In our make system for TinyOS, we've had to disable the gcc hardware multiplier support because in complex programs, the arbitration for the multipler has bugs that yield invalid computations and failed program execution. We've tried to do this with a simple test case, but it appears the problem only occurs in complex programs. I believe Gilman Tolle wrote about this problem a few months back and was also unable to produce a simple test case.
Has anyone else seen this problem or found a way around it for gcc 3.2.3? I have "complex" examples that I can share, but no simple test cases despite trying. Seems that using TinyOS and nesC's machine generated C code can wreak havoc with msp430-gcc. Steve--No worries, just wanted to make sure someone was aware of the bug. thanks for fixing it. -Joe On Sat, 12 Mar 2005 13:10:48 +0800, Steve Underwood <[email protected]> wrote: > Hi Joe, > > Sorry about that. I should be OK in developer CVS now. I don't know how > long that will take to reach the public CVS. The error was just the last > underscore missing in this line in compa.h: > > #if defined(__msp430_have_compa_extended_mux) > > This addition provides for new comparator features in the msp430f21x1 > devices. > > Regards, > Steve >
