Subject: Re: bad performance?
To: None <>
From: Mike Long <>
List: current-users
Date: 12/07/1993 18:05:56
>From: "Chris G. Demetriou" <cgd@nobozo.CS.Berkeley.EDU>
>Date: Tue, 07 Dec 93 14:18:52 -0800
>> It took 330 seconds on my DX50 (NetBSD current).
>i just tried the program on sun-lamp. in doing so,
>i noticed that it needs libm.  that's the problem right there.
>NetBSD's libm uses C functions to do all of the math operations,
>because the emulator does poorly enough on some of them that using
>the FP instructions is "dangerous"...

So, even if I have a coprocessor and remove 'option MATH_EMULATE' from
my kernel config, I'm *still* not using the coprocessor?

>what really needs to happen is the emulator needs to be improved,
>so that we can use a rational libm (i.e. one that acutally
>uses the math coprocessor)...

Does anyone have an alternate libm that does use the coprocessor?


Mike Long                               
VLSI Design Engineer                              voice: (617)461-4030
Analog Devices, SPD Div.                            FAX: (617)461-3010
Norwood, MA 02062                            *this = !opinion(Analog);