[BOLT] Delta-encode function start addresses in BAT (#76902)

Further reduce the size of BAT section:
- large binary: to 12716312 bytes (0.33x original),
- medium binary: to 1649472 bytes (0.28x original),
- small binary: to 428 bytes (0.30x original).

Test Plan: Updated bolt/test/X86/bolt-address-translation.test

3 files changed, 10 insertions, 4 deletions

diff --git a/bolt/docs/BAT.md b/bolt/docs/BAT.md
index b44a51b..8a3b2fc 100644
--- a/bolt/docs/BAT.md
+++ b/bolt/docs/BAT.md
@@ -64,9 +64,11 @@ Header:
 | `NumFuncs` | ULEB128 | Number of functions in the functions table |
 
 The header is followed by Functions table with `NumFuncs` entries.
+Output binary addresses are delta encoded, meaning that only the difference with
+the previous output address is stored. Addresses implicitly start at zero.
 | Entry  | Encoding | Description |
 | ------ | ------| ----------- |
-| `Address` | ULEB128 | Function address in the output binary |
+| `Address` | Delta, ULEB128 | Function address in the output binary |
 | `NumEntries` | ULEB128 | Number of address translation entries for a function |
 
 Function header is followed by `NumEntries` pairs of offsets for current
diff --git a/bolt/lib/Profile/BoltAddressTranslation.cpp b/bolt/lib/Profile/BoltAddressTranslation.cpp
index 6079b97..19b63d4 100644
--- a/bolt/lib/Profile/BoltAddressTranslation.cpp
+++ b/bolt/lib/Profile/BoltAddressTranslation.cpp
@@ -106,13 +106,15 @@ void BoltAddressTranslation::write(const BinaryContext &BC, raw_ostream &OS) {
   const uint32_t NumFuncs = Maps.size();
   encodeULEB128(NumFuncs, OS);
   LLVM_DEBUG(dbgs() << "Writing " << NumFuncs << " functions for BAT.\n");
+  uint64_t PrevAddress = 0;
   for (auto &MapEntry : Maps) {
     const uint64_t Address = MapEntry.first;
     MapTy &Map = MapEntry.second;
     const uint32_t NumEntries = Map.size();
     LLVM_DEBUG(dbgs() << "Writing " << NumEntries << " entries for 0x"
                       << Twine::utohexstr(Address) << ".\n");
-    encodeULEB128(Address, OS);
+    encodeULEB128(Address - PrevAddress, OS);
+    PrevAddress = Address;
     encodeULEB128(NumEntries, OS);
     uint64_t InOffset = 0, OutOffset = 0;
     // Output and Input addresses and delta-encoded
@@ -160,8 +162,10 @@ std::error_code BoltAddressTranslation::parse(StringRef Buf) {
   Error Err(Error::success());
   const uint32_t NumFunctions = DE.getULEB128(&Offset, &Err);
   LLVM_DEBUG(dbgs() << "Parsing " << NumFunctions << " functions\n");
+  uint64_t PrevAddress = 0;
   for (uint32_t I = 0; I < NumFunctions; ++I) {
-    const uint64_t Address = DE.getULEB128(&Offset, &Err);
+    const uint64_t Address = PrevAddress + DE.getULEB128(&Offset, &Err);
+    PrevAddress = Address;
     const uint32_t NumEntries = DE.getULEB128(&Offset, &Err);
     MapTy Map;
 
diff --git a/bolt/test/X86/bolt-address-translation.test b/bolt/test/X86/bolt-address-translation.test
index fc57668..a232f78 100644
--- a/bolt/test/X86/bolt-address-translation.test
+++ b/bolt/test/X86/bolt-address-translation.test
@@ -37,7 +37,7 @@
 # CHECK:      BOLT: 3 out of 7 functions were overwritten.
 # CHECK:      BOLT-INFO: Wrote 6 BAT maps
 # CHECK:      BOLT-INFO: Wrote 3 BAT cold-to-hot entries
-# CHECK:      BOLT-INFO: BAT section size (bytes): 436
+# CHECK:      BOLT-INFO: BAT section size (bytes): 428
 #
 # usqrt mappings (hot part). We match against any key (left side containing
 # the bolted binary offsets) because BOLT may change where it puts instructions