(More on compiling MPI)
So I’ve gotten past the problem with types, but now the code wants the
header sys/shm.h which seems to deal with shared libraries. Does this
exist, or do we not need it?
TIA
Aaron
Ok, so this problem was fixed by removing the offending file, I think I
don’t need for the manner in which I compile.
Next Q, the software uses semaphores, but in the linking step complains that
semctl cannot be found
any thoughts
Aaron
Aaron Macks <aaronmNOSPAM@cs.brandeis.edu> wrote:
(More on compiling MPI)
So I’ve gotten past the problem with types, but now the code wants the
header sys/shm.h which seems to deal with shared libraries. Does this
exist, or do we not need it?
TIA
Aaron
You need to replace SVS5 semaphores with POSIX ones. That’s a bit of
hacking.
- igor
Aaron Macks wrote:
Ok, so this problem was fixed by removing the offending file, I think I
don’t need for the manner in which I compile.Next Q, the software uses semaphores, but in the linking step complains that
semctl cannot be found
any thoughts
AaronAaron Macks <> aaronmNOSPAM@cs.brandeis.edu> > wrote:
(More on compiling MPI)
So I’ve gotten past the problem with types, but now the code wants the
header sys/shm.h which seems to deal with shared libraries. Does this
exist, or do we not need it?
TIA
Aaron
I know a fair bit of C and C++, but I’ve never used semaphores, how
difficult do you think the change will be?
Aaron
In article
<3ADB60FA.B5A5BF6A@motorola.com>, Igor Kovalenko
<Igor.Kovalenko@motorola.com> wrote:
You need to replace SVS5 semaphores with POSIX ones. That’s a bit of
hacking.
- igor
Aaron Macks wrote:
Ok, so this problem was fixed by removing the offending file, I think I
don’t need for the manner in which I compile.Next Q, the software uses semaphores, but in the linking step complains
that semctl cannot be found any thoughts AaronAaron Macks <> aaronmNOSPAM@cs.brandeis.edu> > wrote:
(More on compiling MPI)
So I’ve gotten past the problem with types, but now the code wants
the header sys/shm.h which seems to deal with shared libraries. Does
this exist, or do we not need it? TIA Aaron
Concept of semaphores has nothing to do with C/C++. It is generic OS service
allowing syncronization of 2 or more processes. SYS5 and POSIX semaphores
have a lot in common but there are differences as well. How hard it would be
to hack depends of how exactly SYS5 semaphores are used.
I’d start with reading docs on both to understand concepts and figure
differences.
- igor
“Aaron Macks” <aaronmNOSPAM@cs.brandeis.edu> wrote in message
news:9bg531$f8v$1@inn.qnx.com…
I know a fair bit of C and C++, but I’ve never used semaphores, how
difficult do you think the change will be?
AaronIn article
3ADB60FA.B5A5BF6A@motorola.com> >, Igor Kovalenko
Igor.Kovalenko@motorola.com> > wrote:
You need to replace SVS5 semaphores with POSIX ones. That’s a bit of
hacking.
- igor
Aaron Macks wrote:
Ok, so this problem was fixed by removing the offending file, I think I
don’t need for the manner in which I compile.Next Q, the software uses semaphores, but in the linking step complains
that semctl cannot be found any thoughts AaronAaron Macks <> aaronmNOSPAM@cs.brandeis.edu> > wrote:
(More on compiling MPI)
So I’ve gotten past the problem with types, but now the code wants
the header sys/shm.h which seems to deal with shared libraries. Does
this exist, or do we not need it? TIA Aaron
I had written a wrapper lib for sysv shared memory which just mapped
the sysv calls onto the posix calls. It wasn’t perfect (not all functionality
there) but for the most common use (ie. create memory, map it onto address
space, destroy it, etc.) it worked fine. It’s not really that hard to do
and it saves a huge amount of porting time because you just link to the
object file instead of rewriting the code. Perhaps I could post this somewhere
for people to see as an example…(if they’re interested in my 2 hour hack…)
cheers,
Kris
Alain Magloire <alain@qnx.com> wrote:
Aaron Macks <> aaronmNOSPAM@cs.brandeis.edu> > wrote:
: I know a fair bit of C and C++, but I’ve never used semaphores, how
: difficult do you think the change will be?Do not know, it depends on how “deep” is the use of SYSV semaphores, how
intertwine it is with the code i.e. do you have a nice porting layer.
Maybe the original author would be interrested in making his/her application
somewhat more portable.
–
Kris Warkentin
kewarken@qnx.com
(613)591-0836 x9368
“You’re bound to be unhappy if you optimize everything” - Donald Knuth
Subject:
Re: replacement for sys/shm.h
Date:
Tue, 1 May 2001 20:05:53 +0000 (GMT)
From:
Kris Warkentin <kewarken@qnx.com>
To:
Georg Ertl <geer@pfaffenhofen.de>
Okay, if you go to http://staff.qnx.com/~kewarken/ you’ll find the
tarball. It’s released under the QNX Open Community License so you can do
pretty much whatever you want with it. Could you do me a favor and
mention this fact on the newsgroup? I forgot where the original post
was…
cheers,
Kris Warkentin
kewarken@qnx.com
(613)591-0836 x9368
“You’re bound to be unhappy if you optimize everything” - Donald Knuth
On Tue, 1 May 2001, Georg Ertl wrote:
Hi,
I’m interested in your wrapper lib too!!
Thanks in advance!
Georg Ertl
email: > geer@pfaffenhofen.de
Kris Warkentin wrote:I had written a wrapper lib for sysv shared memory which just mapped
the sysv calls onto the posix calls. It wasn’t perfect (not all functionality
there) but for the most common use (ie. create memory, map it onto address
space, destroy it, etc.) it worked fine. It’s not really that hard to do
and it saves a huge amount of porting time because you just link to the
object file instead of rewriting the code. Perhaps I could post this somewherefor people to see as an example…(if they’re interested in my 2 hour hack…)
cheers,
Kris
Alain Magloire <> alain@qnx.com> > wrote:
Aaron Macks <> aaronmNOSPAM@cs.brandeis.edu> > wrote:
: I know a fair bit of C and C++, but I’ve never used semaphores, how
: difficult do you think the change will be?Do not know, it depends on how “deep” is the use of SYSV semaphores, how
intertwine it is with the code i.e. do you have a nice porting layer.
Maybe the original author would be interrested in making his/her application
somewhat more portable.–
Kris Warkentin
kewarken@qnx.com
(613)591-0836 x9368
“You’re bound to be unhappy if you optimize everything” - Donald Knuth
Kris you don’t mind to tip Garry so he can add MIT-SHM support into
Xphoton finally
Georg Ertl wrote:
Subject:
Re: replacement for sys/shm.h
Date:
Tue, 1 May 2001 20:05:53 +0000 (GMT)
From:
Kris Warkentin <> kewarken@qnx.com
To:
Georg Ertl <> geer@pfaffenhofen.deOkay, if you go to > http://staff.qnx.com/~kewarken/ > you’ll find the
tarball. It’s released under the QNX Open Community License so you can do
pretty much whatever you want with it. Could you do me a favor and
mention this fact on the newsgroup? I forgot where the original post
was…cheers,
Kris Warkentin
kewarken@qnx.com
(613)591-0836 x9368
“You’re bound to be unhappy if you optimize everything” - Donald KnuthOn Tue, 1 May 2001, Georg Ertl wrote:
Hi,
I’m interested in your wrapper lib too!!
Thanks in advance!
Georg Ertl
email: > geer@pfaffenhofen.de
Kris Warkentin wrote:I had written a wrapper lib for sysv shared memory which just mapped
the sysv calls onto the posix calls. It wasn’t perfect (not all functionality
there) but for the most common use (ie. create memory, map it onto address
space, destroy it, etc.) it worked fine. It’s not really that hard to do
and it saves a huge amount of porting time because you just link to the
object file instead of rewriting the code. Perhaps I could post this somewherefor people to see as an example…(if they’re interested in my 2 hour hack…)
cheers,
Kris
Alain Magloire <> alain@qnx.com> > wrote:
Aaron Macks <> aaronmNOSPAM@cs.brandeis.edu> > wrote:
: I know a fair bit of C and C++, but I’ve never used semaphores, how
: difficult do you think the change will be?Do not know, it depends on how “deep” is the use of SYSV semaphores, how
intertwine it is with the code i.e. do you have a nice porting layer.
