i.MX Processors中的主题 Re: Hardware floating point support?

Hardware floating point support?

WayneRadochonsk — Wed, 03 Oct 2012 15:23:30 GMT

I'm porting an application from an x86 Atom to an i.MX53 QSB for a wave energy project.

The i.MX53 QSB optimizes the transfer of mechanical power into electrical current to charge

batteries. Using a i.MX53 QSB should drop our HOTEL load from 12 watts to 2 watts

Having worked through installing a SATA drive and getting a cross tool chain working

with Eclipse CDT/RSE, I have stumbled into another potential pot hole.

Our existing app makes extensive use of floating point arithmetic. I just discovered that

Linux kernels come into two flavors for this class of ARM processors. A "armhf" kernel

seems to imply support for hardware floating point.

Does the kernel on the micro-SD shipped with i.MX53 QSB support floating point natively?

We need just standard IEEE-754 double precision floating point.

If not, is there kernel and tool chain with hardware floating point supported? Where

can it be downloaded?

Thanks,

Wayne

Re: Hardware floating point support?

Yuri — Fri, 20 Sep 2013 10:32:01 GMT

Official Freescale Linux BSP for i.MX53 (QSB) does not support hardware floating point.

Re: Hardware floating point support?

daiane_angolini — Fri, 20 Sep 2013 12:52:51 GMT

If you only want kernel and will not need GPU support or VPU support, you can use kernel mainline and built everything (all the system and your application) for armhf on yocto.

The problem regarding GPU is that it does need several user space libraries that are provided as pre-built binary files, so you cannot "re built" it.

The VPU support does need firmware that is pre-built, and some user-space library for decoders are pre-built binaries as well.

I'm sure all the pieces you need are not linked "as is", so some work will be needed. But you can take a look on this, if you think it worse.

Yocto Training - HOME

https://lists.yoctoproject.org/listinfo/meta-freescale