Bind the gap: compiling real software to hardware FFT accelerators

Woodruff, Jackson; Armengol-Estapé, Jordi; Ainsworth, Sam; O’Boyle, Michael

doi:10.1145/3519939.3523439

Cited by 8 publications

(3 citation statements)

References 109 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Because we cannot express general computation in TACO, there is a need to identify the code regions that can be lifted and accelerated. We use prior work in neural program classification [69] to determine which parts of the program represent tensor operations.…”

Section: Classificationmentioning

confidence: 99%

See 1 more Smart Citation

C2TACO: Lifting Tensor Code to TACO

Magalhães,

Woodruff,

Polgreen

et al. 2023

Proceedings of the 22nd ACM SIGPLAN International Conference on Generative Programming: Concepts and Experiences

View full text Add to dashboard Cite

Domain-specific languages (DSLs) promise a significant performance and portability advantage over traditional languages. DSLs are designed to be high-level and platformindependent, allowing an optimizing compiler significant leeway when targeting a particular device. Such languages are particularly popular with emerging tensor algebra workloads. However, DSLs present their own challenge: they require programmers to learn new programming languages and put in significant effort to migrate legacy code.We present C2TACO, a synthesis tool for synthesizing TACO, a well-known tensor DSL, from C code. We develop a guided enumerative synthesizer that uses automatically generated IO examples and source-code analysis to efficiently generate dense tensor algebra code. C2TACO is able to synthesize 95% benchmarks from a tensor benchmark suite, outperforming an alternative neural machine translation technique, and demonstrates substantially higher levels of accuracy when evaluated against two state-of-the-art existing schemes, TF-Coder and ChatGPT. Our synthesized TACO programs are, by design, portable achieving significant performance improvement when evaluated on a multi-core and GPU platform.

show abstract

Section: Classificationmentioning

confidence: 99%

“…We extend the method set out in FACC [69], where inputs are randomly generated according to manually given constraints dictating the length of arrays and favoring smaller values to make evaluation faster. We constrain arrays to be of size 4096, and fix tensor-dimensions to be equal (e.g., a 2-dimensional tensor is of size 64 × 64).…”

Section: Specificationmentioning

confidence: 99%

C2TACO: Lifting Tensor Code to TACO

Magalhães,

Woodruff,

Polgreen

et al. 2023

Proceedings of the 22nd ACM SIGPLAN International Conference on Generative Programming: Concepts and Experiences

View full text Add to dashboard Cite

show abstract

“…KernelFarer [23] works at the program level and restricts its attention to just GEMM API targets, but is more robust than IDL matching significantly more user code. This robustness is extended further in [55], [40] which uses behavioral equivalence to match code. Such approaches, however, are intrinsically limited as they focus on fixed APIs rather than the open-ended nature of DSLs and their IRs.…”

Section: Related Workmentioning

confidence: 99%

mlirSynth: Automatic, Retargetable Program Raising in Multi-Level IR Using Program Synthesis

Brauckmann,

Polgreen,

Grosser

et al. 2023

2023 32nd International Conference on Parallel Architectures and Compilation Techniques (PACT)

Self Cite

View full text Add to dashboard Cite

MLIR is an emerging compiler infrastructure for modern hardware, but existing programs cannot take advantage of MLIR's high-performance compilation if they are described in lower-level general purpose languages. Consequently, to avoid programs needing to be rewritten manually, this has led to efforts to automatically raise lower-level to higher-level dialects in MLIR. However, current methods rely on manually-defined raising rules, which limit their applicability and make them challenging to maintain as MLIR dialects evolve.We present mlirSynth -a novel approach which translates programs from lower-level MLIR dialects to high-level ones without manually defined rules. Instead, it uses available dialect definitions to construct a program space and searches it effectively using type constraints and equivalences. We demonstrate its effectiveness by raising C programs to two distinct high-level MLIR dialects, which enables us to use existing high-level dialect specific compilation flows. On Polybench, we show a greater coverage than previous approaches, resulting in geomean speedups of 2.5x (Intel) and 3.4x (AMD) over state-of-the-art compilation flows. mlirSynth also enables retargetability to domain-specific accelerators, resulting in a geomean speedup of 21.6x on a TPU.

show abstract

Code Detection for Hardware Acceleration Using Large Language Models

Martínez,

Bernabé,

García

2024

IEEE Access

View full text Add to dashboard Cite

Large language models (LLMs) have been massively applied to many tasks, often surpassing state-of-the-art approaches. While their effectiveness in code generation has been extensively studied (e.g., AlphaCode), their potential for code detection remains unexplored. This work presents the first analysis of code detection using LLMs. Our study examines essential kernels, including matrix multiplication, convolution, fast-fourier transform and LU factorization, implemented in C/C++. We propose both a preliminary, naive prompt and a novel prompting strategy for code detection. Results reveal that conventional prompting achieves great precision but poor accuracy (67.5%, 22.5%, 79.5% and 64% for GEMM, convolution, FFT and LU factorization, respectively) due to a high number of false positives. Our novel prompting strategy substantially reduces false positives, resulting in excellent overall accuracy (91.2%, 98%, 99.7% and 99.7%, respectively). These results pose a considerable challenge to existing state-of-the-art code detection methods.

show abstract

Bind the gap: compiling real software to hardware FFT accelerators

Cited by 8 publications

References 109 publications

C2TACO: Lifting Tensor Code to TACO

C2TACO: Lifting Tensor Code to TACO

mlirSynth: Automatic, Retargetable Program Raising in Multi-Level IR Using Program Synthesis

Code Detection for Hardware Acceleration Using Large Language Models

Contact Info

Product

Resources

About