arXiv subjects

Alex Aiken

Publications and source records attributed to Alex Aiken.

At least 37 records · Page 2Linked to original sources

arXiv2024

Automatic Tracing in Task-Based Runtime Systems

Implicitly parallel task-based runtime systems often perform dynamic analysis to discover dependencies in and extract parallelism from sequential programs. Dependence analysis becomes expensive as task granularity drops below a threshold. Tracing techniques have been developed where programmers annotate repeated program fragments (traces) issued by the application, and the runtime system memoizes the dependence analysis for those fragments, greatly reducing overhead when the fragments are executed again. However, manual trace annotation can be brittle and not easily applicable to complex programs built through the composition of independent components. We introduce Apophenia, a system that automatically traces the dependence analysis of task-based runtime systems, removing the burden of manual annotations from programmers and enabling new and complex programs to be traced. Apophenia identifies traces dynamically through a series of dynamic string analyses, which find repeated program fragments in the stream of tasks issued to the runtime system. We show that Apophenia is able to come between 0.92x--1.03x the performance of manually traced programs, and is able to effectively trace previously untraced programs to yield speedups of between 0.91x--2.82x on the Perlmutter and Eos supercomputers.

cs.DC↗

arXiv2024

HiCCL: A Hierarchical Collective Communication Library

HiCCL (Hierarchical Collective Communication Library) addresses the growing complexity and diversity in high-performance network architectures. As GPU systems have envolved into networks of GPUs with different multilevel communication hierarchies, optimizing each collective function for a specific system has become a challenging task. Consequently, many collective libraries struggle to adapt to different hardware and software, especially across systems from different vendors. HiCCL's library design decouples the collective communication logic from network-specific optimizations through a compositional API. The communication logic is composed using multicast, reduction, and fence primitives, which are then factorized for a specified network hieararchy using only point-to-point operations within a level. Finally, striping and pipelining optimizations applied as specified for streamlining the execution. Performance evaluation of HiCCL across four different machines$\unicode{x2014}$two with Nvidia GPUs, one with AMD GPUs, and one with Intel GPUs$\unicode{x2014}$demonstrates an average 17$\times$ higher throughput than the collectives of highly specialized GPU-aware MPI implementations, and competitive throughput with those of vendor-specific libraries (NCCL, RCCL, and OneCCL), while providing portability across all four machines.

cs.DC↗

arXiv2024

Efficient Fault Tolerance for Pipelined Query Engines via Write-ahead Lineage

Modern distributed pipelined query engines either do not support intra-query fault tolerance or employ high-overhead approaches such as persisting intermediate outputs or checkpointing state. In this work, we present write-ahead lineage, a novel fault recovery technique that combines Spark's lineage-based replay and write-ahead logging. Unlike Spark, where the lineage is determined before query execution, write-ahead lineage persistently logs lineage at runtime to support dynamic task dependencies in pipelined query engines. Since only KB-sized lineages are persisted instead of MB-sized intermediate outputs, the normal execution overhead is minimal compared to spooling or checkpointing based approaches. To ensure fast fault recovery times, tasks only consume intermediate outputs with persisted lineage, preventing global rollbacks upon failure. In addition, lost tasks from different stages can be recovered in a pipelined parallel manner. We implement write-ahead lineage in a distributed pipelined query engine called Quokka. We show that Quokka is around 2x faster than SparkSQL on the TPC-H benchmark with similar fault recovery performance.

cs.DC↗

arXiv2023

Training with Mixed-Precision Floating-Point Assignments

When training deep neural networks, keeping all tensors in high precision (e.g., 32-bit or even 16-bit floats) is often wasteful. However, keeping all tensors in low precision (e.g., 8-bit floats) can lead to unacceptable accuracy loss. Hence, it is important to use a precision assignment -- a mapping from all tensors (arising in training) to precision levels (high or low) -- that keeps most of the tensors in low precision and leads to sufficiently accurate models. We provide a technique that explores this memory-accuracy tradeoff by generating precision assignments for convolutional neural networks that (i) use less memory and (ii) lead to more accurate convolutional networks at the same time, compared to the precision assignments considered by prior work in low-precision floating-point training. We evaluate our technique on image classification tasks by training convolutional networks on CIFAR-10, CIFAR-100, and ImageNet. Our method typically provides > 2x memory reduction over a baseline precision assignment while preserving training accuracy, and gives further reductions by trading off accuracy. Compared to other baselines which sometimes cause training to diverge, our method provides similar or better memory reduction while avoiding divergence.

cs.LG↗

arXiv2023

On the Correctness of Automatic Differentiation for Neural Networks with Machine-Representable Parameters

Recent work has shown that forward- and reverse- mode automatic differentiation (AD) over the reals is almost always correct in a mathematically precise sense. However, actual programs work with machine-representable numbers (e.g., floating-point numbers), not reals. In this paper, we study the correctness of AD when the parameter space of a neural network consists solely of machine-representable numbers. In particular, we analyze two sets of parameters on which AD can be incorrect: the incorrect set on which the network is differentiable but AD does not compute its derivative, and the non-differentiable set on which the network is non-differentiable. For a neural network with bias parameters, we first prove that the incorrect set is always empty. We then prove a tight bound on the size of the non-differentiable set, which is linear in the number of non-differentiabilities in activation functions, and give a simple necessary and sufficient condition for a parameter to be in this set. We further prove that AD always computes a Clarke subderivative even on the non-differentiable set. We also extend these results to neural networks possibly without bias parameters.

cs.LG↗

arXiv2023

Putting People in Their Place: Affordance-Aware Human Insertion into Scenes

We study the problem of inferring scene affordances by presenting a method for realistically inserting people into scenes. Given a scene image with a marked region and an image of a person, we insert the person into the scene while respecting the scene affordances. Our model can infer the set of realistic poses given the scene context, re-pose the reference person, and harmonize the composition. We set up the task in a self-supervised fashion by learning to re-pose humans in video clips. We train a large-scale diffusion model on a dataset of 2.4M video clips that produces diverse plausible poses while respecting the scene context. Given the learned human-scene composition, our model can also hallucinate realistic people and scenes when prompted without conditioning and also enables interactive editing. A quantitative evaluation shows that our method synthesizes more realistic human appearance and more natural human-scene interactions than prior work.

cs.CV↗

arXiv2022

Optimizing Mixture of Experts using Dynamic Recompilations

The Mixture of Experts architecture allows for outrageously large neural networks by scaling model parameter size independently from computational demand (FLOPs). However, current DNN frameworks cannot effectively support the dynamic data flow in Mixture of Experts, and implementations on top of these frameworks need to use workarounds that introduce significant overheads. To address the limitation of these frameworks, we present DynaMoE, a DNN library that uses dynamic recompilations to optimize and adapt the use of computational resources to the dynamic needs of Mixture of Experts models. Our evaluation shows that DynaMoE achieves a 1.8x speedup and supports 2.3x larger model sizes when compared to existing MoE systems, even when not using recompilations. We then present further optimizations enabled by dynamic recompilations that yield an additional 1.7x speedup while simultaneously reducing memory pressure and improving model quality.

cs.LG↗

arXiv2022

SpDISTAL: Compiling Distributed Sparse Tensor Computations

We introduce SpDISTAL, a compiler for sparse tensor algebra that targets distributed systems. SpDISTAL combines separate descriptions of tensor algebra expressions, sparse data structures, data distribution, and computation distribution. Thus, it enables distributed execution of sparse tensor algebra expressions with a wide variety of sparse data structures and data distributions. SpDISTAL is implemented as a C++ library that targets a distributed task-based runtime system and can generate code for nodes with both multi-core CPUs and multiple GPUs. SpDISTAL generates distributed code that achieves performance competitive with hand-written distributed functions for specific sparse tensor algebra expressions and that outperforms general interpretation-based systems by one to two orders of magnitude.

cs.DC↗

arXiv2022

Programmatic Concept Learning for Human Motion Description and Synthesis

We introduce Programmatic Motion Concepts, a hierarchical motion representation for human actions that captures both low-level motion and high-level description as motion concepts. This representation enables human motion description, interactive editing, and controlled synthesis of novel video sequences within a single framework. We present an architecture that learns this concept representation from paired video and action sequences in a semi-supervised manner. The compactness of our representation also allows us to present a low-resource training recipe for data-efficient learning. By outperforming established baselines, especially in the small data regime, we demonstrate the efficiency and effectiveness of our framework for multiple applications.

cs.CV↗

arXiv2022

Quartz: Superoptimization of Quantum Circuits (Extended Version)

