[clang] Optimize storage and lookup of analyzer options

[jansvoboda11]

Summary

This patch moves llvm::sort() from AnalyzerOptions constructor to initialization of local static variable in isUnknownAnalyzerConfig().

This avoids unnecessary work, which can speed up Clang tools that initialize lots of CompilerInvocations (and therefore AnalyzerOptions).

[steakhal]

Are you sure about the speedup?
When I looked at the implementation of the StringSwitch, it boiled down to an if-else chain under the hood.

[steakhal]

BTW the change itself looks good to me.

[jansvoboda11]

In D137258#3947181, @steakhal wrote:

Are you sure about the speedup?
When I looked at the implementation of the StringSwitch, it boiled down to an if-else chain under the hood.

You're right. I assumed StringSwitch was more sophisticated than that. This patch provides speedup for my use-case, which involves constructing lots of AnalyzerOptions instances, but no calls to AnalyzerOptions::isUnknownAnalyzerConfig(). I'll spend more time on this and make sure I don't regress the lookup. (Unfortunately, std::sort is only constexpr in C++20 onwards, so I can't just sort the options at compile-time.)

[jansvoboda11]

Move from llvm::StringSwitch back to binary search on sorted vector

[steakhal]

I liked the idea of sorting these options at compile time.
I think the most dummy sort should work just fine inthis case, as the number of elements are not that large.
Do you think worth it?

Anyway , I think moving the runtime sort out of the ctor is a good middleground. Thank you.

[jansvoboda11]

Sorting at compile-time would be nicer, but I don't have the bandwidth to implement something like that. I'll leave a FIXME in case someone wants to pick it up. Thanks for the review!

clang/include/clang/StaticAnalyzer/Core/AnalyzerOptions.h

 //===- AnalyzerOptions.h - Analysis Engine Options --------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //

 
 #define ANALYZER_OPTION(TYPE, NAME, CMDFLAG, DESC, DEFAULT_VAL)                \
   TYPE NAME;
 
 #include "clang/StaticAnalyzer/Core/AnalyzerOptions.def"
 #undef ANALYZER_OPTION
 #undef ANALYZER_OPTION_DEPENDS_ON_USER_MODE
 
-  // Create an array of all -analyzer-config command line options. Sort it in
-  // the constructor.
+  bool isUnknownAnalyzerConfig(llvm::StringRef Name) {
+    static std::vector<llvm::StringLiteral> AnalyzerConfigCmdFlags = []() {
+      // Create an array of all -analyzer-config command line options.
       std::vector<llvm::StringLiteral> AnalyzerConfigCmdFlags = {
 #define ANALYZER_OPTION_DEPENDS_ON_USER_MODE(TYPE, NAME, CMDFLAG, DESC,        \
                                              SHALLOW_VAL, DEEP_VAL)            \
   ANALYZER_OPTION(TYPE, NAME, CMDFLAG, DESC, SHALLOW_VAL)
 
 #define ANALYZER_OPTION(TYPE, NAME, CMDFLAG, DESC, DEFAULT_VAL)                \
   llvm::StringLiteral(CMDFLAG),
 
 #include "clang/StaticAnalyzer/Core/AnalyzerOptions.def"
 #undef ANALYZER_OPTION
 #undef ANALYZER_OPTION_DEPENDS_ON_USER_MODE
       };
-
-  bool isUnknownAnalyzerConfig(StringRef Name) const {
-    assert(llvm::is_sorted(AnalyzerConfigCmdFlags));
+      llvm::sort(AnalyzerConfigCmdFlags);
+      return AnalyzerConfigCmdFlags;
+    }();
 
     return !std::binary_search(AnalyzerConfigCmdFlags.begin(),
                                AnalyzerConfigCmdFlags.end(), Name);
   }
 
   AnalyzerOptions()
       : DisableAllCheckers(false), ShowCheckerHelp(false),
         ShowCheckerHelpAlpha(false), ShowCheckerHelpDeveloper(false),
         ShowCheckerOptionList(false), ShowCheckerOptionAlphaList(false),
         ShowCheckerOptionDeveloperList(false), ShowEnabledCheckerList(false),
         ShowConfigOptionsList(false),
         ShouldEmitErrorsOnInvalidConfigValue(false), AnalyzeAll(false),
         AnalyzerDisplayProgress(false), eagerlyAssumeBinOpBifurcation(false),
         TrimGraph(false), visualizeExplodedGraphWithGraphViz(false),
         UnoptimizedCFG(false), PrintStats(false), NoRetryExhausted(false),
-        AnalyzerWerror(false) {
-    llvm::sort(AnalyzerConfigCmdFlags);
-  }
+        AnalyzerWerror(false) {}
 
   /// Interprets an option's string value as a boolean. The "true" string is
   /// interpreted as true and the "false" string is interpreted as false.
   ///
   /// If an option value is not provided, returns the given \p DefaultVal.
   /// @param [in] CheckerName The *full name* of the checker. One may retrieve
   /// this from the checker object's field \c Name, or through \c
   /// CheckerManager::getCurrentCheckerName within the checker's registry