This is the mail archive of the
newlib@sources.redhat.com
mailing list for the newlib project.
Re: libm performance
- To: "Silverio Diquigiovanni" <silverio dot di at qem dot it>, "newlib read hat" <newlib at sources dot redhat dot com>
- Subject: Re: libm performance
- From: "Tim Prince" <tprince at computer dot org>
- Date: Tue, 3 Oct 2000 06:25:23 -0700
- References: <39D981E3.1419AF80@qem.it>
Consider the example of glibc for x86; the <mathinline.h> has been in use
for some time, and the linkable functions are written in .s. No one has
seen fit to correct bugs such as those which limit the accuracy of the long
double versions, or produce un-warranted NaN's in limiting cases. Yet, you
are free to supply your own <mathinline.h> or link any of your own function
versions before linking against the library, so that you may easily take
responsibility yourself for the performance of your libm.
----- Original Message -----
From: "Silverio Diquigiovanni" <silverio.di@qem.it>
To: "newlib read hat" <newlib@sources.redhat.com>
Sent: Monday, October 02, 2000 11:51 PM
Subject: libm performance
> Hi to all,
> I'm a newbie in this area.
>
> I'm a little worry for libm performance.
> By reading the source code I have see which all libm is in C.
>
> In future is planed the possibility of rewrite part of libm in asm
> for some micro ? (I think to Hitachi SH )
>
> Ciao,
> Silverio Diquigiovanni