Existing quantum compilers optimize quantum circuits by applying circuit transformations designed by experts. This approach requires significant manual effort to design and implement circuit transformations for different quantum devices, which use different gate sets, and can miss optimizations that are hard to find manually. We propose Quartz, a quantum circuit superoptimizer that automatically generates and verifies circuit transformations for arbitrary quantum gate sets. For a given gate set, Quartz generates candidate circuit transformations by systematically exploring small circuits and verifies the discovered transformations using an automated theorem prover. To optimize a quantum circuit, Quartz uses a cost-based backtracking search that applies the verified transformations to the circuit. Our evaluation on three popular gate sets shows that Quartz can effectively generate and verify transformations for different gate sets. The generated transformations cover manually designed transformations used by existing optimizers and also include new transformations. Quartz is therefore able to optimize a broad range of circuits for diverse gate sets, outperforming or matching the performance of hand-tuned circuit optimizers.

cs.PL↗

arXiv2022

DISTAL: The Distributed Tensor Algebra Compiler

We introduce DISTAL, a compiler for dense tensor algebra that targets modern distributed and heterogeneous systems. DISTAL lets users independently describe how tensors and computation map onto target machines through separate format and scheduling languages. The combination of choices for data and computation distribution creates a large design space that includes many algorithms from both the past (e.g., Cannon's algorithm) and the present (e.g., COSMA). DISTAL compiles a tensor algebra domain specific language to a distributed task-based runtime system and supports nodes with multi-core CPUs and multiple GPUs. Code generated by DISTAL is competitive with optimized codes for matrix multiply on 256 nodes of the Lassen supercomputer and outperforms existing systems by between 1.8x to 3.7x (with a 45.7x outlier) on higher order tensor operations.

cs.PL↗

arXiv2021

Inferring Invariants with Quantifier Alternations: Taming the Search Space Explosion

We present a PDR/IC3 algorithm for finding inductive invariants with quantifier alternations. We tackle scalability issues that arise due to the large search space of quantified invariants by combining a breadth-first search strategy and a new syntactic form for quantifier-free bodies. The breadth-first strategy prevents inductive generalization from getting stuck in regions of the search space that are expensive to search and focuses instead on lemmas that are easy to discover. The new syntactic form is well-suited to lemmas with quantifier alternations by allowing both limited conjunction and disjunction in the quantifier-free body, while carefully controlling the size of the search space. Combining the breadth-first strategy with the new syntactic form results in useful inductive bias by prioritizing lemmas according to: (i) well-defined syntactic metrics for simple quantifier structures and quantifier-free bodies, and (ii) the empirically useful heuristic of preferring lemmas that are fast to discover. On a benchmark suite of primarily distributed protocols and complex Paxos variants, we demonstrate that our algorithm can solve more of the most complicated examples than state-of-the-art techniques.

cs.PL↗

arXiv2021

Quanto: Optimizing Quantum Circuits with Automatic Generation of Circuit Identities

Existing quantum compilers focus on mapping a logical quantum circuit to a quantum device and its native quantum gates. Only simple circuit identities are used to optimize the quantum circuit during the compilation process. This approach misses more complex circuit identities, which could be used to optimize the quantum circuit further. We propose Quanto, the first quantum optimizer that automatically generates circuit identities. Quanto takes as input a gate set and generates provably correct circuit identities for the gate set. Quanto's automatic generation of circuit identities includes single-qubit and two-qubit gates, which leads to a new database of circuit identities, some of which are novel to the best of our knowledge. In addition to the generation of new circuit identities, Quanto's optimizer applies such circuit identities to quantum circuits and finds optimized quantum circuits that have not been discovered by other quantum compilers, including IBM Qiskit and Cambridge Quantum Computing Tket. Quanto's database of circuit identities could be applied to improve existing quantum compilers and Quanto can be used to generate identity databases for new gate sets.

quant-ph↗

arXiv2021

Hierarchical Motion Understanding via Motion Programs

Current approaches to video analysis of human motion focus on raw pixels or keypoints as the basic units of reasoning. We posit that adding higher-level motion primitives, which can capture natural coarser units of motion such as backswing or follow-through, can be used to improve downstream analysis tasks. This higher level of abstraction can also capture key features, such as loops of repeated primitives, that are currently inaccessible at lower levels of representation. We therefore introduce Motion Programs, a neuro-symbolic, program-like representation that expresses motions as a composition of high-level primitives. We also present a system for automatically inducing motion programs from videos of human motion and for leveraging motion programs in video synthesis. Experiments show that motion programs can accurately describe a diverse set of human motions and the inferred programs contain semantically meaningful motion primitives, such as arm swings and jumping jacks. Our representation also benefits downstream tasks such as video interpolation and video prediction and outperforms off-the-shelf models. We further demonstrate how these programs can detect diverse kinds of repetitive motion and facilitate interactive video editing.

