henry@utzoo.UUCP (Henry Spencer) (10/18/84)
Now running a new /unix, the same one that was suspected bad during the
confusion at noon today. It seems to be working fine; the earlier problems
were almost certainly due to the write-protect problem and not the kernel.
The major changes are hashing for the inode table (formerly accessed by a
horribly-inefficient sequential search) and a freelist for the file table
(formerly searched for free entries by, you guessed it, a sequential search).
Minor changes at the same time were movement of the Sccsid strings from the
data space (which is tight) to the text space (which isn't), and the deletion
(finally!) of the floating-point simulator from the kernel. The simulator
code actually hasn't been executed in over two years, but it was still
hanging around taking up space. The net result of all this is that, despite
some extra data structures added to support the inode/file table changes,
the new kernel is noticeably smaller than the old one.
--
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henry