Why I Don't Think the Number of Processors Affects the Wall Clock Time of My Tests...

Using the taskset command I demonstrate that my benchmarks run on one processor.

Recently Russ raised a question: are my wall clock times wrong because I have many processors on my machine. I believe that the answer is no.

Firstly, without special efforts, linux has no way of breaking a process onto more than one processor. When I run my benchmarking program, tst_ar4, I see from the top command than one processor goes to > 90% use, and all the others remain at 0%.

Secondly, I confirmed this with the taskset command, which I had never heard of before. It limits a process to one (or any other number) of processors. In fact, it lets you pick which processors are used. Here's some timing results that show I get about the same times using the taskset command as I do without it, on a compressed file read:

bash-3.2$ sudo ./clear_cache.sh && ./tst_ar4 -h -c 35000000 \
bash-3.2$ pr_A1_z1_64_128_256.nc
cs[0] cs[1] cs[2]  cache(MB) deflate shuffle 1st_read_hor(us) avg_read_hor(us)
64    128   256    33.4      1       0       304398    4568

bash-3.2$ sudo ./clear_cache.sh && taskset -c 4 ./tst_ar4 -h -c 35000000 \
bash-3.2$ pr_A1_z1_64_128_256.nc
cs[0] cs[1] cs[2]  cache(MB) deflate shuffle 1st_read_hor(us) avg_read_hor(us)
64    128   256    33.4      1       0       306810   4553

bash-3.2$ sudo ./clear_cache.sh && ./tst_ar4 -h -c 35000000 \
bash-3.2$ pr_A1_z1_64_128_256.nc
cs[0] cs[1] cs[2]  cache(MB) deflate shuffle 1st_read_hor(us) avg_read_hor(us)
64    128   256    33.4      1       0       292707   4616

bash-3.2$ sudo ./clear_cache.sh && taskset -c 4 ./tst_ar4 -h -c 35000000 \
bash-3.2$ pr_A1_z1_64_128_256.nc
cs[0] cs[1] cs[2]  cache(MB) deflate shuffle 1st_read_hor(us) avg_read_hor(us)
64    128   256    33.4      1       0       293713   4567
Comments:

I stumbled on this post when I was trying to find related data on coding and I find this quite interesting though quite complicated to me.

Posted by 75.71.9.118 on February 03, 2010 at 06:14 AM MST #

Post a Comment:
Comments are closed for this entry.
Unidata Developer's Blog
A weblog about software development by Unidata developers*
Unidata Developer's Blog
A weblog about software development by Unidata developers*

Welcome

FAQs

News@Unidata blog

Take a poll!

What if we had an ongoing user poll in here?

Browse By Topic
Browse by Topic
« November 2024
SunMonTueWedThuFriSat
     
1
2
3
4
5
6
7
8
9
10
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
       
Today