[PATCH] D137101: [CodeView] Replace GHASH hasher by BLAKE3

[Alexandre Ganea via Phabricator via llvm-commits]

aganea created this revision.
aganea added reviewers: rnk, hans, thieta.
Herald added a subscriber: hiraditya.
Herald added a project: All.
aganea requested review of this revision.
Herald added a project: LLVM.
Herald added a subscriber: llvm-commits.

We already discussed this in the past: SHA1 in `GloballyHashedType::hashType()` is coming top in the profiles. By simply replacing with BLAKE3, the link time is reduced from 15 sec to 13 sec. I am only using MSVC .OBJs in this case. As a reference, the resulting .PDB is approx 2.1GiB and .EXE is approx 250MiB.

Before:
F25138992: image.png <https://reviews.llvm.org/F25138992>

After:
F25138976: image.png <https://reviews.llvm.org/F25138976>


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D137101

Files:
  lld/COFF/DebugTypes.cpp
  llvm/include/llvm/DebugInfo/CodeView/TypeHashing.h
  llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
  llvm/lib/DebugInfo/CodeView/TypeHashing.cpp


Index: llvm/lib/DebugInfo/CodeView/TypeHashing.cpp
===================================================================
--- llvm/lib/DebugInfo/CodeView/TypeHashing.cpp
+++ llvm/lib/DebugInfo/CodeView/TypeHashing.cpp
@@ -9,7 +9,7 @@
 #include "llvm/DebugInfo/CodeView/TypeHashing.h"
 
 #include "llvm/DebugInfo/CodeView/TypeIndexDiscovery.h"
-#include "llvm/Support/SHA1.h"
+#include "llvm/Support/BLAKE3.h"
 
 using namespace llvm;
 using namespace llvm::codeview;
@@ -35,7 +35,7 @@
                              ArrayRef<GloballyHashedType> PreviousIds) {
   SmallVector<TiReference, 4> Refs;
   discoverTypeIndices(RecordData, Refs);
-  SHA1 S;
+  TruncatedBLAKE3<8> S;
   S.init();
   uint32_t Off = 0;
   S.update(RecordData.take_front(sizeof(RecordPrefix)));
@@ -76,6 +76,6 @@
   auto TrailingBytes = RecordData.drop_front(Off);
   S.update(TrailingBytes);
 
-  std::array<uint8_t, 20> Hash = S.final();
-  return {ArrayRef<uint8_t>(Hash).take_back(8)};
+  std::array<uint8_t, 8> Hash = S.final();
+  return {ArrayRef<uint8_t>(Hash)};
 }
Index: llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
===================================================================
--- llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
+++ llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
@@ -760,7 +760,7 @@
   OS.AddComment("Section Version");
   OS.emitInt16(0);
   OS.AddComment("Hash Algorithm");
-  OS.emitInt16(uint16_t(GlobalTypeHashAlg::SHA1_8));
+  OS.emitInt16(uint16_t(GlobalTypeHashAlg::BLAKE3));
 
   TypeIndex TI(TypeIndex::FirstNonSimpleIndex);
   for (const auto &GHR : TypeTable.hashes()) {
Index: llvm/include/llvm/DebugInfo/CodeView/TypeHashing.h
===================================================================
--- llvm/include/llvm/DebugInfo/CodeView/TypeHashing.h
+++ llvm/include/llvm/DebugInfo/CodeView/TypeHashing.h
@@ -61,7 +61,8 @@
 
 enum class GlobalTypeHashAlg : uint16_t {
   SHA1 = 0, // standard 20-byte SHA1 hash
-  SHA1_8    // last 8-bytes of standard SHA1 hash
+  SHA1_8,   // last 8-bytes of standard SHA1 hash
+  BLAKE3,   // trucated 8-bytes BLAKE3
 };
 
 /// A globally hashed type represents a hash value that is sufficient to
Index: lld/COFF/DebugTypes.cpp
===================================================================
--- lld/COFF/DebugTypes.cpp
+++ lld/COFF/DebugTypes.cpp
@@ -280,7 +280,7 @@
   debugH = debugH.drop_front(sizeof(object::debug_h_header));
   return header->Magic == COFF::DEBUG_HASHES_SECTION_MAGIC &&
          header->Version == 0 &&
-         header->HashAlgorithm == uint16_t(GlobalTypeHashAlg::SHA1_8) &&
+         header->HashAlgorithm == uint16_t(GlobalTypeHashAlg::BLAKE3) &&
          (debugH.size() % 8 == 0);
 }
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D137101.472031.patch
Type: text/x-patch
Size: 2671 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20221031/806c9466/attachment.bin>