cs.CV↗

arXiv2020

Task Bench: A Parameterized Benchmark for Evaluating Parallel Runtime Performance

We present Task Bench, a parameterized benchmark designed to explore the performance of parallel and distributed programming systems under a variety of application scenarios. Task Bench lowers the barrier to benchmarking multiple programming systems by making the implementation for a given system orthogonal to the benchmarks themselves: every benchmark constructed with Task Bench runs on every Task Bench implementation. Furthermore, Task Bench's parameterization enables a wide variety of benchmark scenarios that distill the key characteristics of larger applications. We conduct a comprehensive study with implementations of Task Bench in 15 programming systems on up to 256 Haswell nodes of the Cori supercomputer. We introduce a novel metric, minimum effective task granularity to study the baseline runtime overhead of each system. We show that when running at scale, 100 μs is the smallest granularity that even the most efficient systems can reliably support with current technologies. We also study each system's scalability, ability to hide communication and mitigate load imbalance.

cs.DC↗

arXiv2019

SPoC: Search-based Pseudocode to Code

We consider the task of mapping pseudocode to long programs that are functionally correct. Given test cases as a mechanism to validate programs, we search over the space of possible translations of the pseudocode to find a program that passes the validation. However, without proper credit assignment to localize the sources of program failures, it is difficult to guide search toward more promising programs. We propose to perform credit assignment based on signals from compilation errors, which constitute 88.7% of program failures. Concretely, we treat the translation of each pseudocode line as a discrete portion of the program, and whenever a synthesized program fails to compile, an error localization method tries to identify the portion of the program responsible for the failure. We then focus search over alternative translations of the pseudocode for those portions. For evaluation, we collected the SPoC dataset (Search-based Pseudocode to Code) containing 18,356 programs with human-authored pseudocode and test cases. Under a budget of 100 program compilations, performing search improves the synthesis success rate over using the top-one translation of the pseudocode from 25.6% to 44.7%.

cs.LG↗

arXiv2019

Redundancy-Free Computation Graphs for Graph Neural Networks

Graph Neural Networks (GNNs) are based on repeated aggregations of information across nodes' neighbors in a graph. However, because common neighbors are shared between different nodes, this leads to repeated and inefficient computations. We propose Hierarchically Aggregated computation Graphs (HAGs), a new GNN graph representation that explicitly avoids redundancy by managing intermediate aggregation results hierarchically, eliminating repeated computations and unnecessary data transfers in GNN training and inference. We introduce an accurate cost function to quantitatively evaluate the runtime performance of different HAGs and use a novel HAG search algorithm to find optimized HAGs. Experiments show that the HAG representation significantly outperforms the standard GNN graph representation by increasing the end-to-end training throughput by up to 2.8x and reducing the aggregations and data transfers in GNN training by up to 6.3x and 5.6x, while maintaining the original model accuracy.

cs.LG↗

arXiv2018

Beyond Data and Model Parallelism for Deep Neural Networks

The computational requirements for training deep neural networks (DNNs) have grown to the point that it is now standard practice to parallelize training. Existing deep learning systems commonly use data or model parallelism, but unfortunately, these strategies often result in suboptimal parallelization performance. In this paper, we define a more comprehensive search space of parallelization strategies for DNNs called SOAP, which includes strategies to parallelize a DNN in the Sample, Operation, Attribute, and Parameter dimensions. We also propose FlexFlow, a deep learning framework that uses guided randomized search of the SOAP space to find a fast parallelization strategy for a specific parallel machine. To accelerate this search, FlexFlow introduces a novel execution simulator that can accurately predict a parallelization strategy's performance and is three orders of magnitude faster than prior approaches that have to execute each strategy. We evaluate FlexFlow with six real-world DNN benchmarks on two GPU clusters and show that FlexFlow can increase training throughput by up to 3.8x over state-of-the-art approaches, even when including its search time, and also improves scalability.

cs.DC↗