RFA: AIX 64-bit mega-patch

David Edelsohn dje@watson.ibm.com
Tue Jun 13 15:06:00 GMT 2000


>>>>> Nick Duffek writes:

Nick> On 13-Jun-2000, David Edelsohn wrote:
>> There are a couple of problems with dummy frames:

Nick> Thanks, I'll check into those.

	If I understand the code, it looks like you are using generic
dummy frames for PowerPC and not for "rs6000" which I think means POWER
architecture.  I don't understand the reason for that difference.
Assuming the meaning of "rs6000", the hard-coded TOC offset may not be a
problem (because POWER is only 32-bit), but it still is very confusing.

	I do not think that "rs6000" refers to AIX PowerOpen / XCOFF while
"ppc-uisa" refers to SVR4 / ELF, but I am not sure.

>> 4) RS/6000 (with slash) is a platform, not an architecture.  Informational
>> messages mentioning "PowerPC / RS6000" seem very confusing to me.

Nick> I encountered two uses of the term "RS/6000" while working on the 64-bit
Nick> AIX support:
Nick> (1) the IBM platform;
Nick> (2) the architecture used to implement the RS/6000 platform prior to the
Nick> introduction of PowerPC.

Nick> Can you be more specific about which messages seem confusing?  Are you
Nick> referring to the variants[].description fields?  They're not actually used
Nick> anywhere at the moment, but I'd like to fix them if they're misleading.

	The confusing uses are things like:

	"PowerPC / RS6000 processor variant is ..." and "GDB knows about
the following PowerPC and RS6000 variants ..."

and

	v = find_variant_by_name (rs6000 ? "rs6000" : "ppc-uisa");

and

	if (rs6000)
	  {
	    set_gdbarch_use_generic_dummy_frames (gdbarch, 0);
	  }
	else
	  {
	    set_gdbarch_use_generic_dummy_frames (gdbarch, 1);
	  }

RS/6000 (WITH a slash) only refers to an IBM platform, not an
architecture.  RMS refused to allow early GCC and GDB to refer to the
architecture as "POWER" because he did not want GNU software providing
subliminal publicity for IBM given the spelling of the architecture name.
That was fine as long as the platform and the architecture were
synonymous, but, with the advent of PowerPC and multiple architectures
used in the RS/6000 platform, things get confusing.

	We probably need to continue to use "rs6000" for the GNU filenames
and variables, but that should not carry over to informational messages.
There is a "power-uisa" and an ISA extension for the Power2 (RIOS2 and
P2SC) processor.  All of the PowerPC processors (including PPC601, PPC603,
PPC604, PPC620, PPC630 (aka Power3), A35, RS64, and Power4) use the
PowerPC UISA (either the 32-bit subset or the complete 64-bit ISA, plus or
minus various optional extensions).

	RS/6000 encompasses all of the POWER and PowerPC processors which
have been used as components.  To me, only a statements like: "RS/6000
processor variant ..." or "... POWER and PowerPC processor variants" makes
sense.

David


More information about the Gdb-patches mailing list