What does 'coder.int​ernal.inde​xShapeChec​k>>errORWa​rnIF .... code generation assumption about size violated' mean?

I am attempting to convert a floating point model into fixed point code. I am recieving the error message presented below:

‘ Error using coder.internal.indexShapeCheck>>errOrWarnIF

Code Generation assumption about size violated. Run-time indexing is vector(vector) with different orientations, but compile-time assumption was matrix(vector) indexing.’

This message is from me using the dsp.FFT system object on a [1 2048] variable. The variable is bounded by coder.varsize in order for the system object to work. This code works within the MATLAB environment but in the HDL Coder and produces the above error.

(Unfortunatley due to security I am unable to attach code)

[Walter Roberson]

Can you show us the coder.varsize call?

[Joshua Ford]

coder.varsize('NAME', [1 2048], [0 1])

ft(NAME)

[Walter Roberson]

That code appears to lock the first dimension as being exactly 1, therefore defining a vector of varying length. I do not understand how that could have resulted in matrix(vector) indexing.

Not unless that object has a matrix varsize? A second coder.varsize that forgot to lock the first dimension?

Answer 1

[Kiran Kintali]

This error is unexpected. Please share a sample project file that reproduces the error or reach out to technical support. HDL Coder in general does not supprt variable dimensions.