# Repository: https://gitlab.com/quantify-os/quantify-scheduler
# Licensed according to the LICENCE file on the main branch
"""QASM program class for Qblox backend."""
from __future__ import annotations
from contextlib import contextmanager
from typing import (
TYPE_CHECKING,
Generator,
Hashable,
Iterator,
Sequence,
)
import numpy as np
from columnar import columnar
from columnar.exceptions import TableOverflowError
from quantify_scheduler.backends.qblox import constants, helpers, q1asm_instructions
from quantify_scheduler.backends.qblox.conditional import (
ConditionalManager,
)
if TYPE_CHECKING:
from quantify_scheduler.backends.qblox.operation_handling.base import (
IOperationStrategy,
)
from quantify_scheduler.backends.qblox.operation_handling.virtual import (
ConditionalStrategy,
)
from quantify_scheduler.backends.qblox.register_manager import RegisterManager
from quantify_scheduler.backends.types.qblox import (
OpInfo,
StaticHardwareProperties,
)
from quantify_scheduler.schedules.schedule import AcquisitionMetadata
[docs]
def get_marker_binary(marker_setting: str | int) -> int:
"""
Sets the marker from a string representing a binary number. Each digit
corresponds to a marker e.g. '0010' sets the second marker to True.
If the marker setting is already an integer, the function checks whether it is a
4-bit integer.
Parameters
----------
marker_setting
The string representing a binary number.
"""
if isinstance(marker_setting, str):
if len(marker_setting) != 4:
raise ValueError("4 marker values are expected.")
return int(marker_setting, 2)
else:
if marker_setting > 0b1111:
raise ValueError(f"Invalid marker setting: {marker_setting=}.")
return marker_setting
[docs]
class QASMProgram:
"""
Class that holds the compiled Q1ASM program that is to be executed by the sequencer.
Apart from this the class holds some convenience functions that auto generate
certain instructions with parameters, as well as update the elapsed time.
Parameters
----------
static_hw_properties
Dataclass holding the properties of the hardware that this program is to be
played on.
register_manager
The register manager that keeps track of the occupied/available registers.
align_fields
If True, make QASM program more human-readable by aligning its fields.
acq_metadata
Provides a summary of the used acquisition protocol, bin mode, acquisition
channels, acquisition indices per channel, and repetitions.
"""
def __init__(
self,
static_hw_properties: StaticHardwareProperties,
register_manager: RegisterManager,
align_fields: bool,
acq_metadata: AcquisitionMetadata | None,
) -> None:
[docs]
self.static_hw_properties = static_hw_properties
"""Dataclass holding the properties of the hardware that this program is to be
played on."""
[docs]
self.register_manager = register_manager
"""The register manager that keeps track of the occupied/available registers."""
[docs]
self.align_fields = align_fields
"""If true, all labels, instructions, arguments and comments
in the string representation of the program are printed on the same indention level.
This worsens performance."""
"""Provides a summary of the used acquisition protocol, bin mode, acquisition
channels, acquisition indices per channel, and repetitions."""
[docs]
self.time_last_acquisition_triggered: int | None = None
"""Time on which the last acquisition was triggered. Is ``None`` if no previous
acquisition was triggered."""
[docs]
self.time_last_pulse_triggered: int | None = None
"""Time on which the last operation was triggered. Is ``None`` if no previous
operation was triggered."""
[docs]
self.instructions: list[list] = list()
"""A list containing the instructions added to the program. The instructions
added are in turn a list of the instruction string with arguments."""
[docs]
self.conditional_manager = ConditionalManager()
"""The conditional manager that keeps track of the conditionals."""
[docs]
self._lock_conditional: bool = False
"""A lock to prevent nested conditionals."""
[docs]
self._elapsed_times_in_loops: list[int] = [0]
"""The time elapsed in its current form.
This is used to keep track of the overall and nested loop timing and necessary waits."""
@property
[docs]
def elapsed_time(self) -> int:
"""
Current elapsed time of all the instructions in ns.
It needs to be manually adjusted after each modifications of the QASM program.
If the QASM program is in a loop,
only one repetition's worth of elapsed time should be registered.
After a loop is ended, ``QASMProgram`` will automatically adjust the correct
elapsed time with all repetitions.
"""
return sum(self._elapsed_times_in_loops)
@elapsed_time.setter
def elapsed_time(self, value: int) -> None:
difference: int = value - self.elapsed_time
self._elapsed_times_in_loops[-1] += difference
[docs]
def _find_qblox_acq_index(self, acq_channel: Hashable) -> int:
"""
Finds the Qblox acq_index corresponding to acq_channel
in the acq_metadata.
"""
# This function is a temporary solution.
# Proper solution: SE-298.
assert self.acq_metadata is not None
for (
qblox_acq_index,
acq_channel_metadata,
) in self.acq_metadata.acq_channels_metadata.items():
if acq_channel_metadata.acq_channel == acq_channel:
return qblox_acq_index
raise ValueError(f"Qblox acquisition index not found for {acq_channel=}.")
@staticmethod
[docs]
def get_instruction_as_list(
instruction: str,
*args: int | str,
label: str | None = None,
comment: str | None = None,
) -> list[str | int]:
"""
Takes an instruction with arguments, label and comment and turns it into the
list required by the class.
Parameters
----------
instruction
The instruction to use. This should be one specified in
:mod:`~quantify_scheduler.backends.qblox.q1asm_instructions`
or the assembler will raise an exception.
args
Arguments to be passed.
label
Adds a label to the line. Used for jumps and loops.
comment
Optionally add a comment to the instruction.
Returns
-------
:
List that contains all the passed information in the valid format for the
program.
Raises
------
SyntaxError
More arguments passed than the sequencer allows.
"""
instr_args = ",".join(str(arg) for arg in args)
label_str = f"{label}:" if label is not None else ""
comment_str = f"# {comment}" if comment is not None else ""
return [label_str, instruction, instr_args, comment_str]
[docs]
def emit(self, *args, **kwargs) -> list[str | int]:
"""
Wrapper around the ``get_instruction_as_list`` which adds it to this program.
Parameters
----------
args
All arguments to pass to `get_instruction_as_list`.
**kwargs
All keyword arguments to pass to `get_instruction_as_list`.
Returns
-------
:
A list containing instructions.
"""
# Translating the acquisition channel to qblox acquisition index
# is intended as a temporary solution.
# TODO: Proper solution: SE-298.
instruction = args[0]
if self.acq_metadata and (
instruction
in (
q1asm_instructions.ACQUIRE,
q1asm_instructions.ACQUIRE_TTL,
q1asm_instructions.ACQUIRE_WEIGHED,
)
):
args = list(args)
args[1] = self._find_qblox_acq_index(acq_channel=args[1])
self.instructions.append(self.get_instruction_as_list(*args, **kwargs))
return self.instructions[-1]
# --- QOL functions -----
[docs]
def set_latch(self, op_strategies: Sequence[IOperationStrategy]) -> None:
"""
Set the latch that is needed for conditional playback.
This assumes that the latch address is present inside the pulses'
`operation_info`. If no latch address is found, nothing is emitted.
Parameters
----------
op_strategies
The op_strategies containing the pulses to search the latch address in.
"""
for op_strategy in op_strategies:
op_info = op_strategy.operation_info
if not op_info.is_acquisition and (
op_info.data.get("feedback_trigger_address") is not None
):
self.emit(q1asm_instructions.FEEDBACK_TRIGGER_EN, 1, 4)
return
[docs]
def auto_wait(
self,
wait_time: int,
count_as_elapsed_time: bool = True,
comment: str | None = None,
) -> None:
"""
Automatically emits a correct wait command. If the wait time is longer than
allowed by the sequencer it correctly breaks it up into multiple wait
instructions. If the number of wait instructions is too high (>4), a loop will
be used.
Parameters
----------
wait_time
Time to wait in ns.
count_as_elapsed_time
If true, this wait time is taken into account when keeping track of timing.
Otherwise, the wait instructions are added but this wait time is ignored in
the timing calculations in the rest of the program.
comment
Allows to override the default comment.
Raises
------
ValueError
If ``wait_time <= 0``.
"""
if wait_time == 0:
return
if wait_time < 0:
raise ValueError(
f"Invalid wait time. Attempting to wait "
f"for {wait_time} ns at t={self.elapsed_time}"
f" ns."
)
comment = comment if comment else f"auto generated wait ({wait_time} ns)"
if wait_time > constants.IMMEDIATE_MAX_WAIT_TIME:
repetitions = wait_time // constants.IMMEDIATE_MAX_WAIT_TIME
# number of instructions where it becomes worthwhile to use a loop.
instr_number_using_loop = 4
if repetitions > instr_number_using_loop:
loop_label = f"wait{len(self.instructions)}"
with self.loop(loop_label, repetitions):
self.emit(
q1asm_instructions.WAIT,
constants.IMMEDIATE_MAX_WAIT_TIME,
comment=comment,
)
if count_as_elapsed_time:
self.elapsed_time += constants.IMMEDIATE_MAX_WAIT_TIME
self.conditional_manager.num_real_time_instructions += 1
else:
for _ in range(repetitions):
self.emit(
q1asm_instructions.WAIT,
constants.IMMEDIATE_MAX_WAIT_TIME,
comment=comment,
)
if count_as_elapsed_time:
self.elapsed_time += constants.IMMEDIATE_MAX_WAIT_TIME
self.conditional_manager.num_real_time_instructions += 1
time_left = wait_time % constants.IMMEDIATE_MAX_WAIT_TIME
else:
time_left = int(wait_time)
if time_left > 0:
self.emit(
q1asm_instructions.WAIT,
time_left,
comment=comment,
)
if count_as_elapsed_time:
self.elapsed_time += time_left
self.conditional_manager.num_real_time_instructions += 1
[docs]
def wait_till_start_operation(self, operation: OpInfo) -> None:
"""
Waits until the start of a pulse or acquisition.
Parameters
----------
operation
The pulse or acquisition that we want to wait for.
Raises
------
ValueError
If wait time < 0.
"""
start_time = helpers.to_grid_time(operation.timing)
wait_time = start_time - self.elapsed_time
if wait_time > 0:
self.auto_wait(wait_time)
elif wait_time < 0 and operation.is_parameter_instruction:
raise ValueError(
f"Invalid timing. {repr(operation)} cannot be started at this order or time. "
f"Please try to reorder your operations by adding this operation "
"before any other operation (possibly at the same time) that happens at that time."
)
elif wait_time < 0 and operation.name != "IdlePulse":
# The idle pulse is a no operation, if any other operation
# is simultaneously running, it is allowed.
raise ValueError(
f"Invalid timing. Attempting to wait for {wait_time} "
f"ns before {repr(operation)}. Please note that a wait time of at least"
f" {constants.MIN_TIME_BETWEEN_OPERATIONS} ns is required between "
f"operations.\nAre multiple operations being started at the same time?"
)
[docs]
def set_gain_from_amplitude(
self,
amplitude_path_I: float, # noqa N803 - uppercase in name
amplitude_path_Q: float, # noqa N803 - uppercase in name
operation: OpInfo | None,
) -> None:
"""
Sets the gain such that a 1.0 in waveform memory corresponds to the full awg gain.
Parameters
----------
amplitude_path_I
Voltage to set on path_I.
amplitude_path_Q
Voltage to set on path_Q.
operation
The operation for which this is done. Used for the exception messages.
