From d5067b62768b0621ad8fe214a9a690aab91050d6 Mon Sep 17 00:00:00 2001
From: JosJuice <josjuice@gmail.com>
Date: Sun, 28 Sep 2025 15:57:10 +0200
Subject: [PATCH] Jit64: Replace MOVSD with MOVAPD in software FMA

Should be a little faster by avoiding false dependencies. Note that
there is one remaining MOVSD that really needs to be a MOVSD.
---
 .../Core/PowerPC/Jit64/Jit_FloatingPoint.cpp   | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/Source/Core/Core/PowerPC/Jit64/Jit_FloatingPoint.cpp b/Source/Core/Core/PowerPC/Jit64/Jit_FloatingPoint.cpp
index 75cfbed3d6..5e7802ff67 100644
--- a/Source/Core/Core/PowerPC/Jit64/Jit_FloatingPoint.cpp
+++ b/Source/Core/Core/PowerPC/Jit64/Jit_FloatingPoint.cpp
@@ -443,6 +443,8 @@ void Jit64::fmaddXX(UGeckoInstruction inst)
       {
         if (round_input)
           Force25BitPrecision(XMM1, Rc, XMM2);
+        else if (Rc.IsSimpleReg())
+          MOVAPD(XMM1, Rc);
         else
           MOVSD(XMM1, Rc);
       }
@@ -461,8 +463,15 @@ void Jit64::fmaddXX(UGeckoInstruction inst)
 
       if (i == 0)
       {
-        MOVSD(XMM0, Ra);
-        MOVSD(XMM2, Rb);
+        if (Ra.IsSimpleReg())
+          MOVAPD(XMM0, Ra);
+        else
+          MOVSD(XMM0, Ra);
+
+        if (Rb.IsSimpleReg())
+          MOVAPD(XMM2, Rb);
+        else
+          MOVSD(XMM2, Rb);
       }
       else
       {
@@ -480,9 +489,14 @@ void Jit64::fmaddXX(UGeckoInstruction inst)
     }
 
     if (packed)
+    {
+      // result_xmm's upper lane has the result of the first loop iteration
       MOVSD(R(result_xmm), XMM0);
+    }
     else
+    {
       DEBUG_ASSERT(result_xmm == XMM0);
+    }
 
     if (madds_accurate_nans)
     {