Maciej Besta

Slim fly: a cost effective low-diameter network topology

We introduce a high-performance cost-effective network topology called Slim Fly that approaches the theoretically optimal network diameter. Slim Fly is based on graphs that approximate the solution to the degree-diameter problem. We analyze Slim Fly and compare it to both traditional and state-of the-art networks. Our analysis shows that Slim Fly has significant advantages over other topologies in latency, bandwidth, resiliency, cost, and power consumption. Finally, we propose deadlock-free routing schemes and physical layouts for large computing centres as well as a detailed cost and power model. Slim Fly enables constructing cost effective and highly resilient data enter and HPC networks that offer low latency and high bandwidth under different HPC workloads such as stencil or graph computations.

Enabling Highly-Scalable Remote Memory Access Programming with MPI-3 One Sided

Modern interconnects offer remote direct memory access (RDMA) features. Yet, most applications rely on explicit message passing for communications albeit their unwanted overheads. The MPI-3.0 standard defines a programming interface for exploiting RDMA networks directly, however, its scalability and practicability has to be demonstrated in practice. In this work, we develop scalable bufferless protocols that implement the MPI-3.0 specification. Our protocols support scaling to millions of cores with negligible memory consumption while providing highest performance and minimal overheads. To arm programmers, we provide a spectrum of performance models for all critical functions and demonstrate the usability of our library and models with several application studies with up to half a million processes. We show that our design is comparable to, or better than UPC and Fortran Coarrays in terms of latency, bandwidth and message rate. We also demonstrate application performance improvements with comparable programming complexity.

Enabling highly-scalable remote memory access programming with MPI-3 one sided

Modern interconnects offer remote direct memory access (RDMA) features. Yet, most applications rely on explicit message passing for communications albeit their unwanted overheads. The MPI-3.0 standard defines a programming interface for exploiting RDMA networks directly, however, its scalability and practicability has to be demonstrated in practice. In this work, we develop scalable bufferless protocols that implement the MPI-3.0 specification. Our protocols support scaling to millions of cores with negligible memory consumption while providing highest performance and minimal overheads. To arm programmers, we provide a spectrum of performance models for all critical functions and demonstrate the usability of our library and models with several application studies with up to half a million processes. We show that our design is comparable to, or better than UPC and Fortran Coarrays in terms of latency, bandwidth, and message rate. We also demonstrate application performance improvements with comparable programming complexity.

Programming Abstractions for Data Locality

The goal of the workshop and this report is to identify common themes and standardize concepts for locality-preserving abstractions for exascale programming models. Current software tools are built on the premise that computing is the most expensive component, we are rapidly moving to an era that computing is cheap and massively parallel while data movement dominates energy and performance costs. In order to respond to exascale systems (the next generation of high performance computing systems), the scientific computing community needs to refactor their applications to align with the emerging data-centric paradigm. Our applications must be evolved to express information about data locality. Unfortunately current programming environments offer few ways to do so. They ignore the incurred cost of communication and simply rely on the hardware cache coherency to virtualize data movement. With the increasing importance of task-level parallelism on future systems, task models have to support constructs that express data locality and affinity. At the system level, communication libraries implicitly assume all the processing elements are equidistant to each other. In order to take advantage of emerging technologies, application developers need a set of programming abstractions to describe data locality for the new computing ecosystem. The new programming paradigm should be more data centric and allow to describe how to decompose and how to layout data in the memory.Fortunately, there are many emerging concepts such as constructs for tiling, data layout, array views, task and thread affinity, and topology aware communication libraries for managing data locality. There is an opportunity to identify commonalities in strategy to enable us to combine the best of these concepts to develop a comprehensive approach to expressing and managing data locality on exascale programming systems. These programming model abstractions can expose crucial information about data locality to the compiler and runtime system to enable performance-portable code. The research question is to identify the right level of abstraction, which includes techniques that range from template libraries all the way to completely new languages to achieve this goal.

Evaluating the Cost of Atomic Operations on Modern Architectures

Atomic operations (atomics) such as Compare-and-Swap (CAS) or Fetch-and-Add (FAA) are ubiquitous in parallel programming. Yet, performance tradeoffs between these operations and various characteristics of such systems, such as the structure of caches, are unclear and have not been thoroughly analyzed. In this paper we establish an evaluation methodology, develop a performance model, and present a set of detailed benchmarks for latency and bandwidth of different atomics. We consider various state-of-the-art x86 architectures: Intel Haswell, Xeon Phi, Ivy Bridge, and AMD Bulldozer. The results unveil surprising performance relationships between the considered atomics and architectural properties such as the coherence state of the accessed cache lines. One key finding is that all the tested atomics have comparable latency and bandwidth even if they are characterized by different consensus numbers. Another insight is that the design of atomics prevents any instruction-level parallelism even if there are no dependencies between the issued operations. Finally, we discuss solutions to the discovered performance issues in the analyzed architectures. Our analysis can be used for making better design and algorithmic decisions in parallel programming on various architectures deployed in both off-the-shelf machines and large compute systems.

