<GitHub162>
[artiq] dhslichter commented on issue #40: I don't want to delay 3.0 any further with this, and as @sbourdeauducq points out there are many more issues to be addressed with higher priority for users. While figuring out the latencies is an important issue to address in the fullness of time, it is also something that can be addressed simply, with the current ARTIQ, by end users -- you just wrap pulse/on/off methods with something that looks at the
sb0 has quit [Quit: Leaving]
<mithro>
rjo: I'm fairly confident the problem is a combinational loop in Ishan's code - is there a good way to figure out where the loop is happening?
<sb0>
mithro, not really; though it would be possible to add a combinatorial loop tracer to the migen simulator
<sb0>
do you need a GSoC project? :)
sb0 has quit [Quit: Leaving]
<rjo>
builing the graph of signal dependencies would allow you to detect loops before even executing. and it would speed up the simulator because you only need to execute a part of the graph on each delta cycle.
<GitHub50>
[artiq] jordens commented on issue #40: That's the scheme we would advocate as well. It would just be nice to have examples and some skeleton code that shows how it can be implemented. https://github.com/m-labs/artiq/issues/40#issuecomment-307341337
rohitksingh_wor1 has quit [Read error: Connection reset by peer]
rohitksingh has joined #m-labs
sb0 has joined #m-labs
<sb0>
rjo, there are some nasty edge cases with that, e.g. with partial assignments to a slice of a signal, but yes
<rjo>
sb0: yes. if that really turns out to be a challenge one can still relax the definition of that graph. it doesn't need to be "optimal". just reducing the set of "potential" sensitivity signals from "all" would help already.
<GitHub132>
[artiq] jbqubit commented on issue #40: > you just wrap pulse/on/off methods with something that looks at the channel being operated on and adds/subtracts a deterministic amount looked up from a calibration dataset... https://github.com/m-labs/artiq/issues/40#issuecomment-307417907
<GitHub65>
[artiq] dhslichter commented on issue #40: Why does the end user wrapping this result in a "terrible mess" in experiment files? Just create a subclass called `TTLOutLC(TTLOut)`, and implement the modified versions of the methods as needed. Then just use `TTLOutLC` instead of `TTLOut` in your experiment code. Seems incredibly clean and simple to me. Now, if you set latencies to be wacky numbers you may get some tricky corner cases (such as the a
<GitHub158>
[artiq] dhslichter commented on issue #40: Why does the end user wrapping this result in a "terrible mess" in experiment files? Just create a subclass called `TTLOutLC(TTLOut)`, and implement the modified versions of the methods as needed. Then just use `TTLOutLC` instead of `TTLOut` in your experiment code. Seems incredibly clean and simple to me. Now, if you set latencies to be wacky numbers you may get some tricky corner cases (such as the
<GitHub128>
[artiq] jbqubit commented on issue #40: So rather than modifying six lines of code in the ARTIQ code base each lab is now creating a new class, overloading 7 members of TTLOut and debugging it. In TTL PHY repeat for ... https://github.com/m-labs/artiq/issues/40#issuecomment-307430994
<GitHub76>
[artiq] jbqubit commented on issue #40: Let's discuss a first order solution that addresses gross latency first. 1) PHY-to-PHY latency (eg TTL to SWAG, currently couple micro-second) 2) cable length. ... https://github.com/m-labs/artiq/issues/40#issuecomment-307478066
<GitHub127>
[artiq] jordens commented on issue #745: I have understood that I should look at the movie. But what in the movie shows "RTIO timing error between SAWG and TTLOut"? If you want decent bug triage, please provide a decent description of the problem.... https://github.com/m-labs/artiq/issues/745#issuecomment-307479547
<GitHub27>
[artiq] jordens commented on issue #745: Are you sure you are not just seeing a beating between the timebase of your oscilloscope and the RTIO/DAC frequency? Again, I don't know what I should be looking at in that movie. https://github.com/m-labs/artiq/issues/745#issuecomment-307479907