Re: [PATCH v3 8/8] target/ppc: add vdiv{su}{wd} vmod{su}{wd} instructions

[Richard Henderson]

On 6/25/20 10:00 AM, Lijun Pan wrote:
> +#define VDIV_MOD_DO(name, op, element, sign, bit)                       \
> +    void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)       \
> +    {                                                                   \
> +        int i;                                                          \
> +                                                                        \
> +                                                                        \
> +        for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
> +            if (unlikely((b->element[i] == 0) ||                        \
> +                (sign &&                                                \
> +                (b->element[i] == UINT##bit##_MAX) &&                   \
> +                (a->element[i] == INT##bit##_MIN))))                    \
> +                continue;                                               \
> +            r->element[i] = a->element[i] op b->element[i];             \
> +        }                                                               \
> +    }

Missing braces for the if.  Extra blank line before the for.

I see that the ISA document says divide-by-zero produces an undefined result,
so leaving the previous contents does seem to be within the letter of the law.

However... Are you able to test what real hardware produces?  It would be nice
(but not required) to match if it is simple to do so.

Whichever way we go with the undefined result, this deserves a comment.


r~