diff options
| author |  Rasmus Munk Larsen <rmlarsen@google.com> | 2019-11-06 14:24:33 -0800 |
|---|---|---|
| committer | Rasmus Munk Larsen <rmlarsen@google.com> | 2019-11-06 14:24:33 -0800 |
| commit | cc3d0e6a407d839487de13ecdd565c26c5294f9e (patch) | |
| tree | a10114887c3f2bb4c8322c95b3299b8a247210b4 /Eigen/src/Core/util/Macros.h | |
| parent | ee404667e2d3ffd60c2b39dfaf9fa5de0413f3bd (diff) | |
Add EIGEN_HAS_INTRINSIC_INT128 macro
Add a new EIGEN_HAS_INTRINSIC_INT128 macro, and use this instead of __SIZEOF_INT128__. This fixes related issues with TensorIntDiv.h when building with Clang for Windows, where support for 128-bit integer arithmetic is advertised but broken in practice.
Diffstat (limited to 'Eigen/src/Core/util/Macros.h')
| -rw-r--r-- | Eigen/src/Core/util/Macros.h | 15 |
1 files changed, 15 insertions, 0 deletions
diff --git a/Eigen/src/Core/util/Macros.h b/Eigen/src/Core/util/Macros.h index 6b5ab4278..e7bf75a81 100644 --- a/Eigen/src/Core/util/Macros.h +++ b/Eigen/src/Core/util/Macros.h @@ -757,6 +757,21 @@ #endif #endif +// Does the compiler support the __int128 and __uint128_t extensions for 128-bit +// integer arithmetic? +// +// Clang and GCC define __SIZEOF_INT128__ when these extensions are supported, +// but we avoid using them in certain cases: +// +// * Building using Clang for Windows, where the Clang runtime library has +// 128-bit support only on LP64 architectures, but Windows is LLP64. +#ifndef EIGEN_HAS_BUILTIN_INT128 +#if defined(__SIZEOF_INT128__) && !(EIGEN_OS_WIN && EIGEN_COMP_CLANG) +#define EIGEN_HAS_BUILTIN_INT128 1 +#else +#define EIGEN_HAS_BUILTIN_INT128 0 +#endif +#endif //------------------------------------------------------------------------------------------ // Preprocessor programming helpers |