Allocate P2WSH/P2TR/P2PK scripts on stack

The current `prevector` size of 28 bytes (chosen to fill the `sizeof(CScript)` aligned size) was introduced in 2015 (https://github.com/bitcoin/bitcoin/pull/6914) before SegWit and TapRoot.
However, the increasingly common `P2WSH` and `P2TR` scripts are both 34 bytes, and are forced to use heap (re)allocation rather than efficient inline storage.

The core trade-off of this change is to eliminate heap allocations for common 34-36 byte scripts at the cost of increasing the base memory footprint of all `CScript` objects by 8 bytes (while still respecting peak memory usage defined by `-dbcache`).

Increasing the `prevector` size allows these scripts to be stored on the stack, avoiding heap allocations, reducing potential memory fragmentation, and improving performance during cache flushes. Massif analysis confirms a lower stable memory usage after flushing, suggesting the elimination of heap allocations outweighs the larger base size for common workloads.

Due to memory alignment, increasing the `prevector` size to 36 bytes doesn't change the overall `sizeof(CScript)` compared to an increase to 34 bytes, allowing us to include `P2PK` scripts as well at no additional memory cost.

Performance benchmarks for AssumeUTXO load and flush show:
- Small dbcache (450MB): ~1% performance penalty due to more frequent flushes
- Large dbcache (4500-4500MB+): ~6-7% performance improvement due to fewer heap allocations

Full IBD and reindex-chainstate with larger `dbcache` values also show an overall ~3% speedup.

Co-authored-by: Ava Chow <github@achow101.com>
Co-authored-by: Andrew Toth <andrewstoth@gmail.com>

src/bench/checkqueue.cpp

@@ -16,7 +16,7 @@
 
 static const size_t BATCHES = 101;
 static const size_t BATCH_SIZE = 30;
-static const int PREVECTOR_SIZE = 28;
+static const int PREVECTOR_SIZE = 36;
 static const unsigned int QUEUE_BATCH_SIZE = 128;
 
 // This Benchmark tests the CheckQueue with a slightly realistic workload,

src/bench/prevector.cpp

@@ -27,22 +27,22 @@ template <typename T>
 static void PrevectorDestructor(benchmark::Bench& bench)
 {
     bench.batch(2).run([&] {
-        prevector<28, T> t0;
+        prevector<36, T> t0;
-        prevector<28, T> t1;
+        prevector<36, T> t1;
-        t0.resize(28);
+        t0.resize(36);
-        t1.resize(29);
+        t1.resize(37);
     });
 }
 
 template <typename T>
 static void PrevectorClear(benchmark::Bench& bench)
 {
-    prevector<28, T> t0;
+    prevector<36, T> t0;
-    prevector<28, T> t1;
+    prevector<36, T> t1;
     bench.batch(2).run([&] {
-        t0.resize(28);
+        t0.resize(36);
         t0.clear();
-        t1.resize(29);
+        t1.resize(37);
         t1.clear();
     });
 }
@@ -50,12 +50,12 @@ static void PrevectorClear(benchmark::Bench& bench)
 template <typename T>
 static void PrevectorResize(benchmark::Bench& bench)
 {
-    prevector<28, T> t0;
+    prevector<36, T> t0;
-    prevector<28, T> t1;
+    prevector<36, T> t1;
     bench.batch(4).run([&] {
-        t0.resize(28);
+        t0.resize(36);
         t0.resize(0);
-        t1.resize(29);
+        t1.resize(37);
         t1.resize(0);
     });
 }
@@ -64,8 +64,8 @@ template <typename T>
 static void PrevectorDeserialize(benchmark::Bench& bench)
 {
     DataStream s0{};
-    prevector<28, T> t0;
+    prevector<36, T> t0;
-    t0.resize(28);
+    t0.resize(36);
     for (auto x = 0; x < 900; ++x) {
         s0 << t0;
     }
@@ -74,7 +74,7 @@ static void PrevectorDeserialize(benchmark::Bench& bench)
         s0 << t0;
     }
     bench.batch(1000).run([&] {
-        prevector<28, T> t1;
+        prevector<36, T> t1;
         for (auto x = 0; x < 1000; ++x) {
             s0 >> t1;
         }
@@ -86,7 +86,7 @@ template <typename T>
 static void PrevectorFillVectorDirect(benchmark::Bench& bench)
 {
     bench.run([&] {
-        std::vector<prevector<28, T>> vec;
+        std::vector<prevector<36, T>> vec;
         vec.reserve(260);
         for (size_t i = 0; i < 260; ++i) {
             vec.emplace_back();
@@ -99,11 +99,11 @@ template <typename T>
 static void PrevectorFillVectorIndirect(benchmark::Bench& bench)
 {
     bench.run([&] {
-        std::vector<prevector<28, T>> vec;
+        std::vector<prevector<36, T>> vec;
         vec.reserve(260);
         for (size_t i = 0; i < 260; ++i) {
             // force allocation
-            vec.emplace_back(29, T{});
+            vec.emplace_back(37, T{});
         }
     });
 }

