--- Quote Start ---
I have never used clRetainMemObject. Are you using these calls in-between the execution of your kernels? These are generally used at the end of execution for cleanup purposes, and in such case, will in no way affect data integrity. Are you using blocking read/write buffer calls? Make sure you are not releasing a device buffer right after a non-blocking read from it (which can obviously cause data corruption).
--- Quote End ---
I am always using blocking reading/writing calls. I am releasing only buffers every time I finish with a set of data and then start again with another set of data. I just don't understand why one works only with GPU and the other only with CPU.
CPU gives me problems of resources when I use the "wrong one"... Is there anyone that can help me figure out this? Thanks.
Honored Contributor
