[llvm] r339122 - [DebugInfo] Reduce debug_str_offsets section size

David Blaikie via llvm-commits llvm-commits at lists.llvm.org
Tue Aug 28 11:37:35 PDT 2018


No worries - thanks for talking it out!

On Tue, Aug 28, 2018 at 11:33 AM Pavel Labath <pavel at labath.sk> wrote:

> On 28/08/18 18:46, David Blaikie wrote:
> >
> >
> > On Tue, Aug 28, 2018 at 6:07 AM Pavel Labath <pavel at labath.sk
> > <mailto:pavel at labath.sk>> wrote:
> >
> >     On 28/08/18 02:08, David Blaikie wrote:
> >     > Any idea if there's work to add str_index support to the name
> tables?
> >     >
> >     > I guess the idea is that it's more efficient to have one fewer
> >     > indirections for strings, and worth the cost in duplicate
> >     > relocations/bytes/etc? *shrug*
> >
> >     Interesting question. I am not aware of anyone looking at this. I
> >     suppose that would trade a bit of performance for a bit of size (but
> >     only for the intermediate .o file and only when using split-dwarf),
> >     though I can't estimate how big would those "bits" be.
> >
> >
> > Hmm - why only when using split-dwarf? I'd have thought the opposite -
> > when not using split-dwarf the strings will be indexed but debug_names
> > won't use that index, instead using its own. So in that case you'd get
> > duplicate indexes, basically. When using split-dwarf it'd be probably
> > unhelpful - just trading one index for another/adding that extra layer
> > of indirection. (because the only reason the strings are in the index is
> > for the debug_names in the split-DWARF situation).
> >
> >
>
> You are correct, of course. I think I got my debug modes the wrong way
> round.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180828/928af440/attachment.html>


More information about the llvm-commits mailing list