Beyond Do Loops: Data Transfer Generation with Convex Array Regions

Guelton, Serge; Amini, Mehdi; Creusillet, Béatrice

doi:10.1007/978-3-642-37658-0_17

Cited by 6 publications

(3 citation statements)

References 17 publications

(17 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar results with data reuse between two successive tiles only were then demonstrated for AutoESL Xilinx tool [29]. Different (and more restricted) forms of inter-tile data reuse were also designed for programmable accelerators such as GPUs [5,18,35]. None of these approaches are parametric w.r.t.…”

Section: Introductionmentioning

confidence: 74%

See 1 more Smart Citation

Exact and Approximated Data-Reuse Optimizations for Tiling with Parametric Sizes

Darte

Isoard

2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Loop tiling is a loop transformation widely used to improve spatial and temporal data locality, to increase computation granularity, and to enable blocking algorithms, which are particularly useful when offloading kernels on computing units with smaller memories. When caches are not available or used, data transfers and local storage must be software-managed, and some useless remote communications can be avoided by exploiting data reuse between tiles. An important parameter of tiling is the sizes of the tiles, which impact the size of the required local memory. However, for most analyses involving several tiles, which is the case for inter-tile data reuse, the tile sizes induce non-linear constraints, unless they are numerical constants. This complicates or prevents a parametric analysis with polyhedral optimization techniques. This paper shows that, when tiles are executed in sequence along tile axes, the parametric (with respect to tile sizes) analysis for inter-tile data reuse is nevertheless possible, i.e., one can determine, at compile-time and in a parametric fashion, the copy-in and copy-out data sets for all tiles, with inter-tile reuse, as well as sizes for the induced local memories. When approximations of transfers are performed, the situation is much more complex, and involves a careful analysis to guarantee correctness when data are both read and written. We provide the mathematical foundations to make such approximations possible. Combined with hierarchical tiling, this result opens perspectives for the automatic generation of blocking algorithms, guided by parametric cost models, where blocks can be pipelined and/or can contain parallelism. Previous work on FPGAs and GPUs already showed the interest and feasibility of such automation with tiling, but in a non-parametric fashion.

show abstract

Section: Introductionmentioning

confidence: 74%

“…Full experiments will be needed to validate the approach and help designing cost models for tile size selection. Nevertheless, the different performance studies with inter-tile data reuse for GPUs [17,18,35] or FPGAs [4,29], for non-parametric tile sizes, already demonstrate its interest.…”

Section: Resultsmentioning

confidence: 99%

Exact and Approximated Data-Reuse Optimizations for Tiling with Parametric Sizes

Darte

Isoard

2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Standard high-level use-def chains and DDG are unaware of these implicit dependencies. However, these are key features when distributed loops [4] or when performing a statement isolation [16].…”

Section: Data Dependence Graphmentioning

confidence: 99%

Effects Dependence Graph: A Key Data Concept for C Source-to-Source Compilers

Lossing¹,

Guillou²,

Irigoin³

2016

2016 IEEE 16th International Working Conference on Source Code Analysis and Manipulation (SCAM)

View full text Add to dashboard Cite

International audienceOptimizations, transformations and analyses are applied to programs by compilers at the intermediate representation level, which usually does not include explicit variable declarations. This description level is fine for middle-ends and for source-to-source optimizers of simple languages. Meanwhile, the C language has become much more flexible since the C99standard, and let variable and type declarations appear almost anywhere in source code.We present in this paper a new concept to manage C99 declarations in a source-to-source compiler: the Effects Dependence Graph, which is an extension of the classical Data Dependence Graph. It deals particularly efficiently with user-defined type declarations or dependent types like Variable-Length Array. It is also interesting because no legal scheduling transformationis hindered and because existing algorithms are either not or slightly modified. Finally it reduces the need for variable, struct and array privatization or live range analyses in automatic parallelizers.To the best of our knowledge, the declaration issue is ignored in the literature: existing C source-to-source compilers either do not support C99, or accept only restricted portions of code, and production compilers use low-level intermediate representations, possibly with annotations. In this way our solution addresses a wider range of compiler analysis issues

show abstract