Re: [PATCH 3/4] gdb/aarch64: record/replay support for RPRFM and PRFM (reg)

[Thiago Jung Bauermann <thiago.bauermann@linaro.org>]

<Ezra.Sitorus@arm.com> writes:

> From: Ezra Sitorus <ezra.sitorus@arm.com>
>
> The PRFM (register) instruction variant was unsupported. This is added
> along with RPRFM.
>
> No testcase has been added as these are hint instructions which don't
> modify registers/memory. I'm happy to add them in if you think these are
> required.
>
> The full testsuite was done on aarch64-none-linux-gnu without RPRFM.
> The gdb.arch and gdb.reverse tests were run on Shrinkwrap with RPRFM
> support. I've observed unrelated regressions that appear to be due to
> Shrinkwrap rather than my patch.
> ---
>  gdb/aarch64-tdep.c | 5 +++++
>  1 file changed, 5 insertions(+)

Reviewed-by: Thiago Jung Bauermann <thiago.bauermann@linaro.org>

This patch made me notice that 1. Arm has a PRFUM instruction and 2.,
it's not supported by aarch64_record_load_store either.

Considering that there will be a v2, would you mind adding support for
it as well, for completeness? I think that would complete support for
the prefetch instructions.

> diff --git a/gdb/aarch64-tdep.c b/gdb/aarch64-tdep.c
> index 81d4f160f8f..61c462979b3 100644
> --- a/gdb/aarch64-tdep.c
> +++ b/gdb/aarch64-tdep.c
> @@ -5914,6 +5914,11 @@ aarch64_record_load_store (aarch64_insn_decode_record *aarch64_insn_r)
>        else
>  	if (size_bits != 0x03)
>  	  ld_flag = 0x01;
> +	else if (size_bits == 0x3 && vector_flag == 0x0 && opc == 0x2)
> +	  {
> +	    /* PRFM (register) or RPRFM */
> +	    return AARCH64_RECORD_SUCCESS;
> +	  }
>  	else
>  	  return AARCH64_RECORD_UNKNOWN;

-- 
Thiago