The svwhilele folder mishandled the degenerate case in which
the second argument is the maximum integer. In that case,
the result is all-true regardless of the first parameter:
If the second scalar operand is equal to the maximum signed integer
value then a condition which includes an equality test can never fail
and the result will be an all-true predicate.
This is because the conceptual "increment the first operand
by 1 after each element" is done modulo the range of the operand.
The GCC code was instead treating it as infinite precision.
whilele_5.c even had a test for the incorrect behaviour.
The easiest fix seemed to be to handle that case specially before
doing constant folding. This also copes with variable first operands.
gcc/
PR target/116999
PR target/117045
* config/aarch64/aarch64-sve-builtins-base.cc
(svwhilelx_impl::fold): Check for WHILELTs of the minimum value
and WHILELEs of the maximum value. Fold them to all-false and
all-true respectively.
gcc/testsuite/
PR target/116999
PR target/117045
* gcc.target/aarch64/sve/acle/general/whilele_5.c: Fix bogus
expected result.
* gcc.target/aarch64/sve/acle/general/whilele_11.c: New test.
* gcc.target/aarch64/sve/acle/general/whilele_12.c: Likewise.
(cherry picked from commit
50e7c51b0a0e9dc1d93f829016ae743b4f2e5070)
: while_comparison (unspec_for_sint, unspec_for_uint), m_eq_p (eq_p)
{}
- /* Try to fold a call by treating its arguments as constants of type T. */
+ /* Try to fold a call by treating its arguments as constants of type T.
+ We have already filtered out the degenerate cases of X .LT. MIN
+ and X .LE. MAX. */
template<typename T>
gimple *
fold_type (gimple_folder &f) const
gimple *
fold (gimple_folder &f) const OVERRIDE
{
+ /* Filter out cases where the condition is always true or always false. */
+ tree arg1 = gimple_call_arg (f.call, 1);
+ if (!m_eq_p && operand_equal_p (arg1, TYPE_MIN_VALUE (TREE_TYPE (arg1))))
+ return f.fold_to_pfalse ();
+ if (m_eq_p && operand_equal_p (arg1, TYPE_MAX_VALUE (TREE_TYPE (arg1))))
+ return f.fold_to_ptrue ();
+
if (f.type_suffix (1).unsigned_p)
return fold_type<poly_uint64> (f);
else
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-O2" } */
+
+#include <arm_sve.h>
+#include <limits.h>
+
+svbool_t
+f1 (volatile int32_t *ptr)
+{
+ return svwhilelt_b8_s32 (*ptr, INT32_MIN);
+}
+
+svbool_t
+f2 (volatile uint32_t *ptr)
+{
+ return svwhilelt_b16_u32 (*ptr, 0);
+}
+
+svbool_t
+f3 (volatile int64_t *ptr)
+{
+ return svwhilelt_b32_s64 (*ptr, INT64_MIN);
+}
+
+svbool_t
+f4 (volatile uint64_t *ptr)
+{
+ return svwhilelt_b64_u64 (*ptr, 0);
+}
+
+/* { dg-final { scan-assembler-times {\tpfalse\tp[0-9]+\.b\n} 4 } } */
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-O2" } */
+
+#include <arm_sve.h>
+#include <limits.h>
+
+svbool_t
+f1 (volatile int32_t *ptr)
+{
+ return svwhilele_b8_s32 (*ptr, INT32_MAX);
+}
+
+svbool_t
+f2 (volatile uint32_t *ptr)
+{
+ return svwhilele_b16_u32 (*ptr, UINT32_MAX);
+}
+
+svbool_t
+f3 (volatile int64_t *ptr)
+{
+ return svwhilele_b32_s64 (*ptr, INT64_MAX);
+}
+
+svbool_t
+f4 (volatile uint64_t *ptr)
+{
+ return svwhilele_b64_u64 (*ptr, UINT64_MAX);
+}
+
+/* { dg-final { scan-assembler {\tptrue\tp[0-9]+\.b(?:, all)\n} } } */
+/* { dg-final { scan-assembler {\tptrue\tp[0-9]+\.h(?:, all)\n} } } */
+/* { dg-final { scan-assembler {\tptrue\tp[0-9]+\.s(?:, all)\n} } } */
+/* { dg-final { scan-assembler {\tptrue\tp[0-9]+\.d(?:, all)\n} } } */
*ptr = svwhilele_b16_s32 (0x7ffffffb, 0x7fffffff);
}
-/* { dg-final { scan-assembler {\tptrue\tp[0-7]\.h, vl5\n} } } */
+/* { dg-final { scan-assembler {\tptrue\tp[0-9]+\.h(?:, all)\n} } } */
void
test4 (svbool_t *ptr)