specializer

The specializer is built along with all the other tooling based on the specification that Ivan mentioned briefly.

The concrete machine code will mostly likely only ever be generated by the specializer in the real world.

So when your run the specializer machinery through itself, you get - an optimized toolset for building JIT compilers.

Instead this seems like something that wants an intermediate code + install time specializer a la mainframes.

For those wondering, it's a specializer, applied to an interpreter, to specialize the interpreter into a jit.

Therefore the specializer has to insert belt spills, which seems to be analogous to register allocation/spills.

Remember, this whole specializer machinery is all written in RPython, which can be interpreted by itself.

The generated binaries are cached in the load module, so subsequent runs skip the specializer step.

My guess is that they will generate such a specializer with their current generic specification tools.

If it's the specializer you're concerned about, it's intended to be very transparent to the typical user, integrated into the system.

By the way, just building the spec (defines belt size, available instructions, etc) generates a fully functioning specializer for that machine.

Consider the specializer to be the static software equivalent of the x86-to-uOp decoding core in the hardware of a modern x86.

All optimization is done in the middle end; the back end (the "specializer" runs at program install (or load) time and does only scheduling and emit.

"Binaries" are bitcode that's not specialized to any particular Mill CPU, and get run through the "specializer" which knows the real belt width and other properties to make a final CPU-specific version.

While I generally agree with your points, I think that its a little strange to sum this up as a "really simple problem" and then declare the solution to be building institutions to specializer in the assistance and care of people with a wide variety of issues.

JITs can use an API to convert generated GenAsm into runnable machine code which includes running it through the Specializer.