Re: [PATCH v3 41/48] tcg/optimize: Sink commutative operand swapping into fold functions

[Alex Bennée]

Richard Henderson <richard.henderson@linaro.org> writes:

> On 10/26/21 9:27 AM, Alex Bennée wrote:
>> I find this a bit magical. I couldn't find anything about TCGArg except
>> it's type:
>>    typedef tcg_target_ulong TCGArg;
>
> For an argument that contains a temp,
>
> static inline TCGArg temp_arg(TCGTemp *ts)
> {
>     return (uintptr_t)ts;
> }
>
> static inline TCGTemp *arg_temp(TCGArg a)
> {
>     return (TCGTemp *)(uintptr_t)a;
> }
>
> i.e. the TCGArg is in fact a pointer.
>
>> so I'm not sure what to make of -1 in this case. I guess it just means
>> we never have a (sum == 0 && dest == a2) leg but it's not obvious
>> reading the fold code.
>
> Indeed.  The use of -1 goes back quite a ways.  How about
>
> #define NO_DEST  temp_arg(NULL)
>
> which will also fail to match in that expression?

Sounds good to me. It would be nice to document the details of what
TCGArg can be in another patch.

-- 
Alex Bennée