|
RE: Pure Eu file i/o
|
Cuvier Christian
|
Feb 20, 2006 02:00 PST
|
One of the possible reasons for such slowness could be the translation
process between Eu sequences and native arrays of chars. Using buffering and
threading will help, but everything that has a name will be processed more
slowly.
I could be interested, but will take my last summers leave next month
hopefully.
CChris
-----Message d'origine-----
De : Pete Lomax [mailto:petel-@blueyonder.co.uk]
Envoye : samedi 18 fevrier 2006 05:25
A : Ope-@topica.com
Objet : Pure Eu file i/o
I have been beavering away, on my compiler/interpreter; and got a good
way in but hit a snag over file i/o, especially performance. What I
want is a pure "Eu" version of file i/o (open, getc, gets, puts, seek,
where, flush ,close) which uses the windows API (ie GetStdHandle,
CreateFile, ReadFile, WriteFile, SetFilePointer, FlushFileBuffers,
CloseHandle, etc).
ie:
function open(sequence filename, sequence mode)
...
r=c_func(xCreateFile,...
Should you accept this challenge, I believe your first attempt will be
over one hundred times slower than native i/o; what I really want is
buffered i/o (perhaps) to get that down to say 4..10 times slower,
before I try again to translate to asm.
If anyone is interested, let me know. If necessary, I can throw you a
few files to get you started, but doing so might lead you up the same
blind ally I have gone down so it might be better not to.
Regards,
Pete
|
|
 |
|
