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.

TMS320C6x C/C++ Codegen v8.1.2 has stopped working

 

Hi there,

I have been able to build our code successfully with C6000 Code Generation Tools 7.4.7.

I am trying to use latest CG Tools v8.1.2 and in between, after building of several files, I am getting the the 'stopped working' message. 

Attached snapshot. 


Other packages version details as below:

XDAIS = xdais_7_23_00_06/

MCSDK = mcsdk_video_2_1_0_8

DSPLIB = dsplib_c66x_3_1_0_0/

XDCTOOLS = xdctools_3_23_04_60/

IPC = ipc_1_25_03_15/

BIOS = bios_6_33_06_50/

PDK = pdk_C6678_1_1_2_5/

FC = framework_components_3_23_02_16

Is there any dependencies of CG Tools v8.1.2 with other packages ?

 

Thanks & Regards,

Rajanikanth.

 

 

  • I presume the problem occurs while compiling one C file, and not while linking.  Please preprocess that C file and attach it to your next post.  Also show all the build options exactly as the compiler sees them.  This should allow me to reproduce the error.  Then I can submit a problem report to the compiler development team, and they can explain what occurred, including a possible workaround.

    Thanks and regards,

    -George

  • George,
    I have a similar issue when compiling a single .cpp file with TMS320C6x C/C++ Codegen v8.1.0. I found this thread and was hoping there was a fix for this issue since it's been posted for ~9 months now.

    I recent attempted moving from running CCS6.1.3.00033 on Linux to Windows and ran into this compiler error. Using same version of tools on Ubuntu/Linux compiles the .cpp file without error.

    Has the issue been fixed when building on Windows?

    Here's the problem details...

    Problem signature:
    Problem Event Name: APPCRASH
    Application Name: cg6x3.exe
    Application Version: 8.1.0.0
    Application Timestamp: 5657d9b0
    Fault Module Name: cg6x3.exe
    Fault Module Version: 8.1.0.0
    Fault Module Timestamp: 5657d9b0
    Exception Code: c0000005
    Exception Offset: 001ce100
    OS Version: 6.1.7601.2.1.0.256.28
    Locale ID: 1033
    Additional Information 1: 9b93
    Additional Information 2: 9b9313e5d60bc02d7ef51309485297fd
    Additional Information 3: b790
    Additional Information 4: b790bc8a29798492c942ac7a96d283db

    Thanks,
    -George
  • George,
    I upgraded to Codegen v8.1.4 on Windows CCS3.1.3.00033 and had the same compiler crash issue on same .cpp file.

    Here's the problem details...

    Problem signature:
    Problem Event Name: APPCRASH
    Application Name: cg6x3.exe
    Application Version: 8.1.4.0
    Application Timestamp: 591f42ac
    Fault Module Name: cg6x3.exe
    Fault Module Version: 8.1.4.0
    Fault Module Timestamp: 591f42ac
    Exception Code: c0000005
    Exception Offset: 001cf610
    OS Version: 6.1.7601.2.1.0.256.28
    Locale ID: 1033
    Additional Information 1: 9b93
    Additional Information 2: 9b9313e5d60bc02d7ef51309485297fd
    Additional Information 3: 900b
    Additional Information 4: 900b830d3460a18371dae8ead2c4bf2c

    Any feedback regarding this compiler crash on Windows would be greatly appreciated. At the moment I'm doing development on Ubuntu/Linux with CCS3.1.3 Codegen v8.1.0, but need to move to Windows for development and release build.

    Thanks,
    -George
  • George,
    Sorry about the typo... I meant CCS6.1.3.00033 and *not* CCS3.1.3.00033 in last reply.

    -George
  • As we never received a test case, there's no way we could know if these problems are related. C6000 compiler version 8.1.4 is the latest version; if the bug exists in that version, it is unlikely that it has been fixed. In order to diagnose the problem, we need a reproducible test case, as noted above by George.
  • George,
    I tried using compiler version TI v7.4.16 and witnessed NO compiler crash when compiling the same problematic .cpp file causing crashes on cgtv8.1.0 and cgtv8.1.4.

