Re: indirect threading for bytecode interpreter

[Dan Nicolaescu]

Tom Tromey <address@hidden> writes:

  > This patch implements indirect threading for the Emacs bytecode
  > interpreter.  It uses a GNU C extension, and falls back to the current
  > switch-based code if Emacs is compiled with some other compiler.
  > 
  > This speeds up the bytecode interpreter a bit.  On my benchmark, it
  > improved the overall time between 4-7% (lots of noise in the runs).
  > That isn't bad when you consider that Fbyte_code was the 3rd biggest
  > time user (the first two being re_match_2_internal and mark_object).

If your patch applies to emacs-22.3 (quite likely), and you are curious,
you might want to test the performance there.  The overhead of
mark_object should be lower in that version (it's higher in 23 because
of the way charsets are represented).