From 7797bc2fb29eedbcf81e97f062da5686f4d171d6 Mon Sep 17 00:00:00 2001
From: Lioncash <mathew1800@gmail.com>
Date: Sun, 15 Jul 2018 22:06:29 -0400
Subject: [PATCH] emit_x64_vector: Use non-scratch Use* variants of registers
 within EmitVectorUnsignedAbsoluteDifference()

In some cases, a register isn't modified, depending on the branch taken,
so we can signify this by using the non-scratch variants in certain
cases.
---
 src/backend_x64/emit_x64_vector.cpp | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)

diff --git a/src/backend_x64/emit_x64_vector.cpp b/src/backend_x64/emit_x64_vector.cpp
index f561e985..56a14d8e 100644
--- a/src/backend_x64/emit_x64_vector.cpp
+++ b/src/backend_x64/emit_x64_vector.cpp
@@ -2188,29 +2188,41 @@ static void EmitVectorUnsignedAbsoluteDifference(size_t esize, EmitContext& ctx,
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
 
     const Xbyak::Xmm temp = ctx.reg_alloc.ScratchXmm();
-    const Xbyak::Xmm x = ctx.reg_alloc.UseScratchXmm(args[0]);
-    const Xbyak::Xmm y = ctx.reg_alloc.UseScratchXmm(args[1]);
 
     switch (esize) {
-    case 8:
+    case 8: {
+        const Xbyak::Xmm x = ctx.reg_alloc.UseXmm(args[0]);
+        const Xbyak::Xmm y = ctx.reg_alloc.UseScratchXmm(args[1]);
+
         code.movdqa(temp, x);
         code.psubusb(temp, y);
         code.psubusb(y, x);
         code.por(temp, y);
         break;
-    case 16:
+    }
+    case 16: {
+        const Xbyak::Xmm x = ctx.reg_alloc.UseXmm(args[0]);
+        const Xbyak::Xmm y = ctx.reg_alloc.UseScratchXmm(args[1]);
+
         code.movdqa(temp, x);
         code.psubusw(temp, y);
         code.psubusw(y, x);
         code.por(temp, y);
         break;
+    }
     case 32:
         if (code.DoesCpuSupport(Xbyak::util::Cpu::tSSE41)) {
+            const Xbyak::Xmm x = ctx.reg_alloc.UseScratchXmm(args[0]);
+            const Xbyak::Xmm y = ctx.reg_alloc.UseXmm(args[1]);
+
             code.movdqa(temp, x);
             code.pminud(x, y);
             code.pmaxud(temp, y);
             code.psubd(temp, x);
         } else {
+            const Xbyak::Xmm x = ctx.reg_alloc.UseScratchXmm(args[0]);
+            const Xbyak::Xmm y = ctx.reg_alloc.UseScratchXmm(args[1]);
+
             code.movdqa(temp, code.MConst(xword, 0x8000000080000000, 0x8000000080000000));
             code.pxor(x, temp);
             code.pxor(y, temp);