As an alternative to the computed gotos, you can use switch/case in Clang or GCC to achieve basically the same thing. :-) It becomes a jump table in most cases. (The article claims that a jump table gives smaller code and fewer branch misses, but it doesn't actually give any numbers, and enough things in there are dubious enough that I'm not convinced they ever measured.)
The switch misses the point. The compiler isn't smart enough to convert it to direct-threading, to the best of my knowledge.
A switch only selects on one character. To continue lexing you need the switch inside a loop. The compiler might optimize the switch itself to a jump table - but what does each case do - jumps back to the start of the loop, after which it enters the jump table again. There are two branches involved.
The point of direct threading is that there is no loop - you simply jump directly to the handler for the next character at the end of each handler.
> The compiler isn't smart enough to convert it to direct-threading, to the best of my knowledge.
If you read the URL I linked to, you will see that it is.
> The point of direct threading is that there is no loop - you simply jump directly to the handler for the next character at the end of each handler.
No, the point of direct threading is that you give the branch predictor more context to work with (effectively, the previous opcode), which was relevant with the branch predictors in typical CPUs 10+ years ago. (Modern ones, ITTAGE-style, have per-branch history also for indirect jumps.)
https://blog.nelhage.com/post/cpython-tail-call/ has made the rounds a lot recently, and explores this for Python's bytecode interpreter.