#pragma section-numbers on ---- <> ---- = Overview = General introduciton to initial GPE project can be found [[http://hpc.desy.de/gpe/|here]]. = GPU Hardware = The current GPU system at DESY (Zeuthen) consists of a single [[http://www.supermicro.com/products/system/4U/7046/SYS-7046GT-TRF.cfm?GPU=FC4|server]] with dual nVidia Tesla C2050 GPU cards. It is hosted on gpu1 and is also used as a testbed for new developments in GPU-to-GPU networking with custom designed interconnects and InfiniBand. = Environment = Currently on the system the newest version of the [[http://developer.nvidia.com/cuda-toolkit-40|CUDA SDK 4.0]] alongside with device drivers and libraries are installed on gpu1. The Software Development Kit provides the following: * CUDA C/C++ Compiler * GPU Debugging & Profiling Tools * GPU-Accelerated Math Libraries * GPU-Accelerated Performance Primitives (Thrust library) * GPU-direct (under tests) = GPU Benchmarks = For evaluation and development a set of common benchmarks, as well as specially designed micro benchmarks were run on the gpu1 system. == Low-level benchmarks == Custom designed benchmarks use OpenMPI and OpenMP for task parallelization and allocation on the host CPUs and evaluate the following performance metrics: * Memory Bandwidth for unpinned memory and synchronous / asynchronous transfers Here the bandwidth of host to device memory copy operations is measured. The host memory areas used are allocated with common malloc() calls and are not pinned to physical page addresses - thus they are subject to page swapping. On the other hand memory is pinned when allocating it via the [[http://www.clear.rice.edu/comp422/resources/cuda/html/group__CUDART__MEMORY_g217d441a73d9304c6f0ccc22ec307dba.html|cudaHostAlloc()]] call. For differentiating between synchronous and asynchronous transfers we used [[http://developer.download.nvidia.com/compute/cuda/2_3/toolkit/docs/online/group__CUDART__MEMORY_g48efa06b81cc031b2aa6fdc2e9930741.html#g48efa06b81cc031b2aa6fdc2e9930741|cudaMemcpy]] and [[http://developer.download.nvidia.com/compute/cuda/2_3/toolkit/docs/online/group__CUDART__MEMORY_ge4366f68c6fa8c85141448f187d2aa13.html|cudaMemcpyasync]]. The effects of both transfer type and memory pinning are scribed [[http://www.ifh.de/~boyanov/GPE/notes-gpu.pdf|here]]. * Bandwidth of host-ot-GPU memory copy operations (for one and two devices parallel configured) Here again the bandwidth for host to device memory copy operations is measured. This time however the memory regions of the host memory are pinned to physical addresses via using the ||<45%>{{http://www.ifh.de/~boyanov/GPEfigures/GPEbench-parallel-stream-ALL-rank-1-tcpu.png|Measurement with gettimeofday() for process with rank 0 and two GPUs working in "parallel"|width=380}}||<45%>{{http://www.ifh.de/~boyanov/GPEfigures/GPEbench-parallel-stream-ALL-rank-1-tcuda.png|Measurement with RDTSC for process with rank 0 and two GPUs working in "parallel"|width=380}}|| ||<45%>{{http://www.ifh.de/~boyanov/GPEfigures/GPEbench-parallel-stream-ALL-rank-0-tcpu.png|Measurement with RDTSC for process with rank 0 and two GPUs working "crossed"|width=380}}||<45%>{{http://www.ifh.de/~boyanov/GPEfigures/GPEbench-parallel-stream-ALL-rank-0-tcuda.png|Measurement with RDTSC for process with rank 0 and two GPUs working "crossed"|width=380}}|| * Bandwidth of GPU-ot-host memory copy operations (for one and two devices cross configured) ||<45%>{{http://www.ifh.de/~boyanov/GPEfigures/GPEbench-cross-stream-ALL-rank-0-tcpu.png|ALD|width=380}}||<45%>{{http://www.ifh.de/~boyanov/GPEfigures/GPEbench-cross-stream-ALL-rank-0-tcuda.png|ALD|width=380}}|| ||<45%>{{http://www.ifh.de/~boyanov/GPEfigures/GPEbench-cross-stream-ALL-rank-1-tcpu.png|ALD|width=380}}||<45%>{{http://www.ifh.de/~boyanov/GPEfigures/GPEbench-cross-stream-ALL-rank-1-tcuda.png|ALD|width=380}}|| * Bandwidth and latency of GPU-to-GPU communication * mpirun options and rankfiles * MPI send/recv vs. CUDA 4.0 peer-to-peer communication primitives * GPU-to-InfiniBand hardware datapath propagation delay * perftest measurements = GPU Applications = /!\ Recent applications utilizing the gpu1 system at DESY (Zeuthen) are [[http://usqcd.jlab.org/usqcd-docs/chroma/|Chroma]]-based LQCD munerical simulationd and applications from the field of Astro Particle Physics. == Application-level benchmarks == For ensuring consistency of performance with real-world applications, the Scalable HeterOgeneous Computing ([[http://ft.ornl.gov/doku/shoc/start|SHOC]]) benchmark suite was run on the gpu1 system too. This benchmark suite gives not only benchmark results for CUDA implementations of key algorithms but also corresponding implementations of the more general OpenCL parallel programming framework. == Debugger and Profiler Tools == * Compute Visual profiler * CUDA Debugger = Monitoring = /!\ * gpu1 in Nagios * GPU cores temperature * host and device free and/or used memory * GPU core frequency * GPU cores utilization/load ---- Sections marked with /!\ need further discussion