5 Replies Latest reply on Dec 7, 2011 11:06 AM by simonimpey

    clCreateEventFromGLsyncKHR

    david_aiken

      I'm currently unable to resolve clCreateEventFromGLsyncKHR() linking against the ati stream 2.2 libatiocl64.so or libOpenCL.so libraries under $ATISTREAMSDKROOT/.../lib/x86_64 on Ubuntu Lucid 64-bit, though it is declared in cl_gl_ext.h. Can anyone confirm that I'm looking in the right place? Tried running nm against the libs, but no symbols. Out of idle curiousity, are debug symbols available?

      thanks!

        • clCreateEventFromGLsyncKHR
          nou

          this will be definitely supported at earliest in next release.

            • clCreateEventFromGLsyncKHR
              simonimpey

              I am trying to use clCreateEventFromGLsyncKHR to synchronise OpenCL and OpenGL as I am finding using glFinish/clFinish quite slow.

              Unfortunately I am experiencing the exact same issue that the originator of this thread was. This is with the APP 2.5 running on a Radeon HD5870. Is this function still unsupported?

                • clCreateEventFromGLsyncKHR
                  nou

                  yes OpenCL implementation must support cl_khr_gl_event extension. try wait for 2.6 which will be released on 13.12.

                    • clCreateEventFromGLsyncKHR
                      himanshu.gautam

                      Does your hardware report the following extension cl_khr_gl_event?

                      Can you be clear whether clCreateEventFromGLSyncKHR is not working at all or is making clFinish call slow.

                        • clCreateEventFromGLsyncKHR
                          simonimpey

                           

                          Originally posted by: himanshu.gautam Does your hardware report the following extension cl_khr_gl_event?

                           

                          Can you be clear whether clCreateEventFromGLSyncKHR is not working at all or is making clFinish call slow.

                           

                          Appologies, re-reading my original post it is not as clear as it should be. To clarify.

                          My issue is that using glFinish/clFinish to wrap a section of code calling OpenCL kernels that access shared OpenCL/OpenGL assets is quite slow and so I wanted to explore the synchronisation events approach. 

                          I did check the extension support and indeed it isn't on my HD 5870. I tried implementing the approach just in case.

                          With the above in mind my question was more regarding future support, whether this extension/functionality is something that is likely to be supported in future and if so, only on future hardware or on current hardware too?