I remember trying to build Anki from source code a long time ago and getting some weird error. Let’s see what happens this time.
EDIT: welp
Well, I’ll reiterate my point again in case I wasn’t clear enough. It’s confusing currently that we are getting different results with the inputs only slightly changed, that too with days to simulate.
I wouldn’t bet against people not getting confused when they set this number to different values and they see graphs that look very different from each other. So clearly, this is an issue demanding something to be done.
Oh, that’s what you mean.
I assume it’s because the PRNG seed is somehow based on those inputs, but I’ve not double checked.
(not related to my earlier issue)
Weird indeed. I never experienced something like that on linux. @rossgb do you happen to know how to solve this?
I was part misremembering but I was part right that you were confused at getting different results when changing that value.
Edit: folks, let’s prevent this thread becoming “helping Expertium build a branch from luc’s anki fork”. discord/pms are better for that.
I swear I didnt infect your anki XD
Anyway, I think we need to decide whether it’s worth it or no. Which depends on the amount of time it takes to run 1 simulation vs 10.
Can you benchmark it, please? I mean, obviously the time will be different on different hardware, but having an estimate at all is better than having none.
@L.M.Sherlock is the seed always the same for any simulator settings? That’s probably why we get very little variance.
Yep. Only CMRR uses different seeds.
Sorry I’m late but I believe if you run the simulation 10x with 10 different seeds it would just take pretty much exactly 10x as long as it currently does unless there were some smart way of doing it that I’m missing.
On my fast computer, 3650 days+40k new cards can already take a second, and it’s going to be slower on a slower computer. If we run multiple simulations in parallel, we may not have a 10x slowdown, but that will depend on the number of CPU cores available, and the amount of work that can be parallelised.