This is the mail archive of the crossgcc@sourceware.org mailing list for the crossgcc project.

See the CrossGCC FAQ for lots more information.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: ARM floating point differences

From: Rick Mann <rmann at latencyzero dot com>
To: Dimitry Andric <dimitry at andric dot com>
Cc: crossgcc at sourceware dot org
Date: Thu, 10 Jan 2008 16:51:26 -0800
Subject: Re: ARM floating point differences
References: <E34AF39A-71FA-4AB8-8151-EB18BDD562CD@latencyzero.com> <4786B0AA.5070407@andric.com>

Hey, Dimitry, thanks for the rapid response!

On Jan 10, 2008, at 3:56 PM, Dimitry Andric wrote:

AFAICS it might be that you configured your "old" tools with
--target=arm-elf, while you used --target=xscale-elf for the "new"
tools.  This will change a LOT for the code generation.

Yes, I'm sorry I didn't make this clear. That is exactly what I did. The "old" tools were arm-elf, the new "xscale-elf" (the processor is a Marvell PXA320, an Xscale proc).

Still, the arm-elf floating point code ran very well on it, and I'd like to understand better how to get that same performance out of my xscale-elf tools.

You might want to try using the same --target options, and see if
there's still such a difference.

You mean, rebuild my toolchain with --target=arm-elf, or can I pass -- target to GCC in my target code makefile?

--
Rick


--
For unsubscribe information see http://sourceware.org/lists.html#faq

References:
- ARM floating point differences
  - From: Rick Mann
- Re: ARM floating point differences
  - From: Dimitry Andric

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]