src/script/script.h

@@ -406,7 +406,7 @@ private:
  * Tests in October 2015 showed use of this reduced dbcache memory usage by 23%
  *  and made an initial sync 13% faster.
  */
-typedef prevector<28, unsigned char> CScriptBase;
+typedef prevector<36, unsigned char> CScriptBase;
 
 bool GetScriptOp(CScriptBase::const_iterator& pc, CScriptBase::const_iterator end, opcodetype& opcodeRet, std::vector<unsigned char>* pvchRet);
 

src/test/script_tests.cpp

@@ -1131,10 +1131,10 @@ BOOST_AUTO_TEST_CASE(script_CHECKMULTISIG23)
 
 BOOST_AUTO_TEST_CASE(script_size_and_capacity_test)
 {
-    BOOST_CHECK_EQUAL(sizeof(prevector<28, unsigned char>), 32);
+    BOOST_CHECK_EQUAL(sizeof(prevector<34, uint8_t>), sizeof(prevector<36, uint8_t>));
-    BOOST_CHECK_EQUAL(sizeof(CScriptBase), 32);
+    BOOST_CHECK_EQUAL(sizeof(CScriptBase), 40);
-    BOOST_CHECK_EQUAL(sizeof(CScript), 32);
+    BOOST_CHECK_EQUAL(sizeof(CScript), 40);
-    BOOST_CHECK_EQUAL(sizeof(CTxOut), 40);
+    BOOST_CHECK_EQUAL(sizeof(CTxOut), 48);
 
     CKey dummyKey;
     dummyKey.MakeNewKey(true);
@@ -1146,7 +1146,7 @@ BOOST_AUTO_TEST_CASE(script_size_and_capacity_test)
         const auto scriptSmallOpReturn{CScript() << OP_RETURN << std::vector<uint8_t>(10, 0xaa)};
         BOOST_CHECK_EQUAL(Solver(scriptSmallOpReturn, dummyVSolutions), TxoutType::NULL_DATA);
         BOOST_CHECK_EQUAL(scriptSmallOpReturn.size(), 12);
-        BOOST_CHECK_EQUAL(scriptSmallOpReturn.capacity(), 28);
+        BOOST_CHECK_EQUAL(scriptSmallOpReturn.capacity(), 36);
         BOOST_CHECK_EQUAL(scriptSmallOpReturn.allocated_memory(), 0);
     }
 
@@ -1155,7 +1155,7 @@ BOOST_AUTO_TEST_CASE(script_size_and_capacity_test)
         const auto scriptP2WPKH{GetScriptForDestination(WitnessV0KeyHash{PKHash{CKeyID{CPubKey{dummyKey.GetPubKey()}.GetID()}}})};
         BOOST_CHECK_EQUAL(Solver(scriptP2WPKH, dummyVSolutions), TxoutType::WITNESS_V0_KEYHASH);
         BOOST_CHECK_EQUAL(scriptP2WPKH.size(), 22);
-        BOOST_CHECK_EQUAL(scriptP2WPKH.capacity(), 28);
+        BOOST_CHECK_EQUAL(scriptP2WPKH.capacity(), 36);
         BOOST_CHECK_EQUAL(scriptP2WPKH.allocated_memory(), 0);
     }
 
@@ -1164,7 +1164,7 @@ BOOST_AUTO_TEST_CASE(script_size_and_capacity_test)
         const auto scriptP2SH{GetScriptForDestination(ScriptHash{CScript{} << OP_TRUE})};
         BOOST_CHECK(scriptP2SH.IsPayToScriptHash());
         BOOST_CHECK_EQUAL(scriptP2SH.size(), 23);
-        BOOST_CHECK_EQUAL(scriptP2SH.capacity(), 28);
+        BOOST_CHECK_EQUAL(scriptP2SH.capacity(), 36);
         BOOST_CHECK_EQUAL(scriptP2SH.allocated_memory(), 0);
     }
 
