| Start...End time | Track name |
|---|---|
|
01:50 - 03:00 UTC |
Lang: ja
Track: TrackGrand Auditorium
Ruby meets WebAssemblyWebAssembly (Wasm) is a binary format for programs written in any language, designed to eventually run everywhere without changes, mainly inside Web browsers. Wasm is now used beyond the Web, from edge computing to smart contracts in blockchain. Now CRuby can be compiled to WebAssembly, which can be run in web browsers and many non-web environments with WASI. This talk will give you how we got there, techniques, and its use cases. Memo |
|
04:30 - 05:00 UTC |
Lang: ja
Track: TrackGrand Auditorium
Making *MaNy* threads on RubyWe will introduce *MaNy* project: supports massive number of threads on Ruby. Concurrent mchanism (threads, processes, etc.) is an important language feature especially for multiple network connections. Ruby supports Ractor/Thread/Fiber for concurrency. However making many Threads (and Ractors) introduce huge overhead. Fiber scheduler was introduced for this purpose but it has some limitations. On the other languages, for example, Go language supports goroutine. They can make many goroutines and they can run concurrently or in parallel. Other languages (Erlang, Rust, ...) also support similar features. They use a well-known technique called M:N threading. In short, M:N threading supports many M threads on N (enough small number) system threads. *MaNy* project introduces M:N threading into Ruby. In this talk, we will show the background and progress of *MaNy* project. Because we are replacing the threading mechanism of current Ruby, we can share some details about it. Memo |
|
05:10 - 05:40 UTC |
Lang: en
Track: TrackGrand Auditorium
Building a Lightweight IR and Backend for YJITLast year, the YJIT compiler was upstreamed into CRuby and released as part of Ruby 3.1. Key advantages of YJIT are that it delivers fast warm up time and significant speedup on typical Ruby workloads such as Rails applications. However, up to this point, YJIT has only been available on x86-64 platforms. This year, our goal is to transition from an experimental compiler towards something more broadly usable and future-proof. In order to make YJIT available on more platforms, we are currently working on the implementation of a new Intermediate Representation (IR) and backend for the compiler. This will make it possible for YJIT to run natively on the Apple M1/M2, AWS Graviton processors and the Raspberry Pi. This talk will cover our objectives and constraints, the design of our new IR, the strategy we took to be able to support both x86-64 and ARM CPUs, the challenges we encountered, how we plan to optimize the quality of the generated machine code, and early performance results. Memo |
|
05:50 - 06:20 UTC |
Lang: ja
Track: TrackGrand Auditorium
Tools for Providing rich user experience in debuggerruby/debug is one of the very active projects, and many users use it. Also, there are some hidden valuable features to enhance your development experience. This talk dives into Chrome integration and History Viewer within them. Chrome integration: Debugging with Chrome DevTools is the one of popular ways, especially in JavaScript. The advantages of Chrome DevTools are easy setup and a rich user experience. You don't have to install any tools except for Chrome. History Viewer: What do you do when you want to step back to the specified line? History Viewer can help you in the situation. History Viewer, which extends stepping back, is the feature in VSCode extension for ruby/debug. You can go back to the location without thinking about how many times "step back" commands are needed. In this talk, I will introduce how to use them and how they work, such as how to communicate between ruby/debug and Chrome DevTools. Memo |
|
06:50 - 07:20 UTC |
Lang: en
Track: TrackGrand Auditorium
Towards Ruby 4 JITFrom Ruby 3.1, we started to see CRuby's JIT compilers optimize real-world applications like Rails. However, you might be still wondering whether you should run it on production or not, given the fact that we're only seeing the beginning of the improvement. How far can we go in the foreseeable future? What would the impact on your production application look like? Let's talk about that together. Memo |
01:50 - 03:00
WebAssembly (Wasm) is a binary format for programs written in any language, designed to eventually run everywhere without changes, mainly inside Web browsers. Wasm is now used beyond the Web, from edge computing to smart contracts in blockchain. Now CRuby can be compiled to WebAssembly, which can be run in web browsers and many non-web environments with WASI. This talk will give you how we got there, techniques, and its use cases.
