Processing workload distribution

[Joost Diepenmaat]

Good to know. Thanks. I didn't consider fork because i was under the
impression that it doesn't work right on Windows.

With regards to issues, fork() and threads should be more or less
equivalent on (native, non-cygwin) windows perls, since on windows fork
is implemented using threads. I vaguely recall that there may still be
some issues with fork() on windows, but they should be subtle as long as
you're not relying on /exact/ POSIX semantics for fork on windows.

In any case, threads (on any platform) and fork() on windows both have
their more-or-less subtle issues, so if it works for you, just use it,
but if you run into unexplained issues, they're definitely one of places
trouble may be looming.

 

[Joost Diepenmaat]

I assumed the former was just a special case of the latter. I was
concerned about the former, though, because at present I am merely
launched child processes using it. But the latter would always
create a nagging doubt, for when I need to make more interesting
threads.

System calls, on Unix, are a fairly small set of calls (a couple of
hundred, exact number depending on the OS) that can be made from the
program into the kernel. Those system calls basically define the minimum
functionality that programs may need to interact with the OS (and via
the os, with each other).

In addition, there are library functions for C language programs, a lot
of which are built on top of the system calls.

For example, fork(), wait() and exec() are always (well, probably
always) system calls on Unix, but system() is generally a library call
built on top of fork(), wait() and exec(), just like in general malloc()
is implemented as a library function built on the sbrk() syscall, which
is why perl can be built with different malloc() implementations on most
systems. Programmers not building compilers or other programs that need
to interact with the system on a very low level usually won't care which
is which.

Note that some of perl's built-in functions that interact directly with
the OS are sometimes called "syscalls" or "system calls" but may be
implemented in one of the standard C libraries or the perl binary itself
instead of directly in the OS kernel, and which is which may be
different on different OSs and perl builds.

 

[Ben Morrow]

I didn't make it up, I looked at previous postings in this NG where
someone has ithreads=define and not usemultiplicity.

You made up the bit about it affecting CPU affinity.

http://groups.google.com/group/comp.lang.perl.misc/browse_frm/thread/
8f27b699aff8a37d/18a233467aaa72d8?hl=en&lnk=gst&q=usemultiplicity#
18a233467aaa72d8

[URL broken. Has the 'net collectively forgotten about msg-ids and urls?]

From that posting:
| useithreads=define usemultiplicity=

I don't really know why that is; probably because the user didn't
explicitly ask for multiplicity. However, the important bits come later:

| Characteristics of this binary (from libperl):
| Compile-time options: DEBUGGING MULTIPLICITY USE_ITHREADS USE_LARGE_FILES
| PERL_IMPLICIT_CONTEXT

That list has both USE_ITHREADS and MULTIPLICITY in, so that perl was
built with both, as you can see from

| archname=i386-linux-thread-multi

right at the top.

Ben

 

[Ben Morrow]

[threads->create(sub { system "" }) vs.
async { system "" } vs.
system 1, "" on Win32
]

Are there performance implications for the various forms you suggest,

No. async {...} is just a shortcut for threads->new(sub {...}).

apart from the obvious that the last just has the overhead of creating
the needed processes while the others create the same processes within
threads that carry their own overhead. But I doubt that this overhead
could be significant when the scripts that are launched take hours to
complete.

I doubt *anything* you do in Perl will be significant compared with that
.

Well, I have MS Visual Studio 2003, so I suppose I could build it here
(with a 64 bit target) and deploy it at the office.

Hmmm... maybe, maybe not. You need to use *exactly* the same compiler;
in particular, using a different version of MSVC from the one perl was
built with causes the msvcrt.dll linking to go haywire. I've no idea
what compiler ActiveState use for Win64: for Win32 they use MSVC 6,
which is compatible with MinGW gcc, but somehow I doubt MSVC 6 has a
64-bit target.

I have yet to get cygwin to work on a 64 bit machine, so the 64 bit
machine does not have a native compiler installed.

Oh, no, a cygwin compiler won't do at all.

But wait a minute. On the 64 bit machine, 'ppm list' includes
libwin32 at the bottom of the list. And, I see in the documentation
Activestate provided with this build that Win32:rocess is available,
along with a long list of other Win32 modules. Mind you, I don't know
if it works since I just found it mere minutes ago.

D'oh! Of course, libwin32 has always been a standard part of ActivePerl
(that's where it came from in the first place). If it's there I'd expect
ActiveState to have tested it.

Ben

 

[Ted]

Hmmm... maybe, maybe not. You need to use *exactly* the same compiler;
in particular, using a different version of MSVC from the one perl was
built with causes the msvcrt.dll linking to go haywire. I've no idea
what compiler ActiveState use for Win64: for Win32 they use MSVC 6,
which is compatible with MinGW gcc, but somehow I doubt MSVC 6 has a
64-bit target.

While I have MSVS6, you could not pay me to install something that
antiquated on my system. Being used to a reasonably standard
compliant C++ compiler, using some that nacient and broken would be
much too painful.

