Fix up Synapse benchmark

[ChrisRackauckas]

https://docs.sciml.ai/SciMLBenchmarksOutput/dev/HybridJumps/Synapse/

Continuing from #673 and #672

It looks like it built funny this time and someone needs to take a look at it. @gzagatti can you check that the manifest is using the right branch?

@isaacsas

[isaacsas]

Looks like the Hawkes one has issues too: https://docs.sciml.ai/SciMLBenchmarksOutput/dev/Jumps/MultivariateHawkes/

Also, why did Synapse appear under the Jumps category too? It seems like it ran there?

[ChrisRackauckas]

No, that's the old markdown. When a benchmark moves I have to manually delete the old one. I just did that earlier before posting so it'll take a few hours before the new build shows it removed from the old section.

[gzagatti]

The manifest of HybridJumps was indeed the culprit.

To get things working, I downgraded OrdinaryDiffEq from 6.55 back to 6.20 and IntervalSets from 0.7.7 back to 0.7.3. Now, the benchmarks can run until the end. I have opened PR #714 to track the solution.

It is not clear to me what changed in OrdinaryDiffEq to cause the Return code = DtLessThanMin that prevented the synapse benchmark to complete in your latest change.

To be honest, previously there were a few runs of PDMP which were unstable raising either one of the following warnings:

┌ Warning: Instability detected. Aborting
└ @ SciMLBase ~/.julia/packages/SciMLBase/l4PVV/src/integrator_interface.jl:606

Or:

┌ Warning: The simulation did not reach requested simulated time.
└ @ Synapse ~/.julia/packages/Synapse/cEAkz/src/Solve.jl:236

So could it be that the new version of OrdinaryDiffEq is more strict?

Next, I downgraded IntervalSets because it was having issues with DomainSets.

┌ DomainSets [5b8099bc-c8ec-5219-889f-1d9e522a28bf]
│  WARNING: Method definition isapprox(IntervalSets.AbstractInterval{T} where T, IntervalSets.A
bstractInterval{T} where T) in module IntervalSets at /home/guilherme/.julia/packages/IntervalS
ets/6RTOk/src/IntervalSets.jl:144 overwritten in module DomainSets at /home/guilherme/.julia/pa
ckages/DomainSets/aafhp/src/domains/interval.jl:52.
│    ** incremental compilation may be fatally broken for this module **
└  

However, I don't think that IntervalSets is the main culprit here.

Finally, the Hawkes benchmark didn't build correctly again. It looks like there was a problem precompiling PiecewiseDeterministicMarkovProcesses.

Error: Failed to precompile PiecewiseDeterministicMarkovProcesses [86206cdf
-4603-54e0-bd58-22a2dcbf57aa] to "/cache/julia-buildkite-plugin/depots/5b30
0254-1738-4989-ae0a-f4d2d937f953/compiled/v1.9/PiecewiseDeterministicMarkov
Processes/jl_Yer1rM".

I just updated the whole Manifest and it now completes successfully.

[ChrisRackauckas]

Can you make an MWE?

[gzagatti]

After some digging, I think the problem is with IntervalSets after all. See my comment on the PR.