mark@mips.COM (Mark G. Johnson) (09/11/90)
In article <1477@marlin.NOSC.MIL> aburto@marlin.nosc.mil.UUCP (Alfred A. Aburto) writes: > >I got a feeling that the 040 will do the transcendental functions with ieee >sp (32-bit) real fast (probably quicker and just as accurate as the 68882 >at the same clock), but the ieeedp (64-bits) is going to require a bit of >igenuity (magic) me thinks. Just my opinion at this time..... I suspect the 040 will be able to keep with, and probably beat, the 882 on IEEE 64bit transcendental functions. Have a look at IEEE Micro, June 1988, p. 54. There's a table showing the cycle counts for doing the trigonometric function "sine", for the '882 and for a RISC machine that only implements four FP operations *,-,/,+. The 882 takes three times as many clocks as the RISC. (Both @ 25MHz -- remember, it was 1988). I suspect the 040 is nowhere near as bad as 3x worse than a RISC :-), meaning that the 040 will also beat the 882. Disclaimer, caveat, and surgeon general's warning: I was one of the authors of the IEEE Micro article. -- -- Mark Johnson MIPS Computer Systems, 930 E. Arques M/S 2-02, Sunnyvale, CA 94086 (408) 524-8308 mark@mips.com {or ...!decwrl!mips!mark}