[cfe-commits] [PATCH] Fix default cpu for -march=armv5e

[James Molloy]

Hi Evgeniy,

It should be fairly simple to add to ARM.td:

def : ProcNoItin<"arm1026ejs",      [HasV5TEOps]>;

But I'll LGTM your patch anyway since this is LLVM changes on a Clang patch.

But I think adding it to ARM.td and ARMTargetInfo would be a worthwhile and extremely simple patch for llvm-commits.

Cheers,

James

-----Original Message-----
From: Evgeniy Stepanov [mailto:eugeni.stepanov at gmail.com] 
Sent: 25 April 2012 09:07
To: James Molloy
Cc: cfe-commits at cs.uiuc.edu
Subject: Re: [cfe-commits] [PATCH] Fix default cpu for -march=armv5e

Hi, James,

I could add arm1026ejs to ARMTargetInfo::getCPUDefineSuffix, but then
it fails anyway because arm1026ejs is not in ARM.td.

I've added a test.

On Tue, Apr 24, 2012 at 6:29 PM, James Molloy <james.molloy at arm.com> wrote:
> Hi Evgeniy,
>
> The patch itself appears fine but you're missing a testcase. Could you please add one? The driver test suite is a bit lacking so for any fixes it would be really nice to know if we regress.
>
> Also you mention ARMTargetInfo::getCPUDefineSuffix, but you didn't submit a patch to fix that?
>
> Cheers,
>
> James
>
> -----Original Message-----
> From: cfe-commits-bounces at cs.uiuc.edu [mailto:cfe-commits-bounces at cs.uiuc.edu] On Behalf Of Evgeniy Stepanov
> Sent: 24 April 2012 14:59
> To: cfe-commits at cs.uiuc.edu
> Subject: [cfe-commits] [PATCH] Fix default cpu for -march=armv5e
>
> Hi,
>
> arm1026ejs is armv5tej, and, what's worse, the mapping of cpu name to
> arch suffix in ARMTargetInfo::getCPUDefineSuffix does not know that it
> exists.
>
> Please review.
>
>
>