[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

fast io, number conversions, etc.

Hello experts!
Any pointers to, or new code for these sorts of things? 

1.  Fast file I/O.  especially for arrays or long vectors.  Perhaps block
moves (a la Pascal?).

2. Conversion of numbers (such as a VAX FORTRAN real*4) read into MCL from
a file as bytes with (read-byte). (VAX FORTRAN has the rather scuzzy but
useful capability to equivalence a byte array to a real number, which is
expressly forbidden with (make-array ... :displaced-to ...).   It's the
replacement for this type of think that I'm looking for - take 4 bytes and
munge them into a floating point number.)

3. A collection of useful declarations that speed up commonly used
operations on arrays of numbers, eg:  (slow code with no declarations),
(fast code with spiffy declarations), comments on speedup.  An example of
what I have in mind would be adding two integer arrays (such as images) to
get a third, element by element -- this typically uses lots of overhead in
calculating the indices.

Any help is appreciated.  
David S. Bright                        bright@enh.nist.gov  
Microanalysis Research Group                                   
National Institute of Standards & Technology (NBS)
Gaithersburg, MD 20899