• State TI Thinks Resolved
  • Locked Locked
  • Replies 2 replies
  • Answers 2 answers
  • Subscribers 99 subscribers
  • Views 814 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.

C674x Matrix Math DGEMM performance

Allen35065
Expert 4345 points

Hi experts,

Customer want to compare C66x and C674x performance on double precision fp algorithm. From this table, we can't find matrix math DGEMM performance on C674x. Do you have any information on that? I suppose double precision performance should be same on C66x and C674x as the intrinsics are same. 

Regards,

Allen

  • 0
    TI__Mastermind 35645 points

    Hi,

    Thanks for your post.

    I believe, Matrix Math DGEMM is not applicable for C674x, so, TI have not published the benchmarks cycles and execution time for the same and thereby, the benchmarking projects for DGEMM are not available as part of the libraries. Also, it doesn't contain little endian ELF version of the projects which is used for benchmarking. May be, you could reproduce the other C674x core benchmarks as mentioned in Table 1

    http://www.ti.com/lit/an/sprac13/sprac13.pdf

    In case, if it is helpful for you, you could refer the basic linear algebra subprogram (BLAS) benchmarking for DGEMM as below:

    http://processors.wiki.ti.com/images/d/d2/LINALG_1_0_benchmark_BLAS.pdf

    Thanks & regards,

    Sivaraj K

    -------------------------------------------------------------------------------------------------------

    Please click the Verify Answer button on this post if it answers your question.

    -------------------------------------------------------------------------------------------------------

  • 0 in reply to Sivaraj Kuppuraj
    TI__Genius 12805 points

    Indeed Sivaraj is correct. If you look at the documentations for DSPLIB for C66 and C674 you can see that C66 has many more Matrx functions (especially double precision)

    Ran