SAFETI_CQKIT: ARM-CLANG-CQKIT test results for AM2634-Q1 cortex-r5 and C99

Hello,

I have brought this thread to the attention of the compiler experts. Please note that due to the local holiday it may take until Wednesday for a response. 

Thank you for your paitence.

ki

David Hodgson said:

nearly all the test results applicable for these core types are with either big-endian option or no optimization enabled or for C++.

That is not the case.  In the file validation_results.pdf from the qualification kit, I see lots of tests that combine -mv7r4 with -o0 through -o4 (all the levels of optimization), with and without -me (little and big endian).

Thanks and regards,

-George

I think you must be referring to ARM-CQKIT and not ARM-CLANG-CQKIT (https://www.ti.com/tool/download/ARM-CLANG-CQKIT).

As far as I am aware the TI ARM-CGT compiler is no longer being developed (bug fixes only) and is not recomended for use with the latest TI Sitara AM263x, hence was asking about the CLANG version.

Thanks

David

David Hodgson said:

I think you must be referring to ARM-CQKIT and not ARM-CLANG-CQKIT

You are correct.  I apologize for the oversight.

But I have the same point to make.  I see lots of tests that combine -mcpu=cortex-r4 with all the variations of optimization, with and without -mbig-endian.  

Thanks and regards,

-George

Looking through the list, and disregarding any results for big-endian, or tests specifically for other language version C89, C90, C11 and C++, and results without optimisation.

This is the entire list of tests:

Test                                     Passed    Compiler Options

PHIL Regression Testing    328          -mcpu=cortex-r4 -mthumb -mfloat-abi=hard -mfpu=vfpv3-d16 -mno-unaligned-access -Og
PHIL Regression Testing    324          -mcpu=cortex-r4 -mfloat-abi=hard -mfpu=vfpv3-d16 -O2
PHIL Regression Testing    322          -mcpu=cortex-r4 -mfloat-abi=hard -mfpu=vfpv3-d16 -O2 -fpack-struct -mno-unaligned-access
PHIL Regression Testing    328          -mcpu=cortex-r4 -mthumb -mfloat-abi=hard -mfpu=vfpv3-d16 -mno-unaligned-access -Oz -finline-functions
PHIL Regression Testing    322          -mcpu=cortex-r4 -mfloat-abi=hard -mfpu=vfpv3-d16 -O1 -fpack-struct -mno-unaligned-access

Can you confirm the 325 tests in "PHIL Regression Testing" provides full coverage?

Please consider another way of testing compliance.  Please view the article Simplify your functionally safe system development with the TI Arm®︎ Clang compiler.  Figure 1 contains a flow chart.  Consider choosing Validation mode.  Among other things, this mode calls for building your code, with your build options, using an instrumented compiler.  This generates some coverage files that you send to TI.  TI processes those coverage files, and issues a report which indicates whether your build, which uses your options, exercised any parts of the compiler that are not exercised when we test it for release.  I think this method is more likely to satisfy your requirements.

Thanks and regards,

-George

Thanks for the response.

I agree that this is the best option available for tool qualification.

One further question is that ideally if we need to carry out the tool qualification before starting writing our code, do we just send what we have developed so far now? Then repeat this activity when the final code has been developed?

Thanks

David

David Hodgson said:

if we need to carry out the tool qualification before starting writing our code

That is up to you.  We cannot tell you how to qualify your code.

Thanks and regards,

-George