Re: [PATCH] sched: fix/optimise calculation of weight-inverse

["Stephan Bärwolf" <stephan.baerwolf@tu-ilmenau.de>]

>I think what we wanted to do was minimize the error:
>  err = weight - inv*WMULT_CONST

Hi all,
thanks for your fast response.
I think what you mean is:

err = | WMULT_CONST - inv*weight |  --> min

right?

But the following table(s) shows the difference in the inverses:
(assuming WMULT_CONST = 2**32 as on x64)

weight                                    oldway
inv_weigth                 new inv_weight

1                                             2147483649 (=1+2**31)    
4294967295 [=(2**32)-1]    
2                                            
1431655766                         2147483647
3 (= WEIGHT_IDLEPRIO)        1073741824                         1431655765
15 (= nice 19)                        
268435456                           286331153
1024 (=nice 0)                       
4190212                               4194304
...
...
88761 (=nice -20)                  
48387                                   48388

----

weight                                    err
oldway                             err newway

1                                             2147483647 (=2**31 - 1)   
1     
2                                             1431655764
                        2
3 (= WEIGHT_IDLEPRIO)        1073741824                         1
15 (= nice 19)                         268435456                           1
1024 (=nice 0)                       
4190208                               0
...
...
88761 (=nice -20)                   88789
                                  28



Thus the "err" of the old way can become very large (up to about  2^31 ).
And of course the new error increases with increasing weight, it is
still alway
smaller than the oldway err (because oldway inv converts to newway inv)...


regards Stephan

--
Dipl.-Inf. Stephan Bärwolf
Ilmenau University of Technology, Integrated Communication Systems Group
Phone: +49 (0)3677 69 2821,  FAX: +49 (0)3677 69 1614
Email: stephan.baerwolf@tu-ilmenau.de,
Web: http://www.tu-ilmenau.de/iks