Maybe the original author would be interrested in making his/her application
somewhat more portable.–
Kris Warkentin
kewarken@qnx.com
(613)591-0836 x9368
“You’re bound to be unhappy if you optimize everything” - Donald Knuth
Actually, Chris McKillop told me that he may be writing a sysv manager
with full support for shm, message queues, etc. I don’t know how high
up it is on his list but that would solve a lot of problems. Mine is a
very basic implementation which lacks support for the more esoteric things
that sysv shm is capable of. As long as you only want to create, attach,
etc. (basic functionality), it seems to work fine. Very limited testing
though so use at your own risk (I’d appreciate problem reports though).
Anyway, hopefully Chris will get that manager implemented and we’ll have
real support.
Cheers,
Kris
Igor Kovalenko <Igor.Kovalenko@motorola.com> wrote:
Kris you don’t mind to tip Garry so he can add MIT-SHM support into
Xphoton finally >Georg Ertl wrote:
Subject:
Re: replacement for sys/shm.h
Date:
Tue, 1 May 2001 20:05:53 +0000 (GMT)
From:
Kris Warkentin <> kewarken@qnx.com
To:
Georg Ertl <> geer@pfaffenhofen.deOkay, if you go to > http://staff.qnx.com/~kewarken/ > you’ll find the
–
Kris Warkentin
kewarken@qnx.com
(613)591-0836 x9368
“You’re bound to be unhappy if you optimize everything” - Donald Knuth
Kris Warkentin wrote:
Actually, Chris McKillop told me that he may be writing a sysv manager
with full support for shm, message queues, etc. I don’t know how high
up it is on his list but that would solve a lot of problems. Mine is a
very basic implementation which lacks support for the more esoteric things
that sysv shm is capable of. As long as you only want to create, attach,
etc. (basic functionality), it seems to work fine. Very limited testing
though so use at your own risk (I’d appreciate problem reports though).Anyway, hopefully Chris will get that manager implemented and we’ll have
real support.
I got impression that some indeed esoteric features of SYSV IPC can only
be supported on kernel level. That’s what stopped me from proceeding
with that idea a while ago.
- igor
Igor Kovalenko <Igor.Kovalenko@motorola.com> wrote:
Kris Warkentin wrote:
Actually, Chris McKillop told me that he may be writing a sysv manager
with full support for shm, message queues, etc. I don’t know how high
up it is on his list but that would solve a lot of problems. Mine is a
very basic implementation which lacks support for the more esoteric things
that sysv shm is capable of. As long as you only want to create, attach,
etc. (basic functionality), it seems to work fine. Very limited testing
though so use at your own risk (I’d appreciate problem reports though).Anyway, hopefully Chris will get that manager implemented and we’ll have
real support.
I got impression that some indeed esoteric features of SYSV IPC can only
be supported on kernel level. That’s what stopped me from proceeding
with that idea a while ago.
- igor
You may be right…certainly the posix shared memory was not capable of doing
some of the things that sysv shm needed…for instance, I couldn’t see any
way of changing permissions and ownership (but I’ve been known to not see things .
Either way, if anyone can do it, Chris can.
–
Kris Warkentin
kewarken@qnx.com
(613)591-0836 x9368
“You’re bound to be unhappy if you optimize everything” - Donald Knuth
Kris Warkentin wrote:
Igor Kovalenko <> Igor.Kovalenko@motorola.com> > wrote:
Kris Warkentin wrote:Actually, Chris McKillop told me that he may be writing a sysv manager
with full support for shm, message queues, etc. I don’t know how high
up it is on his list but that would solve a lot of problems. Mine is a
very basic implementation which lacks support for the more esoteric things
that sysv shm is capable of. As long as you only want to create, attach,
etc. (basic functionality), it seems to work fine. Very limited testing
though so use at your own risk (I’d appreciate problem reports though).Anyway, hopefully Chris will get that manager implemented and we’ll have
real support.
