On Sat, Jul 2, 2011 at 1:02 PM, Andreas Färber <andreas.faer...@web.de> wrote: > Am 02.07.2011 um 11:04 schrieb Blue Swirl: > >> On Sat, Jul 2, 2011 at 12:06 AM, malc <av1...@comtv.ru> wrote: >>> >>> On Sat, 2 Jul 2011, Blue Swirl wrote: >>> >>>> On Tue, Jun 28, 2011 at 1:51 AM, malc <av1...@comtv.ru> wrote: >>>>> >>>>> On Sun, 26 Jun 2011, Blue Swirl wrote: >>>>> >>>>>> Use TCG_REG_CALL_STACK instead of TCG_REG_R1 etc. for consistency. >>>>>> >>>>> >>>>> This i'd rather avoid. >>>> >>>> Why? In addition to the consistency among targets, a magic constant is >>>> replaced with a symbol which improves the documenting abilities and >>>> readability of the code. >>> >>> Makes it harder to read for me personally. >> >> What about other people reading the code? With TCG_REG_CALL_STACK it's >> pretty clear what is the purpose of the register, with '1' it is >> assumed that the reader happens to know it. I don't think the exact >> value is so interesting, or is register 1 somehow different from >> others? > > I second malc's notion. This is ppc-specific code, and GPR1 (r1) is the > designated stack pointer, that has nothing to do with TCG. Using some > generic TCG define hides from the reader which reg is being used, > potentially leading to nonsensical moves from r1 to TCG_... or vice versa.
OK, I'll drop the patch. > If we do want a define, I would suggest something ppc-specific like r1 or > sp. > > Andreas >