Re: [PATCH] tree-optimization/108522 Use COMPONENT_REF offset when available

[Jakub Jelinek <jakub@redhat.com>]

On Wed, Jan 25, 2023 at 12:27:13PM +0100, Jakub Jelinek via Gcc-patches wrote:
> On Wed, Jan 25, 2023 at 06:22:56AM -0500, Siddhesh Poyarekar wrote:
> > On 2023-01-25 02:44, Richard Biener wrote:
> > > >         t = TREE_OPERAND (expr, 1);
> > > > -      off = size_binop (PLUS_EXPR, DECL_FIELD_OFFSET (t),
> > > > +      off = size_binop (PLUS_EXPR,
> > > > +                       (TREE_OPERAND (expr, 2) ? TREE_OPERAND (expr, 2)
> > > > +                        : DECL_FIELD_OFFSET (t)),
> > > 
> > > That isn't correct - operand 2 is the field offset in units of
> > > DECL_OFFSET_ALIGN (t) / BITS_PER_UNIT.
> > > See component_ref_filed_offset (), maybe you should be using that
> > > function instead?
> > 
> > Ahh, and it passed my testing only because I was testing a char. Thanks,
> > I'll test and send an update with additional tests.
> 
> I think you want something like:
>   struct S {
>     char a[n];
>     unsigned long long b;
>     int d;
>     char e[2 * n];
>   } s;
> and test say bdos of &s.d, 0 and &s.e[n - 2], 0

And in the caller compared that to offsetof/sizeof based expressions
for a similar structure which just uses constants instead of the n
in there.

	Jakub