🚀 go-pugleaf

>There are some improvements that can be made (particularly in register
>allocation for functions that have more than one boolean variable with auto
>storage duration),

And, perhaps, for what it's worth, in choosing representations that fit
what some architectures might prefer; e.g., the 68K has an instruction
to set a byte based on the condition codes, but the "true" value is all
ones, so C compilers have to stuff a "negate" instruction after the
"seq" in something like

	a = (b == c);

I tend to doubt the win of avoiding that "neg" instruction would be
large, and suspect it would, in most cases, not even be measurable,
however.