Re: MIPS R3900 'bgt' with gcc 3.0/binutils 2.11.2

[Jonathan Larmour <jlarmour at redhat dot com>]

mike stump wrote:
> 
> > Date: Mon, 13 Aug 2001 01:27:19 +0100
> > From: Jonathan Larmour <jlarmour@redhat.com>
> 
> > Certainly at least. But I'm still unsure about bgtu or bgt in
> > opcodes which, according to opcodes, are meant to be in MIPS I, and
> > I'm not sure they really are.
> 
> If I'm not mistaken, they are macros.  Is it a problem to have them as
> macros?

No, but in that case, why is (for example) bgtu okay in MIPS 1, but bgtul
is only in MIPS 2? Perhaps opcodes should allow bgtul for the R3900 after
all. I'm not a compiler person, and don't know (read: try to avoid knowing)
the complexities of MIPS ISAs and their relevance to particular CPUs :-).

> > I don't think this fixed it - The test for TARGET_MIPS3900 was moved
> > into ISA_HAS_BRANCHLIKELY, that's all. This patch by itself would
> > remove all branch likelys for R3900s.
> 
> ?
> 
> /* ISA has branch likely instructions (eg. mips2). */
> /* Disable branchlikely for tx39 until compare rewrite.  They haven't
>    been generated up to this point.  */
> #define ISA_HAS_BRANCHLIKELY    (mips_isa != 1                          \
>                                  /* || TARGET_MIPS3900 */)
> 
> I don't see how they wind up with branch likely instruction support.
> mips I didn't have them, and the 3900 is a mips I part (mostly).
> 
> I'm sure Eric has more state than I.

No, looks like I read too much into the cvs log entry: 
        (GENERATE_BRANCHLIKELY): Move TARGET_MIPS3900.
        (ISA_HAS_BRANCHLIKELY): To here.

So you're probably right. Still, I don't know whether it's correct to fix
bgtul in opcodes for the R3900, or to apply the change to
GENERATE_BRANCHLIKELY. Hopefully someone can say definitely which so I can
send a patch to the right place, hopefully in time for 3.0.1 if GCC.

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine