• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 31 subscribers
  • Views 359 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

Effort to compile SDK with GCC instead of CLANG

Eli Mordel
Expert 1345 points
Other Parts Discussed in Thread: MCU-PLUS-SDK-AM243X

Hi,

I consider using the MCU-PLUS-SDK-AM243x.

I've noticed it was originally compiled with CLANG.

Is it easy to re-compile it wit GCC ?

Thanks,

Eli

 

  • 0
    TI__Mastermind 45084 points

    Hi Eli,

    MCU SDK officially supports CLANG compiler only. GCC compiler support is an experimental feature. 

    Could you please tell me why do you want to compile with GCC.

  • 0 in reply to Nilabh Anand
    Expert 1345 points

    Hi,

    I'm working on a very high performance requirements project.

    I looked at some benchmarks and here is a quote:

    "In terms of 600.perlbench_s and 602.gcc_s/O2, GCC has a great performance advantage (more than 10%). "

    The thing is I'm not sure what happens on the AN2432 , so I would like to compile with both compilers and compare the results.

    Thanks,

    Eli.

  • 0 in reply to Eli Mordel
    TI__Mastermind 45084 points

    Hi Eli,

    Eli Mordel said:
    "In terms of 600.perlbench_s and 602.gcc_s/O2, GCC has a great performance advantage (more than 10%). "

    Could you please tell me if the comparison is against the Clang compiler. Please do mention source of this info too.

    Eli Mordel said:
    The thing is I'm not sure what happens on the AN2432 , so I would like to compile with both compilers and compare the results.

    Could you tell me if you are trying some SDK example or you want to compile and run some benchmark code which you have?

  • 0 in reply to Eli Mordel
    TI__Expert 4680 points

    Hi Eli,

    I am also curious what is being compared, and what compiler options are being used (you mention GCC -O2). It would be good to measure against tiarmclang with -O3 and with LTO enabled (-flto) in the application as well as libraries.

    We measured tiarmclang 2.1.3.LTS on AM65x/R5F for Coremark and achieved a score of 3.76 (-O3 -mthumb) and 3.56 (-O3 -marm) owing to smarter loop optimizations.  This is greater than the Arm Ltd. reported score of 3.47.  For Dhrystone, the next tiarmclang release (3.0.0.STS) for achieves 1.58 DMIPS/MHz (no inlining) and 2.05 (with inlining).

    We can also focus on how to improve tiarmclang to help if you can provide us with your benchmark.

    Thanks

    -Alan

  • 0 in reply to Alan Phipps
    Expert 1345 points

    Hi,

    I don't have my own benchmark yet , and I suppose that the tiarmclang is fast enough 

    The thing is that I'm familiar with the GCC only and according to some publications (which I've never checked myself) - GCC is a bit faster.

    I'll proceed with the Clang as TI uses it on the SDK and I'm also not sure if I can handle the porting.

    Thanks,

    Eli