diff --git a/Makefile b/Makefile
index 8be0a33..badde0b 100644
--- a/Makefile
+++ b/Makefile
@@ -1,6 +1,6 @@
-CXXFLAGS = -Wall -O3
+CXXFLAGS = -Wall -O3 -std=c++11 -msse4.2
 
-all: split1 split2 split6 split7 split8 split9 splitc1 splitc2 splitc3 split_subparser
+all: split1 split2 split6 split7 split8 split9 splitc1 splitc2 splitc3 split_subparser splithb1 splithb2
 
 split7: split7.cpp | deps/strtk
 	$(CXX) $(LDFLAGS) -Ideps/strtk/ $(CXXFLAGS) split7.cpp -o split7
diff --git a/run_all.bash b/run_all.bash
index 2e3818c..9d87823 100755
--- a/run_all.bash
+++ b/run_all.bash
@@ -14,7 +14,8 @@ g++ --version | head -n1
 python --version
 echo "=== End System info"
 echo
-
+# warming up file
+wc test_data
 for i in `find . -maxdepth 1 -executable -iname 'split*.py' | sort ; find . -maxdepth 1 -executable \! -iname '*.py' -iname 'split*' | sort`
 do
 	printf "%-18s " $i
diff --git a/splithb1.cpp b/splithb1.cpp
new file mode 100644
index 0000000..094870a
--- /dev/null
+++ b/splithb1.cpp
@@ -0,0 +1,54 @@
+// Contributed by Henrique Bucher (http://www.vitorian.com/)
+// Requires SSE4.2
+#include <iostream>
+#include <iomanip>
+#include <string>
+#include <array>
+#include <unistd.h>
+
+int main()
+{
+   long count = 0;
+   timespec start;
+   clock_gettime(CLOCK_MONOTONIC, &start);
+
+   std::cin.sync_with_stdio(false); //disable synchronous IO
+
+   std::size_t numWords = 0;
+   std::size_t numChars = 0;
+
+   unsigned bufsize = 1024*1024;
+   char* buf = (char*)aligned_alloc( 16, bufsize );
+
+   while ( true ) {
+       int nb = ::fread( buf, 1, bufsize, stdin );
+       if ( nb<=0 ) break;
+       for ( int j=0; j<nb; ++j ) {
+           if ( buf[j] == '\n' ) {
+               numWords++;
+               count++;
+           }
+           else if ( buf[j] == ' ' ) {
+               numWords++;
+           }
+           else {
+               numChars++;
+           }
+       }
+   }
+
+   timespec end;
+   clock_gettime(CLOCK_MONOTONIC, &end);
+   const double sec = (end.tv_sec - start.tv_sec) + (end.tv_nsec - start.tv_nsec) * 1e-9;
+   std::cerr << "C++   : Saw " << count << " lines (" << numWords << " words/" << numChars << " chars) in " << std::fixed << std::setprecision(1) << sec << " seconds." ;
+
+   if (sec > 0)
+   {
+      const double lps = count / sec;
+      std::cerr << "  Crunch speed: " << std::fixed << std::setprecision(1) << lps << std::endl;
+   }
+   else
+      std::cerr << std::endl;
+
+   return 0;
+}
diff --git a/splithb2.cpp b/splithb2.cpp
new file mode 100644
index 0000000..490e043
--- /dev/null
+++ b/splithb2.cpp
@@ -0,0 +1,67 @@
+// Contributed by Henrique Bucher (http://www.vitorian.com/)
+// Requires SSE4.2
+#include <iostream>
+#include <iomanip>
+#include <string>
+#include <array>
+#include <unistd.h>
+#include <smmintrin.h>
+
+int main()
+{
+   long count = 0;
+   timespec start;
+   clock_gettime(CLOCK_MONOTONIC, &start);
+
+   std::size_t numWords = 0;
+   std::size_t numChars = 0;
+
+   const __m128i LF  = _mm_set1_epi8( '\n' );
+   const __m128i SPC = _mm_set1_epi8( ' ' );
+
+   unsigned bufsize = 1024*1024;
+   char* buf = (char*)aligned_alloc( 16, bufsize );
+
+   while ( true ) {
+       int nb = ::fread( buf, 1, bufsize, stdin );
+       if ( nb<=0 ) break;
+
+       int pos = 0;
+       while ( pos<nb ) {
+           union { __m128i res; uint16_t resu[8]; };
+           __m128i u = _mm_loadu_si128 ((__m128i *) &buf[pos] );
+           unsigned sz = nb-pos;
+           if ( sz>16 ) sz = 16;
+           res = _mm_cmpestrm( SPC, 1, u, sz,
+                               (_SIDD_UBYTE_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK) );
+           uint16_t spcbits = resu[0];
+           res = _mm_cmpestrm( LF, 1, u, sz,
+                               (_SIDD_UBYTE_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK) );
+           uint16_t lfbits = resu[0];
+
+           unsigned numspc = __builtin_popcount( spcbits );
+           unsigned numlf  = __builtin_popcount( lfbits );
+
+           numWords += numspc + numlf;
+           count += numlf;
+           numChars += sz - (numspc+numlf);
+
+           pos += sz;
+       }
+   }
+
+   timespec end;
+   clock_gettime(CLOCK_MONOTONIC, &end);
+   const double sec = (end.tv_sec - start.tv_sec) + (end.tv_nsec - start.tv_nsec) * 1e-9;
+   std::cerr << "C++   : Saw " << count << " lines (" << numWords << " words/" << numChars << " chars) in " << std::fixed << std::setprecision(1) << sec << " seconds." ;
+
+   if (sec > 0)
+   {
+      const double lps = count / sec;
+      std::cerr << "  Crunch speed: " << std::fixed << std::setprecision(1) << lps << std::endl;
+   }
+   else
+      std::cerr << std::endl;
+
+   return 0;
+}