I got impression that some indeed esoteric features of SYSV IPC can only
be supported on kernel level. That’s what stopped me from proceeding
with that idea a while ago.
- igor
You may be right…certainly the posix shared memory was not capable of doing
some of the things that sysv shm needed…for instance, I couldn’t see any
way of changing permissions and ownership (but I’ve been known to not see things > > .
Either way, if anyone can do it, Chris can.
Permissions can be handled by SYSV resmgr. But things like ‘undoing’
semaphore operations are real pain in the neck. May be that can be
simulated by mutexes+condvars, but then it won’t be real semaphore and
it won’t be signal safe which almost defeats purpose. There are other
oddballs too …
On 3 May 2001, Kris Warkentin wrote:
Actually, Chris McKillop told me that he may be writing a sysv manager
with full support for shm, message queues, etc. I don’t know how high
up it is on his list but that would solve a lot of problems. Mine is a
very basic implementation which lacks support for the more esoteric things
that sysv shm is capable of. As long as you only want to create, attach,
etc. (basic functionality), it seems to work fine. Very limited testing
though so use at your own risk (I’d appreciate problem reports though).Anyway, hopefully Chris will get that manager implemented and we’ll have
real support.
Speaking of Chris, has he finished the installation of RTP on linux
yet? Last I checked, get.qnx.com only had qnxrtp.exe, qnxrtp.iso
and qnxrtp.tar.F.
Cheers,
Kris
Igor Kovalenko <> Igor.Kovalenko@motorola.com> > wrote:
Kris you don’t mind to tip Garry so he can add MIT-SHM support into
Xphoton finally >Georg Ertl wrote:
Subject:
Re: replacement for sys/shm.h
Date:
Tue, 1 May 2001 20:05:53 +0000 (GMT)
From:
Kris Warkentin <> kewarken@qnx.com
To:
Georg Ertl <> geer@pfaffenhofen.deOkay, if you go to > http://staff.qnx.com/~kewarken/ > you’ll find the
snip
\Kris Warkentin
kewarken@qnx.com
(613)591-0836 x9368
“You’re bound to be unhappy if you optimize everything” - Donald Knuth
Frank Liu <liug@mama.indstate.edu> wrote:
Speaking of Chris, has he finished the installation of RTP on linux
yet? Last I checked, get.qnx.com only had qnxrtp.exe, qnxrtp.iso
and qnxrtp.tar.F.
Eeek! My face is burning.
The install was finished ages ago but I saw some very strange things
happening on my ext2 drive after the install and so we didn’t release
it. I hope that it was just my machine but I honestly have not tried
it in sometime.
chris
\
cdm@qnx.com > “The faster I go, the behinder I get.”
Chris McKillop – Lewis Carroll –
Software Engineer, QSSL
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
On 4 May 2001, Chris McKillop wrote:
Frank Liu <> liug@mama.indstate.edu> > wrote:
Speaking of Chris, has he finished the installation of RTP on linux
yet? Last I checked, get.qnx.com only had qnxrtp.exe, qnxrtp.iso
and qnxrtp.tar.F.
Eeek! My face is burning. >The install was finished ages ago but I saw some very strange things
happening on my ext2 drive after the install and so we didn’t release
it. I hope that it was just my machine but I honestly have not tried
it in sometime.
Could it be that RTP doesn’t support some of the ext2 features?
btw, is “filetype” supported now?
http://qnx.iaware.org/pipermail/qdn.public.qnxrtp.os/2000-October/000078.html
anyway, if you aren’t making it public, how can other people test it?
Keep the good work chris!
frank
chris
\cdm@qnx.com > “The faster I go, the behinder I get.”
Chris McKillop – Lewis Carroll –
Software Engineer, QSSL