Use a more conservative method to detect non-finite inputs to cbrt.
diff --git a/Eigen/src/Core/arch/Default/GenericPacketMathFunctions.h b/Eigen/src/Core/arch/Default/GenericPacketMathFunctions.h
index 56744aa..e9f564b 100644
--- a/Eigen/src/Core/arch/Default/GenericPacketMathFunctions.h
+++ b/Eigen/src/Core/arch/Default/GenericPacketMathFunctions.h
@@ -333,12 +333,11 @@
const Packet x_sign = pand(sign_mask, x);
Packet root = por(x_sign, abs_root);
- // Handle non-finite and zero values of x.
- // constexpr Scalar kInf = NumTraits<Scalar>::infinity();
- const Packet is_not_finite = psub(x,x);;
+ // Pass non-finite and zero values of x straight through.
+ const Packet is_not_finite = por(pisinf(x), pisnan(x));
const Packet is_zero = pcmp_eq(pzero(x), x);
- const Packet use_root = por(is_not_finite, is_zero);
- return pselect(use_root, x, root);
+ const Packet use_x = por(is_not_finite, is_zero);
+ return pselect(use_x, x, root);
}
// Generic implementation of cbrt(x) for float.
