1 files changed, 60 insertions, 0 deletions
diff --git a/volk/lib/qa_32f_divide_aligned16.cc b/volk/lib/qa_32f_divide_aligned16.cc
new file mode 100644
index 000000000..b20999beb
--- /dev/null
+++ b/volk/lib/qa_32f_divide_aligned16.cc
@@ -0,0 +1,60 @@
+#include <volk/volk.h>
+#include <qa_32f_divide_aligned16.h>
+#include <volk/volk_32f_divide_aligned16.h>
+#include <cstdlib>
+
+//test for sse
+
+#ifndef LV_HAVE_SSE
+
+void qa_32f_divide_aligned16::t1() {
+  printf("sse not available... no test performed\n");
+}
+
+#else
+
+void qa_32f_divide_aligned16::t1() {
+  
+  volk_environment_init();
+  clock_t start, end;
+  double total;
+  const int vlen = 3201;
+  const int ITERS = 100000;
+  float input0[vlen] __attribute__ ((aligned (16)));
+  float input1[vlen] __attribute__ ((aligned (16)));
+  
+  float output0[vlen] __attribute__ ((aligned (16)));
+  float output01[vlen] __attribute__ ((aligned (16)));
+
+  for(int i = 0; i < vlen; ++i) {   
+    input0[i] = ((float) (rand() - (RAND_MAX/2))) / static_cast<float>((RAND_MAX/2));
+    input1[i] = ((float) (rand() - (RAND_MAX/2))) / static_cast<float>((RAND_MAX/2));
+  }
+  printf("32f_divide_aligned\n");
+
+  start = clock();
+  for(int count = 0; count < ITERS; ++count) {
+    volk_32f_divide_aligned16_manual(output0, input0, input1, vlen, "generic");
+  }
+  end = clock();
+  total = (double)(end-start)/(double)CLOCKS_PER_SEC;
+  printf("generic_time: %f\n", total);
+  start = clock();
+  for(int count = 0; count < ITERS; ++count) {
+    volk_32f_divide_aligned16_manual(output01, input0, input1, vlen, "sse");
+  }
+  end = clock();
+  total = (double)(end-start)/(double)CLOCKS_PER_SEC;
+  printf("sse_time: %f\n", total);
+  for(int i = 0; i < 1; ++i) {
+    //printf("inputs: %d, %d\n", input0[i*2], input0[i*2 + 1]);
+    //printf("generic... %d, ssse3... %d\n", output0[i], output1[i]);
+  }
+  
+  for(int i = 0; i < vlen; ++i) {
+    //printf("%d...%d\n", output0[i], output01[i]);
+    CPPUNIT_ASSERT_EQUAL(output0[i], output01[i]);
+  }
+}
+
+#endif