CRZ gate decomposition error #103
Comments
|
So a problem here is that different libraries and people use different standards for this gate. I believe I based the current definition on the qasm spec, but I can check |
|
Yeah, I was right: the official QASM2.0 spec says that the current definition we have is the correct one. See page 12 of https://arxiv.org/pdf/1707.03429.pdf. By the way, I do agree that your definition is more sensible, and it is also how I think of a crz gate. But that is not how it is defined in the spec. |
|
Maybe it's the version of OpenQASM that causes the misunderstanding! According to OpenQASM3.0, the |
|
Yes QASM3.0 seems to use your definition. Which is annoying because now the specs aren't compatible apparently? |
… in pyzx.
The OpenQASM 2 spec defines the following:
```
gate rz(phi) a { u1(phi) a; }
gate crz(lambda) a,b
{
u1(lambda/2) b;
cx a,b;
u1(-lambda/2) b;
cx a,b;
}
gate cu1(lambda) a,b
{
u1(lambda/2) a;
cx a,b;
u1(-lambda/2) b;
cx a,b;
u1(lambda/2) b;
}
```
pyzx's `ZPhase` gate and `CRZ` gates are equivalent to, respectively, OpenQASM 2's `rz` (a synonyn for `u1`) and `crz` gates. Confusingly, the `crz` gate is not simply a controlled version of the `rz` gate; that gate is named `cu1`.
Meanwhile, the OpenQASM 3 spec defines:
```
gate rz(λ) a { gphase(-λ/2); U(0, 0, λ) a; }
gate crz(θ) a, b { ctrl @ rz(θ) a, b; }
```
This has the advantage that `crz` is now simply a controlled version of `rz`, at the expense of changing the meaning of `rz` so that it no longer matches `u1`/`ZPhase`.
This difference in conventions leads to user confusion and coding errors, as seen in issues zxcalc#102 and zxcalc#103. As well, changing the behaviour of `qasmparser` depending on the version of OpenQASM used is a blocker for issue zxcalc#116.
… in pyzx.
The OpenQASM 2 spec defines the following:
```
gate rz(phi) a { u1(phi) a; }
gate crz(lambda) a,b
{
u1(lambda/2) b;
cx a,b;
u1(-lambda/2) b;
cx a,b;
}
gate cu1(lambda) a,b
{
u1(lambda/2) a;
cx a,b;
u1(-lambda/2) b;
cx a,b;
u1(lambda/2) b;
}
```
pyzx's `ZPhase` gate and `CRZ` gates are equivalent to, respectively, OpenQASM 2's `rz` (a synonym for `u1`) and `crz` gates. Confusingly, the `crz` gate is not simply a controlled version of the `rz` gate; that gate is named `cu1`.
Meanwhile, the OpenQASM 3 spec defines:
```
gate rz(λ) a { gphase(-λ/2); U(0, 0, λ) a; }
gate crz(θ) a, b { ctrl @ rz(θ) a, b; }
```
This has the advantage that `crz` is now simply a controlled version of `rz`, at the expense of changing the meaning of `rz` so that it no longer matches `u1`/`ZPhase`.
This difference in conventions leads to user confusion and coding errors, as seen in issues zxcalc#102 and zxcalc#103. As well, changing the behaviour of `qasmparser` depending on the version of OpenQASM used is a blocker for issue zxcalc#116.
|
The definitions in the OpenQASM 3 header file are: If I understand this correctly, |
|
PR #156 specifically adds a test to document the behaviour of |
|
This issue can be closed. |
Hi, according to
pyzx/circuit/gates.py, your decomposition and tensor of crz(pi/2) is shown below:However, the correct form of crz(pi/2) should be like this:
Refer to this transformation:
May I submit a PR to modify this bug?
The text was updated successfully, but these errors were encountered: