root / gnuradio-core / src / lib / general / README @ 5d69a524
History | View | Annotate | Download (4 kB)
| 1 | Files beginning with Gr* define classes that inherit from VrSigProc. |
|---|---|
| 2 | These are high level signal processing modules that can be glued |
| 3 | together in your signal processing chain. |
| 4 | |
| 5 | All the others are either lower level routines which implement the |
| 6 | functionality of the Gr* modules, but are easier to test (fewer |
| 7 | dependencies), or they are general purpose. |
| 8 | |
| 9 | gr_fir_???.{h,cc}, where ??? are in F, S or C are low level Finite
|
| 10 | Impulse Response Filters. These turn out to be where the bulk of the |
| 11 | cycles are burned in many applications. The ??? suffix specifies the |
| 12 | input type, output type and tap type of the arguments. We've |
| 13 | implemented the most frequently used ones. |
| 14 | |
| 15 | [Once upon a time this stuff was done with templates |
| 16 | (gr_fir<iType,oType,tapType>), but this turned out to be a headache. |
| 17 | The code appeared to trigger a bug in GCC where we were getting |
| 18 | multiple definitions of unrelated stuff when we started subclassing |
| 19 | partially specialized templates. It was also not obvious as to to |
| 20 | what combinations of iType, oType and tapType actually worked. We're |
| 21 | now explicit, and the world is a safer place to live...] |
| 22 | |
| 23 | The top level routines for FIR filtering are: |
| 24 | |
| 25 | GrFIRfilterFFF : Float input, Float output, Float taps |
| 26 | -- general purpose |
| 27 | |
| 28 | GrFIRfilterCCF : Complex input, Complex output, Float taps |
| 29 | -- applying real filter to a complex signal |
| 30 | |
| 31 | GrFIRfilterFCC : Float input, Complex output, Complex taps |
| 32 | -- applying complex filter to float input |
| 33 | |
| 34 | GrFIRfilterSCC : Short input, Complex output, Complex taps |
| 35 | -- applying complex filter to short input. Quantizes complex |
| 36 | coefficients to 16 bits and uses MMX or SSE2 as appropriate |
| 37 | |
| 38 | |
| 39 | The combination of down conversion (frequency translation) and channel |
| 40 | selection (filtering) is performed with: |
| 41 | |
| 42 | GrFreqXlatingFIRfilterSFC : Short input, Float taps, Complex baseband output |
| 43 | -- quantizes complex coefficents to 16 bits and uses MMX or |
| 44 | SSE2 (128-bit MMX) as appropriate [optimization to be done]. |
| 45 | |
| 46 | GrFreqXlatingFIRfilterFFC : Float input, Float taps, Complex baseband output |
| 47 | -- 3dnow or SSE as appropriate. |
| 48 | |
| 49 | |
| 50 | [ The stuff described from here down is used to implement the routines |
| 51 | above. This info is only relevant to those who are hacking the internals ] |
| 52 | |
| 53 | |
| 54 | A bit of indirection is involved in selecting the fastest |
| 55 | implementation for any given platform. The lower level classes |
| 56 | gr_fir_FFF.h, gr_fir_CCF, gr_fir_FCC and gr_fir_SCC have i/o |
| 57 | signatures similar to the high level clases above. These |
| 58 | should be considered the abstract base classes that you |
| 59 | work with. Note that they are not actually abstract (they've got a |
| 60 | default implementation; this might be considered a bug), but they |
| 61 | should not be directly instantiated by user code. |
| 62 | |
| 63 | Instead of directly instantiating a gr_fir_FFF, for example, your code |
| 64 | should actually: |
| 65 | |
| 66 | #include <gr_fir_util.h> |
| 67 | |
| 68 | // let the system pick the best implementation for you |
| 69 | gr_fir_FFF *filter = gr_fir_util::create_gr_fir_FFF (my_taps); |
| 70 | |
| 71 | Clear? The same for all the other gr_fir_XXX's. |
| 72 | |
| 73 | |
| 74 | |
| 75 | Performance hacking can be done by subclassing the appropriate |
| 76 | base class. For example, on the x86 platform, there are two |
| 77 | additional classes derived from gr_fir_FFF, gr_fir_FFF_sse and |
| 78 | gr_fir_FFF_3dnow. These classes are then made available to the rest |
| 79 | of the system by virtue of being added to gr_fir_sysconfig_x86.cc, |
| 80 | along with any guards (CPUID checks) needed to ensure that only |
| 81 | compatible code is executed on the current hardware. |
| 82 | |
| 83 | |
| 84 | TO DO |
| 85 | ------ |
| 86 | |
| 87 | * Move all the machine specific code to a subdirectory, then have |
| 88 | configure symlink to the right directory. This will allow us to build on |
| 89 | any platform without choking. There is generic code for all routines, |
| 90 | only the machine dependent speedup will be lacking. |
| 91 | |
| 92 | * Add an interface to gr_fir_util that will return a vector of all |
| 93 | valid constructors with descriptive names for each i/o signature. |
| 94 | This will allow the test code and benchmarking code to be blissfully |
| 95 | ignorant of what platform they're running on. The actual building of |
| 96 | the vectors should be done bottom up through the gr_fir_sysconfig |
| 97 | hierarchy. |
| 98 |