changed: avoid long vectors of shared ptrs

[akva2]

this use a lot of memory with many elements.

For a case I'm looking at this brought memory usage down from 58.9GB (61GB peak) to 47.3GB (which is also the peak).

[akva2]

jenkins build this please

[alfbr]

benchmark please

[blattms]

I guess the numbers are for a serial run or are we doing the computation on the unloadbalanced grid?

[akva2]

it's all serial. even though parallel does this for local cells only, the overhead is still there of course.

[akva2]

Added some more commits that brings it down to 46.5949GB. Another big gain can be had by disabling the non-conditional diffusion storage in the intensive quantities, but that is separate.

[akva2]

jenkins build this opm-simulators=3743 please

[ytelses]

Benchmark result overview:

Test	Configuration	Relative
opm-git	OPM Benchmark: flow_mpi_extra - Threads: 1	0.99
opm-git	OPM Benchmark: flow_mpi_norne - Threads: 1	1
opm-git	OPM Benchmark: flow_mpi_norne - Threads: 8	1.014
opm-git	OPM Benchmark: flow_mpi_norne_4c_msw - Threads: 1	1.001
opm-git	OPM Benchmark: flow_mpi_norne_4c_msw - Threads: 8	1.013

• 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=1446

[alfbr]

These are very impressive gains! Needless to say, we really look forward to seeing these improvements in master :-)

[atgeirr]

Read through, looks good, merging!