search

taijusti / sleep_apnea

0 stars 0 forks source link

Implement global manager #4

Closed taijusti closed 9 years ago

taijusti commented 9 years ago

how is this going patrick? do you need help from ibrahim/me?

taijusti commented 9 years ago

i'm trying to implement this as a separate entity. targetting tomorrow evening for completion. if you have a solution already implemented please let me know ASAP so i can work on something else.

taijusti commented 9 years ago

currently have host calling device. im running the whole thing against ibrahim's smo. i had to port over his smo to use our variable types. for some reason it's not returning. have not even invoked the hw module.

****if anyone has spare cycles, i think im going to need help on this.

p-wu commented 9 years ago

what kind of help do you need, I can look at it over the next few days

2015-05-06 18:11 GMT-04:00 Justin notifications@github.com:

currently have host calling device. im running the whole thing against ibrahim's smo. i had to port over his smo to use our variable types. for some reason it's not returning. have not even invoked the hw module.

****if anyone has spare cycles, i think im going to need help on this.

— Reply to this email directly or view it on GitHub https://github.com/taijusti/sleep_apnea/issues/4#issuecomment-99626008.

taijusti commented 9 years ago

today i just relaxed. i'll continue debugging tomorrow.

take a look at host_tb.cpp and host.cpp--right now i've ported over smo.c into host_tb.cpp. and made host call device directly. basically, host_tb currently tests everything--checks if our hw solution comes up with the same alphas as the sw. tomorrow i'm going to continue debugging--just stepping through the sw and hw implementations and checking at every step if the results line up within a few percent. i think this is going to take me a long time, so i was thinking if someone could do the same thing in parallel, it would go faster.

taijusti commented 9 years ago

ok, i have it at the point of each iteration looks good for a while, then at some point the bias 'b' grows unbounded, causing the errors and alphas to also become weird. i'm having difficulty solving this bug so if anyone has spare cycles, additional eyes to look at this would be helpful.

ghost commented 9 years ago

Bug found and exterminated :dancer: