[LLVMdev] More Encoding Ideas

[Reid Spencer]

I mis-spoke when I said ..

> I don't think that happens today. We handle string constants (as opposed 
> to character constants) separately and they are always written as a 
> length followed by the characters in the string.

Actually, they're written as a type slot number (from which the array length is 
derived), followed by the character data in the string.

and another goof ..

> Again, I'm not sure what you're getting at here. Have a look at the 
> function outputConstantStrings in lib/Bytecode/Writer/Writer.cpp. 
> Strings currently occupy the "void" type plane, plane 0 (which is a bit 
> of a hack), but we write out the type slot number (index), then the 
> character data, then a terminating null (0) byte. 

There's no terminating null byte. The length is inferred from the array type.

Sorry for the inaccuracies .. I shouldn't rely on memory so much :)

Reid.