Hi, The backend I am working on has various different register sizes (i16,i32,i64). The datapath in the architecture is 64bit. When a register is needed for an ALU operation, it is sign extended to the full 64bits, all operations are carried out in 64bits and then the result is truncated to the register size. For example, if an A register is i16 and is loaded with -1 it's value would be FFFF . If you were then to copy it's value to a B register of size i32 then it's value would be FFFFFFFF etc. I'm a bit lost on what the idiomatic way to implement this in LLVM would be. Cheers, Johnny -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140724/ce530244/attachment.html>