Currently debugging LLVM-CBE code is horrible. One issue is that all the locals generated have names that don't correspond to the original source's names at all.
To improve this, we could name generated locals based off information from the debugging info, for example if the debugging info indicates that the value we are generating corresponds with the local index, then maybe we should name the generated local _1_index instead of _1.
This would require keeping track of these generated names so that downstream instructions will use the correct local name.
Currently debugging LLVM-CBE code is horrible. One issue is that all the locals generated have names that don't correspond to the original source's names at all.
To improve this, we could name generated locals based off information from the debugging info, for example if the debugging info indicates that the value we are generating corresponds with the local
index
, then maybe we should name the generated local_1_index
instead of_1
.This would require keeping track of these generated names so that downstream instructions will use the correct local name.