Re: PA-RISC 2.0 fldw instruction (formats 43 and 44)

[Alan Modra]

On Sun, May 19, 2002 at 07:47:57PM -0400, John David Anglin wrote:
> The problem is the R_PARISC_DPREL14R relocation clobbers the "t" bit
> in the insn.  We have this in the .o file
> 
>   14:   5c 36 00 02     fldw 0(r1),fr22R
>                       14: R_PARISC_DPREL14R   a
> 
> but we lose the "t" bit (bit 30) in the insn when the code is linked
> 
>    1043c:       5c 36 02 68     fldw 134(r1),fr22
> 
> Note this code is intended to run in narrow 32-bit mode.

Well firstly, the assembler needs to be fixed to emit the correct
relocation for this insn, R_PARISC_DPREL14WR.  Then we need to support
the reloc and others like it in elf32-hppa.c.  That's quite a bit of
work, and I'm rather busy.  Should gcc be generating this code for
hppa-linux in the first place?

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre