For dsymutil we want to store offsets in the accelerator table entries
rather than DIE pointers. In addition, we need a way to communicate
which CU a DIE belongs to. This patch provides support for both of these
issues. It supersedes r336529 which will be reverted in favor of this.
Details
Details
Diff Detail
Diff Detail
- Repository
- rL LLVM
Event Timeline
Comment Actions
Minor drive-by comment.
llvm/lib/CodeGen/AsmPrinter/AccelTable.cpp | ||
---|---|---|
573 ↗ | (On Diff #154706) | I'm guessing this wants llvm:: on it, same as the other overload. |
Comment Actions
This looks good to me. There are some different tradeoffs we could make (speed/templates vs. code size/virtualization), but I don't have an opinion on that, so if noone else does either, I guess this is fine.
llvm/include/llvm/CodeGen/AccelTable.h | ||
---|---|---|
319 ↗ | (On Diff #154808) | If we wanted to optimize things, this could be an llvm::function_ref. |
Comment Actions
It seemed reasonable to pick the same trade-offs as we did for the Apple-style accel tables. The motivation then was that we didn't want to pay for a virtual call for every entry in the list.