Adjoint op can handle multiple registers#2147
Closed
paul0403 wants to merge 24 commits into
Closed
Conversation
6 tasks
paul0403
added a commit
that referenced
this pull request
Nov 7, 2025
…2130) **Context:** Higher order primitives can take in dynamically allocated registers **Description of the Change:** - [x] changelog - [x] remove the restriction from docs - [x] cond - [x] for loop - [x] while loop - ~[ ] adjoint~ involves changing adjoint op's mlir definition, will do in another PR #2147 - [x] subroutine - ~[ ] decomposition?~ decomp wires will always be allocated from within the scope Essentially, we need to do the following: 1. The subregion kernels need to parse their arguments not as `*args, qreg`, but as `*args, *dynamically_allocated_qregs, global_qreg` 2. The subregion's qubit index recorder needs to have the same global wire indices as the outer interpreter, and the values need to be the corresponding qreg managers from the subregion's qreg tracer arguments. We implement this mapping by the new `qreg_map` in the subregion kernels. 3. The high-order primitives need to be called with the dynamic qregs. **Benefits:** Higher order primitives can take in dynamically allocated registers **Possible Drawbacks:** Don't know why, but while other subregions have arg signature `(args, qreg)`, subroutine has signature `(qreg, args)`, i.e. it has args in reverse order. This means the subregion wrapper (the thing to be `make_jaxpr`'d) cannot be reused for subroutines. This is probably something we can just reorder? Also, subroutine takes in wire indices through tracer args of the subroutine, even when the indices are static. This is expected, since subroutines are really a seperate function call, instead of just a enclosed subregion. This means its `qreg_map` logic between inner qregs and outer qregs is different than the others. **Related GitHub Issues:** [sc-99734]
Member
Author
|
closing in favor of #2720 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Context:
As part of the work to allow higher order primitives to take in dynamically allocated registers, the underlying
quantum.adjointop needs to be able to take in multiple registers.I am de-prioritizing this for the moment, as it involves more changes at the mlir layer, and is less important than control flow and subroutines.
Description of the Change:
Benefits:
Possible Drawbacks:
Related GitHub Issues:
[sc-102216]