This is the mail archive of the
newlib@sourceware.org
mailing list for the newlib project.
Re: libm support for ARMv7 and VSQRT?
- From: "Richard Earnshaw (lists)" <Richard dot Earnshaw at arm dot com>
- To: Sebastian Huber <sebastian dot huber at embedded-brains dot de>, "newlib at sourceware dot org" <newlib at sourceware dot org>
- Date: Mon, 20 Mar 2017 13:31:21 +0000
- Subject: Re: libm support for ARMv7 and VSQRT?
- Authentication-results: sourceware.org; auth=none
- References: <58CF7C40.3060609@embedded-brains.de>
On 20/03/17 06:52, Sebastian Huber wrote:
> Hello,
>
> there is some ARM-specific libm support for ARMv8:
>
> https://sourceware.org/git/gitweb.cgi?p=newlib-cygwin.git;a=tree;f=newlib/libm/machine/arm;h=5457f11619b2d67e8f56f352afb404c06a12c2bb;hb=HEAD
>
>
> The VFPv2+ supports a VSQRT instruction. Would it make sense to use it
> for the sqrt() implementation, similar to
>
> https://sourceware.org/git/gitweb.cgi?p=newlib-cygwin.git;a=blob;f=newlib/libm/machine/aarch64/e_sqrt.c;h=e4b7554ac1811bf04f726da198b913c0c30595ca;hb=HEAD
>
> https://sourceware.org/git/gitweb.cgi?p=newlib-cygwin.git;a=blob;f=newlib/libm/machine/aarch64/ef_sqrt.c;h=19fec952517769ad3851efef3b11dc6c02fb7582;hb=HEAD
>
>
> ?
>
I'm sure a patch would be welcome :-)
Don't forget, however, that you also need to deal with the soft-float
case (something the aarch64 code does not have to worry about).
R.