• State TI Thinks Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 1 answer
  • Subscribers 31 subscribers
  • Views 873 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.

TMS570LC4357: TMS570LC cache benchmark

Etienne Alepins
Intellectual 775 points

Part Number: TMS570LC4357

Hi,

We are wondering what is the performance impact of selecting write-through or write-back cache modes for TMS570LC4357 (Cortex-R5F core). We would like to know if write-through decreases a lot the performance. Do you have a benchmark of that microcontroller?

Thanks.

  • 0
    TI__Guru 59540 points
    Hello Etienne,

    There is a significant impact to performance for selecting write through option but I am uncertain if we have any benchmarks on it. I will check and get back to you.
  • 0 in reply to Chuck Davenport
    Intellectual 775 points
    Our benchmark shows only a 2,64% performance degredation when changing from write-back to write-through. This doesn't seem coherent for us. Our benchmark runs in internal Flash and all the data is in internal RAM. It is a big application using around half of each memory. We made sure we really were in write-back by verifying we lose recently updated data when invalidating the cache without first having flushed it. Any idea?

    Although, our benchmark hopefully shows that having the cache enabled or disabled has a huge impact on performance.
  • 0 in reply to Etienne Alepins
    Intellectual 775 points

    Any explanation for our results?

  • 0 in reply to Etienne Alepins
    TI__Guru 59540 points
    Hello Etienne,

    My apologies for the delay in response. There are several factors that come into play when comparing results. First and foremost will be the number of hits and misses for data in Cache. i.e., if the data is such that there are alot of misses, there won't be as much of an impact as when there is a predominance of hits where data residing in Cache is used repeatedly. Also, there are two types of write back options; either write back with allocation and write back non-allocated. This too would cause a difference in how the wriet back vs. write through would benchmark against each other.