🚀 go-pugleaf

In article <1989Oct12.162236.24239@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes:
>Bootstraps written before the days of decent-sized EPROMs often employ
>all kinds of dirty tricks to minimize size. 
 describes PDP8 bootstrap as...
>..  two instructions.  The first was "kick disk
>controller" (the controller handily came up in a fairly reasonable state
>after bus reset).  The second was "branch to self".[...]
Eons ago we had an IBM 1130 to boot which, you inserted a special binary
punched card in the reader and pressed "Program Load". The 1130 was a
16bit machine with an instruction format like (I think, it's been a while)
 opcode  Long Indirect Index_Reg rest
[4 bits]  [1]    [1]       [2]   [8]
But, an 80col card has only 12 rows, so the load read 12 bits into each
16bit word as [opcode] [4 zeros] [rest]. All very well, except that the XIO
(eXecute Input Output) instruction had to be Long, so the bootstrap program
spent its first few instructions franticly modifying itself to create the
reuquired instructions.
The IBM supplied card used most of the 80 cols and a friend of mine
decided he could do better. The result was a program that modified itself
to create an XIO to read in sector 0 of the disk, which it then
executed... at about which point the program ended. Sector 0 was read in,
containing, among other things, the interrupt handling routines for the
interrupt that resulted when the I/O completed, and some code that was in
the right place by the time the bootstrap program reached it... The 1130
was not noted for its speediness. The new bootstrap card only needed about
1/3 of a card. 

--
Michael Newbery<newbery@rata.vuw.ac.nz> (...!uunet!vuwcomp!newbery if you must)
Abstain from wine, women and song. Mostly song.