[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
Re: Fast thread-local storage for OpenGL drivers
- From: Richard Henderson <rth redhat com>
- To: Roland McGrath <roland redhat com>
- Cc: phil-list redhat com, Gareth Hughes <gareth nvidia com>,Jakub Jelinek <jakub redhat com>
- Subject: Re: Fast thread-local storage for OpenGL drivers
- Date: Sat, 22 Feb 2003 22:22:20 -0800
On Sat, Feb 22, 2003 at 08:13:08PM -0800, Roland McGrath wrote:
> > * Should be able to be reduced to:
> > * movl %gs:__foo_dispatch ntpoff, %eax
> > * jmpl *(%eax)
Indeed, further reduced to
jmpl *%gs:__foo_dispatch ntpoff
> People seem to think it is a sufficient difference to be worth doing.
Yep. You can't go wrong with anything that cuts down the
size of the generated code that much.
> IIRC, someone took a brief crack at making the direct-from-%gs
> access work; it turned out to be more than an afternoon's quick hack...
That was me. My last attempt achieved decent (though not perfect)
results, but wasn't stable so it never got checked in. I'd since
forgotten about it. Should get that back on my schedule...
r~
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]