This is the mail archive of the cygwin-apps mailing list for the Cygwin 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: std::arg() bug : not repetitive ?


Eric Blake wrote:

> According to Dave Korn on 9/2/2009 10:07 PM:
>>   If we turn on SSE in the distro, we block anyone using Pentium2 or early
>> (pre-XP) Athlon CPUs from using Cygwin.  I think that might be a step too far.
> 
> Recent glibc has started providing function overloads, where the library
> contains both SSE and non-SSE variants, then decides at runtime what
> hardware it is on, and calls the appropriate overload.  But I don't know
> how easy that would be to port to PE semantics.

  Well that's an entirely suitable solution for individual package maintainers
to adopt for their packages, and we could build it into some of the core DLL
functions as well (e.g. SSE memcpy would probably be nice), but I don't see we
can just throw the switch on the compiler defaults unless the compiler was
going to auto-generate all the variants and the runtime selection code for us.

  PE semantics?  Well, we'd probably do it by setting a flag in the DLL's
private data at startup and just having a test and branch at the start of the
affected routines, I don't suppose we'd try and auto-select one set of
function variants or another using import table trickery.

    cheers,
      DaveK


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