• [2310.06770] SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
• [2405.15793] SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering
• SWE-bench
• Introducing SWE-bench Verified | OpenAI
• The new Claude 3.5 Sonnet, Computer Use, and Building SOTA Agents — with Erik Schluntz, Anthropic

• An Empirical Study of Rust-for-Linux: The Success, Dissatisfaction, and Compromise | USENIX
• Rust for Linux

• SQL Has Problems. We Can Fix Them: Pipe Syntax In SQL
• PRQL
• Pipe syntax  |  BigQuery  |  Google Cloud
• SQLite Forum: Interesting paper from Google on pipe syntax on SQL

• [2410.05229] GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models
• GitHub – openai/grade-school-math

• NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
• NeRF Tutorial ECCV 2022
• illuminate.google.com

• [1503.02531] Distilling the Knowledge in a Neural Network

• [1603.05346v2] Optimal Quantile Approximation in Streams
• Small Summaries for Big Data
• Data Types – Presto 0.288 Documentation
• Estimating Percentile Values | Snowflake Documentation
• KLL sketch vs t-digest – DataSketches

• [2301.10191] Distinct Elements in Streams: An Algorithm for the (Text) Book
• The CVM Algorithm for Estimating Distinct Elements in Streams
• Computer scientists invent an efficient new way to count | Hacker News

• [1704.04760] In-Datacenter Performance Analysis of a Tensor Processing Unit
• Systolic Arrays – an overview | ScienceDirect Topics
• Pallas: a JAX kernel language — JAX documentation
• About Groq – Fast AI Inference
• The Design Process for Google’s Training Chips: TPUv2 and TPUv3 | IEEE Journals & Magazine | IEEE Xplore

• [2106.09685] LoRA: Low-Rank Adaptation of Large Language Models

• Triton: An Intermediate Language and Compiler for Tiled Neural Network Computations
• Introducing Triton: Open-source GPU programming for neural networks | OpenAI
• Welcome to Triton’s documentation! — Triton documentation
• Hello Triton.ipynb – Colab

• #01: Tensor Comprehensions, Rust Belt – Misreading Chat
• #23 – Halide: Decoupling Algorithms from Schedules for High-Performance Image Processing – Misreading Chat
• #27 – Julia: A Fresh Approach to Numerical Computing – Misreading Chat

• [2112.10752] High-Resolution Image Synthesis with Latent Diffusion Models
• [2105.05233] Diffusion Models Beat GANs on Image Synthesis
• Classifier-Free Diffusion Guidance | OpenReview
• What are Diffusion Models? | Lil’Log

• [2205.14135] FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
• GitHub – Dao-AILab/flash-attention: Fast and memory-efficient exact attention
• GitHub – NVIDIA/apex: A PyTorch Extension: Tools for easy mixed precision and distributed training in Pytorch
• [2307.08691] FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning
• [2112.05682] Self-attention Does Not Need $O(n^2)$ Memory
• GitHub – tspeterkim/flash-attention-minimal: Flash Attention in ~100 lines of CUDA (forward pass only)

• Diffusion models from scratch
• [1503.03585] Deep Unsupervised Learning using Nonequilibrium Thermodynamics
• [2006.11239] Denoising Diffusion Probabilistic Models

• Programming Massively Parallel Processors: A Hands-on Approach (Amazon.co.jp, Elsevier)
• NVIDIA H100 Tensor Core GPU Architecture Overview

• facebookresearch/faiss: A library for efficient similarity search and clustering of dense vectors.
• Home · facebookresearch/faiss Wiki
• Optimized Product Quantization | IEEE Journals & Magazine | IEEE Xplore
• Product Quantization for Nearest Neighbor Search
• [1102.3828] Searching in one billion vectors: re-rank with source coding
• The inverted multi-index | IEEE Conference Publication | IEEE Xplore
• A Survey of Product Quantization

• Programming Massively Parallel Processors: A Hands-on Approach (Amazon.co.jp, Elsevier)
• Evolution of the Graphics Processing Unit (GPU) | IEEE Journals & Magazine | IEEE Xplore
• David Kirk (scientist) – Wikipedia

• [2309.11322] Vector database management systems: Fundamental concepts, use-cases, and current challenges
• [2310.11703] A Comprehensive Survey on Vector Database: Storage and Retrieval Technique, Challenge
• [2310.14021] Survey of Vector Database Management Systems
  
• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• Databases in 2023: A Year in Review | OtterTune

• Always-on Vision Processing Unit for Mobile Applications | IEEE Journals & Magazine | IEEE Xplore (Google Scholar)
  
• Intel Core Ultra Ushers in the Age of the AI PC
• Intel® Movidius Vision Processing Units (VPUs)
• Amazon.com: Intel NCS2 Movidius Neural Compute Stick 2, Perfect for Deep Neural Network Applications (DNN) : Electronics
  
• Movidius – Wikipedia
• Very long instruction word – Wikipedia
• Single instruction, multiple data – Wikipedia
• Google’s Clips camera is powered by a tailor-made AI chip – The Verge
• The revolutionary chipmaker behind Google’s project Tango is now powering DJI’s autonomous drone – The Verge
  
• Qualcomm’s Hexagon DSP, and now, NPU – Chips and Cheese
  
• AMD XDNA Architecture
• AI Engine Technology
  
• Neural Engine | Apple Wiki | Fandom

• [2311.12983] GAIA: a benchmark for General AI Assistants
• gaia-benchmark/GAIA · Datasets at Hugging Face

GPGPU ワナビーの森田が C++ 拡張の GPGPU 標準を眺めました。ご意見感想などは Reddit やおたより投書箱にお寄せください。iTunes のレビューや星もよろしくね。

• SYCL Overview – The Khronos Group Inc
• SYCL.tech – Find out the latest SYCL news, videos, learning materials and projects.
• GitHub – Tohoku-University-Takizawa-Lab/neoSYCL: A SYCL Implementation for CPU and SX-Aurora TSUBASA
• DPC++ Documentation — oneAPI DPC++ Compiler documentation
• GitHub – AdaptiveCpp/AdaptiveCpp: Implementation of SYCL and C++ standard parallelism for CPUs and GPUs from all vendors: The independent, community-driven compiler for C++-based heterogeneous programming models. Lets applications adapt themselves to all the hardware in the system – even at runtime!
• GitHub – triSYCL/triSYCL: Generic system-wide modern C++ for heterogeneous platforms with SYCL from Khronos Group
• Codeplay Software Ltd
• Intel Proposes Adding Full SYCL Programming Model Support To Upstream LLVM – Phoronix
• Migrate scripts to OpenGL ES 3.1  |  Android Developers
• Migrate scripts to Vulkan  |  Android Developers
• tensorflow/tensorflow/lite/delegates/gpu/README.md at master · tensorflow/tensorflow · GitHub

• Lifting the veil on Meta’s microservice architecture: Analyses of topology and request workflows | USENIX
• GitHub – facebook/fbthrift: Facebook’s branch of Apache Thrift, including a new C++ server.
• Canopy: An End-to-End Performance Tracing and Analysis System – Meta Research
• Yuri Shkuro
  
• Mastodon ID: [email protected]

• The SPACE of Developer Productivity – ACM Queue
• DORA | DevOps Research and Assessment
• Accelerate: The Science of Lean Software and DevOps: Building and Scaling High Performing Technology Organizations
• LeanとDevOpsの科学[Accelerate] テクノロジーの戦略的活用が組織変革を加速する

• On-demand Container Loading in AWS Lambda | USENIX
• Marc Brooker’s Blog – Marc’s Blog

• #44: Firecracker
• #91: Serverless in the Wild
• #99: The Prusti Project: Formal Verification for Rust

• Overlapping Experiment Infrastructure: More, Better, Faster Experimentation – Google Research
• Putting a Bolder Face on Google – The New York Times
• Goodbye, Google | Stopdesign

• [2203.12533] Pathways: Asynchronous Distributed Dataflow for ML
• [2204.02311] PaLM: Scaling Language Modeling with Pathways
• System Architecture  |  Cloud TPU  |  Google Cloud

• DuckDB: an Embeddable Analytical Database
• SQLite: Past, Present, and Future
• An in-process SQL OLAP database management system – DuckDB
• 25 Years of Python at CWI

• [1311.2540] Asymmetric numeral systems: entropy coding combining speed of Huffman coding with compression rate of arithmetic coding
• facebook/zstd: Zstandard – Fast real-time compression algorithm
• From Project Management to Data Compression Innovator With Yann Collet – CoRecursive Podcast

• [2203.02155] Training language models to follow instructions with human feedback
• [2212.08073] Constitutional AI: Harmlessness from AI Feedback
• It looks like GPT-4-32k is rolling out | Hacker News
• Anthropic | Introducing Claude
• Goodhart’s law – Wikipedia

• Invisinets: Removing Networking from Cloud Networks | USENIX

• Improving Language Understanding by Generative Pre-Training
• Language Models are Unsupervised Multitask Learners
• [2005.14165] Language Models are Few-Shot Learners
• [2201.11903] Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
• NLP’s ImageNet moment has arrived
• In-Context Learning, In Context
• ChatGPT Prompt Engineering for Developers – DeepLearning.AI

• SkyPilot: An Intercloud Broker for Sky Computing | USENIX
• GitHub – skypilot-org/skypilot: SkyPilot is a framework for easily running machine learning workloads on any cloud through a unified interface.
• From cloud computing to sky computing | Proceedings of the Workshop on Hot Topics in Operating Systems

今回は録音に際し Adobe Podcast (beta) のバグを引き当ててしまい、向井と森田の音声トラックがずれてしまいました。ごめんなさい。次回からは non-beta の手堅いツールで録音しようと思います・・・。

• [2207.09238] Formal Algorithms for Transformers
• #15 – Neural Machine Translation by Jointly Learning to Align and Translate
• #38 – Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates
• #51 – Attention Is All You Need
• #53 – BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
• Jay Alammar – YouTube
• GitHub – openai/tiktoken: tiktoken is a fast BPE tokeniser for use with OpenAI’s models.
• GitHub – karpathy/nanoGPT: The simplest, fastest repository for training/finetuning medium-sized GPTs.
• Let’s build GPT: from scratch, in code, spelled out. – YouTube

• Perseus: A Fail-Slow Detection Framework for Cloud Storage Systems | USENIX

• CockroachDB is 10x more scalable than Amazon Aurora for OLTP workloads
• TPC-Homepage (TLS の証明書が切れてますよ…)
• The History of DebitCredit and the TPC
• A Measure of Transaction Processing Power
• The Evolution of TPC Benchmarks: Why TPC-A and TPC-B are Obsolete
• tpc-c_v5.11.0.pdf
• Cooper: Benchmarking cloud serving systems with YCSB – Google Scholar
• brianfrankcooper/YCSB: Yahoo! Cloud Serving Benchmark
• TPC-E vs. TPC-C: Characterizing the New TPC-E Benchmark via an I/O Comparison Study
• A Study of Database Performance Sensitivity to Experiment Settings
• In 2020s, Don’t Listen to Java Haters – Spinach Forest

• The Verse Calculus: a core calculus for functional logic programming – Simon Peyton Jones
• Beyond Functional Programming: The Verse Programming Language (Simon Peyton Jones) – YouTube

• PostScript: A Digital Printing Press – CHM
• Founding and Growing Adobe Systems, Inc. (paywall)
• Simple Ideas That Changed Printing and Publishing on JSTOR (paywall)
• The Origins of PostScript (paywall)
• LaserWriter – Wikipedia
• TrueType hinting – Typography | Microsoft Learn
• Instructing Fonts – TrueType Reference Manual – Apple Developer
• Adobe Photoshop Source Code – CHM

• Human-level play in the game of Diplomacy by combining language models with strategic reasoning | Science
• CICERO: An AI agent that negotiates, persuades, and cooperates with people
• Diplomacy (game) – Wikipedia

• Volcano – An Extensible and Parallel Query Evaluation System
• Apache Iceberg

• [2210.00714] It’s Time to Replace TCP in the Datacenter
  • Netdev | Netdev 0x16 – Sessions
  • GitHub – PlatformLab/Homa: Low-Latency Data Center Network Transport
• A Philosophy of Software Design, 2nd Edition

• GPU Accelerated Path Rendering | NVIDIA Developer
• Resolution Independent Curve Rendering using Programmable Graphics Hardware – Microsoft Research
• Bezier quadratic (Loop/Blinn) – WebGL demo
• Accelerating vector graphics rendering using the graphics hardware pipeline | ACM Transactions on Graphics

• Do Developers Really Know How to Use Git Commands? A Large-scale Study Using Stack Overflow (Paper)
• Do Developers Really Know How to Use Git Commands? · It Will Never Work in Theory

• Lakehouse: A New Generation of Open Platforms that Unify Data Warehousing and Advanced Analytics – Databricks
• Delta Lake: High-Performance ACID Table Storage over Cloud Object Stores – Databricks
• Photon: A High-Performance Query Engine for the Lakehouse

• Google announces updates for BigQuery data warehouse • The Register

• On Misreading Chat – Jun Mukai’s blog
• On Misreading Chat – 2021-01-06 / steps to phantasien
• Rebuild – Podcast by Tatsuhiko Miyagawa
• ByteCast Archive
• Linear Digressions
• CoRecursive Podcast – The Stories Behind The Code
• Spilled Milk
• the morning paper | a random walk through Computer Science research, by Adrian Colyer
• Lambda the Ultimate | Programming Languages Weblog
• Papers We Love
• CACM Homepage | Communications of the ACM
• ACM COMPUTING SURVEYS Home
• 第84回 ゲストを招いて、労働とC++の不安をカウンセリングする by プログラム雑談

そこで皆様からの質問などをお待ちしております。投稿先はメール: [email protected], Twitter hashtag #misreading, あるいは Google Forms https://forms.gle/aDU21GzBvqpnQTcS7　です。

• Prusti – Programming Methodology Group | ETH Zurich

• Amazon’s Dynamo | All Things Distributed
• Chord (peer-to-peer) – Wikipedia

• Vector clock – Wikipedia
• Dynamo | Apache Cassandra Documentation

• Using Lightweight Formal Methods to Validate a Key-Value Storage Node in Amazon S3 | Proceedings of the ACM SIGOPS 28th Symposium on Operating Systems Principles
• Prusti – Programming Methodology Group | ETH Zurich
• tokio-rs/loom: Concurrency permutation testing tool for Rust.
• awslabs/shuttle: Shuttle is a library for testing concurrent Rust code

• Amazon DynamoDB: A Scalable, Predictably Performant, and Fully Managed NoSQL Database Service | USENIX
• Fast NoSQL Key-Value Database – Amazon DynamoDB – Amazon Web Services
• Going Head-to-Head: ScyllaDB vs Amazon DynamoDB – ScyllaDB
• Amazon’s Dynamo | All Things Distributed

• CockroachDB: The Resilient Geo-Distributed SQL Database | Proceedings of the 2020 ACM SIGMOD International Conference on Management of Data
• Spanner: Google’s Globally Distributed Database: ACM Transactions on Computer Systems: Vol 31, No 3
• CockroachDB: The Definitive Guide — Learn to Build What Can’t Be Killed
• Spencer Kimball (computer programmer) – Wikipedia
• Peter Mattis – Wikipedia
• Ben Darnell – Wikipedia

• Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center | USENIX

• SEQUEL: A structured English query language
• A Relational Model of Data for Large Shared Data Banks
• Specifying queries as relational expressions: The SQUARE data sublanguage

森田の PC ノイズが聞き苦しくなってしまいましたがご容赦ください。夏のせいです。

• Software Development Waste (PDF)

• It Will Never Work in Theory
  • Making Software: What Really Works, and Why We Believe It
• Lean Software Development: An Agile Toolkit
• xkcd: Compiling
• On Pivotal:
  • This Company Believes You Should Never Hack Alone | WIRED
  • VMware completes $2.7 billion Pivotal acquisition | TechCrunch
  • Tanzu Portfolio: Products, Services, and Resources | VMware Tanzu

• Serverless in the Wild: Characterizing and Optimizing the Serverless Workload at a Large Cloud Provider | USENIX
• Chapter 8 ARIMA models | Forecasting: Principles and Practice
• Coefficient of variation – Wikipedia
• Apache OpenWhisk is a serverless, open source cloud platform

• Evaluation of splittable pseudo-random generators* | Journal of Functional Programming | Cambridge Core
• System.Random
• Fix quality and performance issues, add monadic API · haskell/random@b45bd28
• Fast Splittable Pseudorandom Number Generators
• jdk/SplittableRandom.java at master · openjdk/jdk

• Conflict-free Replicated Data Types – Archive ouverte HAL
• A commutative replicated data type for cooperative editing – Archive ouverte HAL
• Dynamo: Amazon’s Highly Available Key-value Store
• About CRDTs • Conflict-free Replicated Data Types
• xi-editor/crdt-details.md at master · xi-editor/xi-editor · GitHub
• xi-editor/crdt-details.md at master · xi-editor/xi-editor · GitHub
• Towards a text editor construction kit · Issue #1187 · xi-editor/xi-editor · GitHub

• A History of Haskell: being lazy with class – Microsoft Research
• Haskell Language
• Home — The Glasgow Haskell Compiler
• https://www.haskell.org/onlinereport/haskell2010/
• Changes since Haskell ’98

• From Laptop to Lambda: Outsourcing Everyday Jobs to Thousands of Transient Functional Containers | USENIX
• https://github.com/StanfordSNR/gg
• Outsourcing Everyday Jobs to Thousands of Cloud Functions with gg [PDF]
• POSH: A Data-Aware Shell | USENIX
• Best of 2019 in Tech Talks. | by Cindy Sridharan | Medium

• JavaScript: the first 20 years | Proceedings of the ACM on Programming Languages

• Continuous profiling: where have all the cycles gone?: ACM Transactions on Computer Systems: Vol 15, No 4

• Google-Wide Profiling: A Continuous Profiling Infrastructure for Data Centers – Google Research
• Profiling a warehouse-scale computer – Google Research
• 2 Billion Devices and Counting: An Industry Perspective on the State of Mobile Computer Architecture – Google Research
• Cap’n Proto: Introduction
• Accelerometer: Understanding Acceleration Opportunities for Data Center Overheads at Hyperscale – Facebook Research

• Dynamic Instrumentation of Production Systems
• dtrace.org » About DTrace
• BPF Performance Tools (Book)

• Glass: A New Media for a New Era? – Microsoft Research
• Project Silica proof of concept stores Warner Bros. ‘Superman’ movie on quartz glass
• GitHub Archive Program: the journey of the world’s open source code to the Arctic – The GitHub Blog

• gprof: a Call Graph Execution Profiler
• Tarjan’s strongly connected components algorithm – Wikipedia
• research!rsc: How To Build a User-Level CPU Profiler

• Elo rating system – Wikipedia
  • Bradley–Terry model – Wikipedia
• Rank – Inkipedia, the Splatoon wiki
• TrueSkill Ranking System – Microsoft Research
• TrueSkill 2: An improved Bayesian skill rating system – Microsoft Research
• dotnet/infer: Infer.NET is a framework for running Bayesian inference in graphical models
  • infer/ChessAnalysis.cs at master · dotnet/infer
• Welcome to Glicko ratings
  • Parameter Estimation in Large Dynamic Paired Comparison Experiments

• [1905.02175] Adversarial Examples Are Not Bugs, They Are Features
• A Discussion of ‘Adversarial Examples Are Not Bugs, They Are Features’

• [1506.01497] Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• [1311.2524] Rich feature hierarchies for accurate object detection and semantic segmentation
• [1504.08083] Fast R-CNN
• [1703.06870] Mask R-CNN
• [1512.02325] SSD: Single Shot MultiBox Detector
• [1506.02640] You Only Look Once: Unified, Real-Time Object Detection

• Orca: Differential Bug Localization in Large-Scale Services | USENIX

• XPC: Architectural Support for Secure and Efficient Cross Process Call
• Huawei Launches New Distributed Operating System, HarmonyOS
  • How Huawei’s Ark Compiler May Improve Android App Performance
• chipsalliance/rocket-chip: Rocket Chip Generator
  • #05: Agile CPU / Versioned Golang – Misreading Chat
• Home | seL4
  • #34 – Trust and Protection in the Illinois Browser Operating System – Misreading Chat
• Zircon  |  Fuchsia
• LiteOS/LiteOS: code and manual

• To unify or not to unify: a case study on unified builds (in WebKit)
• On Compiling WebKit (now twice as fast!) – Michael Catanzaro
• Jumbo / Unity builds
  • We’re removing support for the jumbo build – Google Groups
• 補足: ちょっと口がすべっていますが、Yusuke Suzukiさんは著者の一人ではあるものの、筆頭著者ではないので、この論文への貢献はどちらかといえば副次的であろうと思います

• [1804.06655] Deep Face Recognition: A Survey
• Recommendations for Facial Recognition Input Images – Amazon Rekognition
• DeepFace: Closing the Gap to Human-Level Performance in Face Verification – Facebook Research
• [1503.03832] FaceNet: A Unified Embedding for Face Recognition and Clustering
• [1801.07698v3] ArcFace: Additive Angular Margin Loss for Deep Face Recognition
• MegaFace
• MS-Celeb-1M: Challenge of Recognizing One Million Celebrities in the Real World – Microsoft Research
  • Microsoft deleted a database with 10M faces over privacy risks
• Dual-Agent GANs for Photorealistic and Identity Preserving Profile Face Synthesis

• Understanding Real-World Concurrency Bugs in Go
  • Via Understanding real-world concurrency bugs in Go – the morning paper

• How (and How Not) to Write a Good Systems Paper | USENIX
• What Makes a (Graphics) Systems Paper Beautiful
• How to Get a Paper Accepted at OOPSLA
• Patterns generate architectures | SpringerLink

• Machine Learning at Facebook: Understanding Inference at the Edge – Facebook Research
• 2 Billion Devices and Counting: An Industry Perspective on the State of Mobile Computer Architecture – Google AI
  • #54 – Tales of the Tail – Misreading Chat
• GitHub – Maratyszcza/NNPACK: Acceleration package for neural networks on multi-core CPUs
• GitHub – pytorch/QNNPACK: Quantized Neural Network PACKage – mobile-optimized implementation of quantized neural network operators

• Reality Engine Graphics
• PlayStation technical specifications – Wikipedia
• PlayStation 2 technical specifications – Wikipedia
• HPE Buys SGI for $275 Million: How Far the Mighty Have Fallen
• コンピューターグラフィックスによる津波現象のバーチャルリアリティシステム (1995)
• SGI graphics team moves to Nvidia | EE Times (1999)
• NVIDIA の GPU
  • Montrym: The geforce 6800 – Google Scholar (2005)
  • Lindholm: NVIDIA Tesla: A unified graphics and computing… – Google Scholar (2008)
  • Wittenbrink: Fermi GF100 GPU architecture – Google Scholar (2011)
  • NVIDIA Turing GPU Architecture (White paper, 2018)
• High-Performance Software Rasterization on GPUs | Research

• Bandana: Using Non-Volatile Memory for Storing Deep Learning Models – Facebook Research
• SysML 19: Assaf Eisenman, Bandana: Using Non-Volatile Memory for Storing Deep Learning Models – YouTube
• 肩透かしと不思議な乱数 – Misread Footnotes #1｜森田創｜note

• Handbook of Markov Chain Monte Carlo
  • Introduction to MCMC
• Amazon | Doing Bayesian Data Analysis, Second Edition: A Tutorial with R, JAGS, and Stan
  • 日本語訳
• Amazon | Physically Based Rendering, Third Edition: From Theory to Implementation

なお録音の不備によりいつもより音が悪いですがご了承ください。

• Towards Federated Learning at Scale: System Design – Google AI
• Towards federated learning at scale: system design – the morning paper
• SysML Conference

• Kajiya: The rendering equation – Google Scholar
• Rendering equation – Wikipedia
  • Bidirectional reflectance distribution function – Wikipedia
• SIGGRAPH 2019: Home
• What Makes a (Graphics) Systems Paper Beautiful
• ACM Transactions on Graphics (TOG) – Special Issue On Production Rendering and Regular Papers
• Amazon | Physically Based Rendering: From Theory to Implementation
  • オンライン HTML 版
  • 紹介ウェブページ
  • GitHub
• 2014 | Oscars.org | Academy of Motion Picture Arts and Sciences

Follow up

• Ten Years of Erlang

• [1602.05629] Communication-Efficient Learning of Deep Networks from Decentralized Data
• [1610.05492] Federated Learning: Strategies for Improving Communication Efficiency

Follow-up

• Arrow

• [1905.04648] Automating chaos experiments in production
• Netflix TechBlog
  • Four Reasons We Choose Amazon’s Cloud as Our Computing Platform (2010)
  • 5 Lessons We’ve Learned Using AWS – Netflix TechBlog – Medium (2010)
  • The Netflix Tech Blog: Chaos Monkey Released Into The Wild (2012, from Internet Archive)
  • FIT: Failure Injection Testing – Netflix TechBlog – Medium (2014)
  • Chaos Engineering Upgraded – Netflix TechBlog – Medium (2015)
  • Netflix Chaos Monkey Upgraded – Netflix TechBlog – Medium (2016)
• OSS Projects
  • Netflix/Hystrix: Hystrix is a latency and fault tolerance library designed to isolate points of access to remote systems, services and 3rd party libraries, stop cascading failure and enable resilience in complex distributed systems where failure is inevitable.
  • Netflix/zuul: Zuul is a gateway service that provides dynamic routing, monitoring, resiliency, security, and more.
• Distributed Tracing
  • AWS X-Ray – Distributed Tracing System
  • Stackdriver Trace – Distributed Tracing from Google – Google Cloud Platform  |  Stackdriver Trace  |  Google Cloud
• Principles of Chaos Engineering
• Game day – AWS Well-Architected Framework
• Resilience Engineering: Learning to Embrace Failure – ACM Queue
• Weathering the Unexpected – ACM Queue
• Gremlin: Chaos Engineering Tools to Break Things on Purpose

• [1203.1539] Programming with Algebraic Effects and Handlers
• Algebraic Effects for Functional Programming
• Programming and Reasoning with Algebraic Effects and Dependent Types
• yallop/effects-bibliography: A collaborative bibliography of work related to the theory and practice of computational effects
• Eff Programming Language
• Koka – Microsoft Research
• Episode 41 – IDRIS — Systems Programming Meets Full Dependent Types – Misreading Chat

follow-up

• What Spectre and Meltdown Mean For WebKit | WebKit
• IBM Research | Technical Paper Search | Let’s Not Speculate: Discovering and Analyzing Speculative Execution Attacks(Search Reports)

• A History of Erlang
• Programming Erlang (2nd edition) by Joe Armstrong | The Pragmatic Bookshelf (初版日本語訳)
• Erlang Garbage Collector | Erlang Solution blog
• HPACK: the silent killer (feature) of HTTP/2
• Concurrent Programming in Erlang (2nd Edition)
• erlang nine nines – Google Search
• The Pragmatic Programmer, 20th Anniversary Edition: your journey to mastery by David Thomas, Andrew Hunt
• Programming Elixir 1.6: Functional |> Concurrent |> Pragmatic |> Fun by Dave Thomas (初版日本語訳)
• Nintendo Switch(TM)向け プッシュ通知システム 「NPNS」 (PDF)
• voluntas/japanese-erlang-elixir-companies: 日本で Erlang/OTP や Elixir を利用している会社一覧

• [1902.05178] Spectre is here to stay: An analysis of side-channels and speculative execution
• A year with Spectre: a V8 perspective · V8

• • [1603.02754] XGBoost: A Scalable Tree Boosting System
  • dmlc/xgboost: Scalable, Portable and Distributed Gradient Boosting (GBDT, GBRT or GBM) Library, for Python, R, Java, Scala, C++ and more. Runs on single machine, Hadoop, Spark, Flink and DataFlow
  • Sparse matrices (scipy.sparse) — SciPy v1.2.1 Reference Guide
  • u++ on Twitter: “[#misreading episode 58] 自分にとっては新鮮な視点が多かった。GBDTのFrameworkとして終始「XGBoost」に言及があるけど、Kaggleでの現在の主流は「LightGBM」なので2019年の情報としては誤解を招きそう。あとアンサンブルの例としてバギングとランダムフォレストを横並びで列挙するのは微妙かも。”
  • Microsoft/LightGBM: A fast, distributed, high performance gradient boosting (GBDT, GBRT, GBM or MART) framework based on decision tree algorithms, used for ranking, classification and many other machine learning tasks. It is under the umbrella of the DMTK(http://github.com/microsoft/dmtk) project of Microsoft.
  • LightGBM: A Highly Efficient Gradient Boosting Decision Tree – Microsoft Research
  • How to train Boosted Trees models in TensorFlow – TensorFlow – Medium