    Hopefully this might help identify the issue.

    -George
  • Unfortunately, we continue to lack a test case which allows us to reproduce the behavior.  And that is the only way to advance the issue.  Please follow the directions in the article How to Submit a Compiler Test Case

    Thanks and regards,

    -George

  • Thank you for sending in a test case through other channels.  I can reproduce the problem.  I filed CODEGEN-2401 in the SDOWP system to have this issue addressed.  You are welcome to follow it with the SDOWP link below in my signature.

    Thanks and regards,

    -George

  • George,

    I went to the SDOWP system and search for CODEGEN-2401 and found your problem report on the cgt v8.1.x compiler crash, but did not see any further comments as to resolution, etc.

    Later in the day yesterday, my CCS6.3.1.00033 on both Linux and Win7x64 said there were updates among them was cgtv8.2.0.

    I proceeded with that update on both Linux and Win7x64 and found that a new compiler v8.2.0 was installed.

    I ran my project build (containing --interrupt_threshold) with this new TI v8.2.0 compiler on both Linux and Win7x64 and witnessed NO errors nor crashes on either.

    I loaded the Win7x64 build artifact on our target and it appears to be running correctly, albeit we haven't performed a full regression test as of yet.

    This TI v8.2.0 flavor of the compiler appears to have resolved the Win7x64 compiler crash when both --interrupt_threshold and -0(>=2 ?) are present on some of our source code.

    Is this an intended fix for cgtv8.2.0 or a pleasant surprise?

    Is this cgtv8.2.0 a safe and stable flavor of the code-gen-tools that I can lock-in on for production releases?

    -Or-

    Is there a better fix coming in a future version of the code-gen-tools?

    Please advise.

    Thanks,

    -George

  • The root cause of CODEGEN-2401 has not been found and addressed.

    George Lathrop said:
    Is this an intended fix for cgtv8.2.0 or a pleasant surprise?

    Unfortunately, it is a pleasant surprise.

    George Lathrop said:
    Is this cgtv8.2.0 a safe and stable flavor of the code-gen-tools that I can lock-in on for production releases?

    That is a decision you make based on your circumstances.  Guidance on this matter is in the article Compiler Version Numbers and What They Mean.

    Thanks and regards,

    -George

  • George,

    Thanks for the update.

    Does the compiler team realize that TI v8.2.0 has this "pleasant-surprise-fix" in regards to --interrupt_threshold ?  If not, they should be so informed as that might help them determine the fundamental issue.

    Regarding my questions about this compiler I read the Wiki you pointed me to, but since this compiler is using the legacy version numbering scheme I don't know whether it is considered STS or LTS.

    Is TI v8.2.0 a Short-Term-Support (STS) or Long-Term-Support (LTS) compiler?  Where is this documented for legacy compilers.

    -George

  • The compiler team has all of the forum details provided.

    C6000 does not use the STS/LTS versioning scheme. The active branches for C6000 are 7.4.x, 8.1.x, and 8.2.x
  • George Lathrop said:
    Is TI v8.2.0 a Short-Term-Support (STS) or Long-Term-Support (LTS) compiler?

    That version numbering scheme is not used by the C6000 compilers.  So the distinction between them does not apply.

    Regarding the difference between 8.1.x and 8.2.x, the line in the article which is most relevant is ...

    The Y digit, the "minor" number, means there is still ABI-level (or object level) compatibility. In addition to bug fixes, new features are typically added. Code which uses the new features is probably not compatible with previous versions that lack those features.

    Thanks and regards,

    -George

  • All,

    Thanks for pointing that out.  I missed that C6000 wasn't explicitly called out.  Now, with that knowledge, that article isn't too useful regarding my interest in C6000 compiler versioning.  I'll have to look at the specific compiler release notes themselves.

    Thanks,

    -George