|
|
|
|
|
|
by anqurvanillapy
475 days ago
|
|
|
Woah I enjoyed the read a lot. I also learned that in LLVM IR, they have the implicit null checks [1] to replace the guard with just a signal handler, so it won't hurt the branch predictor too much. So I believe there are many options upon "debug" and "release" profiles for the bound and null checking here. Very good design space. [1]: https://llvm.org/docs/FaultMaps.html#the-implicitnullchecks-... |
|
|