Hi all,
Just an update to my problems with mqueues with the ppc405. thanks for all
the advice from Igor, David and others. I think that the problem lies in
the fact that the way that QNX uses posix queues. We have found that the
mq_timedsend and mq_receive functions opertae okay until the rate at which
packets are put onto the queue is increased.
mq_timedsend then fails and continues to fail until the target board
crashes. I do not think that the problem lies in the mips capacity of the
pp5405(266Mhz) because we have the same software running efficiently well
using 2 previous operating systems.
Has antone else noticed similar processor efficiency problems using QNX and
imparticular with relation to how queues are handled?
Any advice would be very appreciated.
Thanks in advance,
Trev.
Trev Brown wrote:
Hi all,
Just an update to my problems with mqueues with the ppc405. thanks for all
the advice from Igor, David and others. I think that the problem lies in
the fact that the way that QNX uses posix queues. We have found that the
mq_timedsend and mq_receive functions opertae okay until the rate at which
packets are put onto the queue is increased.
Could you be more specific about the test case? If by chance you have a
test utility which demonstrates such behavior, please, post it here. I
would give it a short on my 200 mghz PPC.
mq_timedsend then fails and continues to fail until the target board
crashes. I do not think that the problem lies in the mips capacity of the
pp5405(266Mhz) because we have the same software running efficiently well
using 2 previous operating systems.
Has antone else noticed similar processor efficiency problems using QNX and
imparticular with relation to how queues are handled?
Any advice would be very appreciated.
Thanks in advance,
Trev.
Could you be more specific about the test case? If by chance you have a
test utility which demonstrates such behavior, please, post it here. I
would give it a short on my 200 mghz PPC.
200 Mega Giga hz, wow that must boot pretty fast ?
Mario Charest postmaster@127.0.0.1 wrote:
Could you be more specific about the test case? If by chance you have a
test utility which demonstrates such behavior, please, post it here. I
would give it a short on my 200 mghz PPC.
200 Mega Giga hz, wow that must boot pretty fast ?
Mega Giga Hz? No, it’s quite clearly milligram hecto z, whatever “z” is. 
My favourites are mS for milli-Siemens (meant milliseconds) and
KHz for Kelvin Hertz (meant kiloHertz).
Cheers,
-RK
–
[If replying via email, you’ll need to click on the URL that’s emailed to you
afterwards to forward the email to me – spam filters and all that]
Robert Krten, PDP minicomputer collector http://www.parse.com/~pdp8/
POSIX message queues are [comparatively] inefficient in QNX because they are
implemented on top of synchronous message passing (that causes unneccessary
overhead). I did use them on PPC, but different CPU (750FX). Did not use
mq_timedsend(), but normal mq_send() worked ok with any data rate (my
benchmark was writing into the queue as fast as it could). So I’d suggest
you try regular mq_send() and see if that works better. I don’t really
understand why mq_timedsend() is needed, given that mq_send() is not
supposed to block (if you set O_NONBLOCK).
Anyway, the issue with performance of POSIX queues will be addressed in the
next release, that supposed to go beta real soon now ™.
– igor
“Trev Brown” <trevorb@ieee.org> wrote in message
news:bq4fps$g4r$1@inn.qnx.com…
Hi all,
Just an update to my problems with mqueues with the ppc405. thanks for
all
the advice from Igor, David and others. I think that the problem lies in
the fact that the way that QNX uses posix queues. We have found that the
mq_timedsend and mq_receive functions opertae okay until the rate at which
packets are put onto the queue is increased.
mq_timedsend then fails and continues to fail until the target board
crashes. I do not think that the problem lies in the mips capacity of the
pp5405(266Mhz) because we have the same software running efficiently well
using 2 previous operating systems.
Has antone else noticed similar processor efficiency problems using QNX
and
imparticular with relation to how queues are handled?
Any advice would be very appreciated.
Thanks in advance,
Trev.