[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [Bug-gsl] gsl version 1.5 - A few small source code issues
From: |
Peter Rockett |
Subject: |
RE: [Bug-gsl] gsl version 1.5 - A few small source code issues |
Date: |
Fri, 6 Aug 2004 14:32:43 +0100 |
Thanks for the reply.
The second point is a subtlety I hadn't appreciated!
On the /utils directory, I will have a look at a fix. Certainly MSVC++
v6 does not like the stro(u)l code.
Peter Rockett
-----Original Message-----
From: Brian Gough [mailto:address@hidden
Sent: 06 August 2004 12:18
To: Peter Rockett
Cc: address@hidden
Subject: Re: [Bug-gsl] gsl version 1.5 - A few small source code issues
Peter Rockett writes:
> A couple of small things in the source code of version 1.5: >
> 2)The functions in the /util sub-directory seem to implement things
like > memcpy() if you don't have a system-defined version of memcpy(),
etc....
The utils directory is old and taken from a standard gnu template. If
it is not causing a problem I would prefer to leave it alone, but if you
have a specific error I will fix it.
> Incidentally, shouldn't functions like gsl_vector_memcpy and >
gsl_matrix_memcpy use the memcpy function for efficiency rather than >
nested for-loops? If the compiler doesn't have memcpy then the version
> in /utils would be called instead. Should the /util functions be >
inlines, again for efficiency?
The memory layout is not contiguous (stride, tda) so memcpy can only be
used in special cases (stride=1, tda=size2). Optimised for loops
perform reasonable well in my experience, but if you have a benchmark
that shows differently I would be interested to look at it.
--
Brian Gough