• Publications
  • Influence
Pin: building customized program analysis tools with dynamic instrumentation
TLDR
We have developed a new instrumentation system called Pin that allows the tool writer to analyze an application at the instruction level without the need for detailed knowledge of the underlying instructions. Expand
  • 3,869
  • 539
  • PDF
Applied Machine Learning at Facebook: A Datacenter Infrastructure Perspective
TLDR
We discovered several key factors that emerge at scale and drive decisions in the design of our datacenter infrastructure: the importance of co-locating data with compute. Expand
  • 230
  • 19
  • PDF
Where is the data? Why you cannot debate CPU vs. GPU performance without the answer
TLDR
General purpose GPU Computing (GPGPU) has taken off in the past few years, with great promises for increased desktop processing power due to the large number of fast computing cores on high-end graphics cards. Expand
  • 279
  • 14
  • PDF
Machine Learning at Facebook: Understanding Inference at the Edge
TLDR
This paper takes a datadriven approach to present the opportunities and design challenges faced by Facebook in order to enable machine learning inference locally on smartphones and other edge platforms. Expand
  • 129
  • 13
  • PDF
SuperPin: Parallelizing Dynamic Instrumentation for Real-Time Performance
TLDR
In this paper, we present a novel approach to dynamic instrumentation where several non-overlapping slices of an application are launched as separate instrumentation threads and executed in parallel in order to approach real-time performance. Expand
  • 95
  • 13
  • PDF
Profiling a warehouse-scale computer
TLDR
We present detailed quantitative analysis of microarchitecture events based on a longitudinal study across tens of thousands of server machines over three years running workloads and services used by billions of users. Expand
  • 144
  • 12
  • PDF
MLPerf Training Benchmark
TLDR
We present MLPerf, an ML benchmark that overcomes these challenges and demonstrates its efficacy at driving performance and scalability improvements. Expand
  • 81
  • 12
  • PDF
Reducing DRAM footprint with NVM in Facebook
TLDR
We design a key-value store, MyNVM, which leverages an NVM block device to reduce DRAM usage, and to reduce the total cost of ownership, while providing comparable latency and queries-per-second of MyRocks on a server with a much larger amount of DRAM. Expand
  • 68
  • 12
  • PDF
Analyzing Parallel Programs with PIN
TLDR
Pin is a software system that performs runtime binary instrumentation of Linux and Microsoft Windows applications. Expand
  • 125
  • 11
  • PDF
The Architectural Implications of Facebook's DNN-Based Personalized Recommendation
TLDR
The widespread application of deep learning has changed the landscape of computation in data centers. Expand
  • 49
  • 9
  • PDF
...
1
2
3
4
5
...