| Message ID | 20231114005555.2139904-1-hongtao.liu@intel.com |
|---|---|
| State | New |
| Headers | show |
| Series | Fix ICE in vectorizable_nonlinear_induction with bitfield. | expand |
On Tue, Nov 14, 2023 at 1:56 AM liuhongt <hongtao.liu@intel.com> wrote: > > if (TREE_CODE (init_expr) == INTEGER_CST) > init_expr = fold_convert (TREE_TYPE (vectype), init_expr); > else > gcc_assert (tree_nop_conversion_p (TREE_TYPE (vectype), > TREE_TYPE (init_expr))); > > and init_expr is a 24 bit integer type while vectype has 32bit components. > > The "fix" is to bail out instead of asserting. > > Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}. > Ok for trunk?(and backport to GCC-13) OK for both. Thanks, Richard. > gcc/ChangeLog: > > PR tree-optimization/112496 > * tree-vect-loop.cc (vectorizable_nonlinear_induction): Return > false when !tree_nop_conversion_p (TREE_TYPE (vectype), > TREE_TYPE (init_expr). > > gcc/testsuite/ChangeLog: > > * gcc.target/i386/pr112496.c: New test. > --- > gcc/testsuite/gcc.target/i386/pr112496.c | 7 +++++++ > gcc/tree-vect-loop.cc | 13 ++++++++++--- > 2 files changed, 17 insertions(+), 3 deletions(-) > create mode 100644 gcc/testsuite/gcc.target/i386/pr112496.c > > diff --git a/gcc/testsuite/gcc.target/i386/pr112496.c b/gcc/testsuite/gcc.target/i386/pr112496.c > new file mode 100644 > index 00000000000..c478fda9cce > --- /dev/null > +++ b/gcc/testsuite/gcc.target/i386/pr112496.c > @@ -0,0 +1,7 @@ > +/* { dg-do compile } */ > +/* { dg-options "-O2" } */ > + > +struct T { int x : 24; } v; > +void f1(int x) { > + while (v.x - ((v.x <<= 1) - v.x)) ; > +} > diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc > index 1304b7ece79..75d69bbc30e 100644 > --- a/gcc/tree-vect-loop.cc > +++ b/gcc/tree-vect-loop.cc > @@ -9429,9 +9429,16 @@ vectorizable_nonlinear_induction (loop_vec_info loop_vinfo, > > if (TREE_CODE (init_expr) == INTEGER_CST) > init_expr = fold_convert (TREE_TYPE (vectype), init_expr); > - else > - gcc_assert (tree_nop_conversion_p (TREE_TYPE (vectype), > - TREE_TYPE (init_expr))); > + else if (!tree_nop_conversion_p (TREE_TYPE (vectype), TREE_TYPE (init_expr))) > + { > + /* INIT_EXPR could be a bit_field, bail out for such case. */ > + if (dump_enabled_p ()) > + dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location, > + "nonlinear induction vectorization failed:" > + " component type of vectype is not a nop conversion" > + " from type of init_expr.\n"); > + return false; > + } > > switch (induction_type) > { > -- > 2.31.1 >
diff --git a/gcc/testsuite/gcc.target/i386/pr112496.c b/gcc/testsuite/gcc.target/i386/pr112496.c new file mode 100644 index 00000000000..c478fda9cce --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr112496.c @@ -0,0 +1,7 @@ +/* { dg-do compile } */ +/* { dg-options "-O2" } */ + +struct T { int x : 24; } v; +void f1(int x) { + while (v.x - ((v.x <<= 1) - v.x)) ; +} diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc index 1304b7ece79..75d69bbc30e 100644 --- a/gcc/tree-vect-loop.cc +++ b/gcc/tree-vect-loop.cc @@ -9429,9 +9429,16 @@ vectorizable_nonlinear_induction (loop_vec_info loop_vinfo, if (TREE_CODE (init_expr) == INTEGER_CST) init_expr = fold_convert (TREE_TYPE (vectype), init_expr); - else - gcc_assert (tree_nop_conversion_p (TREE_TYPE (vectype), - TREE_TYPE (init_expr))); + else if (!tree_nop_conversion_p (TREE_TYPE (vectype), TREE_TYPE (init_expr))) + { + /* INIT_EXPR could be a bit_field, bail out for such case. */ + if (dump_enabled_p ()) + dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location, + "nonlinear induction vectorization failed:" + " component type of vectype is not a nop conversion" + " from type of init_expr.\n"); + return false; + } switch (induction_type) {
if (TREE_CODE (init_expr) == INTEGER_CST) init_expr = fold_convert (TREE_TYPE (vectype), init_expr); else gcc_assert (tree_nop_conversion_p (TREE_TYPE (vectype), TREE_TYPE (init_expr))); and init_expr is a 24 bit integer type while vectype has 32bit components. The "fix" is to bail out instead of asserting. Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}. Ok for trunk?(and backport to GCC-13) gcc/ChangeLog: PR tree-optimization/112496 * tree-vect-loop.cc (vectorizable_nonlinear_induction): Return false when !tree_nop_conversion_p (TREE_TYPE (vectype), TREE_TYPE (init_expr). gcc/testsuite/ChangeLog: * gcc.target/i386/pr112496.c: New test. --- gcc/testsuite/gcc.target/i386/pr112496.c | 7 +++++++ gcc/tree-vect-loop.cc | 13 ++++++++++--- 2 files changed, 17 insertions(+), 3 deletions(-) create mode 100644 gcc/testsuite/gcc.target/i386/pr112496.c