text stringlengths 0 1.49k |
|---|
**Johnny Boursiquot:** Bye! |
**Carlisia Thompson:** Bye, this was fun! |
**Ramya Rao:** Bye! |
• Introduction of David Chase and his background in programming languages compilers and runtimes |
• Explanation of the components involved in a compiler and its phases (scanning, parsing, semantic analysis, etc.) |
• Description of the Go compiler process (characters -> AST -> SSA) and optimization |
• Discussion of escape analysis and its role in identifying memory allocation |
• Overview of contributing to the Go compiler (importance of benchmarks and starting with lowering code from generic SSA) |
• Clarification on the current state of the intermediate Go Assembler step |
• Discussion of compiler internals and recommended reading |
• Overview of various compiler books, including those by Andrew-Appel |
• Reference to the LCC compiler as a relatively small and easy-to-understand example |
• Explanation of the Go compiler's complexity due to its success and portability |
• Mention of potential improvements to the Go compiler, including lazy import and mid-stack inlining |
• Discussion of upcoming features for 1.10 and 2.0, including: |
+ Improved debugging experience for optimized code |
+ Cooperative scheduling enforcement within tight loops |
+ Generational collection work and write barrier optimizations |
• Go language vs implementation discussion |
• Advantages of multidimensional slices in Go |
• Generics as a potential feature for Go |
• Compiler optimizations and speed considerations |
• Importance of simplicity and readability in compiler design |
• Formal verification of compilers and its challenges |
• Trusting Trust paper mentioned as relevant to compiler security |
• Verified code and testing |
• Compiler verification for different targets (e.g., x86/64, ARM) |
• Balance between compiler speed and optimization complexity |
• Potential for using LLVM as a backend for Go |
• Pie preferences of the Go compiler team |
• Discussion about pies and pie preferences |
• Recommendation for Emeril Lagasse's banana cream pie at his restaurant |
• Proposal to consider hosting GopherCon at a location with good food options, specifically including the pie mentioned earlier |
• Review of recent Go projects and news |
• Discussion of Oak game engine and its potential for use on PocketCHIP |
• Overview of PocketCHIP as a small, portable computer with Linux capabilities and discussion of its uses |
• Mention of new book "Concurrency in Go" by Katherine Cox-Buday |
• Discussion about a well-written post on cluster schedulers |
• Shoutouts for open-source projects and individuals, including Dave Cheney's Go blog and errors package |
• Review of GopherCon 2019 and its contributor day |
• Stats from Jess Frazelle on new open CLs |
• Thank you to Steve Francia and others who contributed to the GopherCon remote event |
• Discussion about MacPorts and its quality compared to Homebrew |
• David Chase shared his experience with a tech toolchain on a laptop 5 years ago |
• The setup included MacPorts, Track, Python, SQLite, Mercurial, Emacs in bash-mode, and code processing |
• The project was for a website that formatted code into a mathematical style |
• The conversation was part of the GoTime.fm podcast |
**Brian Ketelsen:** Welcome back, everybody. This is GoTime episode number 52, where we're joined by David Chase. I'm Brian Ketelsen, standing in for Erik St. Martin who is out today. I'm joined by Carlisia Pinto... |
**Carlisia Thompson:** Hi, everybody. |
**Brian Ketelsen:** And Ashley McNamara... |
**Ashley McNamara:** If you're filling in for Erik, who am I filling in for? |
**Brian Ketelsen:** You're filling in for me. |
**Ashley McNamara:** Perfect. |
**Brian Ketelsen:** Yeah, I'm standing in for Erik, you're standing in for me, Carlisia in standing in for Carlisia, and David is standing in for David. |
**David Chase:** Right. |
**Ashley McNamara:** Perfect. |
**Brian Ketelsen:** So David, why don't you start off by introducing yourself, kind of giving us some background and telling us what you do? |
**David Chase:** I guess my background is in programming languages compilers and programming language runtimes. I started out way back with Fortran, but now I work on Go, mostly on the Go compiler... But there is a pretty healthy runtime component to the compiler work. |
The garbage collector, for instance, requires write barriers; the compiler has to be aware of them, and it can do optimizations involving them. The scheduling in Go is cooperative, and the compiler enforces the cooperation. And I'm already pretty much deep in the weeds with what I do with this example stuff, what the w... |
**Brian Ketelsen:** Nice. I'm more of a business information guy rather than a computer science guy... Can you give us the rundown of what components are involved in a compiler, what does a compiler do when I type in source code? |
**Ashley McNamara:** Stop stealing my questions, Brian. |
**Brian Ketelsen:** Hey, be nice. \[laughter\] |
**David Chase:** So it varies somewhat from compiler to compiler, but in general, they tend to have phases, and a sort of mix and match. So you start with the characters - they tend to transform those into a tree-based representation. Usually they do scanning to break it into tokens and then parsing to combine the toke... |
Once you have the tree, you can do a certain amount of - I think you would call it semantic analysis, so enforcing a lot of the rules of the language: can you look at this, can you look at that? Do the types match, and so on? |
Some compilers go straight from tree to code generation, and in fact the Go compiler used to do that about -- God, I can't even remember which release it was. I guess we lit up SSA in 1.7, we added a phase... Keith Randall talked about this at GopherCon, and that talk will be online in a little while. But we added a lo... |
\[04:23\] So the Go compilers, again - characters come in, scan and parse into an [AST](https://en.wikipedia.org/wiki/Abstract_syntax_tree), do semantic analysis to enforce rules, transform AST to [SSA](https://en.wikipedia.org/wiki/Static_single_assignment_form) and do a certain amount of optimization. Then we interfa... |
That's a crude description. There are certain important analyses that we do on the tree-based form still, and it would actually be a lot of work to get some of them out... [Escape analysis](https://en.wikipedia.org/wiki/Escape_analysis) is actually interprocedural within the package that's being compiled, and on top of... |
**Brian Ketelsen:** So when you say you do escape analysis against the AST, does that mean that there are notations added to the source code somehow or to the abstract syntax tree that show the results of that escape analysis? How does that analysis move from the AST into the final generated code? |
**David Chase:** So the AST nodes are sort of an artifact; if you go look at the Go source tree, you will find two ASTs, and one of them is sort of very visible and it's for public consumption, for source code transformation. There's also the one that the compiler uses, and the one that the compiler uses, the AST nodes... |
Escape analysis runs over all the identifiers that have their address taken, or all the expressions too, because at the tree level, everything can have a name, even if it doesn't have a source code name. So everything that could have its address taken, escape analysis looks to see where those addresses go, and not only... |
**Brian Ketelsen:** No, that's where we wanna go, we wanna go deep. |
**David Chase:** Okay. |
**Ashley McNamara:** We have several listener questions. We're gonna start with Matt first. He says, "What is a good place to get started if you wanna contribute to the Go compiler?" |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.