<lofty[m]>
ysionneau: so, I uncovered something interesting while researching this
<lofty[m]>
the default seed is 3549216002486605715, which is more meaningfully 0x3141592653589793.
<lofty[m]>
however, nextpnr runs the rng a few times on a seed to get sufficient randomness when seeded, but not by default
<lofty[m]>
so --seed 3549216002486605715 will not give you the same result as a run without --seed
<lofty[m]>
but anyway, a run without --seed is deterministic.
<Adrien[m]>
Thanks lofty, good to know !
<janrinze>
lofty[m]: That's good to know, and it's a nicely chosen seed :-D the digits of Pi in 64 bits hex.
krispaul has quit [Read error: Connection reset by peer]
krispaul has joined #yosys
<ysionneau>
thanks lofty[m] for the research
<ysionneau>
I didn't even notice that the result was deterministic o_o
<Adrien[m]>
I'm not sure it is strictly deterministic 🤔
<Adrien[m]>
The algorithms probably are, but their implementations (per-uarch contributions, etc) could well use containers sensitive to pointer values, so to virtual memory allocation, which is not reproducible. I have not checked that everywhere in the code, though.
buhman has quit [Ping timeout: 244 seconds]
mithro has quit [Ping timeout: 245 seconds]
buhman has joined #yosys
sorear has quit [Ping timeout: 245 seconds]
mithro has joined #yosys
sorear has joined #yosys
<lofty[m]>
which is why we generally make a point of not storing pointers as keys.
<Adrien[m]>
I wasn't aware of this policy, nice, thanks
tpb has quit [Read error: Connection reset by peer]