<div dir="ltr"><br><div class="gmail_extra"><div class="gmail_quote">On Wed, Aug 28, 2013 at 4:03 AM, David Blaikie <span dir="ltr"><<a href="mailto:dblaikie@gmail.com" target="_blank">dblaikie@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im">On Tue, Aug 27, 2013 at 1:45 AM, Dmitry Vyukov <<a href="mailto:dvyukov@google.com">dvyukov@google.com</a>> wrote:<br>
> On Tue, Aug 27, 2013 at 12:19 PM, Alexey Samsonov <<a href="mailto:samsonov@google.com">samsonov@google.com</a>> wrote:<br>
>><br>
>> On Tue, Aug 27, 2013 at 9:21 AM, David Blaikie <<a href="mailto:dblaikie@gmail.com">dblaikie@gmail.com</a>> wrote:<br>
>>><br>
>>> Author: dblaikie<br>
>>> Date: Tue Aug 27 00:21:11 2013<br>
>>> New Revision: 189320<br>
>>><br>
>>> URL: <a href="http://llvm.org/viewvc/llvm-project?rev=189320&view=rev" target="_blank">http://llvm.org/viewvc/llvm-project?rev=189320&view=rev</a><br>
>>> Log:<br>
>>> PR14569: Omit debug info for thunks<br>
>>><br>
>>> This was added in r166676 based on PR13942 on the basis that tools may<br>
>>> need debug information for any executable code/function for some fairly<br>
>>> broad/non-specific purposes. It seems to me (as noted in PR14569) that<br>
>>> the major/only purpose is in backtraces, which should generally not<br>
>>> apply to thunks as they won't appear in the stack themselves.<br>
>><br>
>><br>
>> Are you sure that thunks may not appear in backtraces? I think I saw them in<br>
>> stack traces,<br>
>> more specifically - in stack traces produced by ThreadSanitizer.<br>
>> It's possible that I'm wrong, though...<br>
><br>
> Debug info for thunks is necessary for sampling-based profilers and<br>
> gdb (if you step or stop in thunk).<br>
<br>
</div>In theory I would think this could just be handled using the symbol<br>
table - the only advantage of emitting debug info is to allow inlining<br>
to correctly attribute the inlined instructions to the original<br>
function, as I understand it (for tasks such as this anyway - where<br>
there's no user code that maps into the function, etc).<br>
<br>
Please let me know what your thoughts on that are.<br></blockquote><div><br></div><div>I guess debug info can be necessary to match generated thunk with</div><div>source file (this is also true for generated functions that initialize global variables etc).</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
In the mean time, given that there was no observable debug info size<br>
savings (which I still don't really understand/believe, but can't<br>
refute), I have reverted & reworked this patch to keep thunk debug<br>
info but to encode the name only in the linkage name in the debug<br>
info, not the plain name. This is sufficient to fix the relevant GDB<br>
test case.<br>
<span class="HOEnZb"><font color="#888888"><br>
- David<br>
</font></span><div class="HOEnZb"><div class="h5"><br>
><br>
><br>
>>> By<br>
>>> removing them we fix PR14569 and reduce the size of Clang's debug info.<br>
>>><br>
>>> Strangely enough this doesn't seem to have a substantial impact on<br>
>>> Clang's self-hosted debug info (at least looking at DWO file size) size<br>
>>> at all. Not sure if I failed to test this correctly but I only observed<br>
>>> a 0.004% change in DWO file size over Clang+LLVM.<br>
>>><br>
>>> With thanks to Dinesh Dwivedi for work on this PR.<br>
>>><br>
>>> Modified:<br>
>>> cfe/trunk/lib/CodeGen/CGVTables.cpp<br>
>>> cfe/trunk/test/CodeGenCXX/debug-info-thunk.cpp<br>
>>><br>
>>> Modified: cfe/trunk/lib/CodeGen/CGVTables.cpp<br>
>>> URL:<br>
>>> <a href="http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGVTables.cpp?rev=189320&r1=189319&r2=189320&view=diff" target="_blank">http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGVTables.cpp?rev=189320&r1=189319&r2=189320&view=diff</a><br>
>>><br>
>>> ==============================================================================<br>
>>> --- cfe/trunk/lib/CodeGen/CGVTables.cpp (original)<br>
>>> +++ cfe/trunk/lib/CodeGen/CGVTables.cpp Tue Aug 27 00:21:11 2013<br>
>>> @@ -292,6 +292,7 @@ void CodeGenFunction::GenerateVarArgsThu<br>
>>> void CodeGenFunction::GenerateThunk(llvm::Function *Fn,<br>
>>> const CGFunctionInfo &FnInfo,<br>
>>> GlobalDecl GD, const ThunkInfo<br>
>>> &Thunk) {<br>
>>> + DebugInfo = NULL; // debug info for thunks is not required or desired<br>
>>> const CXXMethodDecl *MD = cast<CXXMethodDecl>(GD.getDecl());<br>
>>> const FunctionProtoType *FPT =<br>
>>> MD->getType()->getAs<FunctionProtoType>();<br>
>>> QualType ThisType = MD->getThisType(getContext());<br>
>>><br>
>>> Modified: cfe/trunk/test/CodeGenCXX/debug-info-thunk.cpp<br>
>>> URL:<br>
>>> <a href="http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenCXX/debug-info-thunk.cpp?rev=189320&r1=189319&r2=189320&view=diff" target="_blank">http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenCXX/debug-info-thunk.cpp?rev=189320&r1=189319&r2=189320&view=diff</a><br>
>>><br>
>>> ==============================================================================<br>
>>> --- cfe/trunk/test/CodeGenCXX/debug-info-thunk.cpp (original)<br>
>>> +++ cfe/trunk/test/CodeGenCXX/debug-info-thunk.cpp Tue Aug 27 00:21:11<br>
>>> 2013<br>
>>> @@ -14,4 +14,4 @@ struct C : A, B {<br>
>>><br>
>>> void C::f() { }<br>
>>><br>
>>> -// CHECK: [ DW_TAG_subprogram ] [line 15] [def] [_ZThn{{4|8}}_N1C1fEv]<br>
>>> +// CHECK-NOT: [ DW_TAG_subprogram ] [line 15] [def]<br>
>>> [_ZThn{{4|8}}_N1C1fEv]<br>
>>><br>
>>><br>
>>> _______________________________________________<br>
>>> cfe-commits mailing list<br>
>>> <a href="mailto:cfe-commits@cs.uiuc.edu">cfe-commits@cs.uiuc.edu</a><br>
>>> <a href="http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits</a><br>
>><br>
>><br>
>><br>
>><br>
>> --<br>
>> Alexey Samsonov, MSK<br>
</div></div></blockquote></div><br><br clear="all"><div><br></div>-- <br><div>Alexey Samsonov, MSK</div>
</div></div>