04:30 - 05:00
We will introduce *MaNy* project: supports massive number of threads on Ruby. Concurrent mchanism (threads, processes, etc.) is an important language feature especially for multiple network connections. Ruby supports Ractor/Thread/Fiber for concurrency. However making many Threads (and Ractors) introduce huge overhead. Fiber scheduler was introduced for this purpose but it has some limitations. On the other languages, for example, Go language supports goroutine. They can make many goroutines and they can run concurrently or in parallel. Other languages (Erlang, Rust, ...) also support similar features. They use a well-known technique called M:N threading. In short, M:N threading supports many M threads on N (enough small number) system threads. *MaNy* project introduces M:N threading into Ruby. In this talk, we will show the background and progress of *MaNy* project. Because we are replacing the threading mechanism of current Ruby, we can share some details about it.
05:10 - 05:40
Last year, the YJIT compiler was upstreamed into CRuby and released as part of Ruby 3.1. Key advantages of YJIT are that it delivers fast warm up time and significant speedup on typical Ruby workloads such as Rails applications. However, up to this point, YJIT has only been available on x86-64 platforms. This year, our goal is to transition from an experimental compiler towards something more broadly usable and future-proof. In order to make YJIT available on more platforms, we are currently working on the implementation of a new Intermediate Representation (IR) and backend for the compiler. This will make it possible for YJIT to run natively on the Apple M1/M2, AWS Graviton processors and the Raspberry Pi. This talk will cover our objectives and constraints, the design of our new IR, the strategy we took to be able to support both x86-64 and ARM CPUs, the challenges we encountered, how we plan to optimize the quality of the generated machine code, and early performance results.
05:50 - 06:20
ruby/debug is one of the very active projects, and many users use it. Also, there are some hidden valuable features to enhance your development experience. This talk dives into Chrome integration and History Viewer within them. Chrome integration: Debugging with Chrome DevTools is the one of popular ways, especially in JavaScript. The advantages of Chrome DevTools are easy setup and a rich user experience. You don't have to install any tools except for Chrome. History Viewer: What do you do when you want to step back to the specified line? History Viewer can help you in the situation. History Viewer, which extends stepping back, is the feature in VSCode extension for ruby/debug. You can go back to the location without thinking about how many times "step back" commands are needed. In this talk, I will introduce how to use them and how they work, such as how to communicate between ruby/debug and Chrome DevTools.
06:50 - 07:20
From Ruby 3.1, we started to see CRuby's JIT compilers optimize real-world applications like Rails. However, you might be still wondering whether you should run it on production or not, given the fact that we're only seeing the beginning of the improvement. How far can we go in the foreseeable future? What would the impact on your production application look like? Let's talk about that together.
| Start...End time | Track name |
|---|---|
|
00:30 - 01:40 UTC |
|
|
01:50 - 02:20 UTC |
Lang: en
Track: TrackGrand Auditorium
Ruby debugger - The best investment for your productivityIn this talk, I want to show you how adopting the Ruby debugger (`ruby/debug`) can be the single best thing to improve your productivity. It will include: - The difference between puts debugging, REPL (`Pry`/`IRB`) debugging and debugger debugging - Comparison with byebug - How you can use debugger to achieve even better puts and REPL debugging result: - Replace REPL debugging with breakpoint commands - Replace puts debugging with tracing and context commands - Automate debugging steps with `do:` and `pre:` options - Remote debugging capability - The debugging tooling spectrum and when you should **not** use the debugger Memo |
|
02:30 - 03:00 UTC |
Lang: ja
Track: TrackMiddle Auditorium
Make RuboCop super fastRuboCop introduces the new server options for super fast mode. RuboCop 2.0 has the catchphrase RuboCop 2x2, like Ruby 3x3. the server mode is a big move towards that. This is perfect for the current era where quick feedback to developer is required, especially when interacting with text editors and IDEs. Through the design and implementation of server options, you will get the essence of how fast RuboCop works. Memo |
|
04:30 - 05:00 UTC |
Lang: ja
Track: TrackGrand Auditorium
