- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I'm currently experiencing an error after invoking a large number of the same OpenCL kernels. The OpenCL kernel itself work fine when run on its own. However I need the host program to launch this kernel repeatedly, one after the other. When I run the host program with a low number of iterations eg 10 (there is 1 kernel invocation per iteration) things work fine. Similarly when I increase the iteration count to 10,000 everything works. However at 100,000 iterations the program crashes with an "Out of host memory" error. I believe it crashes at around the 20,000th kernel invocation. I've removed everything from my host program that may growing (in terms of memory footprint), however I still get the same problem. Does the Altera OpenCL save the kernel or buffer object or similar during its execution. I'm just trying to understand what might be causing this problem. A possible work around I can think off is to add the iteration loop to the kernel (fpga side) itself however I'd rather not have to do this. Therefore I'd like to understand more about why this problem is occurring before I go down this route. I'd be grateful for any assistance solving this problem. Many thanksLink Copied
1 Reply
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Normally when I see people running into these problems it's because they allocate more resources inside a loop. The resources could be buffers, events, etc... which all have limitations. Knowing which API you are calling when the error code returns would help us identify what you might be running into.
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page