Re: [squirrelfish] Native code generation

Maybe my question will be naive, but has anyone an idea/opinion about using llvm as a medium to get multiple native architecture support in SF?

Indeed, LLVM supports all the architectures we're interested in. Before putting our eggs in the LLVM basket, though, I'd want to consider the following: * How does LLVM's compile time compare to CTI? How about other JITs and JS engines? * How would LLVM's optimization passes work with function calls to the JS runtime? * How would we specify our custom calling conventions to the LLVM infrastructure? Do you have any data / suggestions in these areas? Cheers, Geoff