Follow-up

• Cloud Run documentation  |  Cloud Run  |  Google Cloud
• 今シーズンはこのエピソードでおしまいです。またしばらくお休みします。

• Do Developers Learn New Tools On The Toilet? – Google AI
• Google Testing Blog: Introducing “Testing on the Toilet”
• Testing on the Toilet – Mike Bland
• Bathroom Reader | Scribd
• Schibsted Eng. on Twitter: “More than 15 @SchibstedGroup’s offices around the world join to Schibsted Testing on the Toilet #SchibstedCulture #STOTT #craftsmanship… https://t.co/hC3lvWleBx”

• Friedman : Greedy function approximation: A gradient boosting machine.
• Elements of Statistical Learning: data mining, inference, and prediction. 2nd Edition
• XGBoost
• A Kaggle Master Explains Gradient Boosting | No Free Hunch
• k-d tree – Wikipedia

Follow up

• 【技術書典6】「PythonとColabでできる-ゼロから作るRAW現像」を出します（書籍版／ダウンロード版） – Moiz’s journal

• Asynchronous Functional Reactive Programming for GUIs
• Elm – A delightful language for reliable webapps
• A Farewell to FRP

Follow up

• Standardizing WASI: A system interface to run WebAssembly outside the web – Mozilla Hacks – the Web developer blog
• Announcing Lucet: Fastly’s native WebAssembly compiler and runtime
• Edge Computing| CDN, Global Serverless Code, Distribution | AWS Lambda@Edge

• Cloud Programming Simplified: A Berkeley View on Serverless Computing – RISE Lab

• Functional Reactive Programming from First Principles
• Arrows, Robots, and Functional Reactive Programming
• Genuinely Functional User Interfaces
• Functional Reactive Programming – HaskellWiki
• Arrow – HaskellWiki

Follow-up

• WebAssembly Troubles part 4: Microwasm

• The Tail at Scale
• Tales of the Tail: Hardware, OS, and Application-level Sources of Tail Latency
• Two Billion Devices and Counting
• Velocity 2011: John Rauser, “Look at Your Data” – YouTube

• [1810.04805] BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
• Improving Language Understanding by Generative Pre-Training
• GitHub – openai/gpt-2: Code for the paper “Language Models are Unsupervised Multitask Learners”
• [1901.11504] Multi-Task Deep Neural Networks for Natural Language Understanding
  • Microsoft’s New MT-DNN Outperforms Google BERT – SyncedReview – Medium
• BERT with SentencePiece を日本語 Wikipedia で学習してモデルを公開しました – 原理的には可能 – データ分析界隈の人のブログ、もとい雑記帳

Follow up

• Jay Alammar جهاد العمار | LinkedIn
• STV

• [1507.00410] Convolutional Color Constancy
• Google AI Blog: Night Sight: Seeing in the Dark on Pixel Phones
• Fast Fourier Color Constancy – Google AI
• YUV – Wikipedia

• [1706.03762] Attention Is All You Need
• The Illustrated Transformer – Jay Alammar – Visualizing machine learning one concept at a time
• The Annotated Transformer
• Episode 15 – Neural Machine Translation by Jointly Learning to Align and Translate – Misreading Chat

• Geodesic Discrete Global Grid Systems
• S2 Geometry | S2Geometry
  • Announcing the S2 Library: Geometry on the Sphere | Google Open Source Blog
• Supercharging geo_point fields in Elasticsearch 2.2 | Elastic
• Space-filling curve – Wikipedia
• HTM Overview
• H3: Uber’s Hexagonal Hierarchical Spatial Index

Follow-up

• My parents are Flat-Earthers

• [1901.09056] Mind the Gap: Analyzing the Performance of WebAssembly vs. Native Code
• Episode 22 – Bringing the Web up to Speed with WebAssembly – Misreading Chat
• SPEC – CPU Benchmark Suites
• Browsix: Unix in the browser tab
• WebAssembly Troubles part 1: WebAssembly Is Not a Stack Machine

Follow-up

• The Operating System in 2018
• January Prize Spotlight: Jeff Bezanson, Steven L. Brunton, Jack Dongarra, Stefan Karpinski, and Viral B. Shah

• KDD 2018 | Real-time Personalization using Embeddings for Search Ranking at Airbnb
• [1810.09591] Applying Deep Learning To Airbnb Search

収録の都合でノイズが多くなってしまいましたがご容赦ください。

• Engineering the Servo Web Browser Engine using Rust
• Servo, the Parallel Browser Engine Project
  • servo/servo: The Servo Browser Engine
  • servo/webrender: A GPU-based renderer for the web
• Mozilla 2 – MozillaWiki (懐古趣味)

Follow up

• bashik7 on Twitter: “#misreading の最新epで話がでてたTLA+の紹介記事がタイムリーなことにHNにあがってた。この記事では(論文ではいい感じだと紹介されていた)Toolboxがイケてないって言ってて面白い。使ったことないからなんともだけどJavaだからawfulっていうのは違う気がするけど。

• An Introduction to Neural Information Retrieval – Microsoft Research
• Manning | Deep Learning for Search
• Introduction to Information Retrieval: Christopher D. Manning, Prabhakar Raghavan, Hinrich Schütze
• Learning to Rank using Gradient Descent – Microsoft Research
  • RankNet: A ranking retrospective – Microsoft Research
• Learning to Rank using Gradient Descent – Microsoft Research
• [1812.00073] TF-Ranking: Scalable TensorFlow Library for Learning-to-Rank
  • tensorflow/ranking: Learning to Rank in TensorFlow
• Learning To Rank | Apache Solr Reference Guide 6.6
  • How Bloomberg Integrated Learning-to-Rank into Apache Solr | Tech At Bloomberg
• We’re Bringing Learning to Rank to Elasticsearch

Follow up

• Kazuyoshi Kato on Twitter: “”We’re using Rust in production in several places, including high-performance metrics capture and communication protocols like CoAP.” https://t.co/orkA3R4AF5 というわけで、Firecracker に限らず AWS で Rust 使われているらしいですよ。 #misreading”

• Why Amazon Chose TLA +  | SpringerLink (Google Scholar)
  • How Amazon Web Services Uses Formal Methods
• Practical TLA+: Planning Driven Development: Hillel Wayne
• Specifying Systems: The TLA+ Language and Tools for Hardware and Software Engineers: Leslie Lamport
• TLA+tlaplus/tlaplus: TLC is an explicit state model checker for specifications written in TLA+. The TLA+Toolbox is an IDE for TLA+.

Follow up

• Grenade! Dependently Typed Neural Networks — Monday Morning Haskell
• ハイパーバイザの作り方
• エンジニアなら知っておきたい仮想マシンのしくみ：連載｜gihyo.jp … 技術評論社

• Firecracker
• firecracker-microvm/firecracker: Secure and fast microVMs for serverless computing.
• Chromium OS Docs – Running Custom Containers Under Chrome OS
• chromiumos/platform/crosvm – Git at Google
• Virtual I/O Device (VIRTIO) Version 1.0
• google/gvisor: Container Runtime Sandbox

Follow up

• Google AI Blog: Learning to Predict Depth on the Pixel 3 Phones
• MIPS Goes Open Source | EE Times
• Per Buer on Twitter: “Listening to a Japanese podcast discussing unikernels. I have no idea what they are saying but I find listening to the language soothing. From time to time they throw in an English word (“hypervisor”) and I light up.”

• Capsicum: practical capabilities for UNIX
• Capsicum – FreeBSD Wiki
• Computer Laboratory: Capsicum: practical capabilities for UNIX
• Capsicum and Casper – a fairy tale about solving security problems
• Capsicum | Hacker News
• Pledge: OpenBSD’s defensive approach to OS Security

Follow up

• Mitsutoshi Aoe/maoe on Twitter: “#misreading ep. 41でIdrisの話をしてた。最近作者は https://t.co/PANlfh8SC1 というIdris 2のプロトタイプを作っているので、これが一番実用的なアプリケーションになるのかなあ。”
  • edwinb/Blodwen: A prototype successor to Idris
• Manning | Type-Driven Development with Idris

• Virtual Machines: Versatile Platforms for Systems and Processes (The Morgan Kaufmann Series in Computer Architecture and Design): Jim Smith, Ravi Nair: 9781558609105: Amazon.com: Books
• QEMU, a Fast and Portable Dynamic Translator
  • Tiny Code Generator
  • TCC : Tiny C Compiler
• Xen and the Art of Virtualization
• x86 virtualization – Wikipedia
• Intel® 64 and IA-32 Architectures Software Developer’s Manual Volume 3C: System Programming Guide, Part 3
• kvm: the Linux Virtual Machine Monitor
• Ten years of KVM [LWN.net]
• Using the KVM API [LWN.net]
• kernel/git/will/kvmtool.git – Stand-alone kvmtool
• Hypervisor | Apple Developer Documentation
• AWS re:Invent 2017: C5 Instances and the Evolution of Amazon EC2 Virtualization (CMP332) – YouTube
• intel/haxm: Intel® Hardware Accelerated Execution Manager (Intel® HAXM)
• virtio: Towards a De-Facto Standard For Virtual I/O Devices
• redox-os / redox · GitLab
• MonaOS – Free Operating System
• KVM/ARM: The Design and Implementation of the Linux ARM Hypervisor

Addendum

• Xen の Linux カーネルへの変更は10年くらい前に大体おわっているようです。(参考)

感想などはハッシュタグ #misreading か [email protected] にお寄せください。

• IDRIS — Systems Programming Meets Full Dependent Types (PDF)
• Idris | A Language with Dependent Types
• Kazuyoshi Kato on Twitter: “最近の言語で dependent types… 多分 Idris じゃないかと https://t.co/tRLdzd0KAr #misreading”

