Re: [PATCH 03/46] tcg/optimize: Add fold_masks_zsa, fold_masks_zs, fold_masks_z

[Richard Henderson]

On 12/17/24 14:03, Pierrick Bouvier wrote:
> > +__attribute__((unused))
> > +static bool fold_masks_zs(OptContext *ctx, TCGOp *op,
> > +                          uint64_t z_mask, uint64_t s_mask)
> > +{
> > +    return fold_masks_zsa(ctx, op, z_mask, s_mask, -1);
> > +}
> > +
> > +__attribute__((unused))
> > +static bool fold_masks_z(OptContext *ctx, TCGOp *op, uint64_t z_mask)
> > +{
> > +    return fold_masks_zsa(ctx, op, z_mask, smask_from_zmask(z_mask), -1);
> > +}
> > +
> > +static bool fold_masks(OptContext *ctx, TCGOp *op)
> > +{
> > +    return fold_masks_zsa(ctx, op, ctx->z_mask, ctx->s_mask, ctx->a_mask);
> > +}
> > +
> >   /*
> >    * Convert @op to NOT, if NOT is supported by the host.
> >    * Return true f the conversion is successful, which will still
>
> I see the direction, but why not simply use a structure for this?
>
> If I understand correctly, we'll only pass the masks to callees, so it's easy to pass the 
> pointer down (without any heap allocation needed), and we can have associated builder 
> functions to create the struct only with a limited set of masks, or directly from an 
> existing "ctx".

Why would we want to use a structure?  I'm confused by the question.


r~