Oh, no, a cygwin compiler won't do at all.

Why not? Isn't it just another build of gcc, with libraries required
for programs to work on Windows?

D'oh! Of course, libwin32 has always been a standard part of ActivePerl
(that's where it came from in the first place). If it's there I'd expect
ActiveState to have tested it.

Good to know.

Thanks

Ted

 

[Joost Diepenmaat]

Why not? Isn't it just another build of gcc, with libraries required
for programs to work on Windows?

Cygwin's compiler is fine, if you're building for cygwin perl. If you're
building modules for activeperl you probably do need a VC compiler that
matches the one used by active state.

Another perl version that ships with a (free) compiler suite is
strawberry perl. It's pretty new, but you might get it to work for your
stuff.

 

[smallpond]

Quoth smallpond <[email protected]>:

I didn't make it up, I looked at previous postings in this NG where
someone has ithreads=define and not usemultiplicity.

You made up the bit about it affecting CPU affinity.

http://groups.google.com/group/comp.lang.perl.misc/browse_frm/thread/
8f27b699aff8a37d/18a233467aaa72d8?hl=en&lnk=gst&q=usemultiplicity#
18a233467aaa72d8

[URL broken. Has the 'net collectively forgotten about msg-ids and urls?]

From that posting:
| useithreads=define usemultiplicity=

I don't really know why that is; probably because the user didn't
explicitly ask for multiplicity. However, the important bits come later:

| Characteristics of this binary (from libperl):
| Compile-time options: DEBUGGING MULTIPLICITY USE_ITHREADS USE_LARGE_FILES
| PERL_IMPLICIT_CONTEXT

That list has both USE_ITHREADS and MULTIPLICITY in, so that perl was
built with both, as you can see from

| archname=i386-linux-thread-multi

right at the top.

Ben

The post is about this exact problem and a poster suggests
checking usemultiplicity. I inferred from that that they
were related since I see nothing in the perl documentation.

I find this in README.cygwin:
"Multiplicity is required when embedding Perl in a C program and using
more than one interpreter instance. This works with the Cygwin port."

It says nothing about ithreads. Maybe you made that up?

 

[Ben Morrow]

The post is about this exact problem and a poster suggests
checking usemultiplicity. I inferred from that that they
were related since I see nothing in the perl documentation.

Then both the suggestion and your inference were incorrect. This is
understandable, as the documentation is a rather unclear in this area.

I find this in README.cygwin:
"Multiplicity is required when embedding Perl in a C program and using
more than one interpreter instance. This works with the Cygwin port."

It says nothing about ithreads. Maybe you made that up?

No. From perl.h:

#ifdef USE_ITHREADS
# if !defined(MULTIPLICITY)
# define MULTIPLICITY
# endif
#endif

so you can't have USE_ITHREADS without MULTIPLICITY. More obviously,
since multiplicity is what lets you have several Perl interpreters in
one process, and ithreads create a new interpreter for each thread, you
obviously can't have one without the other. And neither has anything to
do with thread scheduling, which, as I said, is entirely under the
control of your OS's thread implementation.

Ben

 

[Ben Morrow]

Cygwin's compiler is fine, if you're building for cygwin perl. If you're
building modules for activeperl you probably do need a VC compiler that
matches the one used by active state.

32-bit ActivePerl is built with MSVC6. 32-bit MinGW gcc is
binary-compatible with MSVC6, and 32-bit ActivePerl ships with support
for using gcc to build extensions. Cygwin gcc, while essentially the
same compiler, has different default options (for instance, it links
with Cygwin's libc.dll by default) so it would be possible to make it
work, but difficult; and of course you'd need the MinGW import libs and
so on.

Another perl version that ships with a (free) compiler suite is
strawberry perl. It's pretty new, but you might get it to work for your
stuff.

I looked at that first . Adam isn't yet shipping a 64-bit build. I
don't even know if gcc will build Win64 binaries.

Ben

 

[xhoster]

The post is about this exact problem and a poster suggests
checking usemultiplicity.

I was that poster, and what I tentatively suggested was that
multiplicity looked kind of odd and perhaps should be looked into
as we had run out of more concrete ideas. Sometimes we bounce
ideas off each other here, which are intended to be less than the Word
of God.

Xho

--
-------------------- http://NewsReader.Com/ --------------------
The costs of publication of this article were defrayed in part by the
payment of page charges. This article must therefore be hereby marked
advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate
this fact.

 

[smallpond]

I was that poster, and what I tentatively suggested was that
multiplicity looked kind of odd and perhaps should be looked into
as we had run out of more concrete ideas. Sometimes we bounce
ideas off each other here, which are intended to be less than the Word
of God.

Xho

--
--------------------http://NewsReader.Com/--------------------
The costs of publication of this article were defrayed in part by the
payment of page charges. This article must therefore be hereby marked
advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate
this fact.

and it looked like a good shot to me since the perl docs weren't
much help.
--S