[Gllug] Kernel compile taking forever...!

[Richard Jones]

On Thu, Feb 03, 2005 at 03:10:16PM +0000, t.clarke wrote:
> Talking about registers, does anyone out of curiosity, like me,
> remember the TI-99(00) family which used in-memort 'registers' with
> a single workspace- pointer in the CPU, which allowed for very fast
> context-switching (at the expense of register access speed)?

No, but the Z80 could do this.  It had two banks of registers and two
separate accumulator/flag registers, and you could toggle between them
using the EXX and EX AF,AF' instructions respectively.  Main use was
for fast interrupt handling, since it was much faster to use the
alternate bank during interrupts than to PUSH affected registers onto
the stack and POP them off at the end.

http://www.z80.info/gfx/fig214.gif

> I think the logic at the time was that the memory which pretty much
> the same speed as the CPU registers.  They later added a 'workspace
> cache' when cpus got noticeably faster than memory.

Memory was pretty much the same speed, but actually still a little bit
slower.  IIRC it took three or four cycles to perform a memory load,
versus one to spill a register over the internal databus into the temp
register at the front of the ALU.  Nothing like today, of course,
where the penalty for going to memory is measured in the hundreds of
cycles!

Rich.

-- 
Richard Jones, CTO Merjis Ltd.
Merjis - web marketing and technology - http://merjis.com
Team Notepad - intranets and extranets for business - http://team-notepad.com
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug