> > where POINTER_PLUS_EXPR last operand has sizetype type, thus unsigned,
> > and in the testcase gimple_assign_rhs2 (def) is thus 0xf000000000000001ULL
> > which multiplied by 8 doesn't fit into signed HWI.  If it would be treated
> > as signed offset instead, it would fit (-0xfffffffffffffffLL, multiplied
> > by 8 is -0x7ffffffffffffff8LL).  Unfortunately with the poly_int obfuscation
> > I'm not sure how to convert it from unsigned to signed poly_int.
>
> mem_ref_offset provides a boiler-plate for this:
>
> poly_offset_int::from (wi::to_poly_wide (TREE_OPERAND (t, 1)), SIGNED);

Thanks, that seems to work.
The test now works on both big-endian and little-endian.

2020-03-05  Richard Biener  <rguenther@suse.de>
	    Jakub Jelinek  <jakub@redhat.com>

	PR tree-optimization/93582
	* tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
	last operand as signed when looking for memset offset.  Formatting
	fix.

	* gcc.dg/tree-ssa/pr93582-11.c: New test.

Co-authored-by: Richard Biener <rguenther@suse.de>