This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Bug in optimizer of TMS320C6x C/C++ Compiler v7.3.1

We've found a bug that causes nondeterministic output, when optimizations are enabled, but not otherwise.  I believe it's a dataflow analysis or scheduling problem, since allocating a temporary array on the heap (instead of stack) also seems to avoid the bug.  I don't believe we're misusing the restrict keyword or doing anything else that would break alias analysis.

I can post full details and a code excerpt that should reproduce the problem, but I'd first like to verify that it hasn't already been reported & fixed.  Is there anywhere I can see a list of known (and fixed) bugs?

Also, we're using EZSDK and SYS/BIOS 6.33 with a DM8127.  Can we upgrade the compiler independent of these?  If so, should we anticipate any complications in doing so?  Are there any restrictions on the latest release we can use?  Unfortunately, it seems we're stuck with EZSDK.

Thanks.

  • Matt Gruenke said:
    I'd first like to verify that it hasn't already been reported & fixed.  Is there anywhere I can see a list of known (and fixed) bugs?

    All things considered, it is probably easier for you to submit the bug and for us to then determine whether it is a duplicate.  All the same, you are welcome to look through the lists of known and fixed bugs.  This text file lists all the fixed and known bugs in the 7.3.x stream of compiler releases.  It is current through version 7.3.23, which is the latest 7.3.x release at this time.  A new defect history file is created for each compiler release.

    Matt Gruenke said:
    I can post full details and a code excerpt that should reproduce the problem

    Please do so.  You probably need to preprocess that code.  And be sure to show the exact compiler build options used.

    Matt Gruenke said:
    Can we upgrade the compiler independent of these?

    Yes.  Please see this page.

    Matt Gruenke said:
    should we anticipate any complications in doing so?

    To better understand which version you should get, please see this wiki article.

    Thanks and regards,

    -George

  • Thanks for the helpful reply.

    I have now tested both 7.3.23 and 7.4.14. Code generated by the latter produces nondeterministic output, while the former (7.3.23) does not. I haven't yet confirmed whether it's entirely due to the same issue we found in 7.3.1.

    I am currently trying 8.0.3.