[clang] 2e4d2d8 - Reapply "[DebugInfo] Alternate (more efficient) MD5 fix"

[Paul Robinson via cfe-commits]

Author: Paul Robinson
Date: 2023-08-18T05:24:00-07:00
New Revision: 2e4d2d800b9ce0924513a2f24e7a1f3d22b52383

URL: https://github.com/llvm/llvm-project/commit/2e4d2d800b9ce0924513a2f24e7a1f3d22b52383
DIFF: https://github.com/llvm/llvm-project/commit/2e4d2d800b9ce0924513a2f24e7a1f3d22b52383.diff

LOG: Reapply "[DebugInfo] Alternate (more efficient) MD5 fix"

D155991 changed the file lookup to do a full string compare on the
filename; however, this added ~0.5% to compile time with -g.
Go back to the previous pointer-based lookup, but capture the main
file's checksum as well as its name to use when creating the extra
DIFile entry. This causes all entries to be consistent and also
avoids computing the checksum twice.

This reverts commit 21e7f73494663814e0296066895dfacdebfac6d4.
I'm unable to find a reason for the memory management issues
that caused the revert, so trying again.

Added: 
    

Modified: 
    clang/lib/CodeGen/CGDebugInfo.cpp
    clang/lib/CodeGen/CGDebugInfo.h
    clang/test/CodeGenCXX/debug-info-function-context.cpp

Removed: 
    


################################################################################
diff  --git a/clang/lib/CodeGen/CGDebugInfo.cpp b/clang/lib/CodeGen/CGDebugInfo.cpp
index 54e31bec0cc5bc..e782b7f906e6e6 100644
--- a/clang/lib/CodeGen/CGDebugInfo.cpp
+++ b/clang/lib/CodeGen/CGDebugInfo.cpp
@@ -391,12 +391,14 @@ llvm::DIFile *CGDebugInfo::getOrCreateFile(SourceLocation Loc) {
   SourceManager &SM = CGM.getContext().getSourceManager();
   StringRef FileName;
   FileID FID;
+  std::optional<llvm::DIFile::ChecksumInfo<StringRef>> CSInfo;
 
   if (Loc.isInvalid()) {
     // The DIFile used by the CU is distinct from the main source file. Call
     // createFile() below for canonicalization if the source file was specified
     // with an absolute path.
     FileName = TheCU->getFile()->getFilename();
+    CSInfo = TheCU->getFile()->getChecksum();
   } else {
     PresumedLoc PLoc = SM.getPresumedLoc(Loc);
     FileName = PLoc.getFilename();
@@ -417,13 +419,13 @@ llvm::DIFile *CGDebugInfo::getOrCreateFile(SourceLocation Loc) {
       return cast<llvm::DIFile>(V);
   }
 
-  SmallString<64> Checksum;
-
-  std::optional<llvm::DIFile::ChecksumKind> CSKind =
+  if (!CSInfo) {
+    SmallString<64> Checksum;
+    std::optional<llvm::DIFile::ChecksumKind> CSKind =
       computeChecksum(FID, Checksum);
-  std::optional<llvm::DIFile::ChecksumInfo<StringRef>> CSInfo;
-  if (CSKind)
-    CSInfo.emplace(*CSKind, Checksum);
+    if (CSKind)
+      CSInfo.emplace(*CSKind, Checksum);
+  }
   return createFile(FileName, CSInfo, getSource(SM, SM.getFileID(Loc)));
 }
 

diff  --git a/clang/lib/CodeGen/CGDebugInfo.h b/clang/lib/CodeGen/CGDebugInfo.h
index ec843793deea6c..b47a4934f0b1ba 100644
--- a/clang/lib/CodeGen/CGDebugInfo.h
+++ b/clang/lib/CodeGen/CGDebugInfo.h
@@ -148,7 +148,7 @@ class CGDebugInfo {
   llvm::BumpPtrAllocator DebugInfoNames;
   StringRef CWDName;
 
-  llvm::StringMap<llvm::TrackingMDRef> DIFileCache;
+  llvm::DenseMap<const char *, llvm::TrackingMDRef> DIFileCache;
   llvm::DenseMap<const FunctionDecl *, llvm::TrackingMDRef> SPCache;
   /// Cache declarations relevant to DW_TAG_imported_declarations (C++
   /// using declarations and global alias variables) that aren't covered

diff  --git a/clang/test/CodeGenCXX/debug-info-function-context.cpp b/clang/test/CodeGenCXX/debug-info-function-context.cpp
index 8d3309f42748dc..63fdf877fdb7db 100644
--- a/clang/test/CodeGenCXX/debug-info-function-context.cpp
+++ b/clang/test/CodeGenCXX/debug-info-function-context.cpp
@@ -1,5 +1,5 @@
 // RUN: %clang_cc1 -emit-llvm -debug-info-kind=limited -triple x86_64-pc-linux-gnu %s \
-// RUN:     -dwarf-version=5 -main-file-name %s  -o - | FileCheck %s
+// RUN:     -dwarf-version=5 -main-file-name debug-info-function-context.cpp  -o - | FileCheck %s
 
 struct C {
   void member_function();
@@ -31,8 +31,8 @@ int global_initialized_variable = C::static_member_function();
 
 // The first DIFile is for the CU, the second is what everything else uses.
 // We're using DWARF v5 so both should have MD5 checksums.
-// CHECK: !DIFile(filename: "{{.*}}context.cpp",{{.*}} checksumkind: CSK_MD5
-// CHECK: ![[FILE:[0-9]+]] = !DIFile(filename: "{{.*}}context.cpp",{{.*}} checksumkind: CSK_MD5
+// CHECK: !DIFile(filename: "{{.*}}context.cpp",{{.*}} checksumkind: CSK_MD5, checksum: [[CKSUM:".*"]]
+// CHECK: ![[FILE:[0-9]+]] = !DIFile(filename: "{{.*}}context.cpp",{{.*}} checksumkind: CSK_MD5, checksum: [[CKSUM]]
 // CHECK: ![[C:[0-9]+]] = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "C",
 // CHECK: ![[NS:.*]] = !DINamespace(name: "ns"
 // CHECK: !DISubprogram(name: "member_function",{{.*}} scope: ![[C]],{{.*}} DISPFlagDefinition