From c692ccdd6de6bc7473100172935cb1d9b9ac5a14 Mon Sep 17 00:00:00 2001
From: Lioncash <mathew1800@gmail.com>
Date: Tue, 1 May 2018 08:19:21 -0400
Subject: [PATCH] emit_x64_vector: Vectorize fallback path of EmitVectorMaxS8()

---
 src/backend_x64/emit_x64_vector.cpp | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/src/backend_x64/emit_x64_vector.cpp b/src/backend_x64/emit_x64_vector.cpp
index 9906ea57..0e9a546a 100644
--- a/src/backend_x64/emit_x64_vector.cpp
+++ b/src/backend_x64/emit_x64_vector.cpp
@@ -1001,9 +1001,19 @@ void EmitX64::EmitVectorMaxS8(EmitContext& ctx, IR::Inst* inst) {
         return;
     }
 
-    EmitTwoArgumentFallback(code, ctx, inst, [](std::array<s8, 16>& result, const std::array<s8, 16>& a, const std::array<s8, 16>& b){
-        std::transform(a.begin(), a.end(), b.begin(), result.begin(), [](auto x, auto y) { return std::max(x, y); });
-    });
+    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
+    const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
+    const Xbyak::Xmm b = ctx.reg_alloc.UseScratchXmm(args[1]);
+
+    const Xbyak::Xmm tmp_b = ctx.reg_alloc.ScratchXmm();
+    code.movdqa(tmp_b, b);
+
+    code.pcmpgtb(tmp_b, a);
+    code.pand(b, tmp_b);
+    code.pandn(tmp_b, a);
+    code.por(tmp_b, b);
+
+    ctx.reg_alloc.DefineValue(inst, tmp_b);
 }
 
 void EmitX64::EmitVectorMaxS16(EmitContext& ctx, IR::Inst* inst) {