Merge branch 'next' into guile.

Passes distcheck.

* next: (32 commits)
  volk: Fix for popcnt's 64/32-bit issues.
  Using a copy of config.guess and config.sub instead of sym links.
  Including time header to qa files.
  Changed python env variable to more globally usable version.
  gr_uhd: Quick fix for make distcheck failures if UHD is not installed.
  PFB resampler: fix it this way to avoid the signed/unsigned warning.
  PFB resampler: fixes bug where filter could be looking past the number of inputs.
  resampler PFB: Sets relative rate when rate is changed.
  Create method to set rate on pfb_arb_resamp after it has been created. Allow it to be called from GRC.
  volk: fix for running 32-bit OS on 64-bit processor. System is correctly identified as 32-bit and compiles with the correct flags.
  volk: changing the path variables again. This works on my various systems tested. Using abs_ path names failed on Ubuntu 8.04 32-bit.
  volk: May be a hack, but it was required for my 32-bit Fedora 13 to work.
  volk: Removing unnecessary shell script; last commit takes care if its functions.
  volk: Fixing build system to handle making volk_mktables, volk_tables.h, and volk_config.h instead of a standalone shell script.
  volk: readding 16sc_magnitude_32f_sse with fix for SSE hadd_ps error.
  volk: Adding a few more generic-only test cases.
  volk: adding generic QA test for 16sc_magnitude_32f.
  volk: modified the configure scripts to output which architectures it will be building based on the configure tests.
  uhd: update notes in grc blocks for addressing scheme
  volk: Removing erroneous SSE function that actually usese an SSE3 intrin (mm_hadd_ps).
  ...

1 files changed, 214 insertions, 0 deletions

diff --git a/volk/lib/qa_32fc_dot_prod_aligned16.cc b/volk/lib/qa_32fc_dot_prod_aligned16.cc
new file mode 100644
index 000000000..bcf9ea954
--- /dev/null
+++ b/volk/lib/qa_32fc_dot_prod_aligned16.cc
@@ -0,0 +1,214 @@
+#include <volk/volk.h>
+#include <qa_32fc_dot_prod_aligned16.h>
+#include <stdlib.h>
+#include <math.h>
+#include <time.h>
+#include <stdio.h>
+
+
+
+#define assertcomplexEqual(expected, actual, delta)			\
+  CPPUNIT_ASSERT_DOUBLES_EQUAL (std::real(expected), std::real(actual), fabs(std::real(expected)) * delta); \
+  CPPUNIT_ASSERT_DOUBLES_EQUAL (std::imag(expected), std::imag(actual), fabs(std::imag(expected))* delta);	
+
+#define	ERR_DELTA	(1e-4)
+
+//test for sse
+static float uniform() {
+  return 2.0 * ((float) rand() / RAND_MAX - 0.5);	// uniformly (-1, 1)
+}
+
+static void
+random_floats (float *buf, unsigned n)
+{
+  for (unsigned i = 0; i < n; i++)
+    buf[i] = uniform ();
+}
+
+
+
+#if LV_HAVE_SSE3
+void qa_32fc_dot_prod_aligned16::t1() {
+
+  const int vlen = 2046;
+
+  volk_environment_init();
+  int ret;
+  clock_t start, end;
+  double total;
+  std::complex<float>* input;
+  std::complex<float>* taps;
+  
+  std::complex<float>* result_generic;
+  std::complex<float>* result_sse3;
+
+  ret = posix_memalign((void**)&input, 16, vlen << 3);
+  ret = posix_memalign((void**)&taps, 16, vlen << 3);
+  ret = posix_memalign((void**)&result_generic, 16, 8);
+  ret = posix_memalign((void**)&result_sse3, 16, 8);
+  
+
+  result_generic[0] = std::complex<float>(0,0);
+  result_sse3[0] = std::complex<float>(0,0);
+
+  random_floats((float*)input, vlen * 2);
+  random_floats((float*)taps, vlen * 2);
+  
+  printf("32fc_dot_prod_aligned16\n");
+  
+  start = clock();
+  volk_32fc_dot_prod_aligned16_manual(result_generic, input, taps, vlen * 8,  "generic");
+  end = clock();
+  total = (double)(end-start)/(double)CLOCKS_PER_SEC;
+  printf("generic_time: %f\n", total);
+
+  
+  start = clock();
+  volk_32fc_dot_prod_aligned16_manual(result_sse3, input, taps, vlen * 8, "sse3");
+  end = clock();
+  total = (double)(end-start)/(double)CLOCKS_PER_SEC;
+  printf("sse3_time: %f\n", total);
+
+  printf("generic: %f +i%f ... sse3: %f +i%f\n", std::real(result_generic[0]), std::imag(result_generic[0]), std::real(result_sse3[0]), std::imag(result_sse3[0]));
+
+  
+  assertcomplexEqual(result_generic[0], result_sse3[0], ERR_DELTA);
+
+  free(input);
+  free(taps);
+  free(result_generic);
+  free(result_sse3);
+  
+}
+
+#else
+void qa_32fc_dot_prod_aligned16::t1() {
+  printf("sse3 not available... no test performed\n");
+}
+
+#endif
+
+#if LV_HAVE_SSE && LV_HAVE_32
+void qa_32fc_dot_prod_aligned16::t2() {
+
+  const int vlen = 2046;
+
+  volk_environment_init();
+  int ret;
+  clock_t start, end;
+  double total;
+  std::complex<float>* input;
+  std::complex<float>* taps;
+  
+  std::complex<float>* result_generic;
+  std::complex<float>* result_sse3;
+
+  ret = posix_memalign((void**)&input, 16, vlen << 3);
+  ret = posix_memalign((void**)&taps, 16, vlen << 3);
+  ret = posix_memalign((void**)&result_generic, 16, 8);
+  ret = posix_memalign((void**)&result_sse3, 16, 8);
+  
+
+  result_generic[0] = std::complex<float>(0,0);
+  result_sse3[0] = std::complex<float>(0,0);
+
+  random_floats((float*)input, vlen * 2);
+  random_floats((float*)taps, vlen * 2);
+  
+  printf("32fc_dot_prod_aligned16\n");
+  
+  start = clock();
+  volk_32fc_dot_prod_aligned16_manual(result_generic, input, taps, vlen * 8,  "generic");
+  end = clock();
+  total = (double)(end-start)/(double)CLOCKS_PER_SEC;
+  printf("generic_time: %f\n", total);
+
+  
+  start = clock();
+  volk_32fc_dot_prod_aligned16_manual(result_sse3, input, taps, vlen * 8, "sse_32");
+  end = clock();
+  total = (double)(end-start)/(double)CLOCKS_PER_SEC;
+  printf("sse_32_time: %f\n", total);
+
+  printf("generic: %f +i%f ... sse_32: %f +i%f\n", std::real(result_generic[0]), std::imag(result_generic[0]), std::real(result_sse3[0]), std::imag(result_sse3[0]));
+
+  
+  assertcomplexEqual(result_generic[0], result_sse3[0], ERR_DELTA);
+
+  free(input);
+  free(taps);
+  free(result_generic);
+  free(result_sse3);
+  
+}
+
+#else
+void qa_32fc_dot_prod_aligned16::t2() {
+  printf("sse_32 not available... no test performed\n");
+}
+
+#endif
+
+#if LV_HAVE_SSE && LV_HAVE_64
+
+void qa_32fc_dot_prod_aligned16::t3() {
+
+  const int vlen = 2046;
+
+  volk_environment_init();
+  int ret;
+  clock_t start, end;
+  double total;
+  std::complex<float>* input;
+  std::complex<float>* taps;
+  
+  std::complex<float>* result_generic;
+  std::complex<float>* result_sse3;
+
+  ret = posix_memalign((void**)&input, 16, vlen << 3);
+  ret = posix_memalign((void**)&taps, 16, vlen << 3);
+  ret = posix_memalign((void**)&result_generic, 16, 8);
+  ret = posix_memalign((void**)&result_sse3, 16, 8);
+  
+
+  result_generic[0] = std::complex<float>(0,0);
+  result_sse3[0] = std::complex<float>(0,0);
+
+  random_floats((float*)input, vlen * 2);
+  random_floats((float*)taps, vlen * 2);
+  
+  printf("32fc_dot_prod_aligned16\n");
+  
+  start = clock();
+  volk_32fc_dot_prod_aligned16_manual(result_generic, input, taps, vlen * 8,  "generic");
+  end = clock();
+  total = (double)(end-start)/(double)CLOCKS_PER_SEC;
+  printf("generic_time: %f\n", total);
+
+  
+  start = clock();
+  volk_32fc_dot_prod_aligned16_manual(result_sse3, input, taps, vlen * 8, "sse_64");
+  end = clock();
+  total = (double)(end-start)/(double)CLOCKS_PER_SEC;
+  printf("sse_64_time: %f\n", total);
+
+  printf("generic: %f +i%f ... sse_64: %f +i%f\n", std::real(result_generic[0]), std::imag(result_generic[0]), std::real(result_sse3[0]), std::imag(result_sse3[0]));
+
+  
+  assertcomplexEqual(result_generic[0], result_sse3[0], ERR_DELTA);
+
+  free(input);
+  free(taps);
+  free(result_generic);
+  free(result_sse3);
+  
+}
+
+#else
+void qa_32fc_dot_prod_aligned16::t3() {
+  printf("sse_64 not available... no test performed\n");
+}
+
+
+
+#endif