[Stretch] Support Duration and Stretch types in classical expressions.
#13844
Conversation
This file contains 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
This reverts commit e2b3221.
Types that have some natural order no longer have an ordering when one of them is strictly greater but has an incompatible const-ness (i.e. when the greater type is const but the other type is not).
We need to reject types with const=True in QPY until it supports them. For now, I've also made the Index and shift operator constructors lift their RHS to the same const-ness as the target to make it less likely that existing users of expr run into issues when serializing to older QPY versions.
This is probably a better default in general, since we don't really have much use for const except for timing stuff.
Since we're going for using a Cast node when const-ness differs, this will be fine.
I wasn't going to have this, but since we have DANGEROUS Float => Int, and we have Int => Bool, I think this makes the most sense.
A Stretch can always represent a Duration (it's just an expression without any unresolved stretch variables, in this case), so we allow implicit conversion from Duration => Stretch. The reason for a separate Duration type is to support things like Duration / Duration => Float. This is not valid for stretches in OpenQASM (to my knowledge).
Also adds support to expr.lift to create a value expression of type types.Duration from an instance of qiskit.circuit.Duration.
Also removes the assumption that a const-type can never be an l-value in favor of just restricting l-values with const types from being added to circuits for now. We will (in a separate PR) add support for adding stretch variables to circuits, which are const. However, we may track those differently, or at least not report them as variable when users query the circuit for variables.
This one I'd added thinking I ought to block store from using a const var target. But since I figured it's better to just restrict adding vars to the circuit that are const (and leave the decision of whether or not a const var can be an l-value till later), this test no longer makes sense.
I can't really test Stretch yet since we can't add them to circuits until a later PR.
The best I can do to test these right now (before Delay is made to accept timing expressions) is to use them in comparison operations (will be in a follow up commit).
|
One or more of the following people are relevant to this code:
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary
Adds two new types
DurationandStretchto the classical expression type system, as well as a new (and separate) coreDurationtype toqiskit.circuit.Details and comments
Based on #13832. Here's a readable diff in the meantime.
Core type
The new core type
Durationis apyclassimplemented in Rust as anenumtype. For now, it's got variants for each of the supported timing kinds defined in OpenQASM (ns,us,ms,s,dt). All of these aref64s, except fordtwhich is au64.Equality of
Durationcurrently follows the same choices made by #13816, in that durations are not first brought to the same unit and compared for magnitude. Perhaps in the future, we can consider equality between walltimes in different units if that feels useful.Expression types
Like in QASM, the
Durationtype represents a known length of time, possibly negative. It is separate fromStretchto support things likeDuration / Duration => Float. It is alwaysconst. These are implicitly convertible toStretch, and just simply won't have an unknown component to them.A
Stretchtype represents a not-yet-known length of time. The plan is to use these inVar.newexpressions. They are alwaysconstas well, so they introduce the first use of a const-typedVarexpression in Qiskit (which will be made not to emit aStoreinstruction when added to a circuit).To-do
Durationrepresentation and expression type (ValueorVar).Durationequality.