32 bits and 64 bits integer status about mir HOT 2 OPEN

You can consider MIR VM is 64-bit register machine. 32-bit insns work with only 32-bit part of 64-bit registers (vars). What happens with another part of 64-bit register is not defined (for some environment another part can be not changed, for other environments another part can be zeroed etc).

That is why integer variables can be declared only as 64-bit integers. Function parameter declaration is an exception used to declare how parameter passing. Still integer parameters are 64-bit too.

For 64-bit registers, you need only 64-bit mov. If you want to move 32-bit value with zeroing another part of 64-bit register you can use UEXT32. To do move with sign extension you can use EXT32.

So MIR currently is oriented for efficient implementation on 64-bit machines. But with small effort implementation on 32-bit machines can be efficient too (of course if you use mostly 32-bit insns).

Originally MIR had no 32-bit insns. But to implement 32-bit arithmetic a lot of extension insns would require (for insn operands and results). So I decided that it is easier to add 32-bit insns.

from mir.

Thanks !
I've got some sound coming from the Faust => MIR translator ((-; more later.

from mir.