@@ -1173,35 +1173,35 @@ BOOST_AUTO_TEST_CASE(script_size_and_capacity_test)
         const auto scriptP2PKH{GetScriptForDestination(PKHash{CKeyID{CPubKey{dummyKey.GetPubKey()}.GetID()}})};
         BOOST_CHECK_EQUAL(Solver(scriptP2PKH, dummyVSolutions), TxoutType::PUBKEYHASH);
         BOOST_CHECK_EQUAL(scriptP2PKH.size(), 25);
-        BOOST_CHECK_EQUAL(scriptP2PKH.capacity(), 28);
+        BOOST_CHECK_EQUAL(scriptP2PKH.capacity(), 36);
         BOOST_CHECK_EQUAL(scriptP2PKH.allocated_memory(), 0);
     }
 
-    // P2WSH is heap allocated
+    // P2WSH is stack allocated
     {
         const auto scriptP2WSH{GetScriptForDestination(WitnessV0ScriptHash{CScript{} << OP_TRUE})};
         BOOST_CHECK(scriptP2WSH.IsPayToWitnessScriptHash());
         BOOST_CHECK_EQUAL(scriptP2WSH.size(), 34);
-        BOOST_CHECK_EQUAL(scriptP2WSH.capacity(), 34);
+        BOOST_CHECK_EQUAL(scriptP2WSH.capacity(), 36);
-        BOOST_CHECK_EQUAL(scriptP2WSH.allocated_memory(), 34);
+        BOOST_CHECK_EQUAL(scriptP2WSH.allocated_memory(), 0);
     }
 
-    // P2TR is heap allocated
+    // P2TR is stack allocated
     {
         const auto scriptTaproot{GetScriptForDestination(WitnessV1Taproot{XOnlyPubKey{CPubKey{dummyKey.GetPubKey()}}})};
         BOOST_CHECK_EQUAL(Solver(scriptTaproot, dummyVSolutions), TxoutType::WITNESS_V1_TAPROOT);
         BOOST_CHECK_EQUAL(scriptTaproot.size(), 34);
-        BOOST_CHECK_EQUAL(scriptTaproot.capacity(), 34);
+        BOOST_CHECK_EQUAL(scriptTaproot.capacity(), 36);
-        BOOST_CHECK_EQUAL(scriptTaproot.allocated_memory(), 34);
+        BOOST_CHECK_EQUAL(scriptTaproot.allocated_memory(), 0);
     }
 
-    // P2PK is heap allocated
+    // P2PK is stack allocated
     {
         const auto scriptPubKey{GetScriptForRawPubKey(CPubKey{dummyKey.GetPubKey()})};
         BOOST_CHECK_EQUAL(Solver(scriptPubKey, dummyVSolutions), TxoutType::PUBKEY);
         BOOST_CHECK_EQUAL(scriptPubKey.size(), 35);
-        BOOST_CHECK_EQUAL(scriptPubKey.capacity(), 35);
+        BOOST_CHECK_EQUAL(scriptPubKey.capacity(), 36);
-        BOOST_CHECK_EQUAL(scriptPubKey.allocated_memory(), 35);
+        BOOST_CHECK_EQUAL(scriptPubKey.allocated_memory(), 0);
     }
 
     // MULTISIG is always heap allocated

src/test/validation_flush_tests.cpp

@@ -26,10 +26,8 @@ BOOST_AUTO_TEST_CASE(getcoinscachesizestate)
     LOCK(::cs_main);
     auto& view = chainstate.CoinsTip();
 
-    // The number of bytes consumed by coin's heap data, i.e. CScript
+    // The number of bytes consumed by coin's heap data, i.e. CScript (prevector<36, unsigned char>)
-    // (prevector<28, unsigned char>) when assigned 56 bytes of data per above.
+    // when assigned 56 bytes of data per above. See also: Coin::DynamicMemoryUsage().
-    //
-    // See also: Coin::DynamicMemoryUsage().
     constexpr unsigned int COIN_SIZE = is_64_bit ? 80 : 64;
 
     auto print_view_mem_usage = [](CCoinsViewCache& view) {