The v2 Rotary Interactivity Favor Scheduler

[kernelOfTruth]

BLD is much simple and perform better than Vanilla load balancing.But now BLD can just work with mainline.

yeah, these are way too oversized/overheaded for desktop or small system's usage


currently trying out RIFS-ES

thanks Chen !

[3766691]

thanks !

I'll probably try it tomorrow or within the next days - currently my box is busy working ...





results:

ok, I put the rifs for 3.4 to some test (the one without ES / enhanced scheduling)

compiling libreoffice + rsyncing data on encrypted partitions (1 TB) to zlib-compressed btrfs + 1080p HD flash/youtube video

everything super-smooth and interactive

== keeper


best recent kernel so far


I combined it with The Barbershop Load Distribution algorithm IMO a winning combination

BLD is much simple and perform better than Vanilla load balancing.But now BLD can just work with mainline.

[kernelOfTruth]

Edit 1:
Sorry for my fault.
The header file sched.h in the patch is not synchronous to my version that I am using.
I will fix it after a few hours.


Edit 2:
Ok, I have posted the new one.

thanks !

I'll probably try it tomorrow or within the next days - currently my box is busy working ...

looks good to me


compiling libreoffice right now

probably will add some more spice to it to see how it works out


thanks ! great job

results:

ok, I put the rifs for 3.4 to some test (the one without ES / enhanced scheduling)

compiling libreoffice + rsyncing data on encrypted partitions (1 TB) to zlib-compressed btrfs + 1080p HD flash/youtube video

everything super-smooth and interactive

== keeper


best recent kernel so far


I combined it with The Barbershop Load Distribution algorithm IMO a winning combination

[3766691]

the patch is monolithic right ?

thanks !


edit:

ok, seemed that the build went wacko - I forgot to make clean mrproper before

d'oh !

2nd compile went fine

my bad


will try now


edit2:

ok doesn't built - same error

seems genkernel has something against it


someone else has to try it


thanks !

Edit 1:
Sorry for my fault.
The header file sched.h in the patch is not synchronous to my version that I am using.
I will fix it after a few hours.


Edit 2:
Ok, I have posted the new one.

[kernelOfTruth]

Thanks for comments.
I want someone to try RIFS-ES V1 also. If you have time could you help me to try? Thanks.
If it works better than RIFS-V2-Bugfix2 then RIFS-ES will be combined to RIFS.

LD arch/x86/built-in.o
CC kernel/sched/rifs.o
kernel/sched/rifs.c: In function ‘sched_fork’:
kernel/sched/rifs.c:1460:3: error: ‘struct task_struct’ has no member named ‘tick_used’
kernel/sched/rifs.c:1460:21: error: ‘struct task_struct’ has no member named ‘tick_used’
kernel/sched/rifs.c: In function ‘wake_up_new_task’:
kernel/sched/rifs.c:1518:3: error: ‘struct task_struct’ has no member named ‘tick_used’
kernel/sched/rifs.c: In function ‘task_running_tick’:
kernel/sched/rifs.c:2389:6: error: ‘struct task_struct’ has no member named ‘tick_used’
kernel/sched/rifs.c:2390:4: error: ‘struct task_struct’ has no member named ‘tick_used’
kernel/sched/rifs.c: In function ‘check_timeslice_end’:
kernel/sched/rifs.c:2509:9: error: ‘struct task_struct’ has no member named ‘sleep_cnt’
kernel/sched/rifs.c:2510:7: error: ‘struct task_struct’ has no member named ‘sleep_cnt’
kernel/sched/rifs.c:2511:8: error: ‘struct task_struct’ has no member named ‘tick_used’
kernel/sched/rifs.c:2511:8: error: ‘struct task_struct’ has no member named ‘sleep_cnt’
kernel/sched/rifs.c:2513:24: error: ‘struct task_struct’ has no member named ‘sleep_jiffy’
kernel/sched/rifs.c:2514:7: error: ‘struct task_struct’ has no member named ‘sleep_cnt’
kernel/sched/rifs.c:2515:6: error: ‘struct task_struct’ has no member named ‘sleep_jiffy’
kernel/sched/rifs.c:2521:37: error: ‘struct task_struct’ has no member named ‘tick_used’
kernel/sched/rifs.c:2522:4: error: ‘struct task_struct’ has no member named ‘tick_used’
make[2]: *** [kernel/sched/rifs.o] Error 1
make[1]: *** [kernel/sched] Error 2
make: *** [kernel] Error 2

the patch is monolithic right ?

thanks !


edit:

ok, seemed that the build went wacko - I forgot to make clean mrproper before

d'oh !

2nd compile went fine

my bad


will try now


edit2:

ok doesn't built - same error

seems genkernel has something against it


someone else has to try it


thanks !

[3766691]

looks good to me


compiling libreoffice right now

probably will add some more spice to it to see how it works out


thanks ! great job

Thanks for comments.
I want someone to try RIFS-ES V1 also. If you have time could you help me to try? Thanks.
If it works better than RIFS-V2-Bugfix2 then RIFS-ES will be combined to RIFS.

[kernelOfTruth]

RIFS V2-Bugfix2 for 3.4 kernel is released

looks good to me


compiling libreoffice right now

probably will add some more spice to it to see how it works out


thanks ! great job

[3766691]

I've sent you an email. I can't compile RC2

RIFS V2-Bugfix2 for 3.4 kernel is released

[3766691]

It was made very clear by CK after his experiments with massive loads that BFS is not meant to be used with -j128. It is meant to be used with high workloads, but not with *ridiculous* workloads. High workload is something like -j10 on a quad core. 128 is ridiculous, and 512 is even beyond ridiculous.

BFS is optimized for real use, not for throwing -j512 kernel builds at it. If you really intend to run such a massive workload (why?), you would run it SCHED_IDLEPRIO or nice 19 (or both). If you have written a new scheduler, posting information about how it handles -j128 "better" than the others says absolutely nothing about the merit of that scheduler. The only thing it says is that it handles something no one needs but doesn't say anything about how it handles stuff people actually *do* need.

See:

http://ck-hack.blogspot.com/2010/11/...y-comment.html

I have released RIFS V2-bugfix2 for 3.4
x kernel too late. Sorry

In the next version I will have a new idea with designing.
Also an O(1) LOWEST DEADLINE PICKING for BFS is released.

[3766691]

where are all these patches ?

link ?

trying to find it but I only see V2 :/


edit:

found it: http://code.google.com/p/rifs-scheduler/downloads/list


could you please also make a version for 3.4 kernel ?

thanks !

RIFS for 3.4 kernel has made(I will start testing it). I have also posted it.
I am planning to make a modular version of RIFS so that I don't need to post RIFS full patch everytime.
Also, BFS-O(1) designing improvement patch for original BFS has been released.
If you can't download RIFS-V2-Bugfix-kernel3.3.4 that means it is still generating.