1 files changed, 64 insertions, 0 deletions
diff --git a/volk/lib/qa_32fc_deinterleave_32f_aligned16.cc b/volk/lib/qa_32fc_deinterleave_32f_aligned16.cc
new file mode 100644
index 000000000..72e084c05
--- /dev/null
+++ b/volk/lib/qa_32fc_deinterleave_32f_aligned16.cc
@@ -0,0 +1,64 @@
+#include <volk/volk.h>
+#include <qa_32fc_deinterleave_32f_aligned16.h>
+#include <volk/volk_32fc_deinterleave_32f_aligned16.h>
+#include <cstdlib>
+#include <ctime>
+
+//test for sse
+
+#ifndef LV_HAVE_SSE
+
+void qa_32fc_deinterleave_32f_aligned16::t1() {
+  printf("sse not available... no test performed\n");
+}
+
+#else
+
+void qa_32fc_deinterleave_32f_aligned16::t1() {
+  
+  volk_environment_init();
+  clock_t start, end;
+  double total;
+  const int vlen = 3201;
+  const int ITERS = 100000;
+  std::complex<float> input0[vlen] __attribute__ ((aligned (16)));
+  
+  float output_generic[vlen] __attribute__ ((aligned (16)));
+  float output_generic1[vlen] __attribute__ ((aligned (16)));
+  float output_sse[vlen] __attribute__ ((aligned (16)));
+  float output_sse1[vlen] __attribute__ ((aligned (16)));
+
+  float* inputLoad = (float*)input0;
+  for(int i = 0; i < 2*vlen; ++i) {   
+    inputLoad[i] = (((float) (rand() - (RAND_MAX/2))) / static_cast<float>((RAND_MAX/2)));
+  }
+  printf("32fc_deinterleave_32f_aligned\n");
+
+  start = clock();
+  for(int count = 0; count < ITERS; ++count) {
+    volk_32fc_deinterleave_32f_aligned16_manual(output_generic, output_generic1, input0, 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_32fc_deinterleave_32f_aligned16_manual(output_sse, output_sse1, input0, 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_DOUBLES_EQUAL(output_generic[i], output_sse[i], fabs(output_generic[i])*1e-4);
+    CPPUNIT_ASSERT_DOUBLES_EQUAL(output_generic1[i], output_sse1[i], fabs(output_generic1[i])*1e-4);
+  }
+}
+
+#endif