[Build] Memory optimization for global allocation

[angehung5]

Huge memory is required by biogenic historic option when run globally (1536 x 3072 grid points).

In file 'canopy_alloc.F90', around line 68: Error allocating 909744537600 bytes: Cannot allocate memory

@MaggieMarvin What do you think? Seems like we need to disable hist_opt at least as default off (=0) in example namelist. I think we could move the historical calculations into the ifcanbio block of canopy_calcs and also move those alloc/dealloc variables inside the ifcanbio block as well though. We can address in future PR with your biogenic updates coming. -- Originally posted by @drnimbusrain

As discussed with @drnimbusrain, we should start with at least warning users about this issue in README.

[drnimbusrain]

@MaggieMarvin Does this further imply we need a restart option like in issue #94 ? Running long time periods for simulation are prohibitive in biogenics considering the historical aspects, and a restart option would facilitate running shorter periods able to restart some variables that are needed historically (e.g., leaf temp, ppfd, age, etc.).

I agree warnings in the short term, and/or maybe other suggestions to reduce number of vertical layers significantly, if this even helps use historical option globally? @angehung5

[angehung5]

Job efficiency for global runs with all options on, ntime=1

10 layers, 5m resolution

Job ID: 3246239
Cluster: hopper
User/Group: whung/users
State: COMPLETED (exit code 0)
Cores: 1
CPU Utilized: 00:04:16
CPU Efficiency: 93.77% of 00:04:33 core-walltime
Job Wall-clock time: 00:04:33
Memory Utilized: 657.92 GB
Memory Efficiency: 82.24% of 800.00 GB

3 layers, 15m resol

Job ID: 3246642
Cluster: hopper
User/Group: whung/users
State: COMPLETED (exit code 0)
Cores: 1
CPU Utilized: 00:00:35
CPU Efficiency: 83.33% of 00:00:42 core-walltime
Job Wall-clock time: 00:00:42
Memory Utilized: 200.04 GB
Memory Efficiency: 25.01% of 800.00 GB

5 layers, 10m resol

Job ID: 3246645
Cluster: hopper
User/Group: whung/users
State: COMPLETED (exit code 0)
Cores: 1
CPU Utilized: 00:01:26
CPU Efficiency: 92.47% of 00:01:33 core-walltime
Job Wall-clock time: 00:01:33
Memory Utilized: 315.49 GB
Memory Efficiency: 39.44% of 800.00 GB 

Much improved after reducing number of vertical layers. 5 layers might be a good number? @drnimbusrain
Also the historical calculation should be moved inside of biogenic canopy_calcs. They are still on even after turning off biogenics.

[drnimbusrain]

Yes reducing layers helps a lot, but still issues when running real cases over many time steps (e.g., hourly for weeks or months). With historical option on this leads to many TBs of memory allocated for the respective temp variables calculated

…

On Tue, Mar 4, 2025, 9:20 PM Wei-Ting Hung ***@***.***> wrote: Job efficiency for global runs with all options on, ntime=1 *10 layers, 5m resolution* Job ID: 3246239 Cluster: hopper User/Group: whung/users State: COMPLETED (exit code 0) Cores: 1 CPU Utilized: 00:04:16 CPU Efficiency: 93.77% of 00:04:33 core-walltime Job Wall-clock time: 00:04:33 Memory Utilized: 657.92 GB Memory Efficiency: 82.24% of 800.00 GB *3 layers, 15m resol* Job ID: 3246642 Cluster: hopper User/Group: whung/users State: COMPLETED (exit code 0) Cores: 1 CPU Utilized: 00:00:35 CPU Efficiency: 83.33% of 00:00:42 core-walltime Job Wall-clock time: 00:00:42 Memory Utilized: 200.04 GB Memory Efficiency: 25.01% of 800.00 GB *5 layers, 10m resol* Job ID: 3246645 Cluster: hopper User/Group: whung/users State: COMPLETED (exit code 0) Cores: 1 CPU Utilized: 00:01:26 CPU Efficiency: 92.47% of 00:01:33 core-walltime Job Wall-clock time: 00:01:33 Memory Utilized: 315.49 GB Memory Efficiency: 39.44% of 800.00 GB Much improved after reducing number of vertical layers. 5 layers might be a good number? @drnimbusrain <https://github.com/drnimbusrain> Also the historical calculation should be moved inside of biogenic canopy_calcs. They are still on even after turning off biogenics. — Reply to this email directly, view it on GitHub <#151 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGLFYNRWUJAMJTH35KLZKY32SZNOHAVCNFSM6AAAAABYC3KPU6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMOJZGU4DCNJVGA> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: angehung5]*angehung5* left a comment (noaa-oar-arl/canopy-app#151) <#151 (comment)> Job efficiency for global runs with all options on, ntime=1 *10 layers, 5m resolution* Job ID: 3246239 Cluster: hopper User/Group: whung/users State: COMPLETED (exit code 0) Cores: 1 CPU Utilized: 00:04:16 CPU Efficiency: 93.77% of 00:04:33 core-walltime Job Wall-clock time: 00:04:33 Memory Utilized: 657.92 GB Memory Efficiency: 82.24% of 800.00 GB *3 layers, 15m resol* Job ID: 3246642 Cluster: hopper User/Group: whung/users State: COMPLETED (exit code 0) Cores: 1 CPU Utilized: 00:00:35 CPU Efficiency: 83.33% of 00:00:42 core-walltime Job Wall-clock time: 00:00:42 Memory Utilized: 200.04 GB Memory Efficiency: 25.01% of 800.00 GB *5 layers, 10m resol* Job ID: 3246645 Cluster: hopper User/Group: whung/users State: COMPLETED (exit code 0) Cores: 1 CPU Utilized: 00:01:26 CPU Efficiency: 92.47% of 00:01:33 core-walltime Job Wall-clock time: 00:01:33 Memory Utilized: 315.49 GB Memory Efficiency: 39.44% of 800.00 GB Much improved after reducing number of vertical layers. 5 layers might be a good number? @drnimbusrain <https://github.com/drnimbusrain> Also the historical calculation should be moved inside of biogenic canopy_calcs. They are still on even after turning off biogenics. — Reply to this email directly, view it on GitHub <#151 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGLFYNRWUJAMJTH35KLZKY32SZNOHAVCNFSM6AAAAABYC3KPU6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMOJZGU4DCNJVGA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[MaggieMarvin]

Yes, I do definitely think we still need to work on a restart option and in the meantime we can turn default hist options off and/or move allocations to ifcanbio block.