@Terry_ALE wrote:
I would go the support route https://www.ni.com/en-us/support.html#requestsupport.
That forum page is good too but not sure if it will get the proper attention.
We (this forum) are pretty sure this is an issue but I think the best way is to go the support route. They will ask you standard questions (blah blah blah) which is good cause we (the forum) could have missed something really simple (don't we always focus on the hard stuff?).
I would tell them that you think this is a bug and that it should get escalated (don't let them have to figure that out).
I recently did this with the MAT file interface (MATLAB files). Patience is a virtue 🙂
Personally I liked digging around this library for this discussion cause I never use it and don't have any background in it.
I submitted a request ticket and got this response:
I was able to test the code, I confirmed that there are errors up to 5%, however, this error is relative to each computation, you could have a mismatch of 1e-9 and it could represent 5% which is a pretty small difference. This is probably caused by the algorithms that use each function (FPGA and LabVIEW in the Host), since the cosine and sine are an approximation of a sum of infinite terms, probably each algorithm is using a different amount of terms and that's why you got that mismatch.
That said, my question is, does this mismatch affects your project/application or similar? If the answer is yes, we could think in escalate the case and check with R&D for a possible workaround. If not, for example, let's say that you are using these values in the order of thousandths, this won't affect your application.
Per our previous discussions - this was more of a learning experiment than a requirement for floating point accuracy in my current project. I asked that they update several NI sites referring to FPGA SGL library with a warning about loss of accuracy against Host blocks, and then asked to close the ticket.
