Re: [PATCH 2/5] target/ppc: Move floating-point compare instructions to decodetree.

[Richard Henderson <richard.henderson@linaro.org>]

On 9/11/25 02:29, Chinmay Rath wrote:
> 
> On 8/27/25 12:49, Richard Henderson wrote:
>> On 6/19/25 19:58, Chinmay Rath wrote:
>>> +static bool do_helper_cmp(DisasContext *ctx, arg_X_bf *a,
>>> +                          void (*helper)(TCGv_env, TCGv_i64, TCGv_i64,
>>> +                                         TCGv_i32))
>>> +    REQUIRE_INSNS_FLAGS(ctx, FLOAT);
>>
>> ...
>>
>>> +TRANS(FCMPU, do_helper_cmp, gen_helper_FCMPU);
>>> +TRANS(FCMPO, do_helper_cmp, gen_helper_FCMPO);
>>
>> It's probably better to standardize on TRANS_FLAGS even though the flags checked is the 
>> same for both of these.
>>
> Hi Richard,
> 
> I did notice in the code that there are quite many instructions using TRANS and doing flag 
> check in the common helper.
> For example, in target/ppc/translate/fixedpoint-impl.c.inc :
> 
> /static bool do_set_bool_cond(DisasContext *ctx, arg_X_bi *a, bool neg, bool rev)
> {
>      REQUIRE_INSNS_FLAGS2(ctx, ISA310);
>      uint32_t mask = 0x08 >> (a->bi & 0x03);
>      ....
>      return true;
> }
> 
> TRANS(SETBC, do_set_bool_cond, false, false)
> TRANS(SETBCR, do_set_bool_cond, false, true)
> TRANS(SETNBC, do_set_bool_cond, true, false)
> TRANS(SETNBCR, do_set_bool_cond, true, true)
> /
> 
> Do we want to standardize the use of TRANS_FLAGS in all such existing insns ?

Yes, I think so.


r~