Help

194 views
Skip to first unread message

Shrikul Joshi

unread,
Mar 6, 2010, 2:39:35 AM3/6/10
to CoreMark
Hello,
I am an engineering undergraduate student.Currently I am working on a
project where I have successfully ported uclinux rtos to a single
processor(here microblaze version 7.10.00a) system on xilinx avnet
xc4vlx60 but am facing some problems with benchmarking this system.Can
anyone please help me with the guidelines as to how should i go ahead
with benchmarking this system.please reply soon.

Shay Gal-On

unread,
Mar 8, 2010, 12:33:40 PM3/8/10
to Shrikul Joshi, CoreMark
Can you give some more details about your setup?
Are you cross compiling or native? How are you loading binaries to the
target? Can you build a hello world program and run it on the board? What
are the steps needed?

- Shay

--
You received this message because you are subscribed to the Google Groups
"CoreMark" group.
To post to this group, send email to core...@googlegroups.com.
To unsubscribe from this group, send email to
coremark+u...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/coremark?hl=en.


Rick Foos

unread,
Mar 6, 2010, 3:04:35 PM3/6/10
to Shrikul Joshi, CoreMark
Shrikul,

You should describe your problem in more detail, and include listings of
errors where applicable.

In general, to develop the port on an embedded device with ethernet, I would
NFS mount the root file system from your development machine.

Work on the port and test runs using tools on your development machine until
you have everything correct, and timers calibrated. At that point copy and
run the binaries from the embedded system. The NFS mounted system could run
the benchmark, but it may be slower due to NFS locking tasks. There may be
no rules, other than reporting, to run over NFS or flash, if that is so,
pick the fastest time.

Essentially the main difference from an embedded run to a full OS run is
that the embedded benchmark depends more on the compiler libraries than the
OS, the degree of this depends on the services the RTOS offers.

Minimizing tasks outside of the benchmark run itself (I.e. NFS) should
improve the score, but you should run in different environments to find out,
and understand the differences.

Benchmarking should lead to a story about how your system performs compared
to others, what is good, what could be done better, and why.

This may not answer your question, so describing the problem would be
helpful.

Rick Foos
SolengTech
rick...@solengtech.com


--------------------------------------------------
From: "Shrikul Joshi" <shriku...@gmail.com>
Sent: Saturday, March 06, 2010 1:39 AM
To: "CoreMark" <core...@googlegroups.com>
Subject: Help

Reply all
Reply to author
Forward
0 new messages