4 ms·

The illusion of a monoculture disappears if you zoom out a bit. Even if clang was a "first-class citizen" compiler for the kernel and was picked up by some big

The illusion of a monoculture disappears if you zoom out a bit. Even if clang was a "first-class citizen" compiler for the kernel and was picked up by some big distros, there's an absolutely gigantic market of companies designing and building embedded systems around GCC. There already is a LLVM/Clang monoculture in some areas (e.g. research, "modern" C++ shops), but while I don't have the numbers or anything, I'd imagine that those fields make up a very small percentage of worldwide C/C++ developers. A lot of those boring embedded systems companies have low margins and don't think particularly highly of software developers, so they'd never consider refactoring all of their tooling and projects to use clang unless they absolutely had to. If the kernel _dropped_ support for GCC, you might see an actual monoculture develop. But I don't see the kernel _supporting_ GCC, or particular distros choosing to build their kernel with it, making any real dent in the huge market share (for lack of a better term) that GCC has built up. It might seem that way if you read HN and talk to people working in SV, but GCC is still pretty ubiquitous.

azinman2 4y ago
All of Apple's toolchain is LLVM, which also means all 3rd party apps running on Apple devices are (almost?) all LLVM. Google Chrome is now built with LLVM, Android is getting there with Clang, and who knows what else inside of Google is built with Clang. These aren't tiny percentages.
- mkipper 4y ago
  I guess it depends on how you look at the the percentages. Lots of people _use_ Apple’s developer tools, Google Chrome and other Google products. But in each of those cases, I believe the actual group of people directly dealing with C++ and Clang are small relative to the global C/C++ developer population. Android has a strong argument since Google can force the hands of OEMs. I don’t see how the size of Chrome’s user base will bring about an LLVM monoculture if 99.999% of C++ developers on earth never build Chrome themselves. The majority of electronic devices on earth have (non-Android) firmware or software written in C/C++, written by people all over the world who don’t work for Google, Apple or anyone you’d even consider a tech company. You will never hear about these projects on HN because they’re closed-source, proprietary and (frankly) boring. But almost all of those projects use whatever toolchain their vendor provides, and more often than not (in my experience), thats GCC. Like I said, I don’t have any numbers to back this up. I’m not claiming that GCC reigns supreme and that LLVM is irrelevant outside of academia. I’m just saying that if you’re really concerned about LLVM _support_ in the kernel leading to an LLVM monoculture, you might not understand how widely used GCC is in “boring” tech. [edit] Adding to this, I guess it also depends on how you view a “monoculture” w.r.t compilers. I could imagine a not-too-distant future where LLVM gets the vast majority of research effort, and you see cool new optimizations brought into it while GCC stagnates. We’re arguably already there. But I don’t see that driving thousands of low-margin hardware OEMs to switch over from GCC to LLVM. Maybe that will happen one day, but I think we’re pretty far away from that.

azinman2 4y ago

All of Apple's toolchain is LLVM, which also means all 3rd party apps running on Apple devices are (almost?) all LLVM. Google Chrome is now built with LLVM, Android is getting there with Clang, and who knows what else inside of Google is built with Clang. These aren't tiny percentages.

mkipper 4y ago
I guess it depends on how you look at the the percentages. Lots of people _use_ Apple’s developer tools, Google Chrome and other Google products. But in each of those cases, I believe the actual group of people directly dealing with C++ and Clang are small relative to the global C/C++ developer population. Android has a strong argument since Google can force the hands of OEMs. I don’t see how the size of Chrome’s user base will bring about an LLVM monoculture if 99.999% of C++ developers on earth never build Chrome themselves. The majority of electronic devices on earth have (non-Android) firmware or software written in C/C++, written by people all over the world who don’t work for Google, Apple or anyone you’d even consider a tech company. You will never hear about these projects on HN because they’re closed-source, proprietary and (frankly) boring. But almost all of those projects use whatever toolchain their vendor provides, and more often than not (in my experience), thats GCC. Like I said, I don’t have any numbers to back this up. I’m not claiming that GCC reigns supreme and that LLVM is irrelevant outside of academia. I’m just saying that if you’re really concerned about LLVM _support_ in the kernel leading to an LLVM monoculture, you might not understand how widely used GCC is in “boring” tech. [edit] Adding to this, I guess it also depends on how you view a “monoculture” w.r.t compilers. I could imagine a not-too-distant future where LLVM gets the vast majority of research effort, and you see cool new optimizations brought into it while GCC stagnates. We’re arguably already there. But I don’t see that driving thousands of low-margin hardware OEMs to switch over from GCC to LLVM. Maybe that will happen one day, but I think we’re pretty far away from that.