Angus Robertson - Magenta Systems Ltd wrote: >> I've uploaded a tiny buffered file stream class as well >> as a simple test program. It is amazing fast when small >> chunks are read/written. Seeking is slower than TFileStream :( > > I'll try in one of my applications, but it won't be until next week. > > For the FTP server, I often have multiple PCs downloading the same > files, in some cases 30 or 40 megs. With the normal TFileStream, > presumably Windows will be effectively buffering the file once, and each > separate stream will be reading from common windows buffers, in the > normal read chunks specified by the server. I've never really > understood whether this was efficient, or just simple. > > So what are the implications of buffering the same file multiple times? > Presumably just the extra memory for the buffer?
You save a lot of API calls. It depends on the ratio of block size and buffer size. The SmtpCli reads files byte by byte hence it benefits from a buffered stream enormously. > > I have a couple of special files that are read thousands of times of > day, both less than 1 meg that I could keep in a memory stream for > efficiency. I think that's the fasted way of caching. > > Angus -- To unsubscribe or change your settings for TWSocket mailing list please goto http://www.elists.org/mailman/listinfo/twsocket Visit our website at http://www.overbyte.be