[lld] r309688 - [ELF] - Use multithreading for building .gdb_index.

[Rafael Avila de Espindola via llvm-commits]

Nice, with 8 cores this is a 7.8% speed improvement of the clang with
gdb index link.

Cheers,
Rafael

George Rimar via llvm-commits <llvm-commits at lists.llvm.org> writes:

> Author: grimar
> Date: Tue Aug  1 07:57:13 2017
> New Revision: 309688
>
> URL: http://llvm.org/viewvc/llvm-project?rev=309688&view=rev
> Log:
> [ELF] - Use multithreading for building .gdb_index.
>
> Enables multithreaded .gdb_index building.
>
> Differential revision: https://reviews.llvm.org/D36090
>
> Modified:
>     lld/trunk/ELF/SyntheticSections.cpp
>
> Modified: lld/trunk/ELF/SyntheticSections.cpp
> URL: http://llvm.org/viewvc/llvm-project/lld/trunk/ELF/SyntheticSections.cpp?rev=309688&r1=309687&r2=309688&view=diff
> ==============================================================================
> --- lld/trunk/ELF/SyntheticSections.cpp (original)
> +++ lld/trunk/ELF/SyntheticSections.cpp Tue Aug  1 07:57:13 2017
> @@ -1813,12 +1813,13 @@ static GdbIndexChunk readDwarf(DWARFCont
>  }
>  
>  template <class ELFT> GdbIndexSection *elf::createGdbIndex() {
> -  std::vector<GdbIndexChunk> Chunks;
> -  for (InputSection *Sec : getDebugInfoSections()) {
> -    ObjFile<ELFT> *F = Sec->getFile<ELFT>();
> +  std::vector<InputSection *> Sections = getDebugInfoSections();
> +  std::vector<GdbIndexChunk> Chunks(Sections.size());
> +  parallelForEachN(0, Chunks.size(), [&](size_t I) {
> +    ObjFile<ELFT> *F = Sections[I]->getFile<ELFT>();
>      DWARFContext Dwarf(make_unique<LLDDwarfObj<ELFT>>(F));
> -    Chunks.push_back(readDwarf(Dwarf, Sec));
> -  }
> +    Chunks[I] = readDwarf(Dwarf, Sections[I]);
> +  });
>    return make<GdbIndexSection>(std::move(Chunks));
>  }
>  
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits