Removed unnecessary SSE4.1 instructions

now purely SSE2 instructions used (in fact it's SSE only)

Author: mawen1250

configure

@@ -29,8 +29,7 @@ options:
 
   --target=TARGET          set target instruction set
                            - sse2
-                           - sse4.1
-                           [sse4.1]
+                           [sse2]
 
 EOF
 exit 1
@@ -96,7 +95,7 @@ CXXFLAGS="-std=c++11 -Wall -Wno-unused-local-typedefs -I. -I$SRCDIR/include -DND
 LDFLAGS=""
 SOFLAGS="-shared"
 DEPLIBS="fftw3f"
-TARGET="-msse4.1"
+TARGET="-msse2"
 
 # -- options ----------------------------------------------------------------
 echo all command lines: > config.log

msvc/Includes.props

@@ -6,7 +6,7 @@
   <ItemDefinitionGroup>
     <ClCompile>
       <AdditionalIncludeDirectories>..\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
-      <PreprocessorDefinitions>__SSE2__;__SSE4_1__;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <PreprocessorDefinitions>__SSE2__;%(PreprocessorDefinitions)</PreprocessorDefinitions>
     </ClCompile>
   </ItemDefinitionGroup>
   <ItemGroup />

source/BM3D_Basic.cpp

@@ -101,11 +101,7 @@ void BM3D_Basic_Process::CollaborativeFilter(int plane,
         const __m128 cmp2 = _mm_cmplt_ps(s1, t1n);
         const __m128 cmp = _mm_or_ps(cmp1, cmp2);
 
-#if defined(__SSE4_1__)
-        const __m128 d1 = _mm_blendv_ps(zero_ps, s1, cmp);
-#else
-        const __m128 d1 = _mm_or_ps(_mm_and_ps(cmp, s1), _mm_andnot_ps(cmp, zero_ps));
-#endif
+        const __m128 d1 = _mm_and_ps(cmp, s1);
         _mm_store_ps(srcp, d1);
         cmp_sum = _mm_sub_epi32(cmp_sum, _mm_castps_si128(cmp));
     }

source/VBM3D_Basic.cpp

@@ -101,11 +101,7 @@ void VBM3D_Basic_Process::CollaborativeFilter(int plane,
         const __m128 cmp2 = _mm_cmplt_ps(s1, t1n);
         const __m128 cmp = _mm_or_ps(cmp1, cmp2);
 
-#if defined(__SSE4_1__)
-        const __m128 d1 = _mm_blendv_ps(zero_ps, s1, cmp);
-#else
-        const __m128 d1 = _mm_or_ps(_mm_and_ps(cmp, s1), _mm_andnot_ps(cmp, zero_ps));
-#endif
+        const __m128 d1 = _mm_and_ps(cmp, s1);
         _mm_store_ps(srcp, d1);
         cmp_sum = _mm_sub_epi32(cmp_sum, _mm_castps_si128(cmp));
     }