Author: Robert Hyatt
Date: 14:24:58 06/03/98
Go up one level in this thread
On June 03, 1998 at 13:11:26, Tom Kerrigan wrote: >Bob, I think you misunderstand the point of VLIW. > >The P6 needs a significant amount of circuitry to schedule instructions, >and it only has two integer execution units. The logic necessary to >handle only three execution units is insane, and four is probably >impossible. > >Compare this to a VLIW design. With VLIW you can scrap all of your >scheduling logic and add execution units until your heart's content. sure you can... but it doesn't work. It is "an old idea that didn't work then and doesn't work now".. and when you look at it under a magnifying glass, it is simply a restricted form of superscalar, as was done in the first chips (old pentiums come to mind) where the compiler had to pair instructions. And there are only so many parallel threads of execution in a typical program... and VLIW fails in those just like the normal super-scalar designs.. except that for most programs, the P6 type instruction pool is far more efficient, because when you can't use all the VLIW operations, you pass noops in those fields, and that consumes bandwidth, and produces no results. So neither really has any big advantage over the other, excepting our current P6 architecture makes the compiler's life a whole lot easier, and it makes binary compatibility about 10 times easier... > >Regarding the VLIW clock speed issue, I think clock speed was confused >with number of execution units. The clock speed doesn't make a >difference with a VLIW processor, but depending on the instruction set, >the number of execution units can. > >With IA64, Intel is basically building a VLIW processor that doesn't >take VLIW instructions. Instead, the instruction stream contains data >about which instructions can be run in parallel, which have what memory >dependancies, etc. Thus, the scheduling logic on the processor dosen't >have to be terribly complicated. which, when you think about it, is how everything since the first 586 has sort of worked... > >Cheers, >Tom
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.