Linux-Based XanMod Kernel Tests

[phoronix]

Phoronix: Linux-Based XanMod Kernel Tests

Following yesterday's tests of the Liquorix 4.3 kernel, a Phoronix reader pointed out another customized kernel I previously hadn't heard of: XanMod...

http://www.phoronix.com/scan.php?page=news_item&px=XanMod-Kernel-Tests

[Azrael5]

Excellent new! What about KDBus!?

[post-factum]

I believe one should note that xanmod kernel is derived from pf-kernel, so one should examine pf-kernel as well.

https://pf.natalenko.name/

[Guest]

Phoronix: Linux-Based XanMod Kernel Tests

Following yesterday's tests of the Liquorix 4.3 kernel, a Phoronix reader pointed out another customized kernel I previously hadn't heard of: XanMod...

http://www.phoronix.com/scan.php?pag...d-Kernel-Tests

http://openbenchmarking.org/result/1601109-GA-XANMODLIN68&obr_sor=y&obr_hgv=4.3.3-xanmod8

How did you obtain the large performance difference in pbzip2 between linux-4.3.0-generic and linux-4.4.0-generic? If you run "time pbzip2 ..." from command line on the 20-thread Xeon CPU, do you see the same difference between 4.3.0 and 4.4.0?

There is no pbzip2 performance difference between 4.3.0 and 4.4.0 on my AMD 4-core CPU.

[Michael]

http://openbenchmarking.org/result/1601109-GA-XANMODLIN68&obr_sor=y&obr_hgv=4.3.3-xanmod8

How did you obtain the large performance difference in pbzip2 between linux-4.3.0-generic and linux-4.4.0-generic? If you run "time pbzip2 ..." from command line on the 20-thread Xeon CPU, do you see the same difference between 4.3.0 and 4.4.0?

There is no pbzip2 performance difference between 4.3.0 and 4.4.0 on my AMD 4-core CPU.

Did you run the test profile with PTS? Or manually set it up similarly? http://openbenchmarking.org/innhold/...955138fbc648ad

[thelongdivider]

If nothing else, this gets me excited for linux 4.4 (maybe with the liquorix mods... but probably not).

[profoundWHALE]

A few interesting things:
- It beat the other kernels with Unigine Tropics despite losing in every other graphics benchmark.
- It did really well in C-Ray, Bzip2, OpenSSL as well as beating everyone in FFmpeg and Apache. Does it have something that 4.4 has as well?
- It got destroyed with the Postgre SQL and Redis benches
- The stress benchmarks had it doing poorly vs 4.3 in memory copying, but did very well or won otherwise in those.

the kernel is spelled wrong a few times on the testing site.

[nils_]

There should probably be a link to Xanmod in the article for the sake of completeness.

[Guest]

Did you run the test profile with PTS? Or manually set it up similarly? http://openbenchmarking.org/innhold/...955138fbc648ad

Phoronix Test Suite v6.2.0m3

Linux kernel 4.3.0-generic:

Parallel BZIP2 Compression 1.1.12:
    pts/compress-pbzip2-1.5.0
    Test 1 of 1
    Estimated Trial Run Count: 3
        Started Run 1 @ 22:33:11
        Started Run 2 @ 22:33:28
        Started Run 3 @ 22:33:44  [Std. Dev: 1.55%]

    Test Results:
        16.985441923141
        16.469793081284
        16.78581404686

    Average: 16.75 Seconds

Linux kernel 4.4.0-generic:

Parallel BZIP2 Compression 1.1.12:
    pts/compress-pbzip2-1.5.0
    Test 1 of 1
    Estimated Trial Run Count:    3
    Estimated Time To Completion: 1 Minute (22:43 CET)
        Started Run 1 @ 22:42:47
        Started Run 2 @ 22:43:04
        Started Run 3 @ 22:43:20  [Std. Dev: 1.69%]

    Test Results:
        17.140381097794
        16.623400211334
        16.676486968994

    Average: 16.81 Seconds

[Xan's]

Two points to consider after seeing the results. The release tested uses, by default, the interactive mode from ck's bfs v0.467. In graphics applications this behavior can have a negative effect on results.
To turn off:

echo 0 > /proc/sys/kernel/interactive

In next release will disable interactive by default.

Now about the uksm. It is not recommended to run benchmarks immediately after the boot because it uses processing peaks in the meantime. I will improve this point as well.

Thank you Michael.