On Wed, Sep 3, 2025 at 10:29 PM Richard Henderson <
[email protected]> wrote:
> On 9/3/25 15:01, Max Chou wrote:
> > +#define OPMVX_VQDOTQ(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2)
> \
> > +static void do_##NAME(void *vd, target_long s1, void *vs2, int i)
> \
> > +{
> \
> > + int idx;
> \
> > + T1 r1;
> \
> > + T2 r2;
> \
> > + TX1 *r1_buf = (TX1 *)&s1;
> \
> > + TX2 *r2_buf = (TX2 *)vs2 + HD(i);
> \
> > + TD acc = ((TD *)vd)[HD(i)];
> \
> > +
> \
> > + for (idx = 0; idx < 4; ++idx) {
> \
> > + r1 = *((T1 *)r1_buf + HS1(idx));
> \
> > + r2 = *((T2 *)r2_buf + HS2(idx));
> \
> > + acc += r1 * r2;
> \
>
> One could argue for a missing widening cast to TD here. You got away with
> it because the
> only uses happen to have small inputs and "int" sized outputs, so C
> arithmetic promotion
> worked for you.
>
> You can move the variable declarations into the loop.
>
Thanks for the suggestions. I overlooked this part. I’ll update version 3
to include this part.
Thanks a lot,
Max
