PENCIL: A Platform-Neutral Compute Intermediate Language for Accelerator Programming

Baghdadi, Riyadh; Beaugnon, Ulysse; Cohen, Albert; Grosser, Tobias; Kruse, Michael; Reddy, Chandan; Verdoolaege, Sven; Betts, Adam; Donaldson, Alastair F.; Ketema, Jeroen; Absar, Javed; Haastregt, Sven van; Kravets, Alexey; Lokhmotov, Anton; Dávid, Róbert; Hajiyev, Elnar

doi:10.1109/pact.2015.17

Cited by 81 publications

(86 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Figure 1 gives an overview of the different approaches related to this paper. Loop-based autoparallelization techniques have been extensively studied for languages like C. Recent work on polyhedral compilation [1] for instance has pushed the boundaries of such techniques for GPU code generation. However, these techniques only operate on loops and requires certain property such as affine indices to work effectively.…”

Section: Related Work and Motivationmentioning

confidence: 99%

See 1 more Smart Citation

LIFT: A functional data-parallel IR for high-performance GPU code generation

Steuwer

Remmelg

Dubach

2017

2017 IEEE/ACM International Symposium on Code Generation and Optimization (CGO)

102

113

View full text Add to dashboard Cite

Parallel patterns (e.g., map, reduce) have gained traction as an abstraction for targeting parallel accelerators and are a promising answer to the performance portability problem. However, compiling high-level programs into efficient lowlevel parallel code is challenging. Current approaches start from a high-level parallel IR and proceed to emit GPU code directly in one big step. Fixed strategies are used to optimize and map parallelism exploiting properties of a particular GPU generation leading to performance portability issues.We introduce the Lift IR, a new data-parallel IR which encodes OpenCL-specific constructs as functional patterns. Our prior work has shown that this functional nature simplifies the exploration of optimizations and mapping of parallelism from portable high-level programs using rewrite-rules. This paper describes how Lift IR programs are compiled into efficient OpenCL code. This is non-trivial as many performance sensitive details such as memory allocation, array accesses or synchronization are not explicitly represented in the Lift IR. We present techniques which overcome this challenge by exploiting the pattern's high-level semantics. Our evaluation shows that the Lift IR is flexible enough to express GPU programs with complex optimizations achieving performance on par with manually optimized code.

show abstract

Section: Related Work and Motivationmentioning

confidence: 99%

“…map parallelism + optimization [19] auto parallelization [1] C Code for loops GPU code GPU code GPU code Figure 1. GPU code generation landscape.…”

Section: Gpu Codementioning

confidence: 99%

LIFT: A functional data-parallel IR for high-performance GPU code generation

Steuwer

Remmelg

Dubach

2017

2017 IEEE/ACM International Symposium on Code Generation and Optimization (CGO)

102

113

View full text Add to dashboard Cite

show abstract

“…An annotation approach is described in [6], based on the Platform-Neutral Compute Intermediate Language [4]. This used the code generator in [35] to generate CUDA and OpenCL code for multiple compute platforms.…”

Section: Related Workmentioning

confidence: 99%

Discovery and exploitation of general reductions: A constraint based approach

Ginsbach¹,

O’Boyle²

2017

2017 IEEE/ACM International Symposium on Code Generation and Optimization (CGO)

View full text Add to dashboard Cite

Discovering and exploiting scalar reductions in programs has been studied for many years. The discovery of more complex reduction operations has, however, received less attention. Such reductions contain compile-time unknown parameters, indirect memory accesses and dynamic control flow, which are challenging for existing approaches.In this paper we develop a new compiler based approach that automatically detects a wide class of reductions. This is based on a constraint formulation of the reduction idiom and has been implemented as an LLVM pass. We use a custom constraint solver to identify program subsets that adhere to the constraint specification. Once discovered, we automatically generate parallel code to exploit the reduction.This approach is robust and was evaluated on C versions of well known benchmark suites: NAS, Parboil and Rodinia. We detected 84 scalar reductions and 6 histograms, outperforming existing approaches. We show that the exploitation of histograms gives significant performance improvement.

show abstract

“…Pencil [3] is an intermediate language defined as a restricted subset of C99. It is intended as an implementation language for libraries and a compilation target by DSLs.…”

Section: Polyhedral Gpu Compilationmentioning

confidence: 99%

Performance portable GPU code generation for matrix multiplication

Remmelg

Lutz

Steuwer

et al. 2016

Proceedings of the 9th Annual Workshop on General Purpose Processing Using Graphics Processing Unit

View full text Add to dashboard Cite

Parallel accelerators such as GPUs are notoriously hard to program; exploiting their full performance potential is a job best left for ninja programmers. High-level programming languages coupled with optimizing compilers have been proposed to attempt to address this issue. However, they rely on device-specific heuristics or hard-coded library implementations to achieve good performance resulting in non-portable solutions that need to be re-optimized for every new device.Achieving performance portability is the holy grail of high-performance computing and has so far remained an open problem even for well studied applications like matrix multiplication. We argue that what is needed is a way to describe applications at a high-level without committing to particular implementations. To this end, we developed in a previous paper a functional data-parallel language which allows applications to be expressed in a device neutral way. We use a set of well-defined rewrite rules to automatically transform programs into semantically equivalent devicespecific forms, from which OpenCL code is generated.In this paper, we demonstrate how this approach produces high-performance OpenCL code for GPUs with a wellstudied, well-understood application: matrix multiplication. Starting from a single high-level program, our compiler automatically generate highly optimized and specialized implementations. We group simple rewrite rules into more complex macro-rules, each describing a well-known optimization like tiling and register blocking in a composable way. Using an exploration strategy our compiler automatically generates 50,000 OpenCL kernels, each providing a differently optimized -but provably correct -implementation of matrix multiplication. The automatically generated code offers competitive performance compared to the manually tuned MAGMA library implementations of matrix multiplication on Nvidia and even outperforms AMD's clBLAS library.

show abstract

PENCIL: A Platform-Neutral Compute Intermediate Language for Accelerator Programming

Cited by 81 publications

References 16 publications

LIFT: A functional data-parallel IR for high-performance GPU code generation

LIFT: A functional data-parallel IR for high-performance GPU code generation

Discovery and exploitation of general reductions: A constraint based approach

Performance portable GPU code generation for matrix multiplication

Contact Info

Product

Resources

About