This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc 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: RFC: TLS improvements for IA32 and AMD64/EM64T


On Sep 16, 2005, John Reiser <jreiser@BitWagon.com> wrote:

> Alexandre Oliva <aoliva@redhat.com>, 16 Sep 2005 02:08:22 -0300:
>> ...  glibc and nptl,
>> aware that the dynamic access models used to be so inefficient, take
>> advantage of the fact that libc is always loaded initially and
>> reference almost all thread-local variables using the more efficient
>> access models.  ...

> (g)libc is not always loaded initially, neither is ld-linux.  Other
> PT_INTERP, or a main program with no PT_INTERP at all, can load glibc
> and perhaps nptl later, perhaps in combination with ld-linux.

Have you tried to do that recently?  With the current use of the IE
access model in libc, ld-linux, nptl et al, it just won't work.

> How will the proposed TLS improvements work in such cases?

They would work better than the current model, that relies on the
Initial Exec models for such libraries.  With the more efficient GD
and LD access models, it might even be reasonable to do away with the
use of IE in libc, nptl et al, so as to make possible what you say
should be possible.

-- 
Alexandre Oliva         http://www.lsd.ic.unicamp.br/~oliva/
Red Hat Compiler Engineer   aoliva@{redhat.com, gcc.gnu.org}
Free Software Evangelist  oliva@{lsd.ic.unicamp.br, gnu.org}


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