Today IBM announced the establishment of a new POWER Acceleration and Design Center in Montpellier, France. Launched in collaboration with Nvidia and Mellanox, the new center will advance the development of data-intensive research, industrial, and commercial applications.
“In Deep Learning what we do is try to minimize the amount of hand engineering and get the neural nets to learn, more or less, everything. Instead of programing computers to do particular tasks, you program the computer to know how to learn. And then you can give it any old task, and the more data and the more computation you provide, the better it will get.”
“This talk will introduce these three debugging techniques and provide some suggestions on selecting the optimal approach for a variety of debugging scenarios such as hangs, numerical errors, and crashes. Specific examples will be given using the TotalView debugger but the concepts covered may apply to other debugging tools such as GDB and the NVIDIA NSIGHT debugger.”
“In this session we describe how GPUs can be used within virtual environments with near-native performance. We begin by showing GPU performance across four hypervisors: VMWare ESXi, KVM, Xen, and LXC. After showing that performance characteristics of each platform, we extend the results to the multi-node case with nodes interconnected by QDR InfiniBand. We demonstrate multi-node GPU performance using GPUDirect-enabled MPI, achieving efficiencies of 97-99% of a non-virtualized system.”
“OpenACC was applied to the a global high-resolution atmosphere model named NICAM. We executed the dynamical core test without re-writing any specific kernel subroutines for GPU execution. Only 5% of the lines of source code were modified, demonstrating good portability. The results showed that the kernels generated by OpenACC achieved good performance, which was appropriate to the memory performance of GPU, as well as weak scalability. A large-scale simulation was carried out using 2560 GPUs, which achieved 60 TFLOPS.”