Age | Commit message (Collapse) | Author |
|
Conflicts:
gnuradio-core/src/lib/filter/gr_fft_filter_ccc.cc
gnuradio-core/src/lib/filter/gr_fft_filter_fff.cc
|
|
|
|
Using the Volk function causes too much of a change in the output values right now. Will have to relook at it for the right thing to do. Keeping the use of vlen and scale, though.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Also adds SSE and AVX and unaligned Volk versions for this.
|
|
|
|
|
|
|
|
|
|
Also updates copyright dates.
|
|
swigdocs).
|
|
|
|
Can set nthreads as last arg to filter; defaults to 1 so no change in default behavior.
|
|
Can set and get nthreads; defaults to 1, so no change in default behavior.
|
|
|
|
Can set nthreads when creating an fftw class; defaults to 1, so there's no change in default behavior.
|
|
|
|
strip-chart features.
|
|
|
|
general so they can make use of volk calls.
QA code now explicitly tests the cc versions of these blocks.
|
|
Works with max_noutput_items and set_output_multiple.
|
|
No need for sse implementation now but keeping code for reference.
|
|
benchmark were used to decide which implementation to use.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
of short_to_float.
|
|
|
|
constructor to set the vector length and a free scaling with the data.
|
|
differences between Volk and non-Volk for rounding issues are made here.
|
|
|
|
|
|
The msvc compiler doesnt like the ambiguous overload of floor.
However, floor should not have an effect on the integer divide anyway.
|
|
documented.
|
|
|
|
|
|
|
|
|
|
indicator on whether or not the buffers are aligned; they can use this to determine which Volk function to use or if to use Volk at all.
|
|
|