ATF: A generic directive‐based auto‐tuning framework

Rasch, Ari; Gorlatch, Sergei

doi:10.1002/cpe.4423

Cited by 25 publications

(21 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Our Kernel Tuning Toolkit focuses on tuning of userdefined code optimization parameters. Most similar to our work are CLTune [35], AUMA [13], ATF [36], and Kernel Tuner [46], which are problem domain-agnostic autotuners designed for heterogeneous computing.…”

Section: Related Workmentioning

confidence: 85%

“…They may select one of the predefined variants of a tuned function [34], or generate and compile implementations according to the values of the tuning parameters. We distinguish between compiler-based tuning, where the space of code transformation is generated automatically [37,43,39] and user-defined code optimization parameters autotuning [35,13,36,15]. User-defined code optimization parameters tuning requires expert programmers to identify and implement tuning possibilities in the source code manually (e. g., by using preprocessor macros).…”

Section: Related Workmentioning

confidence: 99%

“…To the best of our knowledge, there is no comprehensive benchmark set for code optimization parameters tuning in heterogeneous computing. In [35,36], two benchmarks are used to evaluate code optimization parameters tuning: GEMM and 2D convolutions. Those benchmarks are also used in our benchmark set.…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

A benchmark set of highly-efficient CUDA and OpenCL kernels and its dynamic autotuning with Kernel Tuning Toolkit

Petrovič

Střelák

Hozzová

et al. 2020

Future Generation Computer Systems

View full text Add to dashboard Cite

In recent years, the heterogeneity of both commodity and supercomputers hardware has increased sharply. Accelerators, such as GPUs or Intel Xeon Phi co-processors, are often key to improving speed and energy efficiency of highly-parallel codes. However, due to the complexity of heterogeneous architectures, optimization of codes for a certain type of architecture as well as porting codes across different architectures, while maintaining a comparable level of performance, can be extremely challenging. Addressing the challenges associated with performance optimization and performance portability, autotuning has gained a lot of interest. Autotuning of performance-relevant source-code parameters allows to automatically tune applications without hard coding optimizations and thus helps with keeping the performance portable. In this paper, we introduce a benchmark set of ten autotunable kernels for important computational problems implemented in OpenCL or CUDA. Using our Kernel Tuning Toolkit, we show that with autotuning most of the kernels reach near-peak performance on various GPUs and outperform baseline implementations on CPUs and Xeon Phis. Our evaluation also demonstrates that autotuning is key to performance portability. In addition to offline tuning, we also introduce dynamic autotuning of code optimization parameters during application runtime. With dynamic tuning, the Kernel Tuning Toolkit enables applications to re-tune performance-critical kernels at runtime whenever needed, for example, when input data changes. Although it is generally believed that autotuning spaces tend to be too large to be searched during application runtime, we show that it is not necessarily the case when tuning spaces are designed rationally. Many of our kernels reach near peak-performance with moderately sized tuning spaces that can be searched at runtime with acceptable overhead. Finally we demonstrate, how dynamic performance tuning can be integrated into a real-world application from cryo-electron microscopy domain.

show abstract

Section: Related Workmentioning

confidence: 85%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

A benchmark set of highly-efficient CUDA and OpenCL kernels and its dynamic autotuning with Kernel Tuning Toolkit

Petrovič

Střelák

Hozzová

et al. 2020

Future Generation Computer Systems

View full text Add to dashboard Cite

show abstract

“…Listing 3 shows two Fireiron strategies expressing the maxwell reference (top) and the wmma reference (bottom). In the maxwell-schedule for example, we use different strategies for moving A (lines [16][17][18][19][20] and B (lines [22][23][24][25][26] to shared memory because considering the storage layouts separately enables coalesced global memory loads for both operands. We use swizzling (line 2) [20], and specify which loops to unroll and where to add or avoid synchronization with refinements.…”

Section: Discussionmentioning

confidence: 99%

Fireiron

Hagedorn

Elliott²,

Barthels

et al. 2020

Proceedings of the ACM International Conference on Parallel Architectures and Compilation Techniques

View full text Add to dashboard Cite

High GPU performance can only be achieved if a kernel efficiently uses the multi-layered compute and memory hierarchies. For example, accelerators such as NVIDIA's Tensor Cores require specific mappings of threads to data that must be considered in data movements to and from registers. Current compilers struggle to match the performance of vendor libraries like cuBLAS, which are developed by experts in assembly. This manual low-level coding is time-consuming and complicates to unlock the full GPU potential, preventing experimentation to achieve even higher performance. In this paper we introduce Fireiron, a scheduling language aimed at performance experts. Fireiron provides high-level abstractions for expressing GPU optimizations that are unavailable to compilers today and which so far must be written in assembly. Our innovation is that both computations and data movements are first class concepts that can be separately mapped to threads, as required for the efficient use of specialized hardware like Tensor Cores. We evaluate Fireiron on three GPU architectures against expertwritten advanced matrix multiplications. First, we show that Fireiron schedules are able to express the strategies of these implementations requiring about 6× less lines of code. Second, we show that the code generated by Fireiron schedules outperforms the fastest implementations (provided by cuBLAS) by more than 2×.

show abstract

“…Rasch and Gorlatch propose a simple‐to‐use framework for automatic program optimization by choosing the most suitable values of program parameters, allowing to tune a broad class of applications written in different programming languages.…”

Section: In This Issuementioning

confidence: 99%