trie: cache collapsed tries node, not rlp blobs (ethereum#16876)

The current trie memory database/cache that we do pruning on stores
trie nodes as binary rlp encoded blobs, and also stores the node
relationships/references for GC purposes. However, most of the trie
nodes (everything apart from a value node) is in essence just a
collection of references.

This PR switches out the RLP encoded trie blobs with the
collapsed-but-not-serialized trie nodes. This permits most of the
references to be recovered from within the node data structure,
avoiding the need to track them a second time (expensive memory wise).

Author: karalabe

core/blockchain.go

@@ -672,7 +672,7 @@ func (bc *BlockChain) Stop() {
 			}
 		}
 		for !bc.triegc.Empty() {
-			triedb.Dereference(bc.triegc.PopItem().(common.Hash), common.Hash{})
+			triedb.Dereference(bc.triegc.PopItem().(common.Hash))
 		}
 		if size, _ := triedb.Size(); size != 0 {
 			log.Error("Dangling trie nodes after full cleanup")
@@ -947,7 +947,7 @@ func (bc *BlockChain) WriteBlockWithState(block *types.Block, receipts []*types.
 					bc.triegc.Push(root, number)
 					break
 				}
-				triedb.Dereference(root.(common.Hash), common.Hash{})
+				triedb.Dereference(root.(common.Hash))
 			}
 		}
 	}

core/blockchain_test.go

@@ -1313,8 +1313,8 @@ func TestTrieForkGC(t *testing.T) {
 	}
 	// Dereference all the recent tries and ensure no past trie is left in
 	for i := 0; i < triesInMemory; i++ {
-		chain.stateCache.TrieDB().