[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20150708141500.GN1173@brightrain.aerifal.cx>
Date: Wed, 8 Jul 2015 10:15:00 -0400
From: Rich Felker <dalias@...c.org>
To: musl@...ts.openwall.com
Subject: Re: [PATCH] Add PowerPC soft-float support
On Wed, Jul 08, 2015 at 01:59:58PM +0200, Felix Fietkau wrote:
> Some PowerPC CPUs (e.g. Freescale MPC85xx) have a completely different
> instruction set for floating point operations (SPE).
> Executing regular PowerPC floating point instructions results in
> "Illegal instruction" errors.
>
> Until support for SPE FPU is added, make it possible to run these
> devices in soft-float mode.
I don't think musl can support it: according to Wikipedia it's not
IEEE conforming, but it seems to use integer registers, so if the
calling convention is soft-float-ABI compatible, you could build musl
with soft-float then use this broken FPU in callers that are able to
cope with "-ffast-math"-like brokenness.
The patch as sent is incomplete though. arch/powerpc/reloc.h would
need to define LDSO_ARCH for a soft-float ABI variant, and we need to
make sure that matches what's going in gcc.
Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
