🚀 go-pugleaf

In article <9raH6#es6@cs.psu.edu> melling@cs.psu.edu (Michael D Mellinger)  
writes:
> 
> In article <1992Apr16.180949.26816@cs.ucla.edu> esky@marathon.cs.ucla.edu  
(Eskandar Ensafi) writes:
> 
>    Does anyone know how many MFLOPS the Motorola 56001 DSP and 88110
>    CPU are capabale of?  I am trying to compare this to the 80 MFLOPS
>    of the 33-MHx i860 and to the 50 MFLOPS of the 33-MHz 96002 DSP.
> 
> The i860 doesn't get 80 MFLOPS in practice.  10 MFLOPS is the number
> that I hear a lot.  Anyway, the i860 isn't used for crunching on the
> ND, it's just used for bitblits.
> 

-Despite Intel's best efforts to pass i860 off as a general purpose chip,
It does not run at anywhere near 80 mflops. It's a great graphics chip,
but the special features that allows 80 mflops in graphics ops makes
building a robust compiler difficult. There isn't a compiler in existance 
that will get faster than 10 Mflops on a simple SAXPY loop, never mind a 
REAL operation. Even people writing hand-coded assembler have a tough time  
getting more than 30 Mflops.  I've yet to get any of the test examples I've  
been working with to run at more than 3 mflops. When you compare that to the
2 mflops available on my lowly 15mhz '040 , it doesn't look that great. 

- Anyway, NeXT should be looking at the new 64-bit chips, although how you'll 
ever get a memory system cheap and fast enough to feed such a beasty is still 
an open question. 

---
- Booker C. Bense                    
prefered: benseb@grumpy.sdsc.edu  "I think it's GOOD that everyone 
NeXT Mail: benseb@next.sdsc.edu            becomes food " - Hobbes

P.S. If you know of an i860 compiler that does better than 10 Mflops 
PLEASE FLAME ME AS AN IGNORANT FOOL AND SEND ME THE ADDRESS !!!!!