Subject: Re: [linux-audio-dev] Float to int conversions
From: Steve Harris (S.W.Harris_AT_ecs.soton.ac.uk)
Date: Thu Nov 01 2001 - 16:55:16 EET
On Thu, Nov 01, 2001 at 03:33:14PM +0100, Baker, Steve wrote:
> > >I have recently written a paper titled "Faster Floating
> > Point to Integer Conve
> > >rsions"
> > >which can be read here:
> > >
> > > http://mega-nerd.com/FPcast/
>
>
> There is some good stuff here. However in audio float to int conversions
> the operation really needs to be a rounding one in most cases as truncating
> the value introduces an error of 0.5 for negative values.
The asm macro at least will be rounding, as the float unit is left in its
defualt state, which is rounding under linux.
- Steve
This archive was generated by hypermail 2b28 : Thu Nov 01 2001 - 16:50:08 EET