Method-based JIT compilation by transpiling to JuliaI will demonstrate a new approach to the method-based Just-In-Time compilation for Ruby. This is used Julia language as an infrastructure of the JIT compilation. I will describe the characteristics of this approach and show you some example results. Memo |
|
05:10 - 05:40 UTC |
Lang: ja
Track: TrackMiddle Auditorium
How fast really is Ruby 3.x?Ruby 3 brings novel improvements such as YJIT and Ractor, but the extent to which these techniques can speed up real applications is unknown. Performance claims should not be accepted solely based on abstract theories or micro benchmarks, but by the supporting evidence from actual applications. We recently ported a large-scale, open-source Ruby application (Fluentd) to Ruby 3, and conducted a survey to obtain an estimate of the speed improvements over Ruby 2.x. This talk will explain and discuss the results. Memo |
|
06:50 - 07:20 UTC |
Lang: en
Track: TrackGrand Auditorium
Caching With MessagePackAn object cache is like magic: put an object into the box and it comes back later, exactly as it went in. The magic is seductive because it makes things so easy. It only takes a bit of thought, however, to see the dangers of this trick. Class names can change; a method can return a different result. Objects exist in a universe, and that universe is constantly evolving. Marshal, Ruby's default binary serialization library, does not handle changes like this gracefully. We discovered this recently when a code change collided with a cached class, leading to a fire in production. To avoid incidents like this, we switched the cache serializer in our core Rails monolith from Marshal to MessagePack, an alternative, generic binary serializer. MessagePack has a more compact format than Marshal, with stricter typing and less magic. In this talk, I'll crack open MessagePack to show how it works, how we migrated to it, and why you might want to do the same. Memo |
|
07:30 - 08:40 UTC |
Lang: ja
Track: TrackGrand Auditorium
Ruby Committers vs The WorldRuby core committers on stage! Memo |
00:30 - 01:40
01:50 - 02:20
In this talk, I want to show you how adopting the Ruby debugger (`ruby/debug`) can be the single best thing to improve your productivity. It will include: - The difference between puts debugging, REPL (`Pry`/`IRB`) debugging and debugger debugging - Comparison with byebug - How you can use debugger to achieve even better puts and REPL debugging result: - Replace REPL debugging with breakpoint commands - Replace puts debugging with tracing and context commands - Automate debugging steps with `do:` and `pre:` options - Remote debugging capability - The debugging tooling spectrum and when you should **not** use the debugger
02:30 - 03:00
RuboCop introduces the new server options for super fast mode. RuboCop 2.0 has the catchphrase RuboCop 2x2, like Ruby 3x3. the server mode is a big move towards that. This is perfect for the current era where quick feedback to developer is required, especially when interacting with text editors and IDEs. Through the design and implementation of server options, you will get the essence of how fast RuboCop works.
04:30 - 05:00
I will demonstrate a new approach to the method-based Just-In-Time compilation for Ruby. This is used Julia language as an infrastructure of the JIT compilation. I will describe the characteristics of this approach and show you some example results.
05:10 - 05:40
Ruby 3 brings novel improvements such as YJIT and Ractor, but the extent to which these techniques can speed up real applications is unknown. Performance claims should not be accepted solely based on abstract theories or micro benchmarks, but by the supporting evidence from actual applications. We recently ported a large-scale, open-source Ruby application (Fluentd) to Ruby 3, and conducted a survey to obtain an estimate of the speed improvements over Ruby 2.x. This talk will explain and discuss the results.
06:50 - 07:20
An object cache is like magic: put an object into the box and it comes back later, exactly as it went in. The magic is seductive because it makes things so easy. It only takes a bit of thought, however, to see the dangers of this trick. Class names can change; a method can return a different result. Objects exist in a universe, and that universe is constantly evolving. Marshal, Ruby's default binary serialization library, does not handle changes like this gracefully. We discovered this recently when a code change collided with a cached class, leading to a fire in production. To avoid incidents like this, we switched the cache serializer in our core Rails monolith from Marshal to MessagePack, an alternative, generic binary serializer. MessagePack has a more compact format than Marshal, with stricter typing and less magic. In this talk, I'll crack open MessagePack to show how it works, how we migrated to it, and why you might want to do the same.