Fault tolerance for remote memory access programming models

Remote Memory Access (RMA) is an emerging mechanism for programming high-performance computers and datacenters. However, little work exists on resilience schemes for RMA-based applications and systems. In this paper we analyze fault tolerance for RMA and show that it is fundamentally different from resilience mechanisms targeting the message passing (MP) model. We design a model for reasoning about fault tolerance for RMA, addressing both flat and hierarchical hardware. We use this model to construct several highly-scalable mechanisms that provide efficient low-overhead in-memory checkpointing, transparent logging of remote memory accesses, and a scheme for transparent recovery of failed processes. Our protocols take into account diminishing amounts of memory per core, one of the major features of future exascale machines. The implementation of our fault-tolerance scheme entails negligible additional overheads. Our reliability model shows that in-memory checkpointing and logging provide high resilience. This study enables highly-scalable resilience mechanisms for RMA and fills a research gap between fault tolerance and emerging RMA programming models.

To Push or To Pull: On Reducing Communication and Synchronization in Graph Computations

We reduce the cost of communication and synchronization in graph processing by analyzing the fastest way to process graphs: pushing the updates to a shared state or pulling the updates to a private state. We investigate the applicability of this push-pull dichotomy to various algorithms and its impact on complexity, performance, and the amount of used locks, atomics, and reads/writes. We consider 11 graph algorithms, 3 programming models, 2 graph abstractions, and various families of graphs. The conducted analysis illustrates surprising differences between push and pull variants of different algorithms in performance, speed of convergence, and code complexity; the insights are backed up by performance data from hardware counters. We use these findings to illustrate which variant is faster for each algorithm and to develop generic strategies that enable even higher speedups. Our insights can be used to accelerate graph processing engines or libraries on both massively-parallel shared-memory machines as well as distributed-memory systems.

SlimSell: A Vectorizable Graph Representation for Breadth-First Search

Vectorization and GPUs will profoundly change graph processing. Traditional graph algorithms tuned for 32- or 64-bit based memory accesses will be inefficient on architectures with 512-bit wide (or larger) instruction units that are already present in the Intel Knights Landing (KNL) manycore CPU. Anticipating this shift, we propose SlimSell: a vectorizable graph representation to accelerate Breadth-First Search (BFS) based on sparse-matrix dense-vector (SpMV) products. SlimSell extends and combines the state-of-the-art SIMD-friendly Sell-C-σ matrix storage format with tropical, real, boolean, and sel-max semiring operations. The resulting design reduces the necessary storage (by up to 50%) and thus pressure on the memory subsystem. We augment SlimSell with the SlimWork and SlimChunk schemes that reduce the amount of work and improve load balance, further accelerating BFS. We evaluate all the schemes on Intel Haswell multicore CPUs, the state-of-the-art Intel Xeon Phi KNL manycore CPUs, and NVIDIA Tesla GPUs. Our experiments indicate which semiring offers highest speedups for BFS and illustrate that SlimSell accelerates a tuned Graph500 BFS code by up to 33%. This work shows that vectorization can secure high-performance in BFS based on SpMV products; the proposed principles and designs can be extended to other graph algorithms.

Active Access: A Mechanism for High-Performance Distributed Data-Centric Computations

Remote memory access (RMA) is an emerging high-performance programming model that uses RDMA hardware directly. Yet, accessing remote memories cannot invoke activities at the target which complicates implementation and limits performance of data-centric algorithms. We propose Active Access (AA), a mechanism that integrates well-known active messaging (AM) semantics with RMA to enable high-performance distributed data-centric computations. AA supports a new programming model where the user specifies handlers that are triggered when incoming puts and gets reference designated addresses. AA is based on a set of extensions to the Input/Output Memory Management Unit (IOMMU), a unit that provides high-performance hardware support for remapping I/O accesses to memory. We illustrate that AA outperforms existing AM and RMA designs, accelerates various codes such as distributed hashtables or logging schemes, and enables new protocols such as incremental checkpointing for RMA. We also discuss how extended IOMMUs can support a virtualized global address space in a distributed system that offers features known from on-node memory virtualization. We expect that AA and other IOMMU features can enhance the design of HPC operating and runtime systems in large computing centers.

Communication-avoiding parallel minimum cuts and connected components

We present novel scalable parallel algorithms for finding global minimum cuts and connected components, which are important and fundamental problems in graph processing. To take advantage of future massively parallel architectures, our algorithms are communication-avoiding: they reduce the costs of communication across the network and the cache hierarchy. The fundamental technique underlying our work is the randomized sparsification of a graph: removing a fraction of graph edges, deriving a solution for such a sparsified graph, and using the result to obtain a solution for the original input. We design and implement sparsification with O(1) synchronization steps. Our global minimum cut algorithm decreases communication costs and computation compared to the state-of-the-art, while our connected components algorithm incurs few cache misses and synchronization steps. We validate our approach by evaluating MPI implementations of the algorithms on a petascale supercomputer. We also provide an approximate variant of the minimum cut algorithm and show that it approximates the exact solutions well while using a fraction of cores in a fraction of time.