nkanazawa1989 / qiskit Goto Github PK

Information

• Qiskit Terra version:
• Python version:
• Operating system:

What is the current behavior?

assemble_schedule uses Schedule.channels and cannot generate lo_frequencies.

Steps to reproduce the problem

What is the expected behavior?

Suggested solutions

use new .device structure

We have to include mem_slots in DeviceSpecification.__init__().

What is the expected enhancement?

This is our PR plan to master.

Guides:

• We make each PR step by step from 1
• If the parent PR is modified during the review, a child should merge terra-master after the parent is merged (to reflect stable code). If the change is critical to develop children branches, we can merge parent branch itself. This process propagate from child to grandchild.
• If you continue to develop during the terra owner's review, you should make PR to below 4 branches or terra-master.
• Test branch accepts these changes only by pull upstream master
• Confirm linting before make PR to master
• All PRs should refer Thomas's issue for API #1713
• Before stating PR, previous PR for frontend should be closed #1901

New:

• Child branch will be rebased by parent to serialize commit history.

1. channels module

https://github.com/nkanazawa1989/qiskit-terra/tree/openpulse-channels
preview #27
reference branch: pulse_channel
This PR adds qiskit/pulse/channels

WIP

• fix DeviceSpecification constructor #23
• proper u_channel creation #14

2. commands rework

https://github.com/nkanazawa1989/qiskit-terra/tree/openpulse-commands-rework
preview #31
reference branch: openpulse-channels
This PR apply reworks for command instruction

status: WIP

• add unittest for command instruction #28

3. schedule module

https://github.com/nkanazawa1989/qiskit-terra/tree/openpulse-schedule
preview #32
reference branch: openpulse-commands-rework
This PR adds qiskit/pulse/schedules

status: under Thomas's review

4. assembler module

https://github.com/nkanazawa1989/qiskit-terra/tree/openpulse-assembler
preview #34
reference branch: openpulse-schedule
This PR update qiskit/compiler/assembler and execute

status: WIP

• add proper LO config #26
• add unittest for LO config

What is the expected enhancement?

In lo-rework branch, there are unrelated commits to the branch name because of some bug fix.
We can apply these commits to proper PR branches in #29

As far as I know:
e699f00
80983fa
48714f8

What is the expected enhancement?

Currently the backend returns n_uchannels = 0.
DeviceSpecification is better to accept different numbers of n_qubit and n_uchannels .

This situation could also happen when a system comprises only 1 qubit (no need of CR).

What is the expected enhancement?

We don't have unittest for

AcquireInstruction
FrameChangeInstruction
PersistentValueInstruction
DriveInstruction

What is the expected enhancement?

ConfiguredSchedule is better not to be exposed. Adding a factory method to Schedule will increase usability.

when bind LO config to each schedule

for setting in settings:
    schedule = Schedule()
    ....
    schedules.extend(schedule.configured(lo_config))

when bind a set of LO configs to a single schedule

schedule = Schedule()
....
schedules.extend([schedule.configured(lo_config) for lo_config in lo_configs])

We can use same interface

execute_schedule(schedules, backend, **kwargs)

Current Acquire command (in #12) requires memory slots as its operands.

Acquire(10)(device.q[0], device.mem[0], device.c[1])

But one qubit has exactly one acquire and one memory slot so far.
And I guess it will not change even in the future.
If so, I think a memory slot should belong to an acquire channel and be hidden from users.

Acquire(10)(device.q[0], device.c[1])

(Note: How to get results also may be changed if this change is accepted.)

Information

• Qiskit Terra version:
• Python version:
• Operating system:

What is the current behavior?

9117541 merged old assembler.

Steps to reproduce the problem

What is the expected behavior?

#8 should be used.

Suggested solutions