[virtmach] Re: Transmeta Crusoe

Manoj Plakal plakal@cs.wisc.edu
Thu, 3 Feb 2000 10:44:05 -0600


David Rush wrote (Thu, Feb 03, 2000 at 03:13:58PM +0000) :
> The 'Bulldog' (IIRC) basic ran anywhere from 4 to 16 CPUs in parallel
> on one chip, splitting the instruction stream at each branch
> instruction, *executing both branches* until it was clear which one
> was actually the correct one for the data. I know this sounds
> confusing, but they claimed that they had a very smart compiler
> technology behind this that made it all automagically work out.

	I think Bulldog was the name of the compiler (it was
	the PhD thesis of John Ellis). The CPU itself was
	either the Multiflow or the Cydrome (unless I am	
	confusing the name of the chip with the name of
	the company).


> If they're going to sell the CPUs, they'll sell the debuggers
> eventually. Just wait.

	They are going to sell the CPUs to OEMs who will
	put it in laptops/webpads/other devices. I'm not	
	sure that these OEMs will do much debugging of
	the morpher, maybe just report bugs back to Transmeta.

	I can see Transmeta supplying tools (but not
	publicly sold) to OS vendors who want to port
	their x86 OSes to run on Crusoe without problems.

	End users will never see the native ISA of Crusoe,	
	all they see is they are running unmodified x86 apps
	with no apparent access to the underlying hardware.
	Why would Transmeta sell a debugger to an end-user?

	Manoj