diff options
| author | Tom Rondeau <trondeau@vt.edu> | 2010-12-07 18:50:28 -0500 |
|---|---|---|
| committer | Tom Rondeau <trondeau@vt.edu> | 2010-12-07 18:50:28 -0500 |
| commit | 239144659b29c0a5ecd83a34e0e57387a1060ed7 (patch) | |
| tree | 3476e1c123da4696c64cc1756ddec5d971bcf9f2 /volk/lib/qa_32f_normalize_aligned16.cc | |
| parent | e13783aeb84a2c3656c3344a8d52fa2c9ee38a00 (diff) | |
Initial checkin for VOLK - Vector-Optimized Library of Kernels. This is a new SIMD library.
It currently stands by itself under the GNU Radio tree and can be used separately. We will integrate the build process into GNU Raio and start building off of its functionality over time.
Diffstat (limited to 'volk/lib/qa_32f_normalize_aligned16.cc')
| -rw-r--r-- | volk/lib/qa_32f_normalize_aligned16.cc | 65 |
1 files changed, 65 insertions, 0 deletions
diff --git a/volk/lib/qa_32f_normalize_aligned16.cc b/volk/lib/qa_32f_normalize_aligned16.cc new file mode 100644 index 000000000..2954fc3ae --- /dev/null +++ b/volk/lib/qa_32f_normalize_aligned16.cc @@ -0,0 +1,65 @@ +#include <volk/volk.h> +#include <qa_32f_normalize_aligned16.h> +#include <volk/volk_32f_normalize_aligned16.h> +#include <cstdlib> +#include <cstring> + +//test for sse + +#ifndef LV_HAVE_SSE + +void qa_32f_normalize_aligned16::t1() { + printf("sse not available... no test performed\n"); +} + +#else + +void qa_32f_normalize_aligned16::t1() { + + volk_environment_init(); + int ret; + clock_t start, end; + double total; + const int vlen = 320001; + const int ITERS = 100; + + float* output0; + float* output01; + ret = posix_memalign((void**)&output0, 16, vlen*sizeof(float)); + ret = posix_memalign((void**)&output01, 16, vlen*sizeof(float)); + + for(int i = 0; i < vlen; ++i) { + output0[i] = ((float) (rand() - (RAND_MAX/2))) / static_cast<float>((RAND_MAX/2)); + } + memcpy(output01, output0, vlen*sizeof(float)); + printf("32f_normalize_aligned\n"); + + start = clock(); + for(int count = 0; count < ITERS; ++count) { + volk_32f_normalize_aligned16_manual(output0, 1.15, 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_normalize_aligned16_manual(output01, 1.15, 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("%e...%e\n", output0[i], output01[i]); + CPPUNIT_ASSERT_DOUBLES_EQUAL(output0[i], output01[i], fabs(output0[i])*1e-4); + } + + free(output0); + free(output01); +} + +#endif |