Follow up

• skip · A programming language to skip the things you have already computed
• Chromeのメモリ割り当てフック機構 – kanejaku.org

Follow up

• California Hasn’t Seen Fires Like This: Pictures of a State in Flames – The New York Times
• subword vocabulary of chromium class names
  • subwords of chromium class names
• The Commons: The Past Is 100% Part of Our Future | Flickr Blog
• 15-463, 15-663, 15-862 Computational Photography, Fall 2018
• SynthCam
• Extreme imaging using cell phones (PDF)
• ゼロから作るRAW現像 – まとめページ – Moiz’s journal

Mukai’s Reading List

• 形態素解析の理論と実装 (実践・自然言語処理シリーズ)
• SRE サイトリライアビリティエンジニアリング ―Googleの信頼性を支えるエンジニアリングチーム
• The Little Typer | The MIT Press

Morrita’s Reading List

• A Tour of C++ (2nd Edition) (C++ In-Depth Series)
• Computer Vision: Algorithms and Applications (Texts in Computer Science)
  • Computer Vision: Algorithms and Applications – PDF 版がダウンロード可能
• The Manager’s Path: A Guide for Tech Leaders Navigating Growth and Change

• Service Fabric: A Distributed Platform for Building Microservices in the Cloud
• Microsoft/service-fabric: Service Fabric is a distributed systems platform for packaging, deploying, and managing stateless and stateful distributed applications and containers at large scale.
• Chord: A Scalable Peer-to-peer Lookup Service for Internet Applications
• Azure Service Fabric—Building Microservices | Microsoft Azure
• Programming Microsoft Azure Service Fabric (2nd Edition) (Developer Reference): Haishi Bai: 9781509307098: Amazon.com: Books
• Episode 241 – Service Fabric & Service Fabric Mesh

• Neural Machine Translation of Rare Words with Subword Units
• [1804.10959] Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates
• google/sentencepiece: Unsupervised text tokenizer for Neural Network-based text generation.
• vocabulary for chromium class names
  • subword examples

• WebAssembly Threads ready to try in Chrome 70
  • WebAssembly/threads: Threads and Atomics in WebAssembly

• [1806.04171] Synthetic Depth-of-Field with a Single-Camera Mobile Phone

• BLeak: Automatically debug memory leaks in web applications
• PLDI ’18 paper

Follow-up

• Python’s New Package Landscape
  • Pipenv ※本編では誤って Pypienv と読んでいますが Pipenv  の間違いです。

• Darkroom: Compiling High-Level Image Processing Code into Hardware Pipelines
  • Rigel: Flexible Multi-Rate Image Processing Hardware
• [1610.09405] Programming Heterogeneous Systems from an Image Processing DSL
• Computer Architecture: A Quantitative Approach (The Morgan Kaufmann Series in Computer Architecture and Design): John L. Hennessy, David A. Patterson
•  Pixel Visual Core: image processing and machine learning on Pixel 2
• iPhone XS – Cameras – Apple

Correction

• Processing Element を間違えて Processing Unit と読んでいる部分があります
• Halo の幅が 4 ピクセル, Processing Element が 16×16 ピクセルで、Stencil Unit 毎の PE/Halo 合計は (20×20 ではなく) 24×24 ピクセルです。

• Trust and Protection in the Illinois Browser Operating System
• Embassies: Radically Refactoring the Web – Microsoft Research
• L4Android: a generic operating system framework for secure smartphones
  • L4Linux
  • L4Android
• L4 Kernel projects
  • On micro-kernel construction
  • From L3 to seL4 what have we learnt in 20 years of L4 microkernels?
  • seL4: formal verification of an OS kernel

follow-ups

• バスケ on Twitter: “WebView は結構前にDeprecated で次ぐらいで撤去のはず。…
  • UIWebView (deprecated)
• Inside look at modern web browser (part 1) | Web | Google Developers
  • (ちなみに前回のshow notesにもしれっとリンクを貼っていましたがまったく言及していませんでした。すいません）
• Creative Selection
• MinRT by WebAssembly
  • NodeでWebAssembly誰得なのか、みたいなことを話していますが、アーキテクチャやABI等に関係なく再コンパイルなしで使えるという面で好まれることはありうるかもしれません

• Distinctive Image Features from Scale-Invariant Keypoints
• Computer Vision: Algorithms and Applications (日本語訳) – 教科書
• [1805.04140] Neural Best-Buddies: Sparse Cross-Domain Correspondence
  • Neural Image Stitching And Morphing | Two Minute Papers #256 – YouTube

Follow-up

• Beyond Interactive: Notebook Innovation at Netflix – Netflix TechBlog – Medium

• Isolating Web Programs in Modern Browser Architectures
  • The Security Architecture of the Chromium Browser
• Chromeのプロセス分離関連
  • Site Isolation
  • Inside look at modern web browser
  • https://twitter.com/kinu/status/1037508916861591552
• Firefox Electrolysis
• Webkit vs Webkit2 API
• 関連研究
  • Tahoma browser: A Safety-Oriented Platform for Web Applications
  • The Multi-Principal OS Construction of the Gazelle Web Browser
• HAL Types  |  Android Open Source Project Treble のプロセスモデルについて
• Android 8.0 Behavior Changes  |  Android Developers WebView のプロセス分離

follow-up

• Kazuyoshi Kato on Twitter: “Rust の場合 out of bounds は実行時に panic, use after free は静的に ownership で防ぐ、はず。ちなみに llvm 経由で各種 san も使えます。たぶんこれは unsafe むけで、レポジトリを持ってる japaric さんは Rust Embedded ワーキンググループのリードです。
• 眼力 玉壱號 on Twitter: “#misreading MSVC に無いのがツライけど、clang-cl 経由で使えるみたいな話を見た様な…

• big.LITTLE Technology: The Future of Mobile
• Single-ISA Heterogeneous Multi-Core Architectures
• Apple-designed processors – Wikipedia
• Snapdragon 808 Processors with X10 LTE Specs and Details (Nexus 5X)
• Snapdragon 821 Mobile Platform with X12 LTE and Adreno 530 GPU (Pixel 1)
• Scheduling for Android devices [LWN.net]
• Energy Aware Scheduling on Android
• Energy Aware Scheduling [LWN.net]
• EAS Overview and Integration Guide
• The Linux Scheduler: a Decade of Wasted Cores (The morning paper)
• Evaluation of Hybrid Run-Time Power Models for the ARM Big.LITTLE Architecture – IEEE Conference Publication
• Documentation/scheduler/sched-tune.txt – kernel/common – Git at Google
• IFA 2018: Huawei Kirin 980 Keynote Live Blog

Follow-up

• I Don’t Like Notebooks – Joel Grus – #JupyterCon 2018 – Google Slides
• lisa/ipynb/tutorial at master · ARM-software/lisa
• Kazuyoshi Kato on Twitter: “Lisp 入ってるのいい話。さがしたら Discourse にスレがあった。 https://t.co/e4N0RQr5jc #misreading”

• AddressSanitizer: A Fast Address Sanity Checker | USENIX
• https://github.com/google/sanitizers
• Clang 8 documentation ADDRESSSANITIZER
• Valgrind
• CPU2006

Follow-ups

• masayoshi takahashi on Twitter: “#misreading 28 赤黒木と左傾赤黒(LLRB)木の話は最近話題の『みんなのデータ構造』にも載ってて、TreapとかScapegoat木との比較もあります(実装は簡単だけど最悪ケースだと赤黒木の方が優秀だそう) https://t.co/I65lJRl0EJ”

• Image Analogies
• Image Style Transfer Using Convolutional Neural Networks
• [1603.08155] Perceptual Losses for Real-Time Style Transfer and Super-Resolution
• Neural Transfer with PyTorch — PyTorch Tutorials 0.4.1 documentation
• Correction: リアルタイム版の Style Transfer は training の際にホワイトノイズと content 画像を与えると言っていますが、実際には content 画像のみが入力です。該当論文の fig.2 を参照ください。要するに様々な写真を与えそれを stylize した画像が生成されるようネットワークを train します。

Follow-up

• I Python, You R, We Julia – Jupyter Blog
• Leading NLP Ninja | Anchor – The easiest way to start a podcast

• Left-leaning Red-Black Trees
• Robert Sedgewick
• Red-black tree
• 2-3-4 tree
• A dichromatic framework for balanced trees — the original paper for red-black tree

• [1411.1607] Julia: A Fresh Approach to Numerical Computing
• The Julia Language
• Julia 1.0
• Home · JuliaDB
• The 5th annual JuliaCon 2018 (London) – YouTube
• https://juliabox.com

Follow-up

• Nvidia unveils Turing architecture and GPUs with dedicated ray-tracing hardware – The Verge
• Introduction to DirectX RayTracing
• nvidia hq – Google Search
• 眼力 玉壱號 on Twitter: “#misreading ep.23 #Halide 編… to FPGA やってる Fixstars の話が無かったのはちょっと悲しい…が、paper を解説するというのなら正しいかな cf. https://t.co/cyyhWcMGtd”
• イベントレポート: 「Halide勉強会」を開催しました – Fixstars Tech Blog /proc/cpuinfo
• Darkroom: Compiling High-Level Image Processing Code into Hardware Pipelines

• DROPS – Mixed Messages: Measuring Conformance and Non-Interference in TypeScript
• Understanding TypeScript
• Safe & Efficient Gradual Typing for TypeScript

