qvm
qvm copied to clipboard
Enhance performance of DQVM
This ticket is a placeholder for isolating and addressing hot spots in DQVM.
The tasks ahead are:
- [ ] Accelerate address-related computations in
apply-distributed-gate
- [x] Allocate offset arrays directly in foreign memory so no copies are necessary when creating MPI datatypes.
- [x] Profile to find bottlenecks in the code.
See also #174.