Thread: __cpu__ defines
As part of my spinlock testing, I noticed that we test for __cpu__ when using gcc, and __cpu when not using gcc. However, I see that my i386 gcc 2.95 defines both (shown using src/tools/ccsym): __GNUC__=2__GNUC_MINOR__=95unix__i386__i386__bsdi__bsdi__ELF____GAS__=2__GAS_MINOR__=10__unix____i386____i386____bsdi____bsdi____ELF____GAS__=2__GAS_MINOR__=10__unix__i386__bsdisystem=unixsystem=bsdcpu=i386machine=i386cpu=i386machine=i386i386__i386__i386__ So, I wonder if we should be testing _just_ for __cpu, perhaps starting in 7.5. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001+ If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania19073
--On Friday, September 12, 2003 09:53:10 -0400 Bruce Momjian <pgman@candle.pha.pa.us> wrote: > As part of my spinlock testing, I noticed that we test for __cpu__ when > using gcc, and __cpu when not using gcc. However, I see that my i386 > gcc 2.95 defines both (shown using src/tools/ccsym): > > __GNUC__=2 > __GNUC_MINOR__=95 > unix > __i386__ > i386 > __bsdi__ > bsdi > __ELF__ > __GAS__=2 > __GAS_MINOR__=10 > __unix__ > __i386__ > __i386__ > __bsdi__ > __bsdi__ > __ELF__ > __GAS__=2 > __GAS_MINOR__=10 > __unix > __i386 > __bsdi > system=unix > system=bsd > cpu=i386 > machine=i386 > cpu=i386 > machine=i386 > i386 > __i386 > __i386__ > > So, I wonder if we should be testing _just_ for __cpu, perhaps starting > in 7.5. I corresponded with Dave Prosser of SCO, and he pointed me at the #assert stuff. That's where the xxx=xxx stuff comes from. Might it make more sense to use #if #cpu(i386) xxx #endif instead of depending on the different flavors of #defines. GCC and at least SCO's cc support this. I sent the details to Tom, since he seems to be the spinlock maintainer. LER -- Larry Rosenman http://www.lerctr.org/~ler Phone: +1 972-414-9812 E-Mail: ler@lerctr.org US Mail: 1905 Steamboat Springs Drive, Garland, TX 75044-6749
Bruce Momjian <pgman@candle.pha.pa.us> writes:
> As part of my spinlock testing, I noticed that we test for __cpu__ when
> using gcc, and __cpu when not using gcc.
> ...
> So, I wonder if we should be testing _just_ for __cpu, perhaps starting
> in 7.5.
I might be all wet on this, but I had the idea that the __cpu__ forms
were considered more standard/common. In any case, I can't see any
good reason not to test for both. The amount of code saved by checking
only one is negligible; why should we take a chance on breaking things
for that?
regards, tom lane
Tom Lane wrote: > Bruce Momjian <pgman@candle.pha.pa.us> writes: > > As part of my spinlock testing, I noticed that we test for __cpu__ when > > using gcc, and __cpu when not using gcc. > > ... > > So, I wonder if we should be testing _just_ for __cpu, perhaps starting > > in 7.5. > > I might be all wet on this, but I had the idea that the __cpu__ forms > were considered more standard/common. In any case, I can't see any > good reason not to test for both. The amount of code saved by checking > only one is negligible; why should we take a chance on breaking things > for that? Yes, that what confuses me --- which is standard. Right now, we aren't consistent. My patch tests for __cpu__ on gcc, and both on non-gcc, which seems safest. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001+ If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania19073
Bruce Momjian writes: > As part of my spinlock testing, I noticed that we test for __cpu__ when > using gcc, and __cpu when not using gcc. However, I see that my i386 > gcc 2.95 defines both (shown using src/tools/ccsym): gcc only documents the __foo__ version, so there is a small reason to lean in that direction if you want to make it uniform. -- Peter Eisentraut peter_e@gmx.net
Peter Eisentraut wrote: > Bruce Momjian writes: > > > As part of my spinlock testing, I noticed that we test for __cpu__ when > > using gcc, and __cpu when not using gcc. However, I see that my i386 > > gcc 2.95 defines both (shown using src/tools/ccsym): > > gcc only documents the __foo__ version, so there is a small reason to lean > in that direction if you want to make it uniform. Oh, that helps, so my logic of testing for only __foo__ on gcc, and both when the compiler isn't gcc or isn't known is correct. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001+ If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania19073