This is the mail archive of the 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: PING: A Far Less Ambitious AltiVec patch

On 20 Feb, 2004, at 15.33, Geoff Keating wrote:

From: Ziemowit Laski <>
Date: Fri, 20 Feb 2004 14:52:12 -0800

On 20 Feb, 2004, at 14.20, Geoff Keating wrote:

From: Ziemowit Laski <>
Date: Fri, 20 Feb 2004 14:05:47 -0800

On 20 Feb, 2004, at 13.49, Geoff Keating wrote:

That's worse. Use reg_names always.

I can't -- non-Darwin assemblers will barf on it.

No, they won't. Try it and see.

Please read for testimony by someone who's tried it.

I don't see any reference to reg_names in that message. You might want to investigate how reg_names gets set.

From the REGISTER_NAMES macro, which in turn points (in our case) at
rs6000_reg_names (just register numbers), which (#if TARGET_REGNAMES)
actually gets clobbered with alt_reg_names (register names, preceded by
Which tells me absolutely nothing about how this is supposed to behave
other platforms.

Please advise on how to proceed.

Well, you can investigate further in order to determine why using reg_names is correct and how it works, or you could just assume that I'm right and it is correct. Your choice.

:-) My bad; I mistakenly took the fact that non-Darwin assembler barf on
full register names (which apparently they do) to mean that the reg_names
mechanism itself is broken (and I don't have any evidence that it is). My

Ziemowit Laski                 1 Infinite Loop, MS 301-2K
Mac OS X Compiler Group        Cupertino, CA USA  95014-2083
Apple Computer, Inc.           +1.408.974.6229  Fax .5477

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