From 81484a47604e1041c437501c265e9fe02bf2e2e7 Mon Sep 17 00:00:00 2001
From: nmeyer-ur <nils.meyer@ur.de>
Date: Wed, 20 May 2020 22:36:45 +0200
Subject: [PATCH] symmetrize Mult and MultAddComplex

---
 Grid/simd/Grid_a64fx-2.h         | 6 ++----
 Grid/simd/Grid_a64fx-fixedsize.h | 8 ++++----
 2 files changed, 6 insertions(+), 8 deletions(-)

diff --git a/Grid/simd/Grid_a64fx-2.h b/Grid/simd/Grid_a64fx-2.h
index cc25faa3..d92f8b40 100644
--- a/Grid/simd/Grid_a64fx-2.h
+++ b/Grid/simd/Grid_a64fx-2.h
@@ -400,7 +400,7 @@ struct MultComplex{
 struct MultAddComplex{
   // Complex a*b+c
   template <typename T>
-  inline vec<T> mac(const vec<T> &a, const vec<T> &b, const vec<T> &c){
+  inline vec<T> mac(const vec<T> &a, const vec<T> b, const vec<T> c){
 
     vec<T> out;
     svbool_t pg1 = acle<T>::pg1();
@@ -412,9 +412,7 @@ struct MultAddComplex{
     typename acle<T>::vt r_v = svcmla_x(pg1, c_v, a_v, b_v, 0);
     r_v = svcmla_x(pg1, r_v, a_v, b_v, 90);
 
-    svst1(pg1, out.v, r_v);
-
-    return out;
+    svst1(pg1, a.v, r_v);
   }
 };
 
diff --git a/Grid/simd/Grid_a64fx-fixedsize.h b/Grid/simd/Grid_a64fx-fixedsize.h
index 3a0066d1..9e0fc51e 100644
--- a/Grid/simd/Grid_a64fx-fixedsize.h
+++ b/Grid/simd/Grid_a64fx-fixedsize.h
@@ -369,18 +369,18 @@ struct MultComplex{
 struct MultAddComplex{
   // Complex a*b+c
   // Complex float
-  inline vecf mac(vecf a, vecf b, vecf c){
+  inline vecf mac(vecf &a, vecf b, vecf c){
     pred pg1 = acle<float>::pg1();
     // using FCMLA
     vecf r_v = svcmla_x(pg1, c, a, b, 0);
-    return svcmla_x(pg1, r_v, a, b, 90);
+    a = svcmla_x(pg1, r_v, a, b, 90);
   }
   // Complex double
-  inline vecd mac(vecd a, vecd b, vecd c){
+  inline vecd mac(vecd &a, vecd b, vecd c){
     pred pg1 = acle<double>::pg1();
     // using FCMLA
     vecd r_v = svcmla_x(pg1, c, a, b, 0);
-    return svcmla_x(pg1, r_v, a, b, 90);
+    a = svcmla_x(pg1, r_v, a, b, 90);
   }
 };