feat: Add Taso optimiser #104
Conversation
|
Don't get too scared by the +10k (I got scared too), it's just an ECC file for testing. |
|
Oh never mind, it did optimise over the course of 3 hours: it's soooo sloooooow |
lmondada
force-pushed
the
feat/update-taso
branch
from
024990b to
da00392
Compare
lmondada
force-pushed
the
feat/update-taso
branch
from
18d3c43 to
27a574f
Compare
|
Turns out it works quite a bit better when using their pre-processed circuits (at within 40 seconds |
Required for some #104 ecc sets. Note that the published `tket-json-rs` dropped the `tket2` feature. But it should be fine once CQCL/tket-json-rs#13 is merged and published.
# Conflicts: # src/json/op.rs
src/passes/taso.rs
Outdated
| seen_hashes.insert(hash); | ||
| } | ||
| // try to send to first available thread | ||
| if let Some(next_ind) = cycle_inds.by_ref().take(n_threads).find(|next_ind| { |
There was a problem hiding this comment.
A crossbeam-channel would simplify this quite a bit.
There was a problem hiding this comment.
I've added a TODO for this.
|
I've reorganised the code quite a bit: mainly moved from the passes module to a new optimiser module. I've also introduced a I apologise for the significant rewrite -- the code logic hasn't changed however, so hopefully reviewing this shouldn't be too time consuming. |
Co-authored-by: Agustín Borgna <[email protected]>
Co-authored-by: Agustín Borgna <[email protected]>
Co-authored-by: Agustín Borgna <[email protected]>
Co-authored-by: Agustín Borgna <[email protected]>
Co-authored-by: Agustín Borgna <[email protected]>
Co-authored-by: Agustín Borgna <[email protected]>
From the
taso-optimiserfolder, you should be able to runand see the Barenco benchmark circuit getting optimised.
However, currently it doesn't actually optimise atm, I don't know why that is the case. We need to look into it. Maybe I just haven't let it run long enough...