tiomap-dev team mailing list archive

[Ricardo Salveti <rsalveti@xxxxxxxxxxxx>]

This is what I'm getting with the 4403 based SGX driver:

With Unity-2d running with a Full-HD resolution:
ubuntu@natty-panda:~/test$ time ./eglinit
real    0m0.838s
user    0m0.039s
sys     0m0.219s
ubuntu@natty-panda:~/test$ time ./eglinit
real    0m0.846s
user    0m0.000s
sys     0m0.258s
ubuntu@natty-panda:~/test$ time ./eglinit
real    0m0.846s
user    0m0.031s
sys     0m0.227s

Just X server with a Full-HD resolution:
ubuntu@natty-panda:~/test$ time ./eglinit
real    0m0.245s
user    0m0.039s
sys     0m0.039s
ubuntu@natty-panda:~/test$ time ./eglinit
real    0m0.245s
user    0m0.039s
sys     0m0.039s
ubuntu@natty-panda:~/test$ time ./eglinit
real    0m0.245s
user    0m0.008s
sys     0m0.078s

Which is not as good as Rogelio Garcia's results but still quite good if
you compare with the initial results from Alexandros.

Interesting now that the strace output shows that almost half of the
time is spent sleeping:

ubuntu@natty-panda:~/test$ strace -c ./eglinit
% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
 54.29    0.000399           0      4573      2310 stat64
 45.71    0.000336           0      2250           nanosleep

Alexandros, I just uploaded the new SGX driver, once the build is done
I'll move to the release PPA and then you can test to see what you'll
get with the SDP.

** Changed in: ubuntu-omap4-extras-graphics
       Status: New => Confirmed

** Changed in: ubuntu-omap4-extras-graphics
   Importance: Undecided => Medium

-- 
You received this bug notification because you are a member of TI OMAP
Developers, which is subscribed to ubuntu-omap4-extras-graphics.
https://bugs.launchpad.net/bugs/690821

Title:
  Calling eglInitialize() takes too long

Status in Ubuntu OMAP4 graphics addons:
  Confirmed

Bug description:
  With the latest drivers on an SDP 4430 board calling eglInitialize
  takes extremely long. Running the example program (attached) on an
  completely idle system produces the following results:

  # time ./eglinit       
  real	0m5.797s
  user	0m0.008s
  sys	0m1.055s

  # time ./eglinit
  real	0m10.523s
  user	0m0.008s
  sys	0m1.047s

  # time ./eglinit
  real	0m30.459s
  user	0m0.008s
  sys	0m1.055s

  # time ./eglinit
  real	0m5.638s
  user	0m0.008s
  sys	0m1.055s

  # time ./eglinit
  real	0m10.088s
  user	0m0.008s
  sys	0m1.047s

  I have attached an strace capture of a sample run that shows that the
  delay is happening in an ioctl() around line 10000.

  The trace also shows that the driver wastes a lot of time trying out
  /dev/dri/* devices, but that is another subject :)