KernelGen - GPU Technology Conference
KernelGen - GPU Technology Conference
KernelGen - GPU Technology Conference
- No tags were found...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>KernelGen</strong> execution workflow.Scenario 2: main kernel approached a point where CPU function is called (no source code, externallibrary, syscall, or inefficient loop). In this case, function is called via host callback, using libffi API.<strong>GPU</strong> data is lazily mapped for using on host, where needed, by pagefault handler.versionswitchcompilemain kernelcallback: requestedhost function launchcallbackCPU<strong>GPU</strong>stream #2(many threads)<strong>GPU</strong>stream #1(single thread)launchmain kernelLaunchhost function<strong>GPU</strong> data is mappedto host, by requestresumemain kernel58 / 75