I thought of to have a LLVM backend to target Glulx. Some things are done easily enough; the ! operator of C corresponds to "ushiftr 1,$,$" in Glulx (this is us

I thought of to have a LLVM backend to target Glulx. Some things are done easily enough; the ! operator of C corresponds to "ushiftr 1,$,$" in Glulx (this is using Glasm syntax). Calling conventions are simple. The Glk functions can be converted directly to Glulx code, with a few exceptions; the Glk type "S" means a different thing in C and in Glulx, although for most of them you can instead use ones that take an array and length, making the function to calculate the length before calling Glk. The C strlen() function becomes "linearsearch 0,1,$,1,-1,0,4,$" in Glulx. However, one thing with both Z-machine and Glulx is that there is no alloca(). However, you might be able to emulate it with a "shadow stack", I thought. (If programming in assembly language, this is not a problem since you can take advantage of those specific virtual machines instead.) Also, does LLVM support targeting 16-bit computers so well? (I don't know how to write a LLVM backend.) For Z-machine, one thing I thought might work better to compile into a Z-machine code might be a variant of BASIC instead. But when programming in Z-machine or Glulx, I just use assembly language; you can then run those programs on other computers anyways, so it wouldn't be as much problem. (Many of the tricks in "Tricky Document" assume the use of assembly language, although some could be implemented by optimizers.)