On Thu, Aug 3, 2023 at 12:23 AM Richard Biener via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> On Thu, Aug 3, 2023 at 4:34 AM Andrew Pinski via Gcc-patches
> <gcc-patches@gcc.gnu.org> wrote:
> >
> > So I didn't expect valueization to cause calling gimple_nop_convert
> > to iterate between 2 different SSA names causing an infinite loop
> > in gimple_bitwise_inverted_equal_p.
> > So we should cause a bound on gimple_bitwise_inverted_equal_p calling
> > gimple_nop_convert and only look through one rather than always.
> >
> > OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.
> >
> > gcc/ChangeLog:
> >
> > PR tree-optimization/110874
> > * gimple-match-head.cc (gimple_bitwise_inverted_equal_p):
> > Add new argument, again with default value of true.
> > Don't try gimple_nop_convert if again is false.
> > Update call to gimple_bitwise_inverted_equal_p for
> > new argument.
> >
> > gcc/testsuite/ChangeLog:
> >
> > PR tree-optimization/110874
> > * gcc.c-torture/compile/pr110874-a.c: New test.
> > ---
> > gcc/gimple-match-head.cc | 14 +++++++++-----
> > .../gcc.c-torture/compile/pr110874-a.c | 17 +++++++++++++++++
> > 2 files changed, 26 insertions(+), 5 deletions(-)
> > create mode 100644 gcc/testsuite/gcc.c-torture/compile/pr110874-a.c
> >
> > diff --git a/gcc/gimple-match-head.cc b/gcc/gimple-match-head.cc
> > index b1e96304d7c..e91aaab86dd 100644
> > --- a/gcc/gimple-match-head.cc
> > +++ b/gcc/gimple-match-head.cc
> > @@ -273,7 +273,7 @@ gimple_bitwise_equal_p (tree expr1, tree expr2, tree
> > (*valueize) (tree))
> > /* Helper function for bitwise_equal_p macro. */
> >
> > static inline bool
> > -gimple_bitwise_inverted_equal_p (tree expr1, tree expr2, tree (*valueize)
> > (tree))
> > +gimple_bitwise_inverted_equal_p (tree expr1, tree expr2, tree (*valueize)
> > (tree), bool again = true)
> > {
> > if (expr1 == expr2)
> > return false;
> > @@ -285,12 +285,16 @@ gimple_bitwise_inverted_equal_p (tree expr1, tree
> > expr2, tree (*valueize) (tree)
> > return false;
> >
> > tree other;
> > - if (gimple_nop_convert (expr1, &other, valueize)
> > - && gimple_bitwise_inverted_equal_p (other, expr2, valueize))
> > + if (again
> > + && gimple_nop_convert (expr1, &other, valueize)
> > + && other != expr1
> > + && gimple_bitwise_inverted_equal_p (other, expr2, valueize, false))
> > return true;
> >
> > - if (gimple_nop_convert (expr2, &other, valueize)
> > - && gimple_bitwise_inverted_equal_p (expr1, other, valueize))
> > + if (again
> > + && gimple_nop_convert (expr2, &other, valueize)
> > + && other != expr2
> > + && gimple_bitwise_inverted_equal_p (expr1, other, valueize, false))
> > return true;
>
> Hmm, I don't think this tests all three relevant combinations? I think the
> way
> gimple_bitwise_equal_p handles this is better (not recursing). I'd split out
> the "tail" matching the BIT_NOT to another helper, I suppose that could
> even be a (match ...) pattern here.
That sounds like a better idea. I am testing the patch right now that
uses a (match ) pattern for
the BIT_NOT and CMP cases. That will remove the recursiveness too.
Thanks,
Andrew
>
> > if (TREE_CODE (expr1) != SSA_NAME
> > diff --git a/gcc/testsuite/gcc.c-torture/compile/pr110874-a.c
> > b/gcc/testsuite/gcc.c-torture/compile/pr110874-a.c
> > new file mode 100644
> > index 00000000000..b314410a892
> > --- /dev/null
> > +++ b/gcc/testsuite/gcc.c-torture/compile/pr110874-a.c
> > @@ -0,0 +1,17 @@
> > +struct S1 {
> > + unsigned f0;
> > +};
> > +static int g_161;
> > +void func_109(unsigned g_227, unsigned t) {
> > + struct S1 l_178;
> > + int l_160 = 0x1FAE99D5L;
> > + int *l_230[] = {&l_160};
> > + if (l_160) {
> > + for (l_178.f0 = -7; l_178.f0;) {
> > + ++g_227;
> > + break;
> > + }
> > + (g_161) = g_227;
> > + }
> > + (g_161) &= t;
> > +}
> > --
> > 2.31.1
> >
