This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Fix 64-bit Solaris 2/x86 IE TLS code sequence (PR target/43309)
On 12/21/2010 08:30 AM, Rainer Orth wrote:
> Uros Bizjak <ubizjak@gmail.com> writes:
>
>> Anyway, attached patch also generates correct code for my small tests.
>> Rainer, can you please test attached patch?
>
> Unfortunately, this doesn't work. E.g. for the g++.dg/tls/static-1.C
> testcase, I get this difference (among others) for static-1a.s:
>
> --- 11-gcc/gcc/testsuite/g++/static-1a.s 2010-12-21 17:18:41.725085343 +0
> 100
> +++ 11-gcc.new/gcc/testsuite/g++/static-1a.s 2010-12-21 17:17:58.242204790 +0
> 100
> @@ -6,14 +6,21 @@
> _Z4testv:
> .LFB0:
> movq %fs:0, %rax
> + movq %rax, %rcx
> addq _ZN1A1iE@gottpoff(%rip), %rax
>
> I.e. there's an intervening insn between the movq %fs:0, %rax and the
> addq, which current Sun ld cannot handle.
I thought that might happen, Uros. You can't split the patterns
for Sun ld.
r~