Re: [PATCH][ARM][GCC][3/x]: MVE ACLE intrinsics framework patch.

[Kyrill Tkachov <kyrylo.tkachov@foss.arm.com>]

Hi Srinath,

On 11/14/19 7:12 PM, Srinath Parvathaneni wrote:
> Hello,
>
> This patch is part of MVE ACLE intrinsics framework.
>
> The patch supports the use of emulation for the double-precision 
> arithmetic
> operations for MVE. This changes are to support the MVE ACLE 
> intrinsics which
> operates on vector floating point arithmetic operations.
>
> Please refer to Arm reference manual [1] for more details.
> [1] 
> https://static.docs.arm.com/ddi0553/bh/DDI0553B_h_armv8m_arm.pdf?_ga=2.102521798.659307368.1572453718-1501600630.1548848914
>
> Regression tested on arm-none-eabi and found no regressions.
>
> Ok for trunk?
>
> Thanks,
> Srinath.
>
> gcc/ChangeLog:
>
> 2019-11-11  Andre Vieira <andre.simoesdiasvieira@arm.com>
>             Srinath Parvathaneni <srinath.parvathaneni@arm.com>
>
>         * config/arm/arm.c (arm_libcall_uses_aapcs_base): Modify 
> function to add
>         emulator calls for dobule precision arithmetic operations for MVE.


I'm a bit confused by the changelog and the comment in the patch....


>
>
> ###############     Attachment also inlined for ease of reply    
> ###############
>
>
> diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
> index 
> 6faed76206b93c1a9dea048e2f693dc16ee58072..358b2638b65a2007d1c7e8062844b67682597f45 
> 100644
> --- a/gcc/config/arm/arm.c
> +++ b/gcc/config/arm/arm.c
> @@ -5658,9 +5658,25 @@ arm_libcall_uses_aapcs_base (const_rtx libcall)
>        /* Values from double-precision helper functions are returned 
> in core
>           registers if the selected core only supports single-precision
>           arithmetic, even if we are using the hard-float ABI.  The 
> same is
> -        true for single-precision helpers, but we will never be using the
> -        hard-float ABI on a CPU which doesn't support single-precision
> -        operations in hardware.  */
> +        true for single-precision helpers except in case of MVE, 
> because in
> +        MVE we will be using the hard-float ABI on a CPU which 
> doesn't support
> +        single-precision operations in hardware.  In MVE the 
> following check
> +        enables use of emulation for the double-precision arithmetic
> +        operations.  */
> +      if (TARGET_HAVE_MVE)
> +       {
> +         add_libcall (libcall_htab, optab_libfunc (add_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (sdiv_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (smul_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (neg_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (sub_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (eq_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (lt_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (le_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (ge_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (gt_optab, SFmode));
> +         add_libcall (libcall_htab, optab_libfunc (unord_optab, SFmode));
> +       }

... this adds emulation for SFmode but you say you want double-precision 
emulation?

Can you demonstrate what this patch wants to achieve with a testcase?

Thanks,

Kyrill




>        add_libcall (libcall_htab, optab_libfunc (add_optab, DFmode));
>        add_libcall (libcall_htab, optab_libfunc (sdiv_optab, DFmode));
>        add_libcall (libcall_htab, optab_libfunc (smul_optab, DFmode));
>