Hello,
How many pulses and messages can a channel enqueue before it starts
throwing away pulses?
Regards, Vasili
William Halchin <bhalchin@syrres.com> wrote:
Hello,
How many pulses and messages can a channel enqueue before it starts
throwing away pulses?
Messages are not queued, per se. Thread that try to send to the channel
are blocked. In theory, all threads in a system could be blocked on
one channel.
For pulses, the memory to hold them is allocated, and there is no
hard limit to this. You will either run out of memory, or run out
of CPU from walking the queue to insert the pulse. In practice, this
will only happen in “bad” situations, where you have a rapid rate of
pulse accrual without any pulse draining.
-David
David Gibbs
QNX Training Services
dagibbs@qnx.com
“David Gibbs” <dagibbs@qnx.com> wrote in message
news:d3k3ns$s4g$1@inn.qnx.com…
William Halchin <> bhalchin@syrres.com> > wrote:
Hello,How many pulses and messages can a channel enqueue before it starts
throwing away pulses?Messages are not queued, per se. Thread that try to send to the channel
are blocked. In theory, all threads in a system could be blocked on
one channel.For pulses, the memory to hold them is allocated, and there is no
hard limit to this. You will either run out of memory, or run out
of CPU from walking the queue to insert the pulse. In practice, this
will only happen in “bad” situations, where you have a rapid rate of
pulse accrual without any pulse draining.
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer half
way through the constructor. I am waiting to get a board so I can put
procnto-instr and use system profiler to see what is going on. Aside: to
make matters worse, the above-mentioned process uses shared-memory as an IPC
to enqueue data generated as a side effect of timer pulse that is dequeued
by another process’s thread that is running constantly at 10.
Vasili
-David
David Gibbs
QNX Training Services
dagibbs@qnx.com
Don’t worry - when shit hits the fan you’ll know it by seeing the non-stop
“Out of interrupt events” message on the screen. At that point that will be
ALL the system is doing 
– igor
“William Halchin” <bhalchin@syrres.com> wrote in message
news:d3k4eh$spe$1@inn.qnx.com…
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
William Halchin <> bhalchin@syrres.com> > wrote:
Hello,How many pulses and messages can a channel enqueue before it
starts
throwing away pulses?Messages are not queued, per se. Thread that try to send to the channel
are blocked. In theory, all threads in a system could be blocked on
one channel.For pulses, the memory to hold them is allocated, and there is no
hard limit to this. You will either run out of memory, or run out
of CPU from walking the queue to insert the pulse. In practice, this
will only happen in “bad” situations, where you have a rapid rate of
pulse accrual without any pulse draining.
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer half
way through the constructor. I am waiting to get a board so I can put
procnto-instr and use system profiler to see what is going on. Aside: to
make matters worse, the above-mentioned process uses shared-memory as an
IPC to enqueue data generated as a side effect of timer pulse that is
dequeued by another process’s thread that is running constantly at 10.Vasili
-DavidDavid Gibbs
QNX Training Services
dagibbs@qnx.com
Hi Vasili…
What you are experiencing is not the worst, would you believe that there
are some people that want to run destroyers with nuclear weapons in them
with Windows as the OS? (Hint: British royal navy - see below). In cases
such as these, the word ‘moron’ comes to mind.
http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stm
Go figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer half
way through the constructor. I am waiting to get a board so I can put
procnto-instr and use system profiler to see what is going on. Aside: to
make matters worse, the above-mentioned process uses shared-memory as an IPC
to enqueue data generated as a side effect of timer pulse that is dequeued
by another process’s thread that is running constantly at 10.Vasili
David Gibbs wrote:
William Halchin <> bhalchin@syrres.com> > wrote:
Hello,
How many pulses and messages can a channel enqueue before it starts
throwing away pulses?
Messages are not queued, per se. Thread that try to send to the channel
are blocked. In theory, all threads in a system could be blocked on
one channel.For pulses, the memory to hold them is allocated, and there is no
hard limit to this. You will either run out of memory, or run out
of CPU from walking the queue to insert the pulse. In practice, this
will only happen in “bad” situations, where you have a rapid rate of
pulse accrual without any pulse draining.
This pulse queue issue occurs only when we can’t compress them (ie.
alternating pulses are queued). The same pulse over and over would not
exhibit that issue, as we compress the same pulse as a reference count.
–
Cheers,
Adam
QNX Software Systems
[ amallory@qnx.com ]
With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
–Peter J. Schoenster <pschon@baste.magibox.net>
“Adam Mallory” <amallory@qnx.com> wrote in message
news:d3lq1n$6rc$1@inn.qnx.com…
David Gibbs wrote:
William Halchin <> bhalchin@syrres.com> > wrote:Hello,
How many pulses and messages can a channel enqueue before it starts
throwing away pulses?
Messages are not queued, per se. Thread that try to send to the channel
are blocked. In theory, all threads in a system could be blocked on
one channel.For pulses, the memory to hold them is allocated, and there is no
hard limit to this. You will either run out of memory, or run out
of CPU from walking the queue to insert the pulse. In practice, this
will only happen in “bad” situations, where you have a rapid rate of
pulse accrual without any pulse draining.This pulse queue issue occurs only when we can’t compress them (ie.
alternating pulses are queued). The same pulse over and over would not
exhibit that issue, as we compress the same pulse as a reference count.
^^^ very nice optimization!! QNX comes through again!–
Cheers,
AdamQNX Software Systems
[ > amallory@qnx.com > ]With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
–Peter J. Schoenster <> pschon@baste.magibox.net
Hilarious: “The new ship is also controlled by state-of-the-art computers
using a Windows NT operating system.”
“Miguel Simon” <simon@ou.edu> wrote in message
news:d3ksmd$f2n$1@inn.qnx.com…
Hi Vasili…
What you are experiencing is not the worst, would you believe that there
are some people that want to run destroyers with nuclear weapons in them
with Windows as the OS? (Hint: British royal navy - see below). In cases
such as these, the word ‘moron’ comes to mind.http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stmGo figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer half
way through the constructor. I am waiting to get a board so I can put
procnto-instr and use system profiler to see what is going on. Aside: to
make matters worse, the above-mentioned process uses shared-memory as an
IPC to enqueue data generated as a side effect of timer pulse that is
dequeued by another process’s thread that is running constantly at 10.Vasili
“William Halchin” <bhalchin@syrres.com> wrote in message
news:d3mar2$if6$1@inn.qnx.com…
Hilarious: “The new ship is also controlled by state-of-the-art computers
using a Windows NT operating system.”
Unless you need 1ms precission, I guess this is fine. The OS is not really
the problem in most case, it’s the skill of the people doing software on
them. Where hardware resource is not an issue, I beleive NT and XP to be
fine OS for this task.
“Miguel Simon” <> simon@ou.edu> > wrote in message
news:d3ksmd$f2n$> 1@inn.qnx.com> …
Hi Vasili…
What you are experiencing is not the worst, would you believe that there
are some people that want to run destroyers with nuclear weapons in them
with Windows as the OS? (Hint: British royal navy - see below). In cases
such as these, the word ‘moron’ comes to mind.http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stmGo figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer
half way through the constructor. I am waiting to get a board so I can
put procnto-instr and use system profiler to see what is going on.
Aside: to make matters worse, the above-mentioned process uses
shared-memory as an IPC to enqueue data generated as a side effect of
timer pulse that is dequeued by another process’s thread that is running
constantly at 10.Vasili
Mario,
The problem in battle situation meeting hard real-time deadlines is
difference between survival and death. E.g., I remember during Falklands
War between UK and Argentina, Argentina air force sunk a British cruiser(??)
was sunk by a French Exercet(sp??) missile. In such situations, the kernel
has to be totally pre-emptible and highest-priority thread run (e.g.
controlling something US Phalanx gun to shoot down missle). With any
“classical” OS like W2K, AIX, Linux, Solaris, etc. you do not get
pre-emptibility and also run-to-completion! Hope this helps.
Regards, Vasili
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:d3mdrr$ksa$1@inn.qnx.com…
“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3mar2$if6$> 1@inn.qnx.com> …
Hilarious: “The new ship is also controlled by state-of-the-art computers
using a Windows NT operating system.”Unless you need 1ms precission, I guess this is fine. The OS is not
really the problem in most case, it’s the skill of the people doing
software on them. Where hardware resource is not an issue, I beleive NT
and XP to be fine OS for this task.
“Miguel Simon” <> simon@ou.edu> > wrote in message
news:d3ksmd$f2n$> 1@inn.qnx.com> …
Hi Vasili…
What you are experiencing is not the worst, would you believe that there
are some people that want to run destroyers with nuclear weapons in them
with Windows as the OS? (Hint: British royal navy - see below). In cases
such as these, the word ‘moron’ comes to mind.http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stmGo figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer
half way through the constructor. I am waiting to get a board so I can
put procnto-instr and use system profiler to see what is going on.
