Results 1 to 3 of 3

Thread: local_work_size question

  1. #1
    Junior Member
    Join Date
    Jul 2011
    Posts
    4

    local_work_size question

    if I understand correctly, local_work_size tells how many GPU cores will process the global_work_size ? That means with local_work_size = 1 and global_work_size=10 I can create a kernell that would do: a=a+1 and not worry about it getting screwed?

  2. #2
    Member
    Join Date
    Jul 2011
    Location
    Moscow, Russia
    Posts
    41

    Re: local_work_size question

    Quote Originally Posted by Noidea
    if I understand correctly, local_work_size tells how many GPU cores will process the global_work_size ? That means with local_work_size = 1 and global_work_size=10 I can create a kernell that would do: a=a+1 and not worry about it getting screwed?
    No, local worksize is actually the size of workgroup. Workgroup is a bunch of workitems which share the same local buffers.
    Blog (in russian)

  3. #3
    Junior Member
    Join Date
    Jul 2011
    Posts
    4

    Re: local_work_size question

    Quote Originally Posted by Maxim Milakov
    Quote Originally Posted by Noidea
    if I understand correctly, local_work_size tells how many GPU cores will process the global_work_size ? That means with local_work_size = 1 and global_work_size=10 I can create a kernell that would do: a=a+1 and not worry about it getting screwed?
    No, local worksize is actually the size of workgroup. Workgroup is a bunch of workitems which share the same local buffers.
    ok, thanks .. I dont understand what you mean by local buffers but atleast I know I wasnt thinking right. Gona google.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •