https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122541
Richard Earnshaw <rearnsha at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|DUPLICATE |FIXED
--- Comment #8 from Richard Earnshaw <rearnsha at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #5)
> From fold-const.cc:
> ```
> /* Canonicalize (X & C1) | C2. */
> ...
> /* Minimize the number of bits set in C1, i.e. C1 := C1 & ~C2,
> unless (C1 & ~C2) | (C2 & C3) for some C3 is a mask of some
> mode which allows further optimizations. */
> c1 &= msk;
> c2 &= msk;
> wide_int c3 = wi::bit_and_not (c1, c2);
> for (w = BITS_PER_UNIT; w <= width; w <<= 1)
> {
> wide_int mask = wi::mask (w, false,
> TYPE_PRECISION (type));
> if (((c1 | c2) & mask) == mask
> && wi::bit_and_not (c1, mask) == 0)
> {
> c3 = mask;
> break;
> }
> }
>
> if (c3 != c1)
> {
> tem = fold_convert_loc (loc, type, TREE_OPERAND (arg0, 0));
> tem = fold_build2_loc (loc, BIT_AND_EXPR, type, tem,
> wide_int_to_tree (type, c3));
> return fold_build2_loc (loc, BIT_IOR_EXPR, type, tem, arg1);
> }
> ```
Rather than minimizing the number of bits set, it would probably be better to
find a constant with the smallest magnitude. On RISC targets that would
probably lead to a smaller number of insns, while on CISC targets it would lead
to a shorter instruction (assuming it can load arbitrary constants in a single
insn).
