On Fri, Oct 11, 2013 at 8:38 PM, H.J. Lu <hongjiu...@intel.com> wrote:
> In x32, when a TLS address is in DImode and Pmode is SImode,
> copy_addr_to_reg will fail. This patch adds ix86_copy_addr_to_reg
> to first copy DImode address into a DImode register and then to generate
> SImode SUBREG in this case. Tested on x32. OK for trunk?
>
> 2013-10-11 H.J. Lu <hongjiu...@intel.com>
>
> PR target/58690
> * config/i386/i386.c (ix86_copy_addr_to_reg): New function.
> (ix86_expand_movmem): Replace copy_addr_to_reg with
> ix86_copy_addr_to_reg.
> (ix86_expand_setmem): Likewise.
>
> gcc/testsuite/
>
> 2013-10-11 H.J. Lu <hongjiu...@intel.com>
>
> PR target/58690
> * gcc.target/i386/pr58690.c: New test
>
> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
> index 37c1bec..e39d63db 100644
> --- a/gcc/config/i386/i386.c
> +++ b/gcc/config/i386/i386.c
> @@ -22076,6 +22076,21 @@ counter_mode (rtx count_exp)
> return SImode;
> }
>
> +/* Copy the address to a register in Pmode. Support x32 TLS address in
> + DImode and Pmode in SImode. */
Copy the address to a Pmode register. This is used for x32 to truncate
DImode TLS address to a SImode register.
> +static rtx
> +ix86_copy_addr_to_reg (rtx addr)
> +{
> + if (GET_MODE (addr) != Pmode)
> + {
> + gcc_assert (GET_MODE (addr) == DImode && Pmode == SImode);
> + return gen_rtx_SUBREG (SImode, copy_to_mode_reg (DImode, addr), 0);
> + }
> + else
> + return copy_addr_to_reg (addr);
> +}
No negative conditions please. Just switch arms of the if clause.
OK with these changes.
Do we also need this patch on 4.8?
Thanks,
Uros.
