uygn / aparapi

Automatically exported from code.google.com/p/aparapi
Other
0 stars 0 forks source link

High total processing/running time on GPU mode w/Aparapi #157

Open GoogleCodeExporter opened 8 years ago

GoogleCodeExporter commented 8 years ago
What steps will reproduce the problem?
1.Run program, usually high process time using Aparapi API (>1700ms)
2.
3.

What is the expected output? What do you see instead?
0ms total processing time, or faster than CPU

What version of the product are you using? On what operating system?
Using Jgrasp, for simplicity. Java JDK 1.8 64-bit.
Samsung notebook, AMD Kabini A4/A6 Quad-Core @1.00ghz(1.40ghz), 4Gb RAM, Radeon 
HD 8250

Please provide any additional information below.

below are the outputs:
 ----jGRASP exec: java Main

GPU mode: 1829ms

 ----jGRASP: operation complete.

Original issue reported on code.google.com by minhquan...@gmail.com on 20 Nov 2014 at 3:04

Attachments:

GoogleCodeExporter commented 8 years ago
Here is my dxdiag info for extra information. I have no error in getting the 
Aparapi to work, just that the performance isnt as good i as want it to be. 

Original comment by minhquan...@gmail.com on 20 Nov 2014 at 3:07

Attachments:

GoogleCodeExporter commented 8 years ago
Hello emporer,
Your "range" or job size is 512? That is not big enough to get any benefit
for offloading a small computation like this -
            sum[gid] = a[gid] + b[gid];
as shown in the big report.
Either you need a much bigger range or more complicated compute algorithm
in the kernel to make offloading it worthwhile.

-Eric

Original comment by ecasp...@gmail.com on 20 Nov 2014 at 8:45