Re: [PATCH v2] MIPS: make mipsisa32 and mipsisa64 link more systematic

[Richard Sandiford <richard.sandiford@arm.com>]

YunQiang Su <yunqiang.su@cipunited.com> writes:
> Introduce `static const struct mips_mach_extension mips_mach_32_64[]`
> and `mips_mach_extends_32_64 (unsigned long base, unsigned long extension)`,
> to make mipsisa32 and mipsisa64 interlink more systemtic.
>
> Normally, the ISA mipsisa64rN has two subset: mipsisa64r(N-1) and
> mipsisa32rN. `mips_mach_extensions` can hold only mipsisa64r(N-1),
> so we need to introduce a new instruction `mips_mach_32_64`, which holds the pair 32vs64.
>
> Note: R6 is not compatible with pre-R6.
>
> bfd/ChangeLog:
>
> 	* elfxx-mips.c (mips_mach_extends_p): make mipsisa32 and
> 	  mipsisa64 interlink more systematic.
> 	  (mips_mach_32_64): new struct added.
> 	  (mips_mach_extends_32_64): new function added.

Sorry for the delay in pushing this.  Could you check whether it's
the right version though?  When I build locally I get:

> ---
>  bfd/elfxx-mips.c | 40 +++++++++++++++++++++++++++++-----------
>  1 file changed, 29 insertions(+), 11 deletions(-)
>
> diff --git a/bfd/elfxx-mips.c b/bfd/elfxx-mips.c
> index 35bbd86044b..21e6396609b 100644
> --- a/bfd/elfxx-mips.c
> +++ b/bfd/elfxx-mips.c
> @@ -14524,6 +14524,16 @@ struct mips_mach_extension
>    unsigned long extension, base;
>  };
>  
> +/* An array that maps 64-bit architectures to the corresponding 32-bit
> +   architectures.  */
> +static const struct mips_mach_extension mips_mach_32_64[] =
> +{
> +  { bfd_mach_mipsisa64r6, bfd_mach_mipsisa32r6 },
> +  { bfd_mach_mipsisa64r5, bfd_mach_mipsisa32r5 },
> +  { bfd_mach_mipsisa64r3, bfd_mach_mipsisa32r3 },
> +  { bfd_mach_mipsisa64r2, bfd_mach_mipsisa32r2 },
> +  { bfd_mach_mipsisa64,   bfd_mach_mipsisa32 }
> +};
>  
>  /* An array describing how BFD machines relate to one another.  The entries
>     are ordered topologically with MIPS I extensions listed last.  */
> @@ -14601,29 +14611,37 @@ static const struct mips_mach_extension mips_mach_extensions[] =
>    { bfd_mach_mips3900, bfd_mach_mips3000 }
>  };
>  
> -/* Return true if bfd machine EXTENSION is an extension of machine BASE.  */
> +/* Return true if bfd machine EXTENSION is the same as BASE, or if
> +   EXTENSION is the 64-bit equivalent of a 32-bit BASE.  */
> +
> +static bool
> +mips_mach_extends_32_64 (unsigned long base, unsigned long extension)
> +{
> +  size_t i;
> +
> +  if (extension == base)
> +    return true;
> +
> +  for (i = 0; i < ARRAY_SIZE (mips_mach_64_32); i++)

error: ‘mips_mach_64_32’ undeclared (first use in this function); did you mean ‘mips_mach_32_64’?

It looks like the previous patch had the same thing (sorry for not
noticing).  Obviously it's a simple fix, but it would be good to have
some reassurance that the patch has been tested beyond my crude sanity
checks.

Thanks,
Richard

> +    if (extension == mips_mach_32_64[i].extension)
> +      return base == mips_mach_32_64[i].base;
> +
> +  return false;
> +}
>  
>  static bool
>  mips_mach_extends_p (unsigned long base, unsigned long extension)
>  {
>    size_t i;
>  
> -  if (extension == base)
> -    return true;
> -
> -  if (base == bfd_mach_mipsisa32
> -      && mips_mach_extends_p (bfd_mach_mipsisa64, extension))
> -    return true;
> -
> -  if (base == bfd_mach_mipsisa32r2
> -      && mips_mach_extends_p (bfd_mach_mipsisa64r2, extension))
> +  if (mips_mach_extends_32_64 (base, extension))
>      return true;
>  
>    for (i = 0; i < ARRAY_SIZE (mips_mach_extensions); i++)
>      if (extension == mips_mach_extensions[i].extension)
>        {
>  	extension = mips_mach_extensions[i].base;



> -	if (extension == base)
> +	if (mips_mach_extends_32_64 (base, extension))
>  	  return true;
>        }