[Gllug] Recommended distro

Nix nix at esperi.org.uk
Mon Dec 13 15:51:30 UTC 2004


On Mon, 13 Dec 2004, Martin A. Brooks mused:
> Nix wrote:
> 
>> i.e., on certain architectures, notably SPARC, OpenSSL benefits
>>*massively* from being built for v8 and above, because of v7's lack of
> 
> Sorry but v7, v8?

SPARC architecture revision numbers, I think. GCC understands the following
-mcpu flags:

sparcv7 -> ancient Suns, Cypress (e.g. SPARCStation IPX, SPARCServer 3xx...)
sparcv8 -> SuperSPARC, HyperSPARC (e.g. SS10, SS2000...)
sparclet -> f930, f934, sparclite86x (all weird clone chips I think)
sparclite -> tsc701 (another weird clone chip)
sparcv9 -> ultrasparc, ultrasparc3 (and all other ultrasparcs)

>From GCC's POV, SparcV8 is just like v7 except for some instruction
timings and the additions of the imul and idiv instructions. Sparclet is
quite like v8, with fewer types of multiply and divide instructions and
different timings (at least, that's what sparc.md seems to say).
Sparclite is much like Sparclet: a sort of cut-down v8.

Sparcv9 is the full-blown 64-bit UltraSPARC. GCC also understands
`v8plus', which is what GCC generates code for if it's asked to generate
32-bit code for an UltraSPARC. It's basically a v8 with the v9's
registers, adn is pretty much a GCC implementation detail.

I don't think anything which isn't an UltraSPARC is being made anymore.


(It was an intentional design decision to leave imul and idiv out of the
v7. It's just that, like register windows in some respects, it turned
out to be a *bad* design decision: it seems that it is possible to be
too RISCy after all.)

-- 
`The sword we forged has turned upon us
 Only now, at the end of all things do we see
 The lamp-bearer dies; only the lamp burns on.'
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list