I was wondering if anyone has experience with using some kind of distributed
make under QNX?
What I would like to do - is to utilize several QNX and possible Win32
machines connected through a LAN to build/compile large projects, and I was
wondering if anyone has tried to do something like that.
I was wondering if anyone has experience with using some kind of
distributed
make under QNX?
What I would like to do - is to utilize several QNX and possible Win32
machines connected through a LAN to build/compile large projects, and I
was
wondering if anyone has tried to do something like that.
I played with that option under QNX4 quite extensively on 100Mbits network
(maybe 1Gig would change something) but I was never able to get it to
compile any faster then on a single machine.
“Jens H Jorgensen” <> jhj@remove-nospam-videk.com> > wrote in message
news:b33eoc$7bl$> 1@inn.qnx.com> …
I was wondering if anyone has experience with using some kind of
distributed
make under QNX?
What I would like to do - is to utilize several QNX and possible Win32
machines connected through a LAN to build/compile large projects, and I
was
wondering if anyone has tried to do something like that.
I played with that option under QNX4 quite extensively on 100Mbits network
(maybe 1Gig would change something) but I was never able to get it to
compile any faster then on a single machine.
Agreed. You can try to convince the makes that the header files are
always on the local filesystem. That will help.
The only time I saw real bebefit was when you haev heavily nested C++
inlineing going on. The compiler would spin for quite a while without
attempting to do any IO. Then multiple CPUs were a benefit.
Otherwise, the slowness of the network just destroys any benefit
otherwise gained.
“Jens H Jorgensen” <> jhj@remove-nospam-videk.com> > wrote in message
news:b33eoc$7bl$> 1@inn.qnx.com> …
I was wondering if anyone has experience with using some kind of
distributed
make under QNX?
What I would like to do - is to utilize several QNX and possible Win32
machines connected through a LAN to build/compile large projects, and I
was
wondering if anyone has tried to do something like that.
I played with that option under QNX4 quite extensively on 100Mbits
network
(maybe 1Gig would change something) but I was never able to get it to
compile any faster then on a single machine.
Agreed. You can try to convince the makes that the header files are
always on the local filesystem. That will help.
The only time I saw real bebefit was when you haev heavily nested C++
inlineing going on. The compiler would spin for quite a while without
attempting to do any IO. Then multiple CPUs were a benefit.
Otherwise, the slowness of the network just destroys any benefit
otherwise gained.
“Jens H Jorgensen” <> jhj@remove-nospam-videk.com> > wrote in message
news:b33eoc$7bl$> 1@inn.qnx.com> …
I was wondering if anyone has experience with using some kind of
distributed
make under QNX?
What I would like to do - is to utilize several QNX and possible Win32
machines connected through a LAN to build/compile large projects, and I
was
wondering if anyone has tried to do something like that.
I played with that option under QNX4 quite extensively on 100Mbits
network
(maybe 1Gig would change something) but I was never able to get it to
compile any faster then on a single machine.
Agreed. You can try to convince the makes that the header files are
always on the local filesystem. That will help.
Note that this was with rather slow machine compare to todays standard so I
would guess the network would become even worst of a bottleneck.
The only time I saw real bebefit was when you haev heavily nested C++
inlineing going on. The compiler would spin for quite a while without
attempting to do any IO. Then multiple CPUs were a benefit.
Otherwise, the slowness of the network just destroys any benefit
otherwise gained.
“Bill Caroselli” <> qtps@earthlink.net> > wrote in message
Agreed. You can try to convince the makes that the header files are
always on the local filesystem. That will help.
Note that this was with rather slow machine compare to todays standard so I
would guess the network would become even worst of a bottleneck.
“Bill Caroselli” <> qtps@earthlink.net> > wrote in message
Agreed. You can try to convince the makes that the header files are
always on the local filesystem. That will help.
Note that this was with rather slow machine compare to todays standard
so I
would guess the network would become even worst of a bottleneck.
Agreed. But it was also with 10-Base-T.
No 100Mbits, that was when 100T would cost a fortune.
Agreed. You can try to convince the makes that the header files are
always on the local filesystem. That will help.
Note that this was with rather slow machine compare to todays standard
so I
would guess the network would become even worst of a bottleneck.
Agreed. But it was also with 10-Base-T.
No 100Mbits, that was when 100T would cost a fortune.
The latest standard is 10GBit ethernet; although the fastest
driver for QNX6 is 1GBit. Don’t think the data throughput is really
a big issue. I suspect that since most source file are small (in
real projects) it is the aggregate latency of the open/close
transactions that really makes distributed makes useless.
Agreed. You can try to convince the makes that the header files are
always on the local filesystem. That will help.
Note that this was with rather slow machine compare to todays standard
so I
would guess the network would become even worst of a bottleneck.
Agreed. But it was also with 10-Base-T.
No 100Mbits, that was when 100T would cost a fortune.
The latest standard is 10GBit ethernet; although the fastest
driver for QNX6 is 1GBit. Don’t think the data throughput is really
a big issue. I suspect that since most source file are small (in
real projects) it is the aggregate latency of the open/close
transactions that really makes distributed makes useless.
That would get worst with QNX6 since open operation are much slower then for
QNX4. I just checked and could find how to use distributed make with gmake.
Since gmake as not concept of qnet it can’t use it.
Did you try it out - if so what is your experience with it?
PVM works very stable (e.g. for an platform independent CAN API) … but
I don’t have experiences with distributed make. IMHO … PVMMAKE looks
realy professional …