On Fri, Jul 21, 2017 at 12:22:00PM +0100, Wilco Dijkstra wrote:
> 
>     
> ping
> 
>     
> As described in PR79041, -mcmodel=large -mpc-relative-literal-loads
> may be used to avoid generating ADRP/ADD or ADRP/LDR.  However both
> trunk and GCC7 may still emit ADRP for some constant pool literals.
> Fix this by adding a aarch64_pcrelative_literal_loads check.
> 
> OK for trunk/GCC7 backport?

OK. Either like this, or with the conditions swapped around as Yvan
suggested to make backporting easier.

Thanks,
James

> 
> ChangeLog:
> 2017-06-27  Wilco Dijkstra  <wdijk...@arm.com>
> 
>         PR target/79041
>         * config/aarch64/aarch64.c (aarch64_classify_symbol):
>         Avoid SYMBOL_SMALL_ABSOLUTE .
>         * testsuite/gcc.target/aarch64/pr79041-2.c: New test.
> --
> diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
> index 
> 060cd8476d2954119daac495ecb059c9be73edbe..329d244e9cf16dbdf849e5dd02b3999caf0cd5a7
>  100644
> --- a/gcc/config/aarch64/aarch64.c
> +++ b/gcc/config/aarch64/aarch64.c
> @@ -10042,7 +10042,7 @@ aarch64_classify_symbol (rtx x, rtx offset)
>            /* This is alright even in PIC code as the constant
>               pool reference is always PC relative and within
>               the same translation unit.  */
> -         if (CONSTANT_POOL_ADDRESS_P (x))
> +         if (CONSTANT_POOL_ADDRESS_P (x) && 
> !aarch64_pcrelative_literal_loads)
>              return SYMBOL_SMALL_ABSOLUTE;
>            else
>              return SYMBOL_FORCE_TO_MEM;
> diff --git a/gcc/testsuite/gcc.target/aarch64/pr79041-2.c 
> b/gcc/testsuite/gcc.target/aarch64/pr79041-2.c
> new file mode 100644
> index 
> 0000000000000000000000000000000000000000..e7899725bad2b770f8488a07f99792113275bdf2
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/aarch64/pr79041-2.c
> @@ -0,0 +1,10 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -mcmodel=large -mpc-relative-literal-loads" } */
> +
> +__int128
> +t (void)
> +{
> +  return (__int128)1 << 80;
> +}
> +
> +/* { dg-final { scan-assembler "adr" } } */
>         

Reply via email to