Use AVX2 by default.

project/vc2017/nvtt/nvtt.vcxproj

@@ -198,6 +198,7 @@
       </PrecompiledHeader>
       <WarningLevel>Level3</WarningLevel>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
+      <EnableEnhancedInstructionSet>AdvancedVectorExtensions2</EnableEnhancedInstructionSet>
     </ClCompile>
     <Link>
       <AdditionalDependencies>cudart.lib;%(AdditionalDependencies)</AdditionalDependencies>
@@ -281,6 +282,7 @@
       </PrecompiledHeader>
       <WarningLevel>Level3</WarningLevel>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
+      <EnableEnhancedInstructionSet>AdvancedVectorExtensions2</EnableEnhancedInstructionSet>
     </ClCompile>
     <Link>
       <AdditionalDependencies>cudart.lib;%(AdditionalDependencies)</AdditionalDependencies>
@@ -343,6 +345,7 @@
       <WarningLevel>Level3</WarningLevel>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
+      <EnableEnhancedInstructionSet>AdvancedVectorExtensions2</EnableEnhancedInstructionSet>
     </ClCompile>
     <Link>
       <AdditionalLibraryDirectories>$(CUDA_LIB_PATH)\..\lib64;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
@@ -400,6 +403,7 @@
       <WarningLevel>Level3</WarningLevel>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
       <MultiProcessorCompilation>true</MultiProcessorCompilation>
+      <EnableEnhancedInstructionSet>AdvancedVectorExtensions2</EnableEnhancedInstructionSet>
     </ClCompile>
     <Link>
       <AdditionalLibraryDirectories>$(CUDA_LIB_PATH)\..\lib64;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>

src/nvtt/icbc.cpp

@@ -1,3 +1,3 @@
 #define ICBC_IMPLEMENTATION
-#define ICBC_SIMD 2 // SSE4.1
+//#define ICBC_SIMD 2 // SSE4.1
 #include "icbc.h"