Aside: to make matters worse, the above-mentioned process uses
shared-memory as an IPC to enqueue data generated as a side effect of
timer pulse that is dequeued by another process’s thread that is
running constantly at 10.Vasili
\
William Halchin wrote:
Mario,
The problem in battle situation meeting hard real-time deadlines is
difference between survival and death. E.g., I remember during Falklands
War between UK and Argentina, Argentina air force sunk a British cruiser(??)
One of the cruisers did launch a anti-aircraft missile into it’s partnering ship because of a crash and re-boot delay of the fire control computer. So, even back then re-boots of more than one second was apparently acceptable to them. 
Evan
“William Halchin” <bhalchin@syrres.com> wrote in message
news:d3megf$l88$1@inn.qnx.com…
Mario,
The problem in battle situation meeting hard real-time deadlines is
difference between survival and death. E.g., I remember during Falklands
War between UK and Argentina, Argentina air force sunk a British
cruiser(??) was sunk by a French Exercet(sp??) missile. In such
situations, the kernel has to be totally pre-emptible and highest-priority
thread run (e.g. controlling something US Phalanx gun to shoot down
missle). With any “classical” OS like W2K, AIX, Linux, Solaris, etc. you
do not get pre-emptibility and also run-to-completion! Hope this helps.
There is a difference between controlling a gun and controling a ship…
Plus if I were to design a gun control system, I would probably offload the
core OS with dedicated hardware to do the positionning
Regards, Vasili
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:d3mdrr$ksa$> 1@inn.qnx.com> …“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3mar2$if6$> 1@inn.qnx.com> …
Hilarious: “The new ship is also controlled by state-of-the-art
computers using a Windows NT operating system.”Unless you need 1ms precission, I guess this is fine. The OS is not
really the problem in most case, it’s the skill of the people doing
software on them. Where hardware resource is not an issue, I beleive NT
and XP to be fine OS for this task.
“Miguel Simon” <> simon@ou.edu> > wrote in message
news:d3ksmd$f2n$> 1@inn.qnx.com> …
Hi Vasili…
What you are experiencing is not the worst, would you believe that
there are some people that want to run destroyers with nuclear weapons
in them with Windows as the OS? (Hint: British royal navy - see below).
In cases such as these, the word ‘moron’ comes to mind.
http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stmGo figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer
half way through the constructor. I am waiting to get a board so I can
put procnto-instr and use system profiler to see what is going on.
Aside: to make matters worse, the above-mentioned process uses
shared-memory as an IPC to enqueue data generated as a side effect of
timer pulse that is dequeued by another process’s thread that is
running constantly at 10.Vasili
\
“Evan Hillas” <evanh@clear.net.nz> wrote in message
news:d3mmmp$r3m$1@inn.qnx.com…
William Halchin wrote:
Mario,The problem in battle situation meeting hard real-time deadlines is
difference between survival and death. E.g., I remember during Falklands
War between UK and Argentina, Argentina air force sunk a British
cruiser(??)
One of the cruisers did launch a anti-aircraft missile into it’s
partnering ship because of
^^^ into it’s partnering ship?? ouch!
a crash and re-boot delay of the fire control computer. So, even back then
re-boots of more than one second was apparently acceptable to them.
Evan
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:d3n6jc$7rv$1@inn.qnx.com…
“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3megf$l88$> 1@inn.qnx.com> …
Mario,The problem in battle situation meeting hard real-time deadlines is
difference between survival and death. E.g., I remember during Falklands
War between UK and Argentina, Argentina air force sunk a British
cruiser(??) was sunk by a French Exercet(sp??) missile. In such
situations, the kernel has to be totally pre-emptible and
highest-priority thread run (e.g. controlling something US Phalanx gun to
shoot down missle). With any “classical” OS like W2K, AIX, Linux,
Solaris, etc. you do not get pre-emptibility and also run-to-completion!
Hope this helps.There is a difference between controlling a gun and controling a ship…
^^^ this part I agree with
Plus if I were to design a gun control system, I would probably offload
the core OS with dedicated hardware to do the positionning >
^^^ I think this just pushes the problem down a level where …
guesss what … you will have an RTOS running!
