On Mon, Mar 12, 2012 at 3:35 PM, H.J. Lu <hjl.to...@gmail.com> wrote:
> On Mon, Mar 12, 2012 at 12:39 PM, Uros Bizjak <ubiz...@gmail.com> wrote:
>> On Sun, Mar 11, 2012 at 10:24 PM, H.J. Lu <hjl.to...@gmail.com> wrote:
>>
>>> Here is the patch which is equivalent to clearing MASK_TLS_DIRECT_SEG_REFS
>>> when Pmode != word_mode. We need to keep
>>>
>>> else if (Pmode == SImode)
>>> {
>>> /* Always generate
>>> movl %fs:0, %reg32
>>> addl xgottpoff(%rip), %reg32
>>> to support linker IE->LE optimization and avoid
>>> fs:(%reg32) as memory operand. */
>>> dest = gen_reg_rtx (Pmode);
>>> emit_insn (gen_tls_initial_exec_x32 (dest, x));
>>> return dest;
>>> }
>>>
>>> to support linker IE->LE optimization. TARGET_TLS_DIRECT_SEG_REFS only
>>> affects
>>> TLS LE access and fs:(%reg) is only generated by combine.
>>>
>>> So the main impact of disabling TARGET_TLS_DIRECT_SEG_REFS is to disable
>>> fs:immediate memory operand for TLS LE access, which doesn't have any
>>> problems
>>> to begin with.
>>>
>>> I would prefer to keep TARGET_TLS_DIRECT_SEG_REFS and disable only
>>> fs:(%reg), which is generated by combine.
>>
>> Please try attached patch. It introduces TARGET_TLS_INDIRECT_SEG_REFS
>> to block only indirect seg references.
>>
>> Uros.
>
> I am testing it.
>
There is no regression.
BTW, this x32 TLS IE optimization:
http://gcc.gnu.org/ml/gcc-patches/2012-03/msg00714.html
is still useful. For
[hjl@gnu-6 tls]$ cat ie2.i
extern __thread long long int x;
extern long long int y;
void
ie2 (void)
{
x = y;
}
[hjl@gnu-6 tls]$
my patch turns
ie2:
.LFB0:
.cfi_startproc
movq y(%rip), %rdx # 6 *movdi_internal_rex64/2 [length = 7]
movl %fs:0, %eax # 5 tls_initial_exec_x32 [length = 16]
addl x@gottpoff(%rip), %eax
movq %rdx, (%eax) # 7 *movdi_internal_rex64/4 [length = 3]
ret # 14 simple_return_internal [length = 1]
.cfi_endproc
into
ie2:
.LFB0:
.cfi_startproc
movq y(%rip), %rax # 6 *movdi_internal_rex64/2 [length = 7]
movq x@gottpoff(%rip), %rdx # 7 *tls_initial_exec_x32_store
[length = 16]
movq %rax, %fs:(%rdx)
ret # 14 simple_return_internal [length = 1]
.cfi_endproc
--
H.J.
