Not a member yet? Why not Sign up today
Create an account  

  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
 
Brainstorming for XanMod Kernel

#61
@Xan, thanks for the advice!

Very interesting results for some I/O schedulers on benchmarking:
http://www.phoronix.com/scan.php?page=ar...2-io&num=1
https://www.phoronix.com/scan.php?page=n...-BFQ-Kyber
It seems that CFQ has become very balanced for 4.12, while mq-deadline is closer.

Edit: very noticeable behaviour of Kyber scheduler for average benchmarking results.
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#62
It seems that recent tweak 'sampling_down_factor'= 100 for Ondemand governor settings is amazing for jitter reduction and CPU overload stability (in other words 'sampling_rate'= 10 ms w/ 'sampling_down_factor'= 100 gives total sampling rate= 1 sec, very good measure). This is a well balanced tweak imho, however I noticed that now Ondemand governor gives better responsiveness for me if 'up_threshold' is set to 50%, with no significative battery cost after two hours on normal office workload plus I/O operations at background. By the way, would latency be reduced with frequency_up_threshold over 63 on systems with modern CPUs? -- good question to test in the future. Sleepy
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#63
Kyber I/O scheduler + MuQSS? Sleepy
https://patchwork.kernel.org/patch/9672023/

Edit: "the Kyber I/O scheduler is a full MQ-aware (multi-queue) scheduler using a scalable token based algorithm. Work on this modern blk-mq geared scheduler was done by Omar Sandoval of Facebook" with the advantage that is easily tunable and very short extension code supported by a power enterprise and also "(...) the Kyber I/O scheduler is a low-overhead scheduler suitable for multiqueue and other fast devices. Given target latencies for reads and synchronous writes, it will self-tune queue depths to achieve that goal (...) the policy Kyber uses is inspired by active queue management techniques for network routing, similar to blk-wbt. The scheduler monitors latencies and scales the number of dispatch tokens accordingly. Queueing tokens are used to prevent starvation of synchronous requests by asynchronous requests."
https://www.phoronix.com/scan.php?page=n...-BFQ-Kyber
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#64
Quote:@tropic, Monday, April 3rd, 2017, 08:46 pm
- Consider to valuate dirty background ratio= 25 and dirty ratio= 50.

I was wrong, new page writeback dirty_background_ratio & dirty_ratio values are better. 
Also thanks Xan for enabling Intel P-State driver by default for whole series. Smile
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#65
Muqss + IO BFQ + Full Preempt + 1000HZ ?

With Xandmod 4.9.33 I have several sound glitches than not happen with liquorix kernel (4.9.28). I think that y because liquorix have full preempt and 1000HZ. (And feels little more desktop responsive).
Reply

#66
Hello @Gsus,
speed values at 800/1000/1250 were suggested in the past, however 500Hz still rules for average systems.
However, it would be a nice idea to have at least one kernel at 1000Hz to give it a chance if needed. 
Please try the next setting to get maximum control over your sound issue, useful for me in the past:
Code:
root@ echo 4096 > /proc/asound/card0/pcm0p/sub0/prealloc
#  to set it permanently please add the echo line before 'exit 0' at /etc/rc.local
Thanks for your idea! Smile
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#67
Dark letters at https://xanmod.org/ are very difficult to be read. Sad
Edit: solved.
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#68
Please, consider to valuate the suggestions from member @GWD72 about:
"Laptop-mode config needs adjustments for xanmod!". Thanks in advance. Sleepy
https://forum.xanmod.org/thread-173.html
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#69
(21-06-2017, 09:45 PM)Gsus Wrote: Muqss + IO BFQ + Full Preempt + 1000HZ ?

With Xandmod 4.9.33 I have several sound glitches than not happen with liquorix kernel (4.9.28). I think that y because liquorix have full preempt and 1000HZ. (And feels little more desktop responsive).

Gsus,

W/ 4.9.37-xanmod44 this problem persists?

Otherwise, I'll compile different versions to test them for us.
Reply

#70
(13-07-2017, 01:55 PM)Xan Wrote:
(21-06-2017, 09:45 PM)Gsus Wrote: Muqss + IO BFQ + Full Preempt + 1000HZ ?

With Xandmod 4.9.33 I have several sound glitches than not happen with liquorix kernel (4.9.28). I think that y because liquorix have full preempt and 1000HZ. (And feels little more desktop responsive).

Gsus,

W/ 4.9.37-xanmod44 this problem persists?

Otherwise, I'll compile different versions to test them for us.

I tested all day and the sound problem not shown, I will continue testing all the week, is looking good.

Thanks!
Reply



[-]
Quick Reply
Message
Type your reply to this message here.

Image Verification
Please enter the text contained within the image into the text box below it. This process is used to prevent automated spam bots.
Image Verification
(case insensitive)

Forum Jump:


Users browsing this thread:
1 Guest(s)