[Fpga-synth] Very Odd Error Message from WebPACK ISE

Scott Gravenhorst music.maker at gte.net
Fri Nov 14 18:26:31 CET 2008


"The making of synthesizers in FPGAs." wrote:
>Scott Gravenhorst wrote:
>
>> Quite honestly - it started when I got the thing to compile and 
>make sound. Until then, there were > issues with the code that 
>either prevented it from finishing the compile or it wouldn't 
>make sound. > I noticed the error after I got sound out of it and 
>honestly hadn't paid much attention to the > summary display. I 
>use a set of scripts to process .syr and other files that give me 
>up-front > information about really bad things I may have done. 
>And of course, this is the only copy of the > source, I don't 
>have previous nonworking attempts archived. > > Thus I shall plod 
>and slog as usual until the beast is revealed... 
>
>I've had pretty good luck in searching for issues by commenting out 
>large blocks of logic until the problem disappears. Bear in mind that 
>this means:
>
>* Problem is in the deleted logic
>* Problem is in the interface between deleted and surrounding logic
>* Problem is resource or overall size related.
>
>> BTW, this is for a Spartan-3A DSP 1800.
>
>Yup. Which means you need to be careful about the way your using the 
>DSP48 blocks. XST is pretty smart about inferring them, but sometimes 
>makes a real mess of it. These new blocks require more I/O pipeline 
>registers than the older MULT18x18 blocks, and if you don't explicitly 
>provide them in your code, XST will try to borrow them from upstream & 
>downstream logic with often hilarious results.
>
>If you're instantiating then you don't need to be concerned about this. 
>I typically don't instantiate though, because it's easier to maintain 
>the code, understand the functions and port to new architectures.

Ok, that's helpful.  I'm using eight DSP48s as 18x18 multipliers with registered
inputs.  That's pretty easy to instantiate, so I can try that.  I can also just
remove the IIR filters where they reside and replace them with wire vectors.  In
either case, if the problem is there, the message will abate.

-- ScottG
________________________________________________________________________
-- Scott Gravenhorst
-- FPGA MIDI Synthesizer Information: home1.gte.net/res0658s/FPGA_synth/
-- FatMan: home1.gte.net/res0658s/fatman/
-- NonFatMan: home1.gte.net/res0658s/electronics/
-- When the going gets tough, the tough use the command line.



More information about the Fpga-synth mailing list