opm-grid icon indicating copy to clipboard operation
opm-grid copied to clipboard

[bugfix?] Add overlap cell even if intersection has zero transmissibility

Open blattms opened this issue 1 year ago • 5 comments

While there might be no flow of oil/gas/water over an intersection with zero transmissibility, there might still be heat conduction. I think in that case we need to see the neighbor in parallel.

I have to admit that I don't have the slightest idea how heat transport is implemented. So this wild guess might very well be wrong.

Hence this is up for discussion,

blattms avatar Jun 07 '24 11:06 blattms

jenkins build this please

blattms avatar Jun 07 '24 11:06 blattms

benchmark please

blattms avatar Jun 07 '24 12:06 blattms

Benchmark result overview:

Test Configuration Relative
opm-git OPM Benchmark: drogon - Threads: 1 1.008
opm-git OPM Benchmark: drogon - Threads: 8 1.212
opm-git OPM Benchmark: punqs3 - Threads: 1 1.013
opm-git OPM Benchmark: punqs3 - Threads: 8 1.012
opm-git OPM Benchmark: smeaheia - Threads: 1 1.014
opm-git OPM Benchmark: smeaheia - Threads: 8 0.996
opm-git OPM Benchmark: spe10_model_1 - Threads: 1 1.01
opm-git OPM Benchmark: spe10_model_1 - Threads: 8 0.993
opm-git OPM Benchmark: flow_mpi_extra - Threads: 1 0.995
opm-git OPM Benchmark: flow_mpi_extra - Threads: 8 1.001
opm-git OPM Benchmark: flow_mpi_norne - Threads: 1 1.003
opm-git OPM Benchmark: flow_mpi_norne - Threads: 8 0.999
opm-git OPM Benchmark: flow_mpi_norne_4c_msw - Threads: 1 0.989
opm-git OPM Benchmark: flow_mpi_norne_4c_msw - Threads: 8 0.988
  • Speed-up = Total time master / Total time pull request. Above 1.0 is an improvement. *

View result details @ https://www.ytelses.com/opm/?page=result&id=2501

ytelses avatar Jun 07 '24 19:06 ytelses

While there might be no flow of oil/gas/water over an intersection with zero transmissibility, there might still be heat conduction. I think in that case we need to see the neighbor in parallel.

Good point. Would it be easy to make this configurable? The original implementation comments that not adding this overlap may improve parallel performance, which sounds like an option is a good idea. Also we do not have any idea if this condition has ever been triggered (or at least I don't), maybe we should try to make a count?

atgeirr avatar Jun 11 '24 08:06 atgeirr

The same would be in the case of mechanics. I think it is to make the grid and or overlap should not depend to much or physics and or discretization.

hnil avatar Mar 07 '25 08:03 hnil