Regards, Vasili
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:d3mdrr$ksa$> 1@inn.qnx.com> …“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3mar2$if6$> 1@inn.qnx.com> …
Hilarious: “The new ship is also controlled by state-of-the-art
computers using a Windows NT operating system.”Unless you need 1ms precission, I guess this is fine. The OS is not
really the problem in most case, it’s the skill of the people doing
software on them. Where hardware resource is not an issue, I beleive NT
and XP to be fine OS for this task.
“Miguel Simon” <> simon@ou.edu> > wrote in message
news:d3ksmd$f2n$> 1@inn.qnx.com> …
Hi Vasili…
What you are experiencing is not the worst, would you believe that
there are some people that want to run destroyers with nuclear weapons
in them with Windows as the OS? (Hint: British royal navy - see
below). In cases such as these, the word ‘moron’ comes to mind.
http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stmGo figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer
half way through the constructor. I am waiting to get a board so I
can put procnto-instr and use system profiler to see what is going
on. Aside: to make matters worse, the above-mentioned process uses
shared-memory as an IPC to enqueue data generated as a side effect of
timer pulse that is dequeued by another process’s thread that is
running constantly at 10.Vasili
\
Not to rain on your parade, but check your facts. Solaris does preemption
and so do modern Linux flavors. They have different issues, but preemtion is
not one of them.
“William Halchin” <bhalchin@syrres.com> wrote in message
news:d3megf$l88$1@inn.qnx.com…
Mario,
The problem in battle situation meeting hard real-time deadlines is
difference between survival and death. E.g., I remember during Falklands
War between UK and Argentina, Argentina air force sunk a British
cruiser(??) was sunk by a French Exercet(sp??) missile. In such
situations, the kernel has to be totally pre-emptible and highest-priority
thread run (e.g. controlling something US Phalanx gun to shoot down
missle). With any “classical” OS like W2K, AIX, Linux, Solaris, etc. you
do not get pre-emptibility and also run-to-completion! Hope this helps.Regards, Vasili
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:d3mdrr$ksa$> 1@inn.qnx.com> …“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3mar2$if6$> 1@inn.qnx.com> …
Hilarious: “The new ship is also controlled by state-of-the-art
computers using a Windows NT operating system.”Unless you need 1ms precission, I guess this is fine. The OS is not
really the problem in most case, it’s the skill of the people doing
software on them. Where hardware resource is not an issue, I beleive NT
and XP to be fine OS for this task.
“Miguel Simon” <> simon@ou.edu> > wrote in message
news:d3ksmd$f2n$> 1@inn.qnx.com> …
Hi Vasili…
What you are experiencing is not the worst, would you believe that
there are some people that want to run destroyers with nuclear weapons
in them with Windows as the OS? (Hint: British royal navy - see below).
In cases such as these, the word ‘moron’ comes to mind.
http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stmGo figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer
half way through the constructor. I am waiting to get a board so I can
put procnto-instr and use system profiler to see what is going on.
Aside: to make matters worse, the above-mentioned process uses
shared-memory as an IPC to enqueue data generated as a side effect of
timer pulse that is dequeued by another process’s thread that is
running constantly at 10.Vasili
\
In article <d3m168$bur$1@inn.qnx.com>, bhalchin@syrres.com says…
will only happen in “bad” situations, where you have a rapid rate of
pulse accrual without any pulse draining.This pulse queue issue occurs only when we can’t compress them (ie.
alternating pulses are queued). The same pulse over and over would not
exhibit that issue, as we compress the same pulse as a reference count.
^^^ very nice optimization!! QNX comes through again!
You must not have a “bad” situation in first place. Pulse compressing is
a very nice optimization indeed. It’s probably time for a new generation
of OS that will analyse the code before actually running it and report
“I won’t run it becasue it’s ******* stupid!\n”
Cheers,
Eduard.
“Igor Kovalenko” <kovalenko@comcast.net> wrote in message
news:d3qh3f$mai$1@inn.qnx.com…
Not to rain on your parade, but check your facts. Solaris does preemption
and so do modern Linux flavors. They have different issues, but preemtion
is not one of them.
I though kernel 2.6 has improved support for these kind of thing, but I
didn’t feel like getting into that since I don’t really now about Linux.
