[lld] r243232 - ELF2: Avoid calling std::sort to make output deterministic.

[Chandler Carruth]

On Mon, Jul 27, 2015 at 2:02 PM Rui Ueyama <ruiu at google.com> wrote:

> On Mon, Jul 27, 2015 at 12:01 PM, Duncan P. N. Exon Smith <
> dexonsmith at apple.com> wrote:
>
>>
>> > On 2015-Jul-25, at 17:50, Rui Ueyama <ruiu at google.com> wrote:
>> >
>> > Author: ruiu
>> > Date: Sat Jul 25 19:50:15 2015
>> > New Revision: 243232
>> >
>> > URL: http://llvm.org/viewvc/llvm-project?rev=243232&view=rev
>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject-3Frev-3D243232-26view-3Drev&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=A7YEbZn5UiSjeArc6rLNc6OXlfY7Hm0OXfSQWGYHAns&s=wmdRPI5Lb9TNc_6lTBnZH0eo6CyF0IJSE6TQ7CrgwlU&e=>
>> > Log:
>> > ELF2: Avoid calling std::sort to make output deterministic.
>>
>> Using `std::stable_sort` should also make it deterministic.  Is that
>> better than switching to a `std::map`?
>>
>
> Yeah, I think the new code is slightly more readable than the previous
> one. At least it's shorter than before.
>


"Yeah" doesn't seem to agree with the rest of your email, so I'm left
confused.

Why wouldn't switching to std::stable_sort be better?

std::map is horribly slow in essentially every way, and so I struggle to
believe it is the best tool for the job even if the code using it is very
compact.


>
>
>> >
>> > Modified:
>> >    lld/trunk/ELF/Writer.cpp
>> >
>> > Modified: lld/trunk/ELF/Writer.cpp
>> > URL:
>> http://llvm.org/viewvc/llvm-project/lld/trunk/ELF/Writer.cpp?rev=243232&r1=243231&r2=243232&view=diff
>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_lld_trunk_ELF_Writer.cpp-3Frev-3D243232-26r1-3D243231-26r2-3D243232-26view-3Ddiff&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=A7YEbZn5UiSjeArc6rLNc6OXlfY7Hm0OXfSQWGYHAns&s=pRVul_ct6AiIzUBKOQnvAVJ__uAbRsd-pdfkjBz9dco&e=>
>> >
>> ==============================================================================
>> > --- lld/trunk/ELF/Writer.cpp (original)
>> > +++ lld/trunk/ELF/Writer.cpp Sat Jul 25 19:50:15 2015
>> > @@ -10,6 +10,7 @@
>> > #include "Writer.h"
>> > #include "Chunks.h"
>> > #include "Driver.h"
>> > +#include <map>
>> >
>> > using namespace llvm;
>> > using namespace llvm::ELF;
>> > @@ -58,24 +59,14 @@ void OutputSection::addChunk(Chunk *C) {
>> >   Header.sh_size = Off;
>> > }
>> >
>> > -static int compare(const Chunk *A, const Chunk *B) {
>> > -  return A->getSectionName() < B->getSectionName();
>> > -}
>> > -
>> > // Create output section objects and add them to OutputSections.
>> > template <class ELFT> void Writer<ELFT>::createSections() {
>> > -  std::vector<Chunk *> Chunks = Symtab->getChunks();
>> > -  if (Chunks.empty())
>> > -    return;
>> > -  std::sort(Chunks.begin(), Chunks.end(), compare);
>> > -
>> > -  Chunk *Prev = nullptr;
>> > -  OutputSection *Sec = nullptr;
>> > -  for (Chunk *C : Chunks) {
>> > -    if (Prev == nullptr || Prev->getSectionName() !=
>> C->getSectionName()) {
>> > +  std::map<StringRef, OutputSection *> Map;
>> > +  for (Chunk *C : Symtab->getChunks()) {
>> > +    OutputSection *&Sec = Map[C->getSectionName()];
>> > +    if (!Sec) {
>> >       Sec = new (CAlloc.Allocate()) OutputSection(C->getSectionName());
>> >       OutputSections.push_back(Sec);
>> > -      Prev = C;
>> >     }
>> >     Sec->addChunk(C);
>> >   }
>> >
>> >
>> > _______________________________________________
>> > llvm-commits mailing list
>> > llvm-commits at cs.uiuc.edu
>> > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>
>> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150727/06d8e340/attachment.html>