[PATCH] Use the integrated assembler by default on OpenBSD/powerpc
Daniel Sanders
Daniel.Sanders at imgtec.com
Mon Jun 16 08:10:21 PDT 2014
> -----Original Message-----
> From: Brad Smith [mailto:brad at comstyle.com]
> Sent: 14 June 2014 00:42
> To: Daniel Sanders
> Cc: llvm cfe
> Subject: Re: [PATCH] Use the integrated assembler by default on
> OpenBSD/powerpc
>
> On 13/06/14 7:03 AM, Daniel Sanders wrote:
> > Hi Brad,
> >
> >> // CHECK-POWERPC-AS-NOT: "-no-integrated-as"
> >
> > That looks right to me. The -NOT suffix causes FileCheck to report
> > failure if the string
> > matches.
>
> Ok, then I will use that.
>
> > That said, I think the behaviour I expected from your patch was wrong.
> > Even though Jeorg
> > committed your patch in r202967, -no-integrated-as still appears in the
> clang subcommand > and it still calls the assembler separately when I run
> 'clang -target powerpc-unknown-openbsd > -### -c
> tools/clang/test/Driver/openbsd.c'.
>
> Yes, without the patch as part of this thread that is expected. The patch
> you're referring to is part of what sounds like a move to have the knowledge
> as to whether to use the integrated assembler within LLVM instead of within
> Clang but that hasn't been fully completed yet from Clang's perspective.
It seems I didn't go back far enough on the thread. I've just found the patch you're referring to. Sorry for the noise.
> > I believe the bit I hadn't realised at the time is that clang has its
> > opinion on whether
> > the integrated assembler is enabled by default and that it currently
> (r210886) thinks it
> > should be off for PowerPC OpenBSD (see
> Generic_GCC::IsIntegratedAssemblerDefault() in >
> tools/clang/lib/Driver/ToolChains.cpp). Adding the 'CHECK-POWERPC-AS-
> NOT' will make sense > when clang's default is changed.
>
> The patch I am submitting changes the default for OpenBSD just as has
> already been done for FreeBSD and NetBSD.
>
> --
> This message has been scanned for viruses and dangerous content by
> MailScanner, and is believed to be clean.
More information about the cfe-commits
mailing list