Intel® Quartus® Prime Software
Intel® Quartus® Prime Design Software, Design Entry, Synthesis, Simulation, Verification, Timing Analysis, System Design (Platform Designer, formerly Qsys)
17258 Discussions

aocx files for 2d fft example

Altera_Forum
Honored Contributor II
‎02-18-2014 08:20 PM
1,624 Views

Hi, 

 

I'm trying to debug an issue with the 2d fft example available for download here 

 

http://www.altera.com/support/examples/opencl/fft-2d.html 

 

I compiled the fft2d.cl file for several fft sizes (modifying the host/inc/fft_config.h file accordingly). clcreateprogramwithbinary fails with error code CL_INVALID_VALUE for any fft size but 1024x1024. I was wondering if the offline compiled kernels for sizes other than 1024x1024 are available for download somewhere. That would allow me to determine whether the problem is with my aocx files or in the OpenCL host code I'm using. I have the bittware s5phq_d8 board. 

 

Any help is appreciated. 

Cheers, 

Dominic
0 Kudos

Link Copied

2 Replies
Altera_Forum
Honored Contributor II
‎02-19-2014 12:55 AM
768 Views

So I assume you modified LOGN in the fft_config.h file to map to the size you want, for example a 512x512 FFT would be LOGN = 9. Just so that this isn't just a compilation thing on the host side I would clean the host application then recompile it since makefiles are often not written to be dependent on include files changing. So it might be that the kernel is recompiled but the host code still thinks LOGN is set to 10 since it wasn't built with an updated macro definition.

0 Kudos
Copy link
Altera_Forum
Honored Contributor II
‎02-19-2014 03:54 PM
768 Views

Thanks for the reply. I did change the LOGN in the config header to generate aocx files for sizes 64x64 through 512x512 (the 1024x1024 file comes with the example). 

 

I'm not using the host code that comes with the example. I wrote a new host application that lets the user choose the FFT size at runtime. It is that application that I'm trying to debug. For some reason I can only get the 1024x1024 example to work. As far as I can tell there are three broad classes of errors that could be causing this: 1) A mistake in the host code, 2) an error in the offline compilation of the kernels, 3) an IO error (aocx file is corrupt or the binary file doesn't get read into the host application correctly). If I had an aocx file that is known to work it would allow me to narrow my search down a bit. That's why I was wondering if you could make available a aocx file for a different FFT size. 

 

I'm currently in the process of regenerating aocx files but the turnaround time is pretty slow on my desktop. If you had an aocx file ready to go and if you could make it available to me that would make my life a lot easier. 

 

Thanks again. 

 

Cheers, 

Dominic
0 Kudos
Copy link
Reply

Community support is provided Monday to Friday. Other contact methods are available here.

Intel does not verify all solutions, including but not limited to any file transfers that may appear in this community. Accordingly, Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.

For more complete information about compiler optimizations, see our Optimization Notice.