrclaw/src/2d/amr_module.f90 to handle it.

Just wanted to check in if increasing the lfdim has any adverse effects?

Thanks
Naveen

[mjberger]

There are now side effects from increasing "lfdim", the size of the free list. In fact, it's on the list of improvements other have it automatically resize the way some of the other arrays do, but we haven't gotten to it yet.

…

On Sep 11, 2025, at 6:03 AM, Naveen ***@***.***> wrote: naveenragur created an issue (clawpack/geoclaw#678) <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_clawpack_geoclaw_issues_678&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=4PEQeJ-eTbHJv3d0pLWzLQ&m=c7cNo4Rh77ddZjUJ9XGFBXTdIsSJRlyuksl2f6JSoxAKTuPTjhMCW7z21FVGkNAq&s=CjN-K5-NP-N1m8ZHhDPRBk8NlraX5wKcDc8bhqCogxc&e=> While running a model for high resolution inundation modelling I ran into a simulation crash. Maybe due to how memory allocation is handled for extensive AMR systems. Simulation log ended with following message: AMRCLAW: level 4 CFL = .680E+00 dt = 0.6073E-01 final t = 0.372212E+01 AMRCLAW: level 4 CFL = .680E+00 dt = 0.6073E-01 final t = 0.378284E+01 AMRCLAW: level 4 CFL = .680E+00 dt = 0.6073E-01 final t = 0.384357E+01 free list full with 5000 items Note: The following floating-point exceptions are signalling: IEEE_UNDERFLOW_FLAG IEEE_DENORMAL ==> runclaw: Done executing /home/nrr/projects/sim_tests/noto/tsunami/geoclaw/xgeoclaw via clawutil.runclaw.py ==> runclaw: Output is in /home/nrr/projects/sim_tests/noto/tsunami/geoclaw/_output come from Solution tried: Increased the lfdim parameter from 5000 default to 50000 in line 236 of clawpack-v5.13.0/amrclaw/src/2d/amr_module.f90 to handle it. Just wanted to check in if increasing the lfdim has any adverse effects? Thanks Naveen — Reply to this email directly, view it on GitHub <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_clawpack_geoclaw_issues_678&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=4PEQeJ-eTbHJv3d0pLWzLQ&m=c7cNo4Rh77ddZjUJ9XGFBXTdIsSJRlyuksl2f6JSoxAKTuPTjhMCW7z21FVGkNAq&s=CjN-K5-NP-N1m8ZHhDPRBk8NlraX5wKcDc8bhqCogxc&e=>, or unsubscribe <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_AAGUGCYFWX33U7MBGXNY4MT3SFCGZAVCNFSM6AAAAACGHEWV66VHI2DSMVQWIX3LMV43ASLTON2WKOZTGQYDKNZZGYYTENI&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=4PEQeJ-eTbHJv3d0pLWzLQ&m=c7cNo4Rh77ddZjUJ9XGFBXTdIsSJRlyuksl2f6JSoxAKTuPTjhMCW7z21FVGkNAq&s=evfzRV95OlkJGwgDBjNBvIlsUR407KfFHho95t3nTZk&e=>. You are receiving this because you are subscribed to this thread.

[mandli]

I think @mjberger may have meant no instead of now, but that's a guess.

I have not run into that limit in a while. Is the simulation look ok even with the underflow? It may be useful to look at the number of grids too to make sure nothing else is going amiss.

[mjberger]

yup, that was a typo. Nothing to worry about.

…

On Sep 11, 2025, at 11:52 AM, Kyle Mandli ***@***.***> wrote: mandli left a comment (clawpack/geoclaw#678) <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_clawpack_geoclaw_issues_678-23issuecomment-2D3281555206&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=4PEQeJ-eTbHJv3d0pLWzLQ&m=i2vkVovLMbUZKCnW9wTjtzdjwv0iMeoK2cHMmTWgfBCRzyw1CAkfVCMq1P5r8UfG&s=mOTygeC1onIGVcQEZr-iBcT0xC-8pXap-v7UIzXbxSM&e=> I think @mjberger <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_mjberger&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=4PEQeJ-eTbHJv3d0pLWzLQ&m=i2vkVovLMbUZKCnW9wTjtzdjwv0iMeoK2cHMmTWgfBCRzyw1CAkfVCMq1P5r8UfG&s=d-SOXFaieabnJoN7bO_6LkIjH8gEFVOxaYgp_QkpgZ0&e=> may have meant no instead of now, but that's a guess. I have not run into that limit in a while. Is the simulation look ok even with the underflow? It may be useful to look at the number of grids too to make sure nothing else is going amiss. — Reply to this email directly, view it on GitHub <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_clawpack_geoclaw_issues_678-23issuecomment-2D3281555206&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=4PEQeJ-eTbHJv3d0pLWzLQ&m=i2vkVovLMbUZKCnW9wTjtzdjwv0iMeoK2cHMmTWgfBCRzyw1CAkfVCMq1P5r8UfG&s=mOTygeC1onIGVcQEZr-iBcT0xC-8pXap-v7UIzXbxSM&e=>, or unsubscribe <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_AAGUGCYJCZRSZ2SSXEXIL533SGLCFAVCNFSM6AAAAACGHEWV66VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTEOBRGU2TKMRQGY&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=4PEQeJ-eTbHJv3d0pLWzLQ&m=i2vkVovLMbUZKCnW9wTjtzdjwv0iMeoK2cHMmTWgfBCRzyw1CAkfVCMq1P5r8UfG&s=68odCJz5nayYXJcl1hR8nbOgUVwWQw1Ln4cU0it3G_4&e=>. You are receiving this because you were mentioned.

[naveenragur]

Thank you for the clarification. Was waiting for the run to get over before responding, it still needed a bit of finetuning with the refinement region otherwise was too slow. Marching front scheme was useful to setup the refinement region.
The simulation results seems good and inundation observations seem to be close to that observed for the 2024 Noto event.
@mandli : could you suggest how I could track down the number of grid cells? from the amr output?

[mandli]

I often use some version of this script. I recently have an updated version, which I just posted as a gist if you would like to try that one as well.