Rogue Wave Software is excited to announce the new releases of TotalView for HPC and CodeDynamics.
Both products now support NVIDIA’s recently released Tesla P100 GPU. The GPU now supports compute preemption, which is instruction-level preemption. As GPUs are used for both graphics and algorithmic applications graphics, updates tend to slow down because the GPU is busy processing data in a tight loop. This problem is even worse for debuggers because they stop at a breakpoint and could wind up stopping the GPU which is updating the debugger itself. This is no longer the case because NVIDIA now supports compute preemption, which is why TotalView and CodeDynamics now support this new chip.
Debugging has gotten much smoother. When you stop a thread on the GPU you no longer stop the whole GPU, enabling interactive debugging on single-GPU systems and debugging multiple processes using the same GPU. TotalView and CodeDynamics have improved architecture in order to support multi-GPU systems at scale, resulting in a seamless debugging experience.