parallel programming survey results

[Andrae Muys]

On 22/02/2006, at 2:48 AM, Luke Gorrie wrote:

> Daniel Luna <daniel@REDACTED> writes:
>
>> "At [insert mailing list THEY sent to] there are probably a lot of
>> [insert programming language] programmers, so these figures cannot be
>> taken too seriously."
>
> I can sympathise with the surveyors. They seem to be interested in
> parallel programming as in "using lots of hardware in parallel to
> solve a problem" like SETI, CGI rendering farms, etc. I bet they got a
> lot of responses from Erlang people writing internet servers & telecom
> systems which are another kettle of fish entirely.
>
> I don't think they did a very good job of spelling out what
> applications they were interested in though.

You touch on two reasons why this paper was a waste of time.

The choice of forums to advertise the survey predicted the outcome.

Newgroups:
comp.parallel : a moderated FAQ and RFP newsgroup supporting the  
other two c.p.* groups.
comp.parallel.mpi : a C/C++/Fortran library for building distributed  
applications in C/C++/Fortran.
comp.parallel.pvm: a C/C++/Fortran library for building distributed  
applications in C/C++/Fortran.
comp.sys.super: another FAQ/RFP newsgroup.
aus.computers.parallel: There have been exactly 8 posts since your  
survey announcement in Oct (6 were spam)!  a dead-group.
comp.programming.threads: an active group discussing the use of  
threads in C/C++ especially pthreads.  If you doubt me, consider that  
the past 20 threads included 115 posts, of these 91 were exclusively  
C/C++/pthreads focused, 11 were lower level that language, 3 were  
spam, 2 were misposts and only 8 were remotely generic - none of  
which mentioned any language except C/C++.

Mailing Lists:
beowulf - good choice, but heavily dominated by MPI/PVM based  
applications.
mpi - C/C++/Fortran
pvm - C/C++/Fortran
OpenMP x2 - C/C++/Fortran
BSP - (from the website) "BSPlib can be used with C, C++, or Fortran."

I don't know anything about the forums,

Slashdot - good choice.

Note that despite the survey's claim to survey 'parallel  
programmers', and its stated desire to test the claim "Java threads  
is growing strong", the survey was restricted to users of C, C++, and  
Fortran.  It did not in any meaning way survey 'parallel  
programmers', or make any attempt to assess the use of Java in this  
field.

So the paper's results should have been summarised as

"If we ask C/C++/Fortran programmers what language they use they  
predominately answer C/C++/Fortran"

Instead you state:

"Obviously the hype around Java has managed to make the language  
known to
many programmers, but has not convinced too many of them to actually  
use it for parallel
programming."

A statement that is simply not supported by the methodology; the  
closest we might legitimately come from these results is:

"Of those programmers still using C, C++, or Fortran parallel  
programming technologies, very few use Java for their parallel  
programming."

Which again is hardly earth-shattering news.

The second reason is that that paper fails on a more fundamental  
level.  It completely fails to define 'parallel programming', or to  
provide any background to identify the nature and constituents of  
'the community' cited in the introduction.  It is this failure of  
definition that causes a problem when the paper hand-waves away the  
inconvenient response rate from the erlang community.  The erlang  
responses are not a problem in themselves, rather they highlight the  
intrinsic bias shown in the survey sample.  A truly representative  
survey would have identified a community of users who write parallel  
programs, and have targeted them with solicitations

Off the top of my head the following all make extensive use of  
parallel programming:
Chemists
Physicists
Biotech
Computational Mathematicians
Web application developers
Telecom Switch developers :)
Embedded system developers (in my experience mostly C++ using  
pthreads or roll-your-own coroutine/user-level-thread libraries)
Research Engineers (especially Mechanical)

Almost none of these were surveyed except those few that both use MPI/ 
PVM/OpenMP and follow the associated newsgroup/mailing-list.

This is appalling research, I am interested in knowing which journal  
accepted this paper so I know which journal I can avoid because any  
peer-review that passes this tautological waste of effort is not peer- 
reviewed in any meaningful sense of the phrase.

Andrae Muys

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20060222/29d71487/attachment.htm>