aarch64: Fix SVE ACLE gimple folds for C++ LTO [PR116629]

The SVE ACLE code has two ways of handling overloaded functions.
One, used by C, is to define a single dummy function for each unique
overloaded name, with resolve_overloaded_builtin then resolving calls
to real non-overloaded functions.  The other, used by C++, is to
define a separate function for each individual overload.

The builtins harness assigns integer function codes programmatically.
However, LTO requires it to use the same assignment for every
translation unit, regardless of language.  This means that C++ TUs
need to create (unused) slots for the C overloads and that C TUs
need to create (unused) slots for the C++ overloads.

In many ways, it doesn't matter whether the LTO frontend itself
uses the C approach or the C++ approach to defining overloaded
functions, since the LTO frontend never has to resolve source-level
overloading.  However, the C++ approach of defining a separate
function for each overload means that C++ calls never need to
be redirected to a different function.  Calls to an overload
can appear in the LTO dump and survive until expand.  In contrast,
calls to C's dummy overload functions are resolved by the front
end and never survive to LTO (or expand).

Some optimisations work by moving between sibling functions, such as _m
to _x.  If the source function is an overload, the expected destination
function is too.  The LTO frontend needs to define C++ overloads if it
wants to do this optimisation properly for C++.

The PR is about a tree checking failure caused by trying to use a
stubbed-out C++ overload in LTO.  Dealing with that by detecting the
stub (rather than changing which overloads are defined) would have
turned this from an ice-on-valid to a missed optimisation.

In future, it would probably make sense to redirect overloads to
non-overloaded functions during gimple folding, in case that exposes
more CSE opportunities.  But it'd probably be of limited benefit, since
it should be rare for code to mix overloaded and non-overloaded uses of
the same operation.  It also wouldn't be suitable for backports.

gcc/
	PR target/116629
	* config/aarch64/aarch64-sve-builtins.cc
	(function_builder::function_builder): Use direct overloads for LTO.

gcc/testsuite/
	PR target/116629
	* gcc.target/aarch64/sve/acle/general/pr106326_2.c: New test.

(cherry picked from commit fee3adbac055c3ff2649fed866c66d44ebfcbe90)

diff --git a/gcc/config/aarch64/aarch64-sve-builtins.cc b/gcc/config/aarch64/aarch64-sve-builtins.cc
index e0458f7c1f6762e9da01ed4131bf4915013b698d..afc0c36e620b87be902866f46627d8cff8e48c52 100644 (file)
--- a/gcc/config/aarch64/aarch64-sve-builtins.cc
+++ b/gcc/config/aarch64/aarch64-sve-builtins.cc
@@ -1259,7 +1259,7 @@ function_builder::function_builder (handle_pragma_index pragma_index,
                                    bool function_nulls)
 {
   m_overload_type = build_function_type (void_type_node, void_list_node);
-  m_direct_overloads = lang_GNU_CXX ();
+  m_direct_overloads = lang_GNU_CXX () || in_lto_p;
 
