From ebb300d0b4340104dcade3afa656a57da2b7660c Mon Sep 17 00:00:00 2001
From: Antonio Sanchez <cantonios@google.com>
Date: Sun, 23 May 2021 12:35:38 -0700
Subject: Modify Unary/Binary/TernaryOp evaluators to work for non-class types.

This used to work for non-class types (e.g. raw function pointers) in
Eigen 3.3.  This was changed in commit 11f55b29 to optimize the
evaluator:

> `sizeof((A-B).cwiseAbs2())` with A,B Vector4f is now 16 bytes, instead of 48 before this optimization.

though I cannot reproduce the 16 byte result.  Both before the change
and after, with multiple compilers/versions, I always get a result of 40 bytes.

https://godbolt.org/z/MsjTc1PGe

This change modifies the code slightly to allow non-class types.  The
final generated code is identical, and the expression remains 40 bytes
for the `abs2` sample case.

Fixes #2251
---
 Eigen/src/Core/CoreEvaluators.h | 29 ++++++++++++++++-------------
 1 file changed, 16 insertions(+), 13 deletions(-)

diff --git a/Eigen/src/Core/CoreEvaluators.h b/Eigen/src/Core/CoreEvaluators.h
index 90c552f13..66e030b1c 100644
--- a/Eigen/src/Core/CoreEvaluators.h
+++ b/Eigen/src/Core/CoreEvaluators.h
@@ -606,13 +606,13 @@ struct unary_evaluator<CwiseUnaryOp<UnaryOp, ArgType>, IndexBased >
 protected:
 
   // this helper permits to completely eliminate the functor if it is empty
-  class Data : private UnaryOp
+  struct Data
   {
-  public:
     EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
-    Data(const XprType& xpr) : UnaryOp(xpr.functor()), argImpl(xpr.nestedExpression()) {}
+    Data(const XprType& xpr) : op(xpr.functor()), argImpl(xpr.nestedExpression()) {}
     EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
-    const UnaryOp& func() const { return static_cast<const UnaryOp&>(*this); }
+    const UnaryOp& func() const { return op; }
+    UnaryOp op;
     evaluator<ArgType> argImpl;
   };
 
@@ -700,12 +700,13 @@ struct ternary_evaluator<CwiseTernaryOp<TernaryOp, Arg1, Arg2, Arg3>, IndexBased
 
 protected:
   // this helper permits to completely eliminate the functor if it is empty
-  struct Data : private TernaryOp
+  struct Data
   {
     EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
-    Data(const XprType& xpr) : TernaryOp(xpr.functor()), arg1Impl(xpr.arg1()), arg2Impl(xpr.arg2()), arg3Impl(xpr.arg3()) {}
+    Data(const XprType& xpr) : op(xpr.functor()), arg1Impl(xpr.arg1()), arg2Impl(xpr.arg2()), arg3Impl(xpr.arg3()) {}
     EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
-    const TernaryOp& func() const { return static_cast<const TernaryOp&>(*this); }
+    const TernaryOp& func() const { return op; }
+    TernaryOp op;
     evaluator<Arg1> arg1Impl;
     evaluator<Arg2> arg2Impl;
     evaluator<Arg3> arg3Impl;
@@ -793,12 +794,13 @@ struct binary_evaluator<CwiseBinaryOp<BinaryOp, Lhs, Rhs>, IndexBased, IndexBase
 protected:
 
   // this helper permits to completely eliminate the functor if it is empty
-  struct Data : private BinaryOp
+  struct Data
   {
     EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
-    Data(const XprType& xpr) : BinaryOp(xpr.functor()), lhsImpl(xpr.lhs()), rhsImpl(xpr.rhs()) {}
+    Data(const XprType& xpr) : op(xpr.functor()), lhsImpl(xpr.lhs()), rhsImpl(xpr.rhs()) {}
     EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
-    const BinaryOp& func() const { return static_cast<const BinaryOp&>(*this); }
+    const BinaryOp& func() const { return op; }
+    BinaryOp op;
     evaluator<Lhs> lhsImpl;
     evaluator<Rhs> rhsImpl;
   };
@@ -858,12 +860,13 @@ struct unary_evaluator<CwiseUnaryView<UnaryOp, ArgType>, IndexBased>
 protected:
 
   // this helper permits to completely eliminate the functor if it is empty
-  struct Data : private UnaryOp
+  struct Data
   {
     EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
-    Data(const XprType& xpr) : UnaryOp(xpr.functor()), argImpl(xpr.nestedExpression()) {}
+    Data(const XprType& xpr) : op(xpr.functor()), argImpl(xpr.nestedExpression()) {}
     EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
-    const UnaryOp& func() const { return static_cast<const UnaryOp&>(*this); }
+    const UnaryOp& func() const { return op; }
+    UnaryOp op;
     evaluator<ArgType> argImpl;
   };
 
-- 
cgit v1.2.3