On Thu, Aug 15, 2024 at 2:16 AM Nathan Bossart <nathandboss...@gmail.com> wrote:
>
+static inline bool
+pg_neg_u64_overflow(uint64 a, int64 *result)
+{
+#if defined(HAVE__BUILTIN_OP_OVERFLOW)
+ return __builtin_sub_overflow(0, a, result);
+#elif defined(HAVE_INT128)
+ uint128 res = -((int128) a);
+
+ if (unlikely(res < PG_INT64_MIN))
+ {
+ *result = 0x5EED; /* to avoid spurious warnings */
+ return true;
+ }
+ *result = res;
+ return false;
+#else
+ if (unlikely(a > (uint64) PG_INT64_MAX + 1))
+ {
+ *result = 0x5EED; /* to avoid spurious warnings */
+ return true;
+ }
+ if (unlikely(a == (uint64) PG_INT64_MAX + 1))
+ *result = PG_INT64_MIN;
+ else
+ *result = -((int64) a);
+ return false;
+#endif
sorry to bother you.
i am confused with
"
+#elif defined(HAVE_INT128)
+ uint128 res = -((int128) a);
"
I thought "unsigned" means non-negative, therefore uint128 means non-negative.
therefore "int128 res = -((int128) a);" makes sense to me.
also in HAVE_INT128 branch
do we need cast int128 to int64, like
*result = (int64) res;