Follow-up

• Kinuko Yasuda on Twitter: “#misreading 24 割とよくみてるとこなのでコメント…SharedArrayBufferは違うドメイン間ではメモリ共有できなくてよいはず。最近のESではAgent Clusterというのがあってそれがメモリ共有できる単位になっており、HTML側では基本origin違うと共有できなくていいとなってます https://t.co/ZFVPSfvpfl”
• human eslint –fix on Twitter: “#misreading SharedArrayBuffer のメモリモデルがなぜ必要という話、Worker同士の共有モデルのためって説明されたけど、Workerはあまり使われておらず、WebAssemblyプロセスが実質Workerなのでメモリ共有に必要、という感じではないか、と思った”
• Meltdown and Spectre
• WebAssembly.compile() – JavaScript | MDN
• Threading proposal for WebAssembly
• Behind the tech podcast
  • Kevin Scott

• Burst photography for high dynamic range and low-light imaging on mobile cameras
• Google AI Blog: HDR+: Low Light and High Dynamic Range photography in the Google Camera App
• Phase correlation – Wikipedia
• Exposure Fusion
• Correction: ダイナミックレンジを解像度と言ったりビットレートと言ったりしている箇所が何箇所もありますが・・・ごめんなさい。今回は解像度やビットレートに関する話をしている部分はなく、それらはみなダイナミックレンジを指しています。

• EMME: A Formal Tool for ECMAScript Memory Model Evaluation | SpringerLink
• SharedArrayBuffer – JavaScript | MDN
• Mitigating Spectre with Site Isolation in Chrome
• SharedArrayBuffer and timing attacks (Meltdown and Spectre) · Issue #3 · tc39/security

Follow-up

• tc39/ecmascript_simd: SIMD numeric type for EcmaScript
• bashik7 on Twitter: “#misreading wasm回。関数テーブル書き換えられない、みたいな話があった気がするけど、書き換えられるような。https://t.co/kqY9KXidmE”

• Halide: Decoupling Algorithms from Schedules for High-Performance Image Processing CACM 2017
• Decoupling Algorithms from Schedules for Easy Optimization of Image Processing Pipelines – SIGGRAPH 2012
• Decoupling Algorithms from the Organization of Computation for High Performance Image Processing – Ph.D Thesis version. いちばんわかりやすい.
• http://halide-lang.org/
• GitHub – halide/Halide: a language for fast, portable data-parallel computation
• OpenTuner – An extensible framework for program autotuning
• Differentiable Programming for Image Processing and Deep Learning in Halide
• Buck: halide_library() – Facebook の build system には halide の native support がある。

Follow-up

• Domain Name from Company With $215m+ in Funding in Auction | DomainInvesting.com
• Kazuyoshi Kato on Twitter: “Lytro 欲しかった好事家の人々は Light L16 を買うといいのでは。論文あるっぽいけど無料アクセスなさそうなので読んでいないです。 https://t.co/FEBJADDswl”
  • Light Home | Light
  • Light Solutions | Light

• Bringing the Web up to Speed with WebAssembly
• webassembly.org
• asm.js
• emscripten
• Performance Testing Web Assembly vs JavaScript
• JavaScript vs WebAssembly easy benchmark – Takahiro
• PNaClについて、ポッドキャスト内で言及するといっておいて言及し忘れたのでかるくふれておきます
  • PNaClはGoogle ChromeにおけるNative Client (NaCl)という拡張をよりポータブルにしたものです。具体的には、NaClはネイティブコードを配布して検証・実行をしますが、PNaClでは配布データがllvm bitcodeとなり、これを検証・実行します
  • 論文中ではasm.jsとならんで直接的な先祖のひとつとされていました。過去にはPNaCl形式のバイナリをウェブページから配布すればChromeは実行できる、といったことを目標にしていました。現在はwebassemblyにフォーカスを移し、chromeからも使えなくなる（かもう使えなくなった）はず
  • webassemblyになぜllvm bitcodeのような既存のIRを使わなかったのか、ということも（おそらくPNaCl関係の著者により）論文中に書かれていますので、興味のある方は論文内related workの節をご参照ください

Follow-up

• Yutaka Matsubara on Twitter: “そーいえば結局vgoがAPIベースのアクセスやめた話出てなかったな。modulesは少し触れられてたけど #misreading”
• Russ Cox on Go dependency management | Hacker News
  • A response about dep and vgo | Hacker News
• SREcon Conferences | USENIX
  • SREcon18 Americas – YouTube

• The Frankencamera: An Experimental Platform for Computational Photography
• FCam API – Home
• android.hardware.camera2  |  Android Developers
• Nokia N900 – Wikipedia
• apple camera Jobs | LinkedIn
• The iPhone’s camera is so good because 800 people are working on it – The Verge
• Correction: API で指定する画像のフォーマットについて JPEG や RAW といってますが YUV や RGB を指定するパラメタです。
• Addendum: エッジケースでデバイス間の差異が大きい Camera2 API ですが、導入が進まない大きな原因は Lollipop 以上というバージョンの制限です。一方で使い方によっては製品のカメラ機能を強化できる可能性もあります。Google I/O 2018 のトークでは Snapchat のエンジニアが同製品内での Camera2 の利用について紹介しています。

Follow-up

• Julien Verlaguet – Reflex: Reactive Programming at Facebook – YouTube
  • Curry On! – YouTube
• YOW! 2017 Julien Verlaguet – Reflex: A New Reactive Programming Language at Facebook – YouTube – 比較的新しい techtalk (2017 年末)

• The Design and Implementation of Hyperupcalls | USENIX
• Hypercall
• Upcall
• Linux Enhanced BPF (eBPF) Tracing Tools

Follow-up:

• Taking Go modules for a spin | Dave Cheney
• Walt Disney Animation Studios – Research publications
• Swallowing the elephant: all the links

• Light Field Rendering
• Light Field Photography with a Hand-Held Plenoptic Camera
  • 画像を見ると一目瞭然なことも多いので一瞥をおすすめします。
  • DIGITAL LIGHT FIELD PHOTOGRAPHY Ph.D Thesis – サンプル画像多数。
• Lytro – Wikipedia
• Photosynth – Wikipedia
• Experimenting with Light Fields – Light Field + VR
• plenoptic – Wiktionary
• Google AI Blog: Lens Blur in the new Google Camera app
• Google AI Blog: Portrait mode on the Pixel 2 and Pixel 2 XL smartphones

Follow-up

• State of Multicore OCaml [pdf] | Hacker News
• Eager Execution  |  TensorFlow
  • 補足: AutoGraph converts Python into TensorFlow graphs – TensorFlow – Medium を使うと eager と non-eager どちらでも動くコードを書けるらしい。
• ARM: “RISC-V Architecture: Understand the Facts” | Hacker News
  • Up in arms! Arm kills off its anti-RISC-V smear site after own staff revolt • The Register
  • ARM Architecture: Understand the Facts
  •  Microsoft kills its ‘Get the Facts’ anti-Linux site | ZDNet

• Project Zero: Exploiting the DRAM rowhammer bug to gain kernel privileges
  • Flipping Bits in Memory Without Accessing Them: An Experimental Study of DRAM Disturbance Errors – もともとrowhammerバグを報告した論文
  • Native Client
• 補足：ポッドキャストでは口頭で説明しているため、細かい発生条件などを省略・簡略化しています。詳細はリンク先の記事を読んだほうが良いでしょう……
  • Native Clientで動的に実行可能なメモリ領域をつくるためには、記事中ではdyncode_create()というAPIを使ったと書いてあります

Follow-up

• USENIX ATC ’18
  • Throwhammer: Rowhammer Attacks over the Network and Defenses
• Migrating Messenger storage to optimize performance – Facebook Code

• Unikernels: Library Operating Systems for the Cloud
• MirageOS
• Drawbridge – Microsoft Research
• About Lwt: concurrency library – Ecosystem – OCaml
• Xen vchan port
• Moby
  • moby/vpnkit: A toolkit for embedding VPN capabilities in your application
  • moby/hyperkit: A toolkit for embedding hypervisor capabilities in your application
  • moby/datakit: Connect processes into powerful data pipelines with a simple git-like filesystem interface
• OSv—Optimizing the Operating System for Virtual Machines
  • cloudius-systems/osv: OSv, a new operating system for the cloud.
• ScyllaDB
• Using functional programming within an industrial product group: perspectives and perceptions
  • Reason · Reason lets you write simple, fast and quality type safe code while leveraging both the JavaScript & OCaml ecosystems.
  • Pyre · A performant type-checker for Python 3
  • Flow: A Static Type Checker for JavaScript
  • BackType | Crunchbase
  • ITA Software – Wikipedia
  • (TuringComplete.fm/14 3:11-でGoogle内で使われているCommon Lispの話をしています）

Follow up

• John Hennessy and David Patterson 2017 ACM A.M. Turing Award Lecture – YouTube
• on Twitter: “HyperLogLogってなんか聞いたことあると思ったらRedisが機能として入れてたからか #misreading”
• wyukawa on Twitter: “big data界隈ではよく出てきますな。prestoだとapprox_distinctでelasticsearchだとcardinality aggregation / “Episode 13 – HyperLogLog in…” https://t.co/sQtewjVnvU”

Announcement

• ちょっとおやすみして、次回は一ヶ月後くらいの予定です。

• A Deep Learning Approach for Generalized Speech Animation – Disney Research
• Active Appearance Model [PDF]

Follow up

• Disney Research
• SIGGRAPH 2018
• Kinuko Yasuda on Twitter: “#misreading 今回は levelDB だ！昔少し調べたのと使ってる側の話を割と聞くので、軽くコメント。Client側でこの設計でいいのか？→普通に制約になり得る。メモリに入り切らない変更（大量のindex一括更新など）はatomicにできないし、そもそもメモリを食う。

