From: Vladimir Prus (ghost_at_[hidden])
Date: 2005-10-21 01:31:48
On Thursday 20 October 2005 20:20, Rene Rivera wrote:
> David Abrahams wrote:
> > Reece Dunn <msclrhd_at_[hidden]> writes:
> >>How about:
> >><architecture> specifies the general CPU type being used. That is, what
> >>architectural design is being used on the chip. For example, x86
> >> specifies the Intel X86 based CPUs such as Pentium 4.
> >><instruction-set> specifies what CPU/assembler instructions are available
> >> for the given architecture. For example, using the instructions
> >> available with the AMD Athalon CPU.
> > So far, I see no advantage in distinguishing these. What's the point
> > of <architecture>, anyway?
> Link compatibility. For example an x86/i486 compiled object would be
> compatible with an x86/athlon. Perhaps someone is wants the Athlon
> instructions for one specific file.
FWIW, we don't have any "link compatibility" notion in V2. It used to be
there, and some remnants still remain, but it never worked good enough to be
useful. Speaking of architecture, it's theoretically possible to link
together x86 and powerpc code, isn't this exactly what Apple's recent file
format does? So, "link compatibility" is a moot thing.
> > Isn't that completely determined by the
> > instruction-set chosen?
> Sure, but we would have to encode someplace what instruction sets are
> compatible with each other.
What for? Where do we do any checks that instructions sets are compatible with
-- Vladimir Prus http://vladimir_prus.blogspot.com Boost.Build V2: http://boost.org/boost-build2
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk