| Message ID | CAO2gOZW4sA+CanEdrvo9ZQSSXcfCRtq6wVsW4TbHtp+_tDt5Sw@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
Does it paper over the real bug? David On Tue, Dec 16, 2014 at 2:38 PM, Dehao Chen <dehao@google.com> wrote: > This patch fixes the bug for undefined symbol in AutoFDO build. > > Testing on going. OK for google-4_9 branch? > > Thanks, > Dehao > > Index: gcc/auto-profile.c > =================================================================== > --- gcc/auto-profile.c (revision 218784) > +++ gcc/auto-profile.c (working copy) > @@ -592,6 +592,8 @@ > continue; > if (!check_ic_target (stmt, node)) > continue; > + if (!node->definition) > + continue; > (*map)[callee] = iter->second->total_count (); > ret += iter->second->total_count (); > }
The real bug is in debug info. The callgraph should look like: foo->D1EV->D4EV While everything is inlined into foo, but in the inline stack D1EV is missing. This patch does not fix that problem, but is also reasonable because if callee's definition is not available, we should not promote the indirect call anyway. Dehao On Tue, Dec 16, 2014 at 2:45 PM, Xinliang David Li <davidxl@google.com> wrote: > Does it paper over the real bug? > > David > > On Tue, Dec 16, 2014 at 2:38 PM, Dehao Chen <dehao@google.com> wrote: >> This patch fixes the bug for undefined symbol in AutoFDO build. >> >> Testing on going. OK for google-4_9 branch? >> >> Thanks, >> Dehao >> >> Index: gcc/auto-profile.c >> =================================================================== >> --- gcc/auto-profile.c (revision 218784) >> +++ gcc/auto-profile.c (working copy) >> @@ -592,6 +592,8 @@ >> continue; >> if (!check_ic_target (stmt, node)) >> continue; >> + if (!node->definition) >> + continue; >> (*map)[callee] = iter->second->total_count (); >> ret += iter->second->total_count (); >> }
On Tue, Dec 16, 2014 at 3:07 PM, Dehao Chen <dehao@google.com> wrote: > The real bug is in debug info. The callgraph should look like: > > foo->D1EV->D4EV > > While everything is inlined into foo, but in the inline stack D1EV is missing. Is it DFEed? > > This patch does not fix that problem, but is also reasonable because > if callee's definition is not available, we should not promote the > indirect call anyway. > It depends. If there is a wrong DFE before the promotion, we will end up missing opportunities. David > Dehao > > On Tue, Dec 16, 2014 at 2:45 PM, Xinliang David Li <davidxl@google.com> wrote: >> Does it paper over the real bug? >> >> David >> >> On Tue, Dec 16, 2014 at 2:38 PM, Dehao Chen <dehao@google.com> wrote: >>> This patch fixes the bug for undefined symbol in AutoFDO build. >>> >>> Testing on going. OK for google-4_9 branch? >>> >>> Thanks, >>> Dehao >>> >>> Index: gcc/auto-profile.c >>> =================================================================== >>> --- gcc/auto-profile.c (revision 218784) >>> +++ gcc/auto-profile.c (working copy) >>> @@ -592,6 +592,8 @@ >>> continue; >>> if (!check_ic_target (stmt, node)) >>> continue; >>> + if (!node->definition) >>> + continue; >>> (*map)[callee] = iter->second->total_count (); >>> ret += iter->second->total_count (); >>> }
Index: gcc/auto-profile.c =================================================================== --- gcc/auto-profile.c (revision 218784) +++ gcc/auto-profile.c (working copy) @@ -592,6 +592,8 @@ continue; if (!check_ic_target (stmt, node)) continue; + if (!node->definition) + continue; (*map)[callee] = iter->second->total_count (); ret += iter->second->total_count (); }