<div dir="ltr">I can try to take some time to benchmark, hopefully this weekend or so.<div><br></div><div>I do want to reiterate that this without this change debugging binaries linked against libstdc++ is badly degraded, since recent libstdc++ uses ABI tags on ~any methods that touch a std::string.<br><div><br></div><div>- Nelson</div><div><br></div><div class="gmail_quote"><div dir="ltr">On Thu, Jan 11, 2018 at 7:18 AM Davide Italiano <<a href="mailto:dccitaliano@gmail.com" target="_blank">dccitaliano@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On Thu, Jan 11, 2018 at 2:01 AM, Pavel Labath via lldb-commits<br>
<<a href="mailto:lldb-commits@lists.llvm.org" target="_blank">lldb-commits@lists.llvm.org</a>> wrote:<br>
> On 10 January 2018 at 22:51, Greg Clayton <<a href="mailto:clayborg@gmail.com" target="_blank">clayborg@gmail.com</a>> wrote:<br>
>> The right solution seems to be adding some sort of custom GNU ABI tag to the DWARF. I know that won't help with existing binaries, but it sounds too expensive to set the ASM name for everything.<br>
>><br>
><br>
> What makes you think it will be expensive? I don't know much about<br>
> clang internals, but I think something has to provide a mangled name<br>
> in any case. So, if we don't specify a mangled name, clang will have<br>
> to do the mangling itself. For all I know, setting the asm attribute<br>
> may actually speed things up, as clang could avoid doing some extra<br>
> work.<br>
><br>
> I think it would be interesting to try to measure the<br>
> performance/memory footprint impact of just setting the asm attribute<br>
> everywhere.<br>
><br>
<br>
Indeed. If somebody has the time to put into this I think it's<br>
something worth benchmarking before declaring victory (or defeat).<br>
<br>
Thanks,<br>
<br>
--<br>
Davide<br>
</blockquote></div></div></div>