The equivalent is doing more dead code analysis at the MIR stage before sending it to llvm. This is where rustc loses most of its perf. LLVM is usually 50-60% of execution time, mostly because rust sends teams of ir to chew through.
Serious question: is there any reason dead code detection would be any faster at the syntax level than the generated code? I mean, I'm no compiler expert but in a straightforward SSA representation "dead code" is just a register that gets assigned but never used, all of which can be detected and dropped globally with just one pass over the data.
I suspect the real reason is going to turn out more complicated, no? Like, the code is really "dead" yet because we can't know if it will be used later without solving the halting problem.
Which, if you think about it, is pretty much exactly the situation C++ is in. All those expanded templates aren't "dead" yet at the point of expansion either, you just don't know if the code will use them. And in practice it won't, so you wasted a ton of cycles to find that out.
> Serious question: is there any reason dead code detection would be any faster at the syntax level than the generated code? I mean, I'm no compiler expert but in a straightforward SSA representation "dead code" is just a register that gets assigned but never used, all of which can be detected and dropped globally with just one pass over the data.
That is only one type of dead code elimination. It is not necessary to solve the halting problem to eliminate larger portions of code than just unused register assignments. There are plenty of ways to eliminate beyond that, such as proving that entire classes, functions, declarations, branches, loop iterations, and data are unreachable. Especially so in a type-resolved format like MIR, some of these higher level optimizations become really cheap and would allow for LLVM to do a lot less work.
In general, you see inflation in code size as the representation becomes lower level and more explicit. Traversing a larger volume of code to detect dead code is generally more expensive than traversing a smaller volume of code. This might not be true if the deadness determination is really expensive, but it's almost always worth doing some kind of lightweight dead code elimination before each stage of lowering to save time in the subsequent stage.
