This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.

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: Bugs in MIPS FUNCTION_ARG_REGNO_P ?

From: Mark Mitchell <mark at codesourcery dot com>
To: Richard Henderson <rth at redhat dot com>
Cc: Eric Christopher <echristo at redhat dot com>, gcc at gcc dot gnu dot org, wilson at specifixinc dot com
Date: Wed, 03 Mar 2004 00:09:42 -0800
Subject: Re: Bugs in MIPS FUNCTION_ARG_REGNO_P ?
Organization: CodeSourcery, LLC
References: <200403030305.i2335S8S028779@sirius.codesourcery.com> <1078297307.3616.8.camel@dzur.sfbay.redhat.com> <40458861.20301@codesourcery.com> <20040303074811.GA8118@redhat.com>

Richard Henderson wrote:

On Tue, Mar 02, 2004 at 11:25:21PM -0800, Mark Mitchell wrote:

What exactly do you believe to be correct? The macro implementation, or the statement that the O64 ABI does not use floating-point registers for arugment-passing?
O64 has no official document describing it.  It is effectively defined by
the macros in gcc.  I would strongly recommend against changing anything.

Oh, I agree!

The point is that the current state is internally inconsistent.

We (a) use fp regs for arg-passing, and (b) tell the middle-end that we do not. That seems like a recipe for wrong-code through some incorrect optimization to me.

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com

References:
- Bugs in MIPS FUNCTION_ARG_REGNO_P ?
  - From: Mark Mitchell
- Re: Bugs in MIPS FUNCTION_ARG_REGNO_P ?
  - From: Eric Christopher
- Re: Bugs in MIPS FUNCTION_ARG_REGNO_P ?
  - From: Mark Mitchell
- Re: Bugs in MIPS FUNCTION_ARG_REGNO_P ?
  - From: Richard Henderson

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