今回は根性不足によりあまり編集しておりません。聞き苦しい部分はご容赦ください。

• [1409.0473] Neural Machine Translation by Jointly Learning to Align and Translate
• The Unreasonable Effectiveness of Recurrent Neural Networks
• [1709.07809] Neural Machine Translation
  • Statistical Machine Translation: Philipp Koehn: 9780521874151: Amazon.com
• Attention and Augmented Recurrent Neural Networks
• The Great A.I. Awakening – The New York Times
• A novel approach to neural machine translation | Engineering Blog | Facebook Code | Facebook
• [1706.03762] Attention Is All You Need 全然わからん
• Translation with a Sequence to Sequence Network and Attention — PyTorch Tutorials 0.4.0 documentation

• google/leveldb: LevelDB is a fast key-value storage library written at Google that provides an ordered mapping from string keys to string values.
• Bigtable: A Distributed Storage System for Structured Data
• Kyoto Cabinet
• dain/leveldb: Port of LevelDB to Java
• Skip List

Follow-up

• research!rsc: The vgo proposal is accepted. Now what? (Go & Versioning, Part 8)
• research!rsc: What is Software Engineering? (Go & Versioning, Part 9)
• 125: GraphQL and Apollo with GDE Mike Nakhimovich (Part 1) – Fragmented
• Reconciling GraphQL and Thrift at Airbnb – Airbnb Engineering & Data Science – Medium

• HyperLogLog in Practice: Algorithmic Engineering of a State of The Art Cardinality Estimation Algorithm – Google AI
• HyperLogLog: the analysis of a near-optimal cardinality estimation algorithm
• Loglog Counting of Large Cardinalities
• Probabilistic Counting Algorithms for Data Base Applications
• Harmonic mean – Wikipedia
• Three and a half degrees of separation – Facebook Research
• Mining of Massive Datasets
  • 大規模データのマイニング

Follow up

• Literate Programming (Lecture Notes): Donald E. Knuth: 9780937073803: Amazon.com: Books
  • Literate programming – Wikipedia
  • 文芸的プログラミング
• PyTorch, Chainer and Forking – Misreading Chat 補足
• Discover dev

• Semantics and Complexity of GraphQL
• GraphQL
• the morning paper / Semantics and Complexy of GraphQL
• NL-complete (日本語)

Correction

• エピソード内でクエリ結果のサイズ評価について議論していますが、論文をちらっと読んだかんじでは向井さんの主張が正しそうです。自分はなんとなくスキーマの ER 図みたいのを想像しながら話を聞いていましたが、この人のいうグラフはオブジェクトDBみたいなグラフで、何かを評価する際にスキーマなどはおまけみたいなものなんですね。思い込みで話がかみ合っていませんでした。(森田)

まず PyTorch のソースコードや Git レポジトリの log を見るかぎり, ライセンスやコメント、コミットログや明らかなコードの重複など、はっきりとしたフォークの証拠はありません。(API の模倣については当事者による論文のドラフトなど何箇所かで言及されています。)

ウェブを検索してみると、以下のツイートが証言としてリンクされていました。

• James Bradbury on Twitter: “Yep, the PyTorch autograd codebase started with a fork from Chainer — but then rewrote it in highly optimized pure C… “

他の関係者の発言は見つかりませんでした。唯一の証言であるこの J 氏のツイートを信じるなら、フォークされたのはどうも autograd だけのようです。一方で、そもそもフォークではないと主張する証言もみつかります。

• PyTorch was never a Chainer fork. The whole codebase are C libs from Lua Torch, … | Hacker News

どちらの主張が正しいのでしょうか。

フォークを主張する J 氏は PyTorch への数個のコミット履歴をもつ、コードレベルでは比較的カジュアルなコントリビュータです。一方フォークでないと主張する A 氏は継続的なコミットを続ける top contributor。件の autograd コードを最初にコミットとしたのもこの A 氏です。発言者の credibility という点では、フォークでないと主張する A 氏に軍配があがるのではないでしょうか。

はっきりとしたコードの類似性がないと書きましたが、実物を比べてみましょう。PyTorch での最初の Autodiff 実装はこのコミットです。一方、同時期の Chainer (1.13.0) の実装はこれ。グラフのトラバース先をキューにつめながら反復的に gradient を解決していく基本のアプローチは同じですが、表層的なコードの様子は随分違うのがわかると思います。

これは無理もない話です。Autograd のコードはどうしてもベースとなる Tensor 計算用の API に引きずられます。Chainer はその計算に CuPy というライブラリを使っています。(先に登場したバージョン 1.13.0 の時点ではまだ Chainer の一部でした。) 一方の PyTorch は Lua Torch 由来の C ライブラリをラップして使っています。

どちらも NumPy を模している点では似ていますが、drop in の互換性があるとは思えません。特に初期の PyTorch に含まれる Tensor 実装は極めて制限されたものでした。PyTorch にチェックインされた時点の autograd はごく短い単純な実装でしたから、互換性の低い CuPy を Torch に入れ替えて使えるようにするよりは Torch 向けに自分で書いてしまった方が簡単に見えます。

フォークというのはソフトウェアとして切り出しやすい単位で行わないと有り難みがなく、Autograd はその単位としてやや筋悪なのです。そういう意味でも “Autograd のコードがフォークである” とする J 氏の証言は疑わしく思えます。なお J 氏は同じツイートで “のちにコードは pure C の高速バージョンに書き直された” と言っていますが、これも厳密には間違いです。現状の PyTorch の autograd は C++ で書かれており、かつ演算単位では Python にフォールバックする可能性もある “pure C” とは程遠いものだからです。

というわけで PyTorch の Chainer フォーク説はよく似せられた API とひとつのツイートが生み出した都市伝説であるというのが自分の結論です。

フォークかどうかはさておき、API のようなソフトウェアデザインの重要な部分で PyTorch が Chainer を真似しているのは事実です。これはサンプルで引き合いにだされる NN モジュールだけでなく、 Variable や Function といったクラス、forward, backward メソッドの在処など autograd のレイヤについても言えます。ただしこれらが純粋に Chainer 由来なのか、Harvard の NumPy Autograd のような他の実装にも共通するパターンなのかについては、もう少し調査しないとわからないですね。

追記

記事を書かれた中田さんからお返事をいただきました。論拠となったのは先のツイートではなく、Facebook AI の中の人の Reddit 上の発言とのことです。

• The Story in the Notebook: Exploratory Data Science using a Literate Programming Tool
• Exploration and Explanation in Computational Notebooks
• Good enough practices in scientific computing

• The Scientific Paper Is Obsolete. Here’s What’s Next. – The Atlantic
• What Is a Computational Essay?—Stephen Wolfram Blog
• Project Jupyter with Jessica Forde, Yuvi Panda and Chris Holdgraf | Google Cloud Platform Podcast
• Hello, Colaboratory – Colaboratory
• CamDavidsonPilon/Probabilistic-Programming-and-Bayesian-Methods-for-Hackers: aka “Bayesian Methods for Hackers”: An introduction to Bayesian methods + probabilistic programming with a computation/understanding-first, mathematics-second point of view. All in pure Python
• Bayesian Methods for Hackers: Probabilistic Programming and Bayesian Inference (Addison-Wesley Data and Analytics)

• [1701.03757] Deep Probabilistic Programming
• [1804.06458] Deep Probabilistic Programming Languages: A Qualitative Study
• Edward
• Pyro
• WebPPL
• Stan
• PyMC

Follow-up

• #misreading だとGANを「ガン」と呼んでるけど弊社のデータサイエンティストは「ギャン」と呼んでたから、結局どっちなのか気になる。
• [1511.06434] Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Algorithms Behind Modern Storage Systems – ACM Queue

• [1502.05767] Automatic differentiation in machine learning: a survey
• Automatic differentiation in PyTorch | OpenReview
• Chainer: a next-generation open source framework for deep learning
• Implementation of Control Flow in TensorFlow
• pytorch/pytorch: Tensors and Dynamic neural networks in Python with strong GPU acceleration
  • Release alpha-1 release · pytorch/pytorch aka. v0.1.1
• Yann LeCun on Twitter: “Dude, there were graph-based deep learning frameworks with AD back in 1991 (SN at AT&T) and its open source successor Lush, released in 2002.… https://t.co/sL2wPnaOJY”
• 深層学習 (Goodfellow の教科書日本語訳)

Follow-up and Correction

• Netflix Research
• How to Adopt Modern C++17 into Your C++ Code – YouTube
  • Amazon.com: A Tour of C++
• Carnegie Mellon Launches Undergraduate Degree in Artificial Intelligence | Carnegie Mellon School of Computer Science
  • Curriculum | Carnegie Mellon School of Computer Science

• PyTorch の論文は 2017 の Autodiff Workshop で発表されていたようです。
• 10:25 あたりで計算途中の変数が Tensor になっていると言っていますが、Variable の間違いです。

• Generative Adversarial Nets
• 深層学習 (Goodfellow の教科書日本語訳)
• Image-to-Image Translation with Conditional Adversarial Nets
• Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

Follow-up

• Plan 9 from User Space
• Kazuyoshi Kato on Twitter: “#misreading Ninja の話は前回ふれられてた AOSA シリーズにあるので気になる人はぜひ。Bazel というか Blaze は Pants (Twitter) や Buck (Facebook) などクローンの存在が根強い人気をうかがわせるけど、Googler じゃないので今ひとつ何が良いのかわからないのだった。 https://t.co/C3hfCNSqMI”
• rsc.epub
• How I Fail – Ian Goodfellow (PhD’14, Computer Science) | veronikach.com
• 今回は音声データの後処理に Turing Complete FM 推薦の Auphonic を試しています。
• なかでVAEのことを言い間違えていますが、正しくはVariational Auto Encoderです。

なぜこの性質が重要なのでしょうか。論文中の例にならい “+” 演算子を実装した AST ノード AddNode の例で考えてみましょう。この AddNode は引数が数値型か文字列か、あるいは独自の “+” を定義したクラスのオブジェクトかで挙動が変わるとします。もしコンパイラがコード単位で、たとえば AddNode#execute() という単位で実行時情報・・・ここでは引数の型・・・を集めるなら、あるプログラムの中に数値のプラスと文字列のプラスが現れた時点で AddNode#execute() は多態的なコードを生成する必要が生まれます。

Ruby, JS など動的なゲスト言語の AST は、ホスト言語である Java から見るとどうしても多態的になってしまうです。

Truffle では、文字列を扱う関数 F と数値を扱う関数 G の中で登場するプラス演算子の AST ノードがそれぞれ独立して実行時の情報を集めます。おかげで F の中では文字列を連結するコードを、G の中では数値を加算するコードを、生成することができます。

一般的なプログラムではあまり重視されていないオブジェクト単位の実行時プロファイルが、どうして Truffle では重要なのでしょうか。それは Truffle の AST がゲスト言語にとっての「コード」だからです。AST インタプリタとして実装されたゲスト言語がふつうの最適化コンパイラがやるようなコード単位の実行時プロファイルを集めたいと思ったら、AST の上にそのデータを置くのは自然なことです。

Truffle はそうしたデザインを前提として、最適化のためのヒントである実行情報をAST のオブジェクトとの間でやりとりする API を提供しました。それらの API をある種の instrincis として特別に解釈し、ふつうのコンパイラにはできないような最適化が可能になるのです。

• One VM to rule them all
• GraalVM
• graal/Publications.md at master · oracle/graal
• LLVM Project Blog: FTL: WebKit’s LLVM based JIT
  • Introducing the WebKit FTL JIT | WebKit
• Two Minute Papers – YouTube – YouTube

Follow-up and Correction

• わかりにくかった部分の補足を書きました。
• 冒頭で episode 6 後半といいましたが混乱しそうなので episode 7 ということにします。
• 何箇所か execuse と呼んでいる部分は execute の間違いです。

• CloudBuild: Microsoft’s Distributed and Caching Build Service
• Travis / Jenkins
• Perforce / TFVC (Team Foundation Version Control)
• Formalizing and Verifying a Modern Build Language
• nmake
• MSBuild
• chromium build systems
  • ninja
  • GN
  • gyp
  • SCons
  • goma
• bazel

Follow-up

• Russ Cox on changelog
• Implementing Regular Expressions
• Monotonic Elapsed Time Measurements in Go

An Agile Approach to Building RISC-V Microprocessors

• RISC-V
  • RISC-V Foundation | Instruction Set Architecture (ISA)
  • Design of the RISC-V Instruction Set Architecture
  • Intel and the x86 Architecture: A Legal Perspective – Harvard Journal of Law & Technology
  • Instruction Sets Want To Be Free: A Case for RISC-V – YouTube
• アジャイルソフトウェア開発宣言
• Chisel: Constructing Hardware in an Scala Embedded Language
  • The fun of programmingは『関数プログラミングの楽しみ』として翻訳あり。8章がLavaというハードウェア記述DSLの話。
  • Verilog is weird
• GitHub – freechipsproject/rocket-chip: Rocket Chip Generator
• SiFive | CPU IP and Custom SoCs from the Inventors of RISC-V
• Esperanto Technologies / Energy-efficient RISC-V solutions for AI
• NVIDIA Is Building Its Next-Gen Falcon Controller Using RISC-V – Phoronix
• Home Page for UCB CS152/CS252, Spring 2018
• RISC-V Day 2017 Tokyo – December 18, 2017

Go & versioning

• Russ Cox
• vgo – GoDoc
• dep
• bundler
• go-get / Remote import paths
• research!rsc: Version SAT

Miscellaneous

• しげみ on Twitter: “Novaは2017年にも後続の論文が出てて、mmapに対しても実装が進んでいる PCOMMITはWPQによっていらない子に #misreading”
  • Hardening the NOVA File System
• Micron Launches New NAND-based DIMMs, Intel Announces Optane DIMMs – ExtremeTechIntel To Launch 3D XPoint DIMMs in 2H 2018
• Carnegie Mellon Solves 12-Year-Old DARPA Grand Challenge Mystery – IEEE Spectrum

Correction

• Maven 3 Processor のクロック切り替え所要時間を 20 秒と言っていますが 20 ナノ秒の間違いです。20 秒で切り替わっても何も嬉しくないですね・・・
• SATの正式名称は Boolean Satisfiability Problemでした。言い間違えました。

NOVA: A Log-structured File System for Hybrid Volatile/Non-volatile Main Memories

• GitHub – NVSL/linux-nova: NOVA is a log-structured file system designed for byte-addressable non-volatile memories, developed at the University of California, San Diego.
• Intel® Optane Technology Intel 3D XPoint
• Direct Access for files (Linux Kernel)
• CLFLUSHOPT—Flush Cache Line Optimized
• CLWB — Cache Line Write Back
• How to Emulate Persistent Memory Using Dynamic Random-access Memory (DRAM) | Intel® Software

Stanley: The Robot that Won the DARPA Grand Challenge

• Driverless milestone: No Hands Across America – San Francisco Chronicle
• [1704.05519] Computer Vision for Autonomous Vehicles: Problems, Datasets and State-of-the-Art
• PROBABILISTIC-ROBOTICS.ORG
• How simulation turns one flashing yellow light into thousands of hours of experience
• Waymo and GM Cruise: How self-driving car companies compare on miles driven — Quartz
• おまけ Carnegie Mellon Solves 12-Year-Old DARPA Grand Challenge Mystery – IEEE Spectrum

Miscellaneous:

• Computer Chip Visionaries Win Turing Award – The New York Times
• ACMチューリング賞講演集 | 赤 摂也 |本 | 通販 | Amazon
• ACM A.M. Turing Award Lectures – YouTube – YouTube
  • Michael Stonebraker 2014 ACM A.M. Turing Lecture, June 13 2015 – YouTube
• Reflection on Trusting Trust [PDF]

Ramp-up Journey of New Hires: Tug of War of Aids and Impediments

• Ramp-up Journey of New Hires: Tug of War of Aids and Impediments 2015
• Ramp-up Journey of New Hires: Do strategic practices of software companies influence productivity? 2017
• Struggles of New College Graduates in their First Software Development Job – Microsoft Research 2008

Data Scientists in Software Teams: State of the Art and Challenges

• Data Scientists in Software Teams: State of the Art and Challenges
• Analysing the Analyzers (PDF)
• SCOPE: Easy and Efficient Parallel Processing of Massive Data Sets (PDF)
• Using the Microsoft TLC Machine Learning Tool | James D. McCaffrey
• Search Quality Meeting: Spelling for Long Queries (Annotated) – YouTube

Misc.

• Microsoft Research Podcast
  • Functional Programming Languages and the Pursuit of Laziness with Dr. Simon Peyton Jones
• Perspectives on Data Science for Software Engineering: Tim Menzies, Laurie Williams, Thomas Zimmermann: 9780128042069: Amazon.com: Books
• The Art and Science of Analyzing Software Data 1, Christian Bird, Tim Menzies, Thomas Zimmermann, eBook – Amazon.com
• It Will Never Work in Theory
• The Architecture of Open Source Applications

Correction

• 雑談の中で The Architecture of Open Source Applications に Subversion に関する章があるという話をしましたが、Beautiful Codeの間違いでした。Subversion の章はオンラインで読むことができ、日本語訳もあります。この二冊はどちらも同じく Greg Wilson 氏による編集。道理で似てるわけです（と言い訳）。

Inter-app Communication in Android:

• Inter-app Communication in Android: Developer Challenges (PDF)
• Intent
• IC3
• F-Droid
• ADW Launcher
• ZXing
• openintents.org
• Inter-App Communication from Apple Developer

OptiX:

• OptiX: A General Purpose Ray Tracing Engine
• NVIDIA® OptiX Ray Tracing Engine | NVIDIA Developer
• Announcing Microsoft DirectX Raytracing! – DirectX Developer Blog
• Introduction to NVIDIA RTX and DirectX Raytracing | NVIDIA Developer Blog
• OpenGL – The Industry Standard for High Performance Graphics
• Graphics Pipeline (Windows)
• Lecture 11 – “GPU Ray Tracing (1)” from Advanced Graphics 2016/2017
• Understanding the Efficiency of Ray Traversal on GPUs|NVIDIA
• Megakernels Considered Harmful: Wavefront Path Tracing on GPUs | Research
• Adaptive Parallel Computation with CUDA Dynamic Parallelism
• Experiments with DirectX Raytracing in Northlight – YouTube
• SEED – Project PICA PICA – Real-time Raytracing Experiment using DXR (DirectX Raytracing) – YouTube
• NVIDIA RTX and GameWorks Ray Tracing Technology Demonstration – YouTube

Miscellaneous:

• MSR 2018
• Thomas Zimmermann — Senior Researcher at Microsoft Research

森田

Tensor Comprehensions:

• [1802.04730] Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions
• Announcing Tensor Comprehensions – Facebook Research
• GitHub – facebookresearch/TensorComprehensions: A domain specific language to express machine learning workloads.
• Integer Set Library
• Halide
• Lush: Lisp Universal SHell

Rust Belt:

• RustBelt: securing the foundations of the rust programming language
• RustBelt – Project page
• Iris Project
• Bug report thanks to this paper
• xi-editor
  • PR by jmuk

購読には RSS, iTunes, Pocket Casts, Spotify, YouTube, YouTube Music などをご利用ください。ご意見ご感想は 投書箱や Reddit にどうぞ。