ITensor / ITensor/ITensorMPS.jl
Set `eager=true` in `exponentiate` backend
Nobody has claimed this yet.
- Dominant language
- Julia
- Stars
- 85
- Forks
- 27
- Avg merge
- 22m
- Merged PRs (30d)
- 1
Description
In ITensorTDVP.jl v0.3, we set eager=true by default when using the KrylovKit.exponentiate solver/updater, while in ITensorTDVP.jl v0.4 we don't override the default that KrylovKit.jl sets, which is eager=false. This can potentially have big performance penalties since it can lead to many unnecessary applications of the reduced operator.
See:
- https://jutho.github.io/KrylovKit.jl/stable/man/matfun/#KrylovKit.exponentiate
- https://itensor.discourse.group/t/newer-version-of-tdvp-seems-to-be-much-slower/2061
- https://github.com/ITensor/ITensorTDVP.jl/pull/74
For users hitting this issue who are using ITensorTDVP.jl v0.4, you can manually pass updater_kwargs=(; eager=true) to tdvp to get the same (more efficient) behavior of ITensorTDVP.jl v0.3.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by tracing the KrylovKit.exponentiate updater backend and the tdvp updater_kwargs entry point. Verify how ITensorTDVP.jl v0.4 handles the eager default compared with v0.3 and the behavior described in PR 74. Done means the backend uses eager=true by default and the performance regression is covered by an appropriate check.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- julia
- Domain
- performance
- Issue type
- Bug
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 38/100