Discussing The LLVM IR Shortcomings

While GCC developers are currently discussing merging the D programming language front-end into their compiler, LLVM developers are currently discussing their IR and its sufficiency.

Dan Gohman, an Apple engineer, volleyed a long and technical e-mail to the LLVM development list about the LLVM IR. Dan argues that the intermediate representation for this Apple-sponsored compiler infrastructure is "a poor system for building a Platform." He and many felt that LLVM IR would work well for multiple architectures, but in the end it's not been so pretty. In his email he argues his case at length.

Dan ends up proposing a new platform IR for LLVM. "In conclusion, consider the task of writing an independent implementation of an LLVM IR Platform. The set of capabilities it provides depends on who you talk to. Semantic details are left to chance. There are features which require a bunch of complicated infrastructure to implement which are rarely used. And if you want light-weight execution, you'll probably need to translate it into something else better suited for it first. This all doesn't sound very appealing."

There's many responses already from his message that was sent yesterday. Most of the developers agree with his points. Chris Lattner, the primary developer behind LLVM/Clang and who's been heading Apple's Compiler Group, also agrees with most points raised by his colleague. However, he doesn't agree with the conclusion of creating a new IR. The email from Chris can be read here.