Subject: Re: soft float and binary compatibility
To: TAKEMURA Shin <takemura@netbsd.org>
From: Jason Thorpe <thorpej@nas.nasa.gov>
List: port-hpcmips
Date: 02/27/2000 12:17:14
On Sun, 27 Feb 2000 15:43:19 +0900 
 "TAKEMURA Shin" <takemura@netbsd.org> wrote:

 > I want to delete SOFTFLOAT option to get compatibility with pmax and
 > arc.
 > 
 > Currentry, hpcmips's user-land binaries are compiled with SOFTFLOAT
 > option. SOFTFLOAT binaries are possibly a little faster than binaryies
 > with kernel floating point emulation (HARDFLOAT). But it lacks binary
 > compatibility with other mips little endian platforms. Because the register
 > usage of function calling is compleately different between SOFTFLAT and
 > HARDFLOAT, dinamic linking liblaries will lose. And I feel that you can
 > scarely profit by the performance advantage.
 > 
 > Any objection ?

No objection from me... but it would be nice to know how much of a
performance difference there really is.

Though, not many thing in userland really use the FPU that much ... except
the compiler may do bad things like emit floating point ops for inlined
memcpy's, etc.  That would be a major lose.

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>