User: Password:
|
|
Subscribe / Log in / New account

ARM kernel floating-point emulation code removed

ARM kernel floating-point emulation code removed

Posted Apr 10, 2013 21:13 UTC (Wed) by brouhaha (subscriber, #1698)
In reply to: ARM kernel floating-point emulation code removed by gb
Parent article: ARM kernel floating-point emulation code removed

Does the user space floating point library use John Hauser's SoftfFloat library? Or does it use something else?


(Log in to post comments)

ARM kernel floating-point emulation code removed

Posted Apr 10, 2013 21:22 UTC (Wed) by brouhaha (subscriber, #1698) [Link]

The GCC wiki page on Software floating point in GCC describes three different soft FP libraries used by GCC and glibc. Unfortunately the page seems quite dated (last updated in 2008). While it explicitly refers to several processor architectures, it doesn't mention ARM at all.

If anyone really needs kernel emulation of the floating point instructions, e.g. in order to run binaries compiled for hardfp, perhaps one of those might be suitable?

ARM kernel floating-point emulation code removed

Posted Apr 10, 2013 21:35 UTC (Wed) by gb (subscriber, #58328) [Link]

I am familiar with debian-based things. Debian comes in 3 favor on arm: arm, armel, and armhf. Armhf is unaffected by any emulation stuff. Armel is eabi port with softfloat, not affected too. Read here: http://wiki.debian.org/ArmEabiPort.

Only affected thing as i understand is arm port in case your arm cpu does not support emulation. And arm port is deprecated long ago, you can see on debian popcon that there are 55 total arm machines reporting. n900 is not affected (armel) openmoko gta02 - armel, gta04 - armhf, who knows may be some ancient machines?

ARM kernel floating-point emulation code removed

Posted Apr 11, 2013 3:45 UTC (Thu) by npitre (subscriber, #5680) [Link]

I wrote the user space soft-float library, all in ARM assembly, and contributed it to gcc:

http://gcc.gnu.org/ml/gcc-patches/2003-08/msg01172.html

http://gcc.gnu.org/ml/gcc-patches/2004-10/msg01018.html

This is not reusable as is for correct FPA/VFP emulation in the kernel as no exceptions are handled nor any rounding modes other than the default round-to-nearest mode. It doesn't provide any support for extended 80-bit floats either.

ARM kernel floating-point emulation code removed

Posted Apr 11, 2013 8:07 UTC (Thu) by asbradbury (guest, #47625) [Link]

Could the opened up fpemulator be used? http://www.riscosopen.org/news/articles/2010/05/26/arm-lt... Though I've got to admit I'm amazed anyone still cares about oabi.


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds