cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

jean-claude
Journeyman III

__kmap() bug still alive ???

Hi Guys,

A couple of months ago I had to put my Brook developments on hold since my program

was stalled due to an infamous __kmap error.

The well known symptom, is that after several correct runs, all of a sudden

the program crashes miserably when trying to execute  __kmap() code.

As an example :

void  get_RGB (::brook::stream RGB,
        ::brook::stream YCbCr)
{
  static const void *__get_RGB_fp[] = {
     "ps30", __get_RGB_ps30,
     "cal", __get_RGB_cal,
     "cpu", (void *) __get_RGB_cpu,
     NULL, NULL };
  static ::brook::kernel  __k(__get_RGB_fp);

  __k->PushOutput(RGB);
  __k->PushStream(YCbCr);
  __k->Map();  <=== THIS IS WHERE THE CRASH OCCURS !!!!

}

When this issue was reported as a bug, AMD's intent was to have it corrected in the new

release of ATI Brook....

Well... I just tried again to run my code on 1.21 beta and unfortunately this damn bug is still alive !!

 

Micah, 2 questions:

(1) is this one being worked on and should we expect to see it fixed in the very next release ?

(2) Is there any work-around possible ??

 

Many thanks for your answer,  I'd really like to progress on the project.

Regards

Jean-Claude

0 Likes
2 Replies

Jean-Claude,
This bug was not fixed in 1.2.1 because it was a fairly large fix and the runtime was rewritten for 1.3. This bug has been confirmed fixed by our testing team in 1.3 which should be out next week.
0 Likes

Thanks Micah,

I'm looking forward to getting 1.3 !

Regards

Jean-Claude

0 Likes