Ernest

I am getting sporadic nana results when multiplying a signal by a fraction. I have tried:

Doesn't seem to stop the nans. What am I doing wrong?

I am getting sporadic nana results when multiplying a signal by a fraction.  I have tried:

Doesn't seem to stop the nans. What am I doing wrong?


denormals-in-codebox

What I find is that the same thing using objects rather than codebox appears to work as expected. It may be because the gen~ object is receiving signals from another gen~ object in a poly~ object, which used to generate a warning message about separate processing on the same DSP chain, and it may be because the output of the poly~ object is going into display objects with 32-bit rather than 64-bit ranges. I can't isolate the problem as of yet.

What I find is that the same thing using objects rather than codebox appears to work as expected.  It may be because the gen~ object is receiving signals from another gen~ object  in  a poly~ object, which used to generate a warning message about separate processing on the same DSP chain, and it may be because the output of the poly~ object is going into display objects with 32-bit rather than 64-bit ranges.  I can't isolate the problem as of yet.


The 32-bit vs 64-bit part sounds likely -- the fixnan implementation in gen~ assumes 64 bit doubles, and what is denormal at 32bit can be fine at 64.

One way to verify that the fixnan is actually in place is to send 'exportcode' to your gen~ and have a quick peek at the c++ code it generates -- look inside the perform method.

The 32-bit vs 64-bit part sounds likely -- the fixnan implementation in gen~ assumes 64 bit doubles, and what is denormal at 32bit can be fine at 64. 

One way to verify that the fixnan is actually in place is to send 'exportcode' to your gen~ and have a quick peek at the c++ code it generates -- look inside the perform method.


Denormals in codebox?