   if (initial_indexes[pragma_index] == 0)
     {

diff --git a/gcc/testsuite/gcc.target/aarch64/sve/acle/general/pr106326_2.c b/gcc/testsuite/gcc.target/aarch64/sve/acle/general/pr106326_2.c
new file mode 100644 (file)
index 0000000..d312d31
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/sve/acle/general/pr106326_2.c
@@ -0,0 +1,200 @@
+/* { dg-do link } */
+/* { dg-options "-O2 -flto -shared -fPIC --save-temps" } */
+/* { dg-require-effective-target shared } */
+/* { dg-require-effective-target fpic } */
+
+#include <arm_sve.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+svint32_t
+add1 (svint32_t x, svint32_t y)
+{
+  return svadd_z (svptrue_b8 (), x, y);
+}
+
+svint32_t
+add2 (svint32_t x, svint32_t y)
+{
+  return svadd_z (svptrue_b16 (), x, y);
+}
+
+svint32_t
+add3 (svint32_t x, svint32_t y)
+{
+  return svadd_z (svptrue_b32 (), x, y);
+}
+
+svint32_t
+add4 (svint32_t x, svint32_t y)
+{
+  return svadd_z (svptrue_b64 (), x, y);
+}
+
+svint32_t
+add5 (svint32_t x, svint32_t y)
+{
+  return svadd_m (svptrue_b8 (), x, y);
+}
+
+svint32_t
+add6 (svint32_t x, svint32_t y)
+{
+  return svadd_m (svptrue_b16 (), x, y);
+}
+
+svint32_t
+add7 (svint32_t x, svint32_t y)
+{
+  return svadd_m (svptrue_b32 (), x, y);
+}
+
+svint32_t
+add8 (svint32_t x, svint32_t y)
+{
+  return svadd_m (svptrue_b64 (), x, y);
+}
+
+svint16_t
+add9 (svint16_t x, svint16_t y)
+{
+  return svadd_m (svptrue_b32 (), x, y);
+}
+
+svint32_t
+and1 (svint32_t x)
+{
+  return svand_z (svptrue_b8 (), x, 1);
+}
+
+svint32_t
+and2 (svint32_t x)
+{
+  return svand_z (svptrue_b16 (), x, 1);
+}
+
+svint32_t
+and3 (svint32_t x)
+{
+  return svand_z (svptrue_b32 (), x, 1);
+}
+
+svint32_t
+and4 (svint32_t x)
+{
+  return svand_z (svptrue_b64 (), x, 1);
+}
+
+svint32_t
+and5 (svint32_t x)
+{
+  return svand_m (svptrue_b8 (), x, 1);
+}
+
+svint32_t
+and6 (svint32_t x)
+{
+  return svand_m (svptrue_b16 (), x, 1);
+}
+
+svint32_t
+and7 (svint32_t x)
+{
+  return svand_m (svptrue_b32 (), x, 1);
+}
+
+svint32_t
+and8 (svint32_t x)
+{
+  return svand_m (svptrue_b64 (), x, 1);
+}
+
+svbool_t
+and9 (svbool_t x, svbool_t y)
+{
+  return svand_z (svptrue_b8 (), x, y);
+}
+
+svint32_t
+not1 (svint32_t x, svint32_t y)
+{
+  return svnot_m (x, svptrue_b8 (), y);
+}
+
+svint32_t
+cvt1 (svfloat16_t x)
+{
+  return svcvt_s32_z (svptrue_b8 (), x);
+}
+
+svint32_t
+cvt2 (svfloat16_t x)
+{
+  return svcvt_s32_z (svptrue_b16 (), x);
+}
+
+svint32_t
+cvt3 (svfloat16_t x)
+{
+  return svcvt_s32_z (svptrue_b32 (), x);
+}
+
+svint32_t
+cvt4 (svfloat16_t x)
+{
+  return svcvt_s32_z (svptrue_b64 (), x);
+}
+
+svfloat16_t
+cvt5 (svfloat32_t x)
+{
+  return svcvt_f16_z (svptrue_b8 (), x);
+}
+
+svfloat16_t
+cvt6 (svfloat32_t x)
+{
+  return svcvt_f16_z (svptrue_b16 (), x);
+}
+
+svfloat16_t
+cvt7 (svfloat32_t x)
+{
+  return svcvt_f16_z (svptrue_b32 (), x);
+}
+
+svfloat16_t
+cvt8 (svfloat32_t x)
+{
+  return svcvt_f16_z (svptrue_b64 (), x);
+}
+
+svfloat16_t
+cvt9 (svint16_t x)
+{
+  return svcvt_f16_z (svptrue_b8 (), x);
+}
+
+svfloat16_t
+cvt10 (svint16_t x)
+{
+  return svcvt_f16_z (svptrue_b16 (), x);
+}
+
+svfloat16_t
+cvt11 (svint16_t x)
+{
+  return svcvt_f16_z (svptrue_b32 (), x);
+}
+
+svfloat16_t
+cvt12 (svint16_t x)
+{
+  return svcvt_f16_z (svptrue_b64 (), x);
+}
+
+#ifdef __cplusplus
+}
+#endif