Closed GoogleCodeExporter closed 9 years ago
Actually, another possibility is that put() is only transferring data the first
time, this would produce the same result with the attached program. I did some
more testing to determine if this might be the case, and I think I can say that
it's the put() method that is to blame. I replaced the line that writes to the
neuronOutputs array in the run() method with:
neuronOutputs[neuronID] = neuronID + neuronOutputs[neuronID] + simStep[0];
and get:
GPU, auto
[0, 1, 3, 6, 10, 15, 21, 28, 36, 45, 55, 66, 78, 91, 105, 120]
[1, 3, 6, 10, 15, 21, 28, 36, 45, 55, 66, 78, 91, 105, 120, 136]
[2, 5, 9, 14, 20, 27, 35, 44, 54, 65, 77, 90, 104, 119, 135, 152]
GPU, explicit
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]
[2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32]
The get() method returns different results after every execution as expected,
but the value of simStep[0] inside the kernel doesn't seem to change.
Original comment by oliver.c...@gmail.com
on 7 Aug 2012 at 5:03
Oliver thanks for reporting this.
This looks bad. Your analysis looks good.. It does appear that put() is not
being honored..
If you have some test code can you post it for me (from the dumps above it
looks like you mights) I want to build a junit test for this.
Was this using Aparapi built from the trunk? Or from a binary distribution.
Gary
Original comment by frost.g...@gmail.com
on 8 Aug 2012 at 2:28
Apologies Oliver. I just saw that you attached your test code.
Original comment by frost.g...@gmail.com
on 8 Aug 2012 at 9:22
So I turned your test code into a junit case and it seems to work just fine
when I use against the trunk. I will pull aparapi-2012-05-06.zip (R#407, May
6) and try in a moment.
I added aparapi/test/runtime junit tests to track this.
Original comment by frost.g...@gmail.com
on 8 Aug 2012 at 9:36
Gary,
This sounds similar/identical to the bug we've been tracking offline...? That
bug was apparently fixed in the the latest trunk code as well.
Oliver,
Can you build the latest Aparapi Trunk code and verify whether it fixes this
bug?
Original comment by ryan.lam...@gmail.com
on 9 Aug 2012 at 5:40
Yep, latest trunk code seems to fix it. :)
Original comment by oliver.c...@gmail.com
on 9 Aug 2012 at 11:27
Thanks Ryan and Oliver.
It's time to build a new binary ;)
I will keep this open until we make a new binary available.
Gary
Original comment by frost.g...@gmail.com
on 9 Aug 2012 at 12:07
Original issue reported on code.google.com by
oliver.c...@gmail.com
on 7 Aug 2012 at 4:28Attachments: