This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: Printing decimal128 types out of registers
> From: Thiago Jung Bauermann <bauerman@br.ibm.com>
> Date: Wed, 30 Jan 2008 13:16:46 -0200
>
> > If this is too complicated, I wouldn't argue against a patch that just
> > added them if the floating point regsiters are present. Seems like it
> > won't be useful on a lot of processors though.
>
> Actually, this option is more appropriate. I have recently learned that
> there are some regular float instructions which can be directly used in
> DFP values stored in FP registers, like fabs, fneg and the other
> floating point move instructions. So it is very useful to have the
> pseudo-registers around even without specific DFP hardware.
>
> The attached patch is a rework of Luis' patch, with the following
> modifications:
>
> - don't use tdesc XML machinery to define the DFP pseudo-regs;
> - fixed logic in rs6000_pseudo_register_reggroup_p as noted by Daniel;
> - put DFP pseudo-regs in float register group;
> - the DFP pseudo-regs don't show up if there's no FPU;
> - reorganized the pseudo_register_{read,write} functions: there's now
> a generic one which concentrates the gdb_asserts and checking, and
> calls either the SPE or DFP as appropriate.
> - documentation: mention the DFP pseudo-regs in the Decimal Floating
> Point subsection, and created a PowerPC architecture subsection to
> talk about them in more detail. I had to rename the PowerPC Embedded
> section to avoid a node name conflict.
>
> This patch is on top of the SPE macro cleanup patch which I just posted:
>
> http://sourceware.org/ml/gdb-patches/2008-01/msg00794.html
>
> Is this ok?
Looks mostly reasonable to me, but I don't see why you need
ppc_dl0_upper_regnum, and ppc_dl15_regnum.