Windows getloadavg moving averages are incorrect

[bolterr]

Platform

• Windows
• 5.6.3

Bug description
Thanks for implementing getloadavg for windows in psutil/psutil/arch/windows/wmi.c

I think that there was a slipup with the factors:

#define LOADAVG_FACTOR_1F 0.9200444146293232478931553241
#define LOADAVG_FACTOR_5F 0.6592406302004437462547604110
#define LOADAVG_FACTOR_15F 0.2865047968601901003248854266

LOADAVG_FACTOR_1F is good but I think LOADAVG_FACTOR_5F and LOADAVG_FACTOR_15F should be as follows:

#define LOADAVG_FACTOR_5F 0.9834714538216174894737477501 // 1/exp(5/300)
#define LOADAVG_FACTOR_15F 0.9944598480048967508795473395 // 1/exp(5/900)

This better matches linux/sched/loadavg.h:

LOADAVG_FACTOR_1F = 1884/2048= 0.92...
LOADAVG_FACTOR_5F = 2014/2048= 0.98...
LOADAVG_FACTOR_14F = 2037/2048= 0.99...

Sorry, I would post a pr if I could.

[giampaolo]

@ammaraskar do you agree with the analysis?

[ammaraskar]

Aah yeah, this seems to be right. The calculations and wolfram links are provided here: #1484 (comment)

were wrong due to the brackets, it should have been 1/exp(5 / (60*5)) but instead I used 1/exp(5 / 60*5) which gives a completely different result.

Sorry about that, I'll make a PR to fix it.

[giampaolo]

Thanks man.