Once the Sequencer ack's a line, the Parser starts preparing the
next one. This includes jumping through the frame table if necessary.
To stall the Parser while the Sequencer executes the last line of a
frame and to ensure that the frame select lines can be set up and their
sampling is synchronized to a trigger, we add a triggered stall line
at the end of the frame.
When that line is triggered the Parser jumps through the table and starts
parsing the first line of the next frame. We let the duration of this
last stall line be 10 cycles (200ns@50MHz) to be able to distinguish this
sampling of the frame select lines from the triggering of the first line
in the next frame.
frame f
parser n f 0
stb __---________---___
trigger ___----_______----_
ack ____-__________-___
sequencer n-1 n 0
* "trigger" now means that the corresponding line will only start
once the trigger line is high.
* "jump" is implicit as the last line in a frame must jump back.
* spline coefficients are now compensated for finite time step size
You can always (under posix) use #!/usr/bin/env artiq_run as
shebang for experiments and make them executable.
Now, you can also do this (portable):
if __name__ == "__main__":
from artiq.frontend.artiq_run import run
run()
to make an experiment executable. The CLI options are all inherited.
Also:
* removed --elf: can be inferred from filename
* did some refactoring and cleanup
* use logging for all messages, except the result printing (use -v to get
parameter changes and dummy scheduler actions)
Explicitly having to disable unittests that require hardware
(ARTIQ_NO_HARDWARE) is cumbersome.
There is not even a sensible default for the
device or serial number of the devices requiring additional
variables (ARTIQ_LDA_DEVICE etc).
This patch reverts the logic by skipping unittests that
can not automatically determine whether the required hardware
is present and where it is.
