Hi Matt
> I'm trying to compile an old design with a newly installed setup with
> matlab 2011b/ise 13.4. The fft_init failed, so I replaced the old fft
> blocks with new ones from the current casper library browser, but the
> fft_init still doesn't compile:
A good way to test if the init scripts have completed is to set
'casper_log_groups = {'trace'};' at the Matlab command prompt before
setting mask parameters/executing the init script. You will then see
debug output that lets you see how scripts are called and when they
finish. If you continue to have problems could you send me this output,
it will help to locate the problem.
You will definitely have to replace any fft blocks you have with new
ones from the library. This is because the mask parameters have changed
(to be specific, a parameter has been removed), and if you use the old
blocks with the new libraries, the new init scripts will be called with
the old set of parameters.
> I tried running the fft_init from the matlab interpreter with only a warning:
>
> >> fft_init(gcb, ...
> 'FFTSize', 9, ...
> 'n_inputs', 1, ...
> 'input_bit_width', 18, ...
> 'coeff_bit_width', 18, ...
> 'unscramble', 1, ...
> 'add_latency', 2, ...
> 'mult_latency', 2, ...
> 'bram_latency', 3, ...
> 'conv_latency', 2, ...
> 'quantization', 'Truncate', ...
> 'overflow', 'Wrap', ...
> 'arch', 'Virtex5', ...
> 'opt_target', 'logic', ...
> 'coeffs_bit_limit', 8, ...
> 'delays_bit_limit', 8, ...
> 'mult_spec', 2, ...
> 'hardcode_shifts', 'off', ...
> 'shift_schedule', 0, ...
> 'dsp48_adders', 'on');
> Warning: Invalid setting in masked block 'chan_550_packet/PFB/fft2'
> for parameter
> 'MaskValues'
> > In backpopulate_mask at 62
> In save_state at 48
> In fft_init at 315
The warning is probably for 'unscramble', 1, ...
The mask parameter is specified as a checkbox, which takes the value
'on' or 'off' (1 or 0 is allowed but is being phased out). The
backpopulate_mask script tries to write the values passed to the script
back into the mask to ensure that the mask parameter settings are
consistent with what has been drawn by the init script. However, the
setting passed ('1') in this case generates a warning.
> Running biplex_core_init does error out, though:
>
> >> biplex_core_init(gcb,...
> 'FFTSize', 9,...
> 'input_bit_width', 18,...
> 'coeff_bit_width', 18,...
> 'add_latency', 2,...
> 'mult_latency', 2,...
> 'bram_latency', 3,...
> 'conv_latency', 2, ...
> 'quantization', 'Truncate',...
> 'overflow', 'Wrap',...
> 'arch', 'Virtex5', ...
> 'opt_target', 'logic', ...
> 'coeffs_bit_limit', 8, ...
> 'delays_bit_limit', 8, ...
> 'specify_mult', 'off', ...
> 'mult_spec', 0, ...
> 'hardcode_shifts', 'off', ...
> 'shift_schedule', 0, ...
> 'dsp48_adders', 'on' );
> Error using biplex_core_init (line 175)
> Invalid Simulink object name: fft_stage_1/1
The 'specify_mult' mask parameter has been removed from the latest
libraries. Are you setting stuff up in your design from a script? If so
check that you are using the latest mask parameters. It seems there is
some version inconsistency going on.
I have just tried setting up the FFT and biplex_core with the settings
above and things go through as expected. zip and send me your design if
you continue to have issues, and let me know which library git revision
you are working off.
Regards
Andrew