07:30 - 08:40
Ruby core committers on stage!
| Start...End time | Track name |
|---|---|
|
00:30 - 01:00 UTC |
Lang: ja
Track: TrackGrand Auditorium
error_highlight: user-friendly error diagnosticsRuby 3.1 has introduced a gem called "error_highlight", which shows a code snippet with an underline to spot where `NameError` or `NoMethodError` was raised. ``` $ ruby test.rb test.rb:1:in `<main>': undefined method `[]' for nil:NilClass (NoMethodError) json[:foo][:bar] ^^^^^^ ``` We will talk about the design and implementation of error_highlight: how it works, why and how we designed it, how we resolved the difficulties we faced during implementation, and the feedback we received. This feature goes further in Ruby 3.2: more error support other than `NameError`/`NoMethodError` and more fine-grained error spotting. We also redesigned its backend by introducing a new API, `Exception#detailed_message`, which allows us to enhance error messages with fewer incompatibilities. In fact, it allowed simple integration with other error message extensions such as did_you_mean and dead_end. We will also talk about what we would like users to help us with to make error_highlight better. Memo |
|
01:50 - 02:20 UTC |
Lang: en
Track: TrackGrand Auditorium
Real World Applications with the Ruby Fiber SchedulerRuby 3 introduced the fiber scheduler which provides transparent context switching between fibers when they perform blocking operations. With minimal changes, web applications can take advantage of the fiber scheduler to improve performance and unlock new techniques for application design and interactivity. We discuss the current state of the fiber scheduler design, its usage in several real world applications, and how you can take advantage of it in your own projects. Memo |
|
02:30 - 03:00 UTC |
Lang: en
Track: TrackGrand Auditorium
A Faster CRuby interpreter with dynamically specialized IRA new CRuby internal representation (IR) dynamically specialized during execution will be presented. The major specialization includes generation of * Hybrid stack-RTL instructions * Type specialized instructions using basic block versioning technique * Speculatively specialized instructions based on profile-based information Deoptimization technique of speculatively specialized instructions when the original assumptions become wrong will be also discussed. Using specialized IR can significantly speed up the interpretation and can permit generation of effective JIT code with static compilers like MIR or GCC. The new IR can be also used to implement Ruby specific optimizations, e.g. inlining. Performance comparison of CRuby interpreter with the specialized IR, the current CRuby interpreter, YJIT, and MJIT will be given for some benchmarks. Memo |
|
04:30 - 05:00 UTC |
Lang: ja
Track: TrackMiddle Auditorium
The Better RuboCop World to enjoy RubyTools like RuboCop are very useful. I am very grateful for contributors. I think it's really helpful to keep our code clean and consistent. However, sometimes I feel there is a gap between 'Cop' culture and Ruby culture. In general, 'Cop' restricts our rights though Ruby gives freedom to us. I understand we have right not to use RuboCop or disable some Cops. But it is not very easy in reality, especially for Ruby beginners in their teams. To make them happier, experts would be able to set up config perfectly. But again, it is not very easy in reality. As a result, even though there is no evil, in some case people wrongfully make detour and give up their creativeness to keep CI green. It's a pity, isn't it? In this talk, I introduce my thoughts on RubyCop and programming and some idea of a bit better RuboCop world, with lower risk to damage productivity and to misguide beginners. Memo |
|
05:50 - 06:20 UTC |
Lang: ja
Track: TrackMiddle Auditorium
String Meets EncodingRuby's String has Encoding, which allows for very flexible character encoding. What is the trade-off for that flexibility? I recently looked at the bottleneck in CSV.read and found that in one file with Encoding CP932, 30% of the processing time was spent on String#split. From the perspective of optimizing String#split, we will explain the relationship between String and Encoding in Ruby, how String knows its own Encoding, and which process is the bottleneck. Then we will discuss approaches toward faster encoding. Memo |
|
06:50 - 07:20 UTC |
Lang: ja
Track: TrackMiddle Auditorium
History of Japanese Ruby reference manual, and futureI will talk about history of Japanese reference manual and future plans. Contributing to current Japanese Ruby reference manual (rurema) is harder than other projects. Because current document format (based on RD) is not familiar to recent Ruby users, and there are a few documents for new contributors. So I will explain historical reasons and how to improve systems for new contributors. Memo |
|
07:30 - 08:40 UTC |
Lang: en
Track: TrackGrand Auditorium
