From bffd267d176410a517a0fe9afa6dde99c213c08a Mon Sep 17 00:00:00 2001
From: Rasmus Munk Larsen <rmlarsen@google.com>
Date: Thu, 24 Jun 2021 18:52:17 -0700
Subject: Small cleanup: Get rid of the macros
 EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD and CJMADD, which were effectively
 unused, apart from on x86, where the change results in identically performing
 code.

---
 Eigen/src/Core/arch/AltiVec/PacketMath.h          |  4 --
 Eigen/src/Core/arch/MSA/PacketMath.h              |  4 --
 Eigen/src/Core/arch/NEON/PacketMath.h             |  4 --
 Eigen/src/Core/arch/SVE/PacketMath.h              |  4 --
 Eigen/src/Core/arch/ZVector/PacketMath.h          |  4 --
 Eigen/src/Core/products/GeneralBlockPanelKernel.h | 46 ++++-------------------
 6 files changed, 7 insertions(+), 59 deletions(-)

diff --git a/Eigen/src/Core/arch/AltiVec/PacketMath.h b/Eigen/src/Core/arch/AltiVec/PacketMath.h
index a3ebf9ec2..8c42f495c 100755
--- a/Eigen/src/Core/arch/AltiVec/PacketMath.h
+++ b/Eigen/src/Core/arch/AltiVec/PacketMath.h
@@ -22,10 +22,6 @@ namespace internal {
 #define EIGEN_HAS_SINGLE_INSTRUCTION_MADD
 #endif
 
-#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#define EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#endif
-
 // NOTE Altivec has 32 registers, but Eigen only accepts a value of 8 or 16
 #ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS
 #define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS  32
diff --git a/Eigen/src/Core/arch/MSA/PacketMath.h b/Eigen/src/Core/arch/MSA/PacketMath.h
index f03cf61ff..afe8f3375 100644
--- a/Eigen/src/Core/arch/MSA/PacketMath.h
+++ b/Eigen/src/Core/arch/MSA/PacketMath.h
@@ -28,10 +28,6 @@ namespace internal {
 #define EIGEN_HAS_SINGLE_INSTRUCTION_MADD
 #endif
 
-#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#define EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#endif
-
 #ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS
 #define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 32
 #endif
diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h
index e1efe9bcb..5e7702a50 100644
--- a/Eigen/src/Core/arch/NEON/PacketMath.h
+++ b/Eigen/src/Core/arch/NEON/PacketMath.h
@@ -24,10 +24,6 @@ namespace internal {
 #define EIGEN_HAS_SINGLE_INSTRUCTION_MADD
 #endif
 
-#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#define EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#endif
-
 #ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS
 #if EIGEN_ARCH_ARM64
 #define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 32
diff --git a/Eigen/src/Core/arch/SVE/PacketMath.h b/Eigen/src/Core/arch/SVE/PacketMath.h
index 4877b6d80..9060b372f 100644
--- a/Eigen/src/Core/arch/SVE/PacketMath.h
+++ b/Eigen/src/Core/arch/SVE/PacketMath.h
@@ -22,10 +22,6 @@ namespace internal
 #define EIGEN_HAS_SINGLE_INSTRUCTION_MADD
 #endif
 
-#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#define EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#endif
-
 #define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 32
 
 template <typename Scalar, int SVEVectorLength>
diff --git a/Eigen/src/Core/arch/ZVector/PacketMath.h b/Eigen/src/Core/arch/ZVector/PacketMath.h
index 2246439cc..1f55a90a5 100755
--- a/Eigen/src/Core/arch/ZVector/PacketMath.h
+++ b/Eigen/src/Core/arch/ZVector/PacketMath.h
@@ -22,10 +22,6 @@ namespace internal {
 #define EIGEN_HAS_SINGLE_INSTRUCTION_MADD
 #endif
 
-#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#define EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-#endif
-
 #ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS
 #define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS  32
 #endif
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
index 8362ecc02..1116321a9 100644
--- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h
+++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
@@ -349,36 +349,6 @@ inline void computeProductBlockingSizes(Index& k, Index& m, Index& n, Index num_
   computeProductBlockingSizes<LhsScalar,RhsScalar,1,Index>(k, m, n, num_threads);
 }
 
-#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD
-  #define CJMADD(CJ,A,B,C,T)  C = CJ.pmadd(A,B,C);
-#else
-
-  // FIXME (a bit overkill maybe ?)
-
-  template<typename CJ, typename A, typename B, typename C, typename T> struct gebp_madd_selector {
-    EIGEN_ALWAYS_INLINE static void run(const CJ& cj, A& a, B& b, C& c, T& /*t*/)
-    {
-      c = cj.pmadd(a,b,c);
-    }
-  };
-
-  template<typename CJ, typename T> struct gebp_madd_selector<CJ,T,T,T,T> {
-    EIGEN_ALWAYS_INLINE static void run(const CJ& cj, T& a, T& b, T& c, T& t)
-    {
-      t = b; t = cj.pmul(a,t); c = padd(c,t);
-    }
-  };
-
-  template<typename CJ, typename A, typename B, typename C, typename T>
-  EIGEN_STRONG_INLINE void gebp_madd(const CJ& cj, A& a, B& b, C& c, T& t)
-  {
-    gebp_madd_selector<CJ,A,B,C,T>::run(cj,a,b,c,t);
-  }
-
-  #define CJMADD(CJ,A,B,C,T)  gebp_madd(CJ,A,B,C,T);
-//   #define CJMADD(CJ,A,B,C,T)  T = B; T = CJ.pmul(A,T); C = padd(C,T);
-#endif
-
 template <typename RhsPacket, typename RhsPacketx4, int registers_taken>
 struct RhsPanelHelper {
  private:
@@ -2060,14 +2030,14 @@ void gebp_kernel<LhsScalar,RhsScalar,Index,DataMapper,mr,nr,ConjugateLhs,Conjuga
 
               B_0 = blB[0];
               B_1 = blB[1];
-              CJMADD(cj,A0,B_0,C0,  B_0);
-              CJMADD(cj,A0,B_1,C1,  B_1);
-              
+              C0 = cj.pmadd(A0,B_0,C0);
+              C1 = cj.pmadd(A0,B_1,C1);
+
               B_0 = blB[2];
               B_1 = blB[3];
-              CJMADD(cj,A0,B_0,C2,  B_0);
-              CJMADD(cj,A0,B_1,C3,  B_1);
-              
+              C2 = cj.pmadd(A0,B_0,C2);
+              C3 = cj.pmadd(A0,B_1,C3);
+
               blB += 4;
             }
             res(i, j2 + 0) += alpha * C0;
@@ -2092,7 +2062,7 @@ void gebp_kernel<LhsScalar,RhsScalar,Index,DataMapper,mr,nr,ConjugateLhs,Conjuga
           {
             LhsScalar A0 = blA[k];
             RhsScalar B_0 = blB[k];
-            CJMADD(cj, A0, B_0, C0, B_0);
+            C0 = cj.pmadd(A0, B_0, C0);
           }
           res(i, j2) += alpha * C0;
         }
@@ -2101,8 +2071,6 @@ void gebp_kernel<LhsScalar,RhsScalar,Index,DataMapper,mr,nr,ConjugateLhs,Conjuga
   }
 
 
-#undef CJMADD
-
 // pack a block of the lhs
 // The traversal is as follow (mr==4):
 //   0  4  8 12 ...
-- 
cgit v1.2.3