1 files changed, 55 insertions, 0 deletions
diff --git a/main.c b/main.c
new file mode 100644
index 0000000..99a0eca
--- /dev/null
+++ b/main.c
@@ -0,0 +1,55 @@
+#ifdef USE_VECTOR
+#include <immintrin.h>
+#endif
+#include <stdio.h>
+#include <time.h>
+#include <stdlib.h>
+
+#define STRIDE (256/32) // How many integers fit in a 256 vector register
+#define VSIZE (STRIDE*200000)
+
+int main(int argc, char *argv[]) {
+
+	float start, end, duration;
+
+	int *v1;
+	posix_memalign((void**)&v1, 32, VSIZE * sizeof(int));
+	int *v2;
+	posix_memalign((void**)&v2, 32, VSIZE * sizeof(int));
+	int result[VSIZE];
+
+#ifndef USE_VECTOR
+	printf("Sequential..");
+	start = (float)clock()/CLOCKS_PER_SEC;
+	for(int i=0;i<VSIZE;i++){
+		result[i]=v1[i]+v2[i];
+	}
+	end = (float)clock()/CLOCKS_PER_SEC;
+	duration= end - start;
+	printf(" done! (duration=%.6fs)\n", duration);
+#else
+	printf("Vectorize...");
+	__m256i a,b;
+	// e == extended (for historical reasons just brandy name)
+	// p == packed (multiple packed elements)
+	// i == integer
+	// 32 == each integer is 32bits
+	__m256i c = _mm256_add_epi32(a, b);
+	start = (float)clock()/CLOCKS_PER_SEC;
+	for(int i=0;i<VSIZE;i+=STRIDE){
+		a = _mm256_load_si256((__m256i*)&v1[i]);
+		b = _mm256_load_si256((__m256i*)&v2[i]);
+		c = _mm256_add_epi32(a,b);
+
+		_mm256_store_si256((__m256i*)&result[i],c);
+	}
+	end = (float)clock()/CLOCKS_PER_SEC;
+	duration= end - start;
+	printf(" done! (duration=%.6fs)\n", duration);
+#endif
+
+	free(v1);
+	free(v2);
+
+	return 0;
+}