Fast thread-local storage for OpenGL drivers

[Gareth Hughes]

Jakub Jelinek wrote:
> 
> > I'm pretty sure all implementations of OpenGL are not compiled as PIC at
> > this point in time.
> 
> AFAIK on x86 only, but it is wrong everywhere.

Can you elaborate?

> On x86, ld supports Local Exec model in shared libraries (while for most
> other targets it does not). R_386_TLS_LE relocation is simply during
> -shared linking changed into R_386_TLS_TPOFF dynamic relocation (the same
> as is used for IE model, but there this reloc is against .got section
> while for LE it is against text section).

To be clear, are you saying that even if you declare a variable as LE, once
you link the relevant code into a shared library it is changed to IE?

> So, if you don't use -fpic anyway, you can just use LE model on IA-32,
> if you finally change it so that -fpic is used for the whole library,
> then those functions (or assembly stubs) can be put into
> some SHF_ALLOC|SHF_WRITE|SHF_EXECINSTR section.

If we don't use PIC, do we always get LE if we declare a variable as LE?

> Which means you should use the default -ftls-model and use
> __attribute__((tls_model("initial-exec"))) or
> __attribute__((tls_model("local-exec")))
> for the variables which are really performance critical.

Yes, that would be the plan.

> On IA-32, you can use
> __asm ("jmp 1f; .section writetext, \"awx\"; 1: movl 
> $foo at ntpoff, %0; jmp 2f; .previous; 2:" : "=r" (foo_offset));
> to query some variable's offset which you can later on use with:
> __asm ("movl %gs:0(%1), %0" : "=r" (foo_value) : "r" (foo_offset));
> Please do something like this only for runtime generated code, not for
> anything else.

Of course, silly me.

-- 
Gareth Hughes (gareth at nvidia.com)
OpenGL Developer, NVIDIA Corporation