“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3megf$l88$> 1@inn.qnx.com> …
Mario,The problem in battle situation meeting hard real-time deadlines is
difference between survival and death. E.g., I remember during Falklands
War between UK and Argentina, Argentina air force sunk a British
cruiser(??) was sunk by a French Exercet(sp??) missile. In such
situations, the kernel has to be totally pre-emptible and
highest-priority thread run (e.g. controlling something US Phalanx gun to
shoot down missle). With any “classical” OS like W2K, AIX, Linux,
Solaris, etc. you do not get pre-emptibility and also run-to-completion!
Hope this helps.Regards, Vasili
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:d3mdrr$ksa$> 1@inn.qnx.com> …“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3mar2$if6$> 1@inn.qnx.com> …
Hilarious: “The new ship is also controlled by state-of-the-art
computers using a Windows NT operating system.”Unless you need 1ms precission, I guess this is fine. The OS is not
really the problem in most case, it’s the skill of the people doing
software on them. Where hardware resource is not an issue, I beleive NT
and XP to be fine OS for this task.
“Miguel Simon” <> simon@ou.edu> > wrote in message
news:d3ksmd$f2n$> 1@inn.qnx.com> …
Hi Vasili…
What you are experiencing is not the worst, would you believe that
there are some people that want to run destroyers with nuclear weapons
in them with Windows as the OS? (Hint: British royal navy - see
below). In cases such as these, the word ‘moron’ comes to mind.
http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stmGo figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer
half way through the constructor. I am waiting to get a board so I
can put procnto-instr and use system profiler to see what is going
on. Aside: to make matters worse, the above-mentioned process uses
shared-memory as an IPC to enqueue data generated as a side effect of
timer pulse that is dequeued by another process’s thread that is
running constantly at 10.Vasili
\
Igor,
I have a friend who works with Linux 2.6. He said that there are large
parts of the kernel which are still not pre-emptible. I didn’t know that
Solaris is pre-emptible. Thnaks.
Vasili
“Igor Kovalenko” <kovalenko@comcast.net> wrote in message
news:d3qh3f$mai$1@inn.qnx.com…
Not to rain on your parade, but check your facts. Solaris does preemption
and so do modern Linux flavors. They have different issues, but preemtion
is not one of them.“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3megf$l88$> 1@inn.qnx.com> …
Mario,The problem in battle situation meeting hard real-time deadlines is
difference between survival and death. E.g., I remember during Falklands
War between UK and Argentina, Argentina air force sunk a British
cruiser(??) was sunk by a French Exercet(sp??) missile. In such
situations, the kernel has to be totally pre-emptible and
highest-priority thread run (e.g. controlling something US Phalanx gun to
shoot down missle). With any “classical” OS like W2K, AIX, Linux,
Solaris, etc. you do not get pre-emptibility and also run-to-completion!
Hope this helps.Regards, Vasili
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:d3mdrr$ksa$> 1@inn.qnx.com> …“William Halchin” <> bhalchin@syrres.com> > wrote in message
news:d3mar2$if6$> 1@inn.qnx.com> …
Hilarious: “The new ship is also controlled by state-of-the-art
computers using a Windows NT operating system.”Unless you need 1ms precission, I guess this is fine. The OS is not
really the problem in most case, it’s the skill of the people doing
software on them. Where hardware resource is not an issue, I beleive NT
and XP to be fine OS for this task.
“Miguel Simon” <> simon@ou.edu> > wrote in message
news:d3ksmd$f2n$> 1@inn.qnx.com> …
Hi Vasili…
What you are experiencing is not the worst, would you believe that
there are some people that want to run destroyers with nuclear weapons
in them with Windows as the OS? (Hint: British royal navy - see
below). In cases such as these, the word ‘moron’ comes to mind.
http://www.theregister.co.uk/2004/09/06/ams_goes_windows_for_warships/
http://www.itweek.co.uk/comment/1158865
http://www.winbeta.org/comments.php?shownews=1115
http://news.bbc.co.uk/1/hi/technology/3724219.stmGo figure.
Miguel.
William Halchin wrote:
“David Gibbs” <> dagibbs@qnx.com> > wrote in message
news:d3k3ns$s4g$> 1@inn.qnx.com> …
^^^ this is exactly(!!) what I am worried about. The author is
“sending” pulses at prio 24 at 1 ms clip plus arms a repeating timer
half way through the constructor. I am waiting to get a board so I
can put procnto-instr and use system profiler to see what is going
on. Aside: to make matters worse, the above-mentioned process uses
shared-memory as an IPC to enqueue data generated as a side effect of
timer pulse that is dequeued by another process’s thread that is
running constantly at 10.Vasili
\