Hardware Performance Variation: A Comparative Study Using Lightweight Kernels

Weisbach, Hannes; Gerofi, Balazs; Kocoloski, Brian; Härtig, Hermann; Ishikawa, Yutaka

doi:10.1007/978-3-319-92040-5_13

Cited by 10 publications

(5 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…1 As desynchronization can be initiated by system or application noise, this insight sheds new light on the vast amount of research that investigated ways to mitigate the impact of noise on the performance of primitives and algorithms. [17][18][19][20][21][22][23][24][25] Figure 2 shows the range of possible scenarios. While scenario (b), that is, overlapping of loop kernel execution with communication, was covered in Reference 1, and scenario (c) is well described by the ECM performance model for multicore processors, 6 scenario (a) was not studied before.…”

Section: Relevance and Generalization Of Bandwidth Sharing Phenomenonmentioning

confidence: 99%

“…Hence, desynchronization is far from irrelevant in parallel programs and sometimes beneficial as it can provoke automatic communication overlap and an according performance gain 1 . As desynchronization can be initiated by system or application noise, this insight sheds new light on the vast amount of research that investigated ways to mitigate the impact of noise on the performance of primitives and algorithms 17‐25 …”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Analytic performance model for parallel overlapping memory‐bound kernels

Afzal

Hager

Wellein

2022

Concurrency and Computation

View full text Add to dashboard Cite

Complex applications running on multicore processors show a rich performance phenomenology. The growing number of cores per ccNUMA domain complicates performance analysis of memory-bound code since system noise, load imbalance, or task-based programming models can lead to thread desynchronization. Hence, the simplifying assumption that all cores execute the same loop can not be upheld. Motivated by observations on plain and modified versions of the HPCG benchmark, we construct a performance model of execution of memory-bound loop kernels. It can predict the memory bandwidth share per kernel on a memory contention domain depending on the number of active cores and which other workload the kernel is paired with.The only code features required are the single-thread memory request fraction per kernel, which is directly related to the single-thread memory bandwidth, and its saturated bandwidth. The former can either be measured directly or predicted using the Execution-Cache-Memory performance model. The computational intensity of the kernels and the detailed structure of the code is of no significance. We validate our model on Intel Broadwell, Intel Cascade Lake, and AMD Rome processors pairing various streaming and stencil kernels. The error in predicting the bandwidth share per kernel is less than 8%.

show abstract

Section: Relevance and Generalization Of Bandwidth Sharing Phenomenonmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Analytic performance model for parallel overlapping memory‐bound kernels

Afzal

Hager

Wellein

2022

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…Extensive research [13], [14], [15], [16], [17], [18], [19], [20], [21] has been conducted for almost two decades to characterize noise, identify sources of noise outside of the control of the application, and pinpoint its influence on collective operations. Explicit techniques for asynchronous communication, mitigation of noise, MPI process placement, dynamic load balancing, synchronization of OS influence, lightweight OS kernels, etc., have been explored.…”

Section: Noisementioning

confidence: 99%

The Role of Idle Waves, Desynchronization, and Bottleneck Evasion in the Performance of Parallel Programs

Afzal¹,

Hager²,

Wellein³

2022

Preprint

View full text Add to dashboard Cite

The performance of highly parallel applications on distributed-memory systems is influenced by many factors. Analytic performance modeling techniques aim to provide insight into performance limitations and are often the starting point of optimization efforts. However, coupling analytic models across the system hierarchy (socket, node, network) fails to encompass the intricate interplay between the program code and the hardware, especially when execution and communication bottlenecks are involved. In this paper we investigate the effect of bottleneck evasion and how it can lead to automatic overlap of communication overhead with computation. Bottleneck evasion leads to a gradual loss of the initial bulk-synchronous behavior of a parallel code so that its processes become desynchronized. This occurs most prominently in memory-bound programs, which is why we choose memory-bound benchmark and application codes, specifically an MPI-augmented STREAM Triad, sparse matrix-vector multiplication, and a collective-avoiding Chebyshev filter diagonalization code to demonstrate the consequences of desynchronization on two different supercomputing platforms. We investigate the role of idle waves as possible triggers for desynchronization and show the impact of automatic asynchronous communication for a spectrum of code properties and parameters, such as saturation point, matrix structures, domain decomposition, and communication concurrency. Our findings reveal how eliminating synchronization points (such as collective communication or barriers) precipitates performance improvements that go beyond what can be expected by simply subtracting the overhead of the collective from the overall runtime.

show abstract

“…Global properties of such waves like damping and velocity, or the interaction with memory-bound code, were ignored, however. There is a vast body of research that targets the characterization of noise as well as its mitigation via explicit techniques, such as dynamic load balancing, MPI process placement, synchronization of OS influence, and lightweight OS kernels [3,14,16,20]. In contrast, the present paper investigates the favorable consequences of noise as an enabling factor for desynchronization and -in case of memory-bound code -automatic partial overlap of communication and computation.…”

Section: Related Workmentioning

confidence: 99%

Desynchronization and Wave Pattern Formation in MPI-Parallel and Hybrid Memory-Bound Programs

Afzal¹,

Hager²,

Wellein³

2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Analytic, first-principles performance modeling of distributed-memory parallel codes is notoriously imprecise. Even for applications with extremely regular and homogeneous computecommunicate phases, simply adding communication time to computation time does often not yield a satisfactory prediction of parallel runtime due to deviations from the expected simple lockstep pattern caused by system noise, variations in communication time, and inherent load imbalance. In this paper, we highlight the specific cases of provoked and spontaneous desynchronization of memory-bound, bulk-synchronous pure MPI and hybrid MPI+OpenMP programs. Using simple microbenchmarks we observe that although desynchronization can introduce increased waiting time per process, it does not necessarily cause lower resource utilization but can lead to an increase in available bandwidth per core. In case of significant communication overhead, even natural noise can shove the system into a state of automatic overlap of communication and computation, improving the overall time to solution. The saturation point, i.e., the number of processes per memory domain required to achieve full memory bandwidth, is pivotal in the dynamics of this process and the emerging stable wave pattern. We also demonstrate how hybrid MPI-OpenMP programming can prevent desirable desynchronization by eliminating the bandwidth bottleneck among processes. A Chebyshev filter diagonalization application is used to demonstrate some of the observed effects in a realistic setting.

show abstract

Hardware Performance Variation: A Comparative Study Using Lightweight Kernels

Cited by 10 publications

References 33 publications

Analytic performance model for parallel overlapping memory‐bound kernels

Analytic performance model for parallel overlapping memory‐bound kernels

The Role of Idle Waves, Desynchronization, and Bottleneck Evasion in the Performance of Parallel Programs

Desynchronization and Wave Pattern Formation in MPI-Parallel and Hybrid Memory-Bound Programs

Contact Info

Product

Resources

About