🚀 go-pugleaf

In article <1992Aug27.102130.9367@cam-orl.co.uk> thg@cam-orl.co.uk (Tim Glauert) writes:
>In article <19920826.074333.779@almaden.ibm.com>, ADUNSMUI@TOROLAB6.VNET.IBM.COM (Al Dunsmuir) writes:
>|>   Having said that, having a math coprocessor (especially a tightly coupled
>|> one as implemented in the 486) can speed up OS/2 applications.  For example,
>|> IBM's X-Windows server (PM screen interface for TCP/IP) achieves __dramatic__
>|> performance gains when a math coprocessor is present.
>
>I wonder why that is?  The MIT X sample server only uses floating-point code
>where the integer equivalent is likely to be slower, and only for primitives
>such as arcs and wide lines.

They are probably determining whether integer code is likely to be
slower on *workstations*, where floating point hardware is the norm
and is probably only slower than integer code by a factor of two or
three.  On a 386 without a 387, floating point code is slower by
several orders of magnitide, worse if the 387 instructions are being
trapped and emulated.
-- 
Richard Krehbiel                                 richk@grebyn.com
OS/2 2.0 will do for me until AmigaDOS for the 386 comes along...