Stories from developing YJITYJIT is CRuby's second just-in-time compiler first released with Ruby 3.1.0. Much sweat and tears were shed during YJIT's development. What are the performance goals of YJIT and how are we going for them? How do CPUs react to YJIT's output? What is it like to retrofit a JIT compiler into a runtime system with a long history? Why is Rust involved now? Memo |
00:30 - 01:00
Ruby 3.1 has introduced a gem called "error_highlight", which shows a code snippet with an underline to spot where `NameError` or `NoMethodError` was raised. ``` $ ruby test.rb test.rb:1:in `<main>': undefined method `[]' for nil:NilClass (NoMethodError) json[:foo][:bar] ^^^^^^ ``` We will talk about the design and implementation of error_highlight: how it works, why and how we designed it, how we resolved the difficulties we faced during implementation, and the feedback we received. This feature goes further in Ruby 3.2: more error support other than `NameError`/`NoMethodError` and more fine-grained error spotting. We also redesigned its backend by introducing a new API, `Exception#detailed_message`, which allows us to enhance error messages with fewer incompatibilities. In fact, it allowed simple integration with other error message extensions such as did_you_mean and dead_end. We will also talk about what we would like users to help us with to make error_highlight better.
01:50 - 02:20
Ruby 3 introduced the fiber scheduler which provides transparent context switching between fibers when they perform blocking operations. With minimal changes, web applications can take advantage of the fiber scheduler to improve performance and unlock new techniques for application design and interactivity. We discuss the current state of the fiber scheduler design, its usage in several real world applications, and how you can take advantage of it in your own projects.
02:30 - 03:00
A new CRuby internal representation (IR) dynamically specialized during execution will be presented. The major specialization includes generation of * Hybrid stack-RTL instructions * Type specialized instructions using basic block versioning technique * Speculatively specialized instructions based on profile-based information Deoptimization technique of speculatively specialized instructions when the original assumptions become wrong will be also discussed. Using specialized IR can significantly speed up the interpretation and can permit generation of effective JIT code with static compilers like MIR or GCC. The new IR can be also used to implement Ruby specific optimizations, e.g. inlining. Performance comparison of CRuby interpreter with the specialized IR, the current CRuby interpreter, YJIT, and MJIT will be given for some benchmarks.
04:30 - 05:00
Tools like RuboCop are very useful. I am very grateful for contributors. I think it's really helpful to keep our code clean and consistent. However, sometimes I feel there is a gap between 'Cop' culture and Ruby culture. In general, 'Cop' restricts our rights though Ruby gives freedom to us. I understand we have right not to use RuboCop or disable some Cops. But it is not very easy in reality, especially for Ruby beginners in their teams. To make them happier, experts would be able to set up config perfectly. But again, it is not very easy in reality. As a result, even though there is no evil, in some case people wrongfully make detour and give up their creativeness to keep CI green. It's a pity, isn't it? In this talk, I introduce my thoughts on RubyCop and programming and some idea of a bit better RuboCop world, with lower risk to damage productivity and to misguide beginners.
05:50 - 06:20
Ruby's String has Encoding, which allows for very flexible character encoding. What is the trade-off for that flexibility? I recently looked at the bottleneck in CSV.read and found that in one file with Encoding CP932, 30% of the processing time was spent on String#split. From the perspective of optimizing String#split, we will explain the relationship between String and Encoding in Ruby, how String knows its own Encoding, and which process is the bottleneck. Then we will discuss approaches toward faster encoding.
06:50 - 07:20
I will talk about history of Japanese reference manual and future plans. Contributing to current Japanese Ruby reference manual (rurema) is harder than other projects. Because current document format (based on RD) is not familiar to recent Ruby users, and there are a few documents for new contributors. So I will explain historical reasons and how to improve systems for new contributors.
07:30 - 08:40
YJIT is CRuby's second just-in-time compiler first released with Ruby 3.1.0. Much sweat and tears were shed during YJIT's development. What are the performance goals of YJIT and how are we going for them? How do CPUs react to YJIT's output? What is it like to retrofit a JIT compiler into a runtime system with a long history? Why is Rust involved now?