We have an neural network running on a 52832 using the floating point unit. We started encountering FPU exceptions so we enabled the FPU IRQ to identify which specific instructions is causing the error. However we started getting thousands of FPU interrupts caused by the IXC exception, which we do not care about. Is there anyway to mask which exceptions will actually result in an FPU_IRQ? The addition of many thousands of interrupts during our processing is slowing our algorithms to the point where the data collection is stalling and overflowing.
I learned that some processors actually do not connect the IXC bit to the interrupt. https://www.st.com/content/ccc/resource/technical/document/application_note/10/6b/dc/ea/5b/6e/47/46/DM00047230.pdf/files/DM00047230.pdf/jcr:content/translations/en.DM00047230.pdf
I definitely see the value in not connecting IXC, since this exception occurs very frequently in certain computations.