Messages in this thread | | | Date | Sun, 6 Jun 1999 18:31:10 +0300 | From | Matti Aarnio <> | Subject | Re: EPIC |
| |
On Sun, Jun 06, 1999 at 04:02:27PM +0200, Erik Corry wrote: > "Matti Aarnio" <19990606155651.R30475@mea.tmt.tele.fi> wrote: > > per single clock cycle. EPIC in itself does not give you > > anything special, likely it will cause more constraints to > > Did you read the PDF file on Merced before saying this?
No, when such has appeared ? ... A week ago ?
Now that I have browsed a bit of that document which you referred here, I still don't think it supplies anything special. In fact I think it is in amazingly lots of ways like SPARCs are.
128 integer registers, 128 FP registers, and lots of other user-level registers, many of which need saving in context switches ... Looks like intel has joined SPARC family in register window technology (there was a note about register windowing exploitable at function calls), and various pains which relate to massive amounts of registers to be saved/loaded during a context switch.
Where ia-32 is register-poor, at least its context switches are fairly fast...
The instruction format expects 3+1 longwords (32-bits each) worth of instruction stream data, and although its ways to explicitely combine integer and float instructions are more versatile (=complicated) than what DEC 21264 Alpha can do, it can barely match said Alpha processor in speed of instructions per clock cycle. (That alpha can do up to 2i+2fp instructions per clock.)
Sure it can do wonders for a speed-daemon usages, like me using pico to edit email, but a 386 can do that too...
Getting the best out of that instruction format in applications does require lots of smarts in the compiler, and (I hazard a guess) application which does lots of things without frequent function calls, or system calls.
There definitely is no amazing magic bullet to speed things up tremendously. It might be a decent tool for heavy computing environments, but it is not alone at it.
> -- > Erik Corry erik@arbat.com Ceterum censeo, Microsoftem esse delendam!
/Matti Aarnio <matti.aarnio@sonera.fi>
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |