skip to main content


Title: Evaluating Gather and Scatter Performance on CPUs and GPUs
This paper describes a new benchmark tool, Spatter, for assessing memory system architectures in the context of a specific category of indexed accesses known as gather and scatter. These types of operations are increasingly used to express sparse and irregular data access patterns, and they have widespread utility in many modern HPC applications including scientific simulations, data mining and analysis computations, and graph processing. However, many traditional benchmarking tools like STREAM, STRIDE, and GUPS focus on characterizing only uniform stride or fully random accesses despite evidence that modern applications use varied sets of more complex access patterns. Spatter is an open-source benchmark that provides a tunable and configurable framework to benchmark a variety of indexed access patterns, including variations of gather / scatter that are seen in HPC mini-apps evaluated in this work. The design of Spatter includes backends for OpenMP and CUDA, and experiments show how it can be used to evaluate 1) uniform access patterns for CPU and GPU, 2) prefetching regimes for gather / scatter, 3) compiler implementations of vectorization for gather / scatter, and 4) trace-driven "proxy patterns" that reflect the patterns found in multiple applications. The results from Spatter experiments show, for instance, that GPUs typically outperform CPUs for these operations in absolute bandwidth but not fraction of peak bandwidth, and that Spatter can better represent the performance of some cache-dependent mini-apps than traditional STREAM bandwidth measurements.  more » « less
Award ID(s):
1710371
NSF-PAR ID:
10199354
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
The International Symposium on Memory Systems (MEMSYS 2020)
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. null (Ed.)
    As multicore systems continue to grow in scale and on-chip memory capacity, the on-chip network bandwidth and latency become problematic bottlenecks. Because of this, overheads in data transfer, the coherence protocol and replacement policies become increasingly important. Unfortunately, even in well-structured programs, many natural optimizations are difficult to implement because of the reactive and centralized nature of traditional cache hierarchies, where all requests are initiated by the core for short, cache line granularity accesses. For example, long-lasting access patterns could be streamed from shared caches without requests from the core. Indirect memory access can be performed by chaining requests made from within the cache, rather than constantly returning to the core. Our primary insight is that if programs can embed information about long-term memory stream behavior in their ISAs, then these streams can be floated to the appropriate level of the memory hierarchy. This decentralized approach to address generation and cache requests can lead to better cache policies and lower request and data traffic by proactively sending data before the cores even request it. To evaluate the opportunities of stream floating, we enhance a tiled multicore cache hierarchy with stream engines to process stream requests in last-level cache banks. We develop several novel optimizations that are facilitated by stream exposure in the ISA, and subsequent exposure to caches. We evaluate using a cycle-level execution-driven gem5-based simulator, using 10 data-processing workloads from Rodinia and 2 streaming kernels written in OpenMP. We find that stream floating enables 52% and 39% speedup over an inorder and OOO core with state of art prefetcher design respectively, with 64% and 49% energy efficiency advantage. 
    more » « less
  2. Many applications are increasingly becoming I/O-bound. To improve scalability, analytical models of parallel I/O performance are often consulted to determine possible I/O optimizations. However, I/O performance modeling has predominantly focused on applications that directly issue I/O requests to a parallel file system or a local storage device. These I/O models are not directly usable by applications that access data through standardized I/O libraries, such as HDF5, FITS, and NetCDF, because a single I/O request to an object can trigger a cascade of I/O operations to different storage blocks. The I/O performance characteristics of applications that rely on these libraries is a complex function of the underlying data storage model, user-configurable parameters and object-level access patterns. As a consequence, I/O optimization is predominantly an ad-hoc process that is performed by application developers, who are often domain scientists with limited desire to delve into nuances of the storage hierarchy of modern computers.This paper presents an analytical cost model to predict the end-to-end execution time of applications that perform I/O through established array management libraries. The paper focuses on the HDF5 and Zarr array libraries, as examples of I/O libraries with radically different storage models: HDF5 stores every object in one file, while Zarr creates multiple files to store different objects. We find that accessing array objects via these I/O libraries introduces new overheads and optimizations. Specifically, in addition to I/O time, it is crucial to model the cost of transforming data to a particular storage layout (memory copy cost), as well as model the benefit of accessing a software cache. We evaluate the model on real applications that process observations (neuroscience) and simulation results (plasma physics). The evaluation on three HPC clusters reveals that I/O accounts for as little as 10% of the execution time in some cases, and hence models that only focus on I/O performance cannot accurately capture the performance of applications that use standard array storage libraries. In parallel experiments, our model correctly predicts the fastest storage library between HDF5 and Zarr 94% of the time, in contrast with 70% of the time for a cutting-edge I/O model. 
    more » « less
  3. Today's high-performance computing (HPC) applications are producing vast volumes of data, which are challenging to store and transfer efficiently during the execution, such that data compression is becoming a critical technique to mitigate the storage burden and data movement cost. Huffman coding is arguably the most efficient Entropy coding algorithm in information theory, such that it could be found as a fundamental step in many modern compression algorithms such as DEFLATE. On the other hand, today's HPC applications are more and more relying on the accelerators such as GPU on supercomputers, while Huffman encoding suffers from low throughput on GPUs, resulting in a significant bottleneck in the entire data processing. In this paper, we propose and implement an efficient Huffman encoding approach based on modern GPU architectures, which addresses two key challenges: (1) how to parallelize the entire Huffman encoding algorithm, including codebook construction, and (2) how to fully utilize the high memory-bandwidth feature of modern GPU architectures. The detailed contribution is four-fold. (1) We develop an efficient parallel codebook construction on GPUs that scales effectively with the number of input symbols. (2) We propose a novel reduction based encoding scheme that can efficiently merge the codewords on GPUs. (3) We optimize the overall GPU performance by leveraging the state-of-the-art CUDA APIs such as Cooperative Groups. (4) We evaluate our Huffman encoder thoroughly using six real-world application datasets on two advanced GPUs and compare with our implemented multi-threaded Huffman encoder. Experiments show that our solution can improve the encoding throughput by up to 5.0x and 6.8x on NVIDIA RTX 5000 and V100, respectively, over the state-of-the-art GPU Huffman encoder, and by up to 3.3x over the multi-thread encoder on two 28-core Xeon Platinum 8280 CPUs. 
    more » « less
  4. The continued growth in the processing power of FPGAs coupled with high bandwidth memories (HBM), makes systems like the Xilinx U280 credible platforms for linear solvers which often dominate the run time of scientific and engineering applications. In this paper, we present Callipepla, an accelerator for a preconditioned conjugate gradient linear solver (CG). FPGA acceleration of CG faces three challenges: (1) how to support an arbitrary problem and terminate acceleration processing on the fly, (2) how to coordinate long-vector data flow among processing modules, and (3) how to save off-chip memory bandwidth and maintain double (FP64) precision accuracy. To tackle the three challenges, we present (1) a stream-centric instruction set for efficient streaming processing and control, (2) vector streaming reuse (VSR) and decentralized vector flow scheduling to coordinate vector data flow among modules and further reduce off-chip memory access latency with a double memory channel design, and (3) a mixed precision scheme to save bandwidth yet still achieve effective double precision quality solutions. To the best of our knowledge, this is the first work to introduce the concept of VSR for data reusing between on-chip modules to reduce unnecessary off-chip accesses and enable modules working in parallel for FPGA accelerators. We prototype the accelerator on a Xilinx U280 HBM FPGA. Our evaluation shows that compared to the Xilinx HPC product, the XcgSolver, Callipepla achieves a speedup of 3.94×, 3.36× higher throughput, and 2.94× better energy efficiency. Compared to an NVIDIA A100 GPU which has 4× the memory bandwidth of Callipepla, we still achieve 77% of its throughput with 3.34× higher energy efficiency. The code is available at https://github.com/UCLA-VAST/Callipepla. 
    more » « less
  5. Because of severe limitations in technology scaling, architects have innovated in specializing general purpose processors for computation primitives (e.g. vector instructions, loop accelerators). The general principle is exposing rich semantics to the ISA. An opportunity to explore is whether richer semantics of memory access patterns could also be used to improve the efficiency of memory and communication. Two important open questions are how to convey higher level memory information and how to take advantage of this information in hardware. We find that a majority of memory accesses follow a small number of simple patterns; we term these streams (e.g. affine, indirect). Streams can often be decoupled from core execution, and patterns persist long enough to express useful behavior. Our approach is therefore to express streams as ISA primitives, which we argue can enable: prefetch stream accesses to hide memory latency, semi-binding decoupled access to remove address computation and optimize the memory interface, and finally inform cache policies. In this work, we propose ISA-extensions for decoupled-streams, which interact with the core using a FIFO-based interface. We implement optimizations for each of the aforementioned opportunities on an aggressive wide-issue OOO core and evaluate with SPEC CPU 2017 and CortexSuite[1, 2]. Across all workloads, we observe about 1.37× speedup and energy efficiency improvement over hardware stride prefetching. 
    more » « less