[PATCH] D115973: [llvm-profgen] Support symbol loading for debug fission

James Henderson via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Feb 16 01:19:02 PST 2022


jhenderson added a comment.

In D115973#3323622 <https://reviews.llvm.org/D115973#3323622>, @dblaikie wrote:

> Is this about file size in general (I'd expect an object file to be smaller than the equivalent yaml or assembly) - or it's about git and non-textual files? (I guess you're saying that it stores textual file updates as diffs rather than whole copies, but binary files it doesn't know how to encode diffs so it encodes the whole file each time?)

My understanding (which admittedly may be incorrect, as I'm by no means a git expert), is as you describe in the parentheses, so yeah, I'm talking about the git repository (specifically the .git folder).

In D115973#3324904 <https://reviews.llvm.org/D115973#3324904>, @wlei wrote:

> Do you know how to reduce the field of split-dwarf-split.dwo.yaml? see the patch, I found the four fields(.debug_str.dwo, .debug_str_offsets.dwo, .debug_info.dwo, .debug_abbrev.dwo) are must-have(removing will cause test failed), but apparently there're hex string in the `Content` which are not ideal.
>
> I run `obj2yaml split-dwarf-split.dwo > split-dwarf-split.dwo.yaml` to get the yaml file, do I miss anything? (I assume /llvm/test/tools/yaml2obj/ELF/DWARF/debug-info.yaml is also created by obj2yaml) or currently obj2yaml/yaml2obj lack to support *.dwo(as you said "it might not/probably doesn't support things like dwp indexes")?

I wouldn't be surprised if .dwo sections aren't supported yet - I know that @Higuoxing didn't finish everything. That being said, I think it's unlikely to be all that tricky to implement, if you wanted to do some experiments, by basing the implementation on the existing debug section implementations.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D115973/new/

https://reviews.llvm.org/D115973



More information about the llvm-commits mailing list