[Lldb-commits] [PATCH] D152476: [lldb] Remove lldb's DWARFAbbreviationDeclarationSet in favor of llvm's

[Alex Langford via Phabricator via lldb-commits]

bulbazord added a comment.

In D152476#4408654 <https://reviews.llvm.org/D152476#4408654>, @fdeazeve wrote:

> This is awesome! I believe you said there was no measurable perf diff?

Yeah, my initial experiments measured no significant perf difference no matter how you built lldb. llvm's and lldb's `DWARFAbbreviationDeclarationSet` implementations both use llvm's `DWARFAbbreviationDeclaration` and have the same semantics around their extraction. I intentionally made both implementations behave the exact same so that this change would be as simple as possible.



================
Comment at: lldb/source/Plugins/SymbolFile/DWARF/DWARFDebugAbbrev.cpp:62
     std::set<dw_form_t> &invalid_forms) const {
-  for (const auto &pair : m_abbrevCollMap)
-    pair.second.GetUnsupportedForms(invalid_forms);
+  for (const auto &pair : m_abbrevCollMap) {
+    for (const auto &decl : pair.second) {
----------------
aprantl wrote:
> nit: LLVM coding style would probably elide the {}
Will update, thanks!


================
Comment at: lldb/source/Plugins/SymbolFile/DWARF/DWARFDebugAbbrev.h:23
 
 typedef std::map<dw_offset_t, DWARFAbbreviationDeclarationSet>
     DWARFAbbreviationDeclarationCollMap;
----------------
aprantl wrote:
> for a later commit: this should probably be a DenseMap for better performance?
Possibly? It would be interesting to switch away from std::map to a different type (like DenseMap) to measure performance. Either way, I also want to switch us to use `llvm::DWARFDebugAbbrev`, so it might be worth doing that work first.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152476/new/

https://reviews.llvm.org/D152476