Hello, folks.
Well, over the weekend I ported "tpr" (an old roff clone, except it's
named "tpr", except I enhanced it for my wife when she was writing a
book, and I named it back to "roff" :->). To benchmark it, I wanted to
redirect output to /dev/null--which I didn't have! So I wrote a /dev/null
server, and it worked fine.
But this got me thinking--you don't *really* need a server to get /dev/null
semantics. So I also added a file descriptor layer which detects that a
file descriptor is open to the null device, and it stubs the operation
right in the C library, without having to send a message to the /dev/null
server.
The result is that my wife book processes in 4 seconds--50 pages/second. :-)
It was a little slower from a DOS filesystem, but under the vstafs, the
file was a single contiguous extent, and was read in 4 disk I/O's (3 X 64K,
plus the remainder in the 4th I/O). Fun!
The filesystem is also coming along fine. My latest bug was traced to a
latent bug in the libc malloc() when freeing large blocks of memory.
Regards,
Andy Valencia
Received on Sun Oct 17 17:01:23 1993
This archive was generated by hypermail 2.1.8 : Wed Sep 21 2005 - 19:37:12 PDT