On 10/24/21 22:52, Richard Henderson wrote:
> On 10/23/21 2:47 PM, Philippe Mathieu-Daudé wrote:
>> +static bool trans_msa_elm_df(DisasContext *ctx, arg_msa_elm *a,
>> + void (*gen_msa_elm_df)(TCGv_ptr, TCGv_i32,
>> + TCGv_i32, TCGv_i32,
>> + TCGv_i32))
>> +{
>> + TCGv_i32 twd;
>> + TCGv_i32 tws;
>> + TCGv_i32 tdf;
>> + TCGv_i32 tn;
>> + uint32_t df, n;
>> +
>> + if (!df_extract(df_elm, a->df, &df, &n)) {
>> + gen_reserved_instruction(ctx);
>> + return true;
>> + }
>
> Again, I think this should happen before the access check.
>
> You could even return false from these extractions, because they
> represent an invalid encoding for this instruction. In a different
> context, there could be another grouped insn that matches.
Yes, you are correct.
