Added assembly sse instruction for long and int adding

src/core/cpu/naive.hpp

@@ -12,10 +12,15 @@ namespace core {
                 static const unsigned int MASK_32=0xFFFFFFFF;
             public:
                 //Attributes
-                static const unsigned char LEN_64   =   2;
+                static const unsigned char INT_LEN_64   =   2;
-                static const unsigned char LEN_128  =   4;
+                static const unsigned char INT_LEN_128  =   4;
-                static const unsigned char LEN_256  =   8;
+                static const unsigned char INT_LEN_256  =   8;
-                static const unsigned char LEN_512  =   8;
+                static const unsigned char INT_LEN_512  =   16;
+
+                static const unsigned char LONG_LEN_64  =   1;
+                static const unsigned char LONG_LEN_128  =   2;
+                static const unsigned char LONG_LEN_256  =   4;
+                static const unsigned char LONG_LEN_512  =   8;
 
                 //Methods
                 static void sum_128_long(int* a, int* b);

src/core/cpu/sse.cpp

@@ -1,7 +1,7 @@
 #include "sse.hpp"
 #include <iostream>
 
-//X86 Assembly to add two 128 bit numbers
+//X86 Assembly to add two 128 bit numbers in the form of packed integers 32bit
 void core::cpu::SSE::sum_128(int *a,int *b) {
     asm(
         "movdqa %0, %%xmm1\n"
@@ -10,4 +10,17 @@ void core::cpu::SSE::sum_128(int *a,int *b) {
         : "=m"(*a)
         : "m"(*b)
     );
-}
+};
+
+
+//X86 Assembly to add two 128 bit numbers in the form of packed long 64bit
+void core::cpu::SSE::sum_128(long *a,long *b) {
+    asm(
+        "movdqa %0, %%xmm1\n"
+        "paddd %1, %%xmm1\n"
+        "movdqa %%xmm1, %0"
+        : "=m"(*a)
+        : "m"(*b)
+    );
+}
+

src/core/cpu/sse.hpp

@@ -10,6 +10,7 @@ namespace core {
         class SSE {
             public:
                 static void sum_128(int *a,int *b);
+                static void sum_128(long *a,long *b);
         };
     };
 };

sse.cpp

@@ -1,24 +1,58 @@
 #include <iostream>
 #include <string>
 #include <iomanip>
+#include <sys/time.h>
 
 #include "src/core/cpu/naive.hpp"
 #include "src/core/cpu/sse.hpp"
 
 using namespace core::cpu;
 
+long int gettime(){
+    struct timeval tp;
+    gettimeofday(&tp, NULL);
+    long int ms = tp.tv_sec * 1000 + tp.tv_usec / 1000;
+    return ms;
+};
+
+#define MAX_ITER 1000*1000*100
+
 int main(int argc, char** argcv){
-    int v1_128[Naive::LEN_128] = { 0x1, 0x2,0x3,0x4 };
+    int v1_128[Naive::INT_LEN_128] = { 0x1, 0x2,0x3,0x4 };
-    int v2_128[Naive::LEN_128] = { 0x1, 0x2,0x3,0x4 };
+    int v2_128[Naive::INT_LEN_128] = { 0x1, 0x2,0x3,0x4 };
 
-    Naive::print(v1_128,Naive::LEN_128);
+    long v1_128_l[Naive::LONG_LEN_128] = { (long)0x2<<32 | 0x2, (long)0x4<<32 | 0x3 };
-    Naive::print(v2_128,Naive::LEN_128);
+    long v2_128_l[Naive::LONG_LEN_128] = { (long)0x2<<32 | 0x2, (long)0x4<<32 | 0x3 };
 
-    Naive::sum_128_long(v1_128,v2_128);
+    long int start,end;
-    SSE::sum_128(v1_128,v2_128);
 
-    Naive::print(v1_128,Naive::LEN_128);
+    Naive::print(v1_128,Naive::INT_LEN_128);
-    Naive::print(v2_128,Naive::LEN_128);
+    Naive::print(v2_128,Naive::INT_LEN_128);
+
+    start = gettime();
+    for(int i=0;i<MAX_ITER;i++){
+        Naive::sum_128_long(v1_128,v2_128);
+    }
+    end = gettime();
+    cout << "Naive Approach: " << end-start << endl;
+
+    start = gettime();
+    for(int i=0;i<MAX_ITER;i++){
+        SSE::sum_128(v1_128,v2_128);
+    }
+    end = gettime();
+    cout << "SSE Approach paddw: " << end-start << endl;
+
+    start = gettime();
+    for(int i=0;i<MAX_ITER;i++){
+        SSE::sum_128(v1_128_l,v2_128_l);
+    }
+    end = gettime();
+    cout << "SSE Approach paddd: " << end-start << endl;
+
+
+    //Naive::print(v1_128,Naive::LEN_128);
+    //Naive::print(v2_128,Naive::LEN_128);
     
 
 }