"""
awg_gain_path_I_immediate = self.expand_awg_from_normalised_range(
amplitude_path_I,
constants.IMMEDIATE_SZ_GAIN,
"awg_gain_0",
operation,
)
awg_gain_path_Q_immediate = self.expand_awg_from_normalised_range(
amplitude_path_Q,
constants.IMMEDIATE_SZ_GAIN,
"awg_gain_1",
operation,
)
comment = f"setting gain for {operation.name}" if operation else ""
self.emit(
q1asm_instructions.SET_AWG_GAIN,
awg_gain_path_I_immediate,
awg_gain_path_Q_immediate,
comment=comment,
)
@staticmethod
[docs]
def expand_awg_from_normalised_range(
val: float,
immediate_size: int,
param: str | None = None,
operation: OpInfo | None = None,
) -> float:
"""
Takes the value of an awg gain or offset parameter
in normalized form (abs(param) <= 1.0),
and expands it to an integer
in the appropriate range required by the sequencer.
Parameters
----------
val
The value of the parameter to expand.
immediate_size
The size of the immediate. Used to find the max int value.
param
The name of the parameter, to make a possible exception message more
descriptive.
operation
The operation this value is expanded for, to make a possible exception
message more descriptive.
Returns
-------
:
The expanded value of the parameter.
Raises
------
ValueError
Parameter is not in the normalized range.
"""
if np.abs(val) > 1.0:
raise ValueError(
f"{param} is set to {val}. Parameter must be in the range "
f"-1.0 <= {param} <= 1.0 for {repr(operation)}."
)
max_gain = immediate_size // 2
return max(-max_gain, min(round(val * max_gain), max_gain - 1))
def __str__(self) -> str:
"""
Returns a string representation of the program. The sequencer expects the program
to be such a string.
The conversion to str is done using `columnar`, which expects a list of lists,
and turns it into a string with rows and columns corresponding to those lists.
Returns
-------
:
The string representation of the program.
"""
if self.align_fields:
try:
instructions_str = columnar(
self.instructions, headers=None, no_borders=True, wrap_max=0
)
# running in a sphinx environment can trigger a TableOverFlowError
except TableOverflowError:
instructions_str = columnar(
self.instructions, headers=None, no_borders=True, terminal_width=120
)
# columnar inserts a newline before all the the instruction rows
return instructions_str.split("\n", 1)[1]
else:
return "\n".join(" ".join(instruction) for instruction in self.instructions) + "\n"
@contextmanager
[docs]
def conditional(self, operation: ConditionalStrategy) -> Generator[None, None, None]:
"""
Defines a conditional block in the QASM program.
When this context manager is entered/exited it will insert additional
``set_cond`` QASM instructions in the program that specify the
conditionality of a set of instructions.
The following example should make it clear what is happening.
.. code-block:: none
set_cond set_enable=1, mask=0, operator=OR, else_duration=4
<50 ns duration of instructions that contains 3 real time instructions>
set_cond set_enable=1, mask=0, operator=NOR, else_duration=4
wait 50-3*4+4 = 42 ns # adding an additional 4 ns to make math work out
set_cond set_enable=0, mask=0, operator=OR, else_duration=4
The `else_duration` is the wait time per real time instruction in the
conditional block. If a trigger happened, the first block runs normally for
50 ns, the second block runs for 4 ns. If there is no trigger, the first
block runs for 3*4 = 12 ns, second block for 42 ns. So the duration in
both cases is 42 ns. Note that `set_cond` itself has zero duration.
The exact values that need to be passed to the ``set_cond``
instructions are determined while the qasm program is generated with the
help of
:class:`~quantify_scheduler.backends.qblox.conditional.FeedbackTriggerCondition`
and
:class:`~quantify_scheduler.backends.qblox.conditional.ConditionalManager`.
Parameters
----------
operation: ConditionalStrategy
The conditional strategy that defines the start of a conditional block.
"""
trigger_condition = operation.trigger_condition
if self._lock_conditional:
raise RuntimeError(
"Nested conditional playback inside schedules is not supported by "
f"the Qblox backend. "
f"This error is caused by the following operation strategy:\n{operation}."
)
self._lock_conditional = True
# This instruction will be replaced when the context manager exits the
# conditional block.
self.emit(
q1asm_instructions.FEEDBACK_SET_COND,
int(trigger_condition.enable),
trigger_condition.mask,
trigger_condition.operator.value,
constants.MIN_TIME_BETWEEN_OPERATIONS,
comment="start conditional playback",
)
self.conditional_manager.reset()
self.conditional_manager.start_time = self.elapsed_time
yield
# When the context manager exits, add an else branch to fill the correct wait time
# and add a stop conditional playback and
# replace the initial FEEDBACK_SET_COND instruction.
self.conditional_manager.end_time = self.elapsed_time
self.emit(
q1asm_instructions.FEEDBACK_SET_COND,
int(trigger_condition.enable),
trigger_condition.mask,
(~trigger_condition.operator).value,
constants.MIN_TIME_BETWEEN_OPERATIONS,
comment="else wait",
)
# autowait now adds an additional duration to elapsed time that we need to compensate.
duration = (
self.conditional_manager.duration
- constants.MIN_TIME_BETWEEN_OPERATIONS
* self.conditional_manager.num_real_time_instructions
+ constants.MIN_TIME_BETWEEN_OPERATIONS
)
self.auto_wait(duration, count_as_elapsed_time=False)
self.emit(
q1asm_instructions.FEEDBACK_SET_COND,
0,
0,
0,
0,
comment="stop conditional playback",
)
self.elapsed_time += constants.MIN_TIME_BETWEEN_OPERATIONS
self.conditional_manager.reset()
self._lock_conditional = False
@contextmanager
[docs]
def loop(self, label: str, repetitions: int = 1) -> Generator[str, None, None]:
"""
Defines a context manager that can be used to generate a loop in the QASM
program.
Parameters
----------
label
The label to use for the jump.
repetitions
The amount of iterations to perform.
Yields
------
:
The register used as loop counter.
Examples
--------
This adds a loop to the program that loops 10 times over a wait of 100 ns.
.. jupyter-execute::
from quantify_scheduler.backends.qblox.qasm_program import QASMProgram
from quantify_scheduler.backends.qblox.instrument_compilers import QCMCompiler
from quantify_scheduler.backends.qblox import register_manager, constants
from quantify_scheduler.backends.types.qblox import (
StaticAnalogModuleProperties,
BoundedParameter
)
qasm = QASMProgram(
static_hw_properties=QCMCompiler.static_hw_properties,
register_manager=register_manager.RegisterManager(),
align_fields=True,
acq_metadata=None,
)
with qasm.loop(label="repeat", repetitions=10):
qasm.auto_wait(100)
qasm.instructions
"""
register = self.register_manager.allocate_register()
comment = f"iterator for loop with label {label}"
self._elapsed_times_in_loops.append(0)
self.emit(q1asm_instructions.MOVE, repetitions, register, comment=comment)
self.emit(q1asm_instructions.NEW_LINE, label=label)
yield register
self.emit(q1asm_instructions.LOOP, register, f"@{label}")
self.register_manager.free_register(register)
last_elapsed_time = self._elapsed_times_in_loops.pop()
self._elapsed_times_in_loops[-1] += last_elapsed_time * repetitions
@contextmanager
[docs]
def temp_registers(self, amount: int = 1) -> Iterator[list[str]]:
"""
Context manager for using a register temporarily. Frees up the register
afterwards.
Parameters
----------
amount
The amount of registers to temporarily use.
Yields
------
:
Either a single register or a list of registers.
"""
registers: list[str] = list()
for _ in range(amount):
registers.append(self.register_manager.allocate_register())
yield registers
for reg in registers:
self.register_manager.free_register(reg)
