updates based on PR feedback

There are two major changes (and a few minor ones which are not listed here...see PR discussion for details)

1. Eigen::half implementations for HIP and CUDA have been merged.
This means that
- `CUDA/Half.h` and `HIP/hcc/Half.h` got merged to a new file `GPU/Half.h`
- `CUDA/PacketMathHalf.h` and `HIP/hcc/PacketMathHalf.h` got merged to a new file `GPU/PacketMathHalf.h`
- `CUDA/TypeCasting.h` and `HIP/hcc/TypeCasting.h` got merged to a new file `GPU/TypeCasting.h`

After this change the `HIP/hcc` directory only contains one file `math_constants.h`. That will go away too once that file becomes a part of the HIP install.

2. new macros EIGEN_GPUCC, EIGEN_GPU_COMPILE_PHASE and EIGEN_HAS_GPU_FP16 have been added and the code has been updated to use them where appropriate.
- `EIGEN_GPUCC` is the same as `(EIGEN_CUDACC || EIGEN_HIPCC)`
- `EIGEN_GPU_DEVICE_COMPILE` is the same as `(EIGEN_CUDA_ARCH || EIGEN_HIP_DEVICE_COMPILE)`
- `EIGEN_HAS_GPU_FP16` is the same as `(EIGEN_HAS_CUDA_FP16 or EIGEN_HAS_HIP_FP16)`

1 files changed, 16 insertions, 13 deletions

diff --git a/Eigen/src/Core/util/Meta.h b/Eigen/src/Core/util/Meta.h
index 7f78cc89c..5a358bc12 100755
--- a/Eigen/src/Core/util/Meta.h
+++ b/Eigen/src/Core/util/Meta.h
@@ -11,16 +11,19 @@
 #ifndef EIGEN_META_H
 #define EIGEN_META_H
 
-#if defined(EIGEN_CUDA_ARCH)
-#include <cfloat>
-#include <math_constants.h>
-#endif
+#if defined(EIGEN_GPU_COMPILE_PHASE)
 
-#if defined(EIGEN_HIP_DEVICE_COMPILE)
-#include <cfloat>
-#include "Eigen/src/Core/arch/HIP/hcc/math_constants.h"
-#endif
+ #include <cfloat>
+
+ #if defined(EIGEN_CUDA_ARCH)
+  #include <math_constants.h>
+ #endif
 
+ #if defined(EIGEN_HIP_DEVICE_COMPILE)
+  #include "Eigen/src/Core/arch/HIP/hcc/math_constants.h"
+  #endif
+
+#endif
 
 #if EIGEN_COMP_ICC>=1600 &&  __cplusplus >= 201103L
 #include <cstdint>
@@ -181,7 +184,7 @@ template<bool Condition, typename T=void> struct enable_if;
 template<typename T> struct enable_if<true,T>
 { typedef T type; };
 
-#if defined(EIGEN_CUDA_ARCH) || defined(EIGEN_HIP_DEVICE_COMPILE)
+#if defined(EIGEN_GPU_COMPILE_PHASE)
 #if !defined(__FLT_EPSILON__)
 #define __FLT_EPSILON__ FLT_EPSILON
 #define __DBL_EPSILON__ DBL_EPSILON
@@ -565,13 +568,13 @@ template<typename T, typename U> struct scalar_product_traits
 
 namespace numext {
   
-#if defined(EIGEN_CUDA_ARCH) || defined(EIGEN_HIP_DEVICE_COMPILE)
+#if defined(EIGEN_GPU_COMPILE_PHASE)
 template<typename T> EIGEN_DEVICE_FUNC   void swap(T &a, T &b) { T tmp = b; b = a; a = tmp; }
 #else
 template<typename T> EIGEN_STRONG_INLINE void swap(T &a, T &b) { std::swap(a,b); }
 #endif
 
-#if defined(EIGEN_CUDA_ARCH) || defined(EIGEN_HIP_DEVICE_COMPILE)
+#if defined(EIGEN_GPU_COMPILE_PHASE)
 using internal::device::numeric_limits;
 #else
 using std::numeric_limits;
@@ -590,7 +593,7 @@ T div_ceil(const T &a, const T &b)
 template<typename X, typename Y> EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC
 bool equal_strict(const X& x,const Y& y) { return x == y; }
 
-#if !defined(EIGEN_CUDA_ARCH)
+#if !defined(EIGEN_GPU_COMPILE_PHASE) 
 template<> EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC
 bool equal_strict(const float& x,const float& y) { return std::equal_to<float>()(x,y); }
 
@@ -601,7 +604,7 @@ bool equal_strict(const double& x,const double& y) { return std::equal_to<double
 template<typename X, typename Y> EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC
 bool not_equal_strict(const X& x,const Y& y) { return x != y; }
 
-#if !defined(EIGEN_CUDA_ARCH)
+#if !defined(EIGEN_GPU_COMPILE_PHASE) 
 template<> EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC
 bool not_equal_strict(const float& x,const float& y) { return std::not_equal_to<float>()(x,y); }