RE: [EXTERNAL] Re: Linux kernel question, pre 2.6.23 vs later kernel versions

[Groom, George]

Ole,

 

Unfortunately, it’s not really practical for me to change kernels as you suggest.  My apologies for not being able to test in this manner.

 

My ask of pre and post CFS (kernel process scheduling changes) you answered. I don't believe you were able to replicate.  I'm good with this answer.  I do, however, remain puzzled how a purportedly more powerful machine (greater cores, threading count) executes a specific parallel script considerably slower than on a machine with substantially less cores and threading.  I was curious if anyone has observed similar behavior and if similar what was identified as a likely cause.

 

Appreciate your time and especially your parallel utility.  I use it daily. J

 

Thank you,

 

George Groom

 

-----Original Message-----
From: Ole Tange [mailto:ole@tange.dk]
Sent: Monday, October 22, 2018 5:58 PM
To: Groom, George
Cc: parallel
Subject: Re: [EXTERNAL] Re: Linux kernel question, pre 2.6.23 vs later kernel versions

 

On Mon, Oct 22, 2018 at 7:10 PM Groom, George <George_Groom@comcast.com> wrote:

 

> Actual output from my 2 HP systems are below.  Note execution time differences between later and older kernels.

 

Are you saying you are comparing two completely different systems (one

of the of the differences being the kernel version) and you then

conclude that it must be due to the kernel version?

 

It could just as well be due to different cache layout in the CPUs or

some other variable. Or that the top speed of the cores on one system

is different.

 

To minimize the variables please test again, but do it on the same

system with the only difference being the version of the kernel. So

boot the system with kernel A, do the test, then boot with kernel B

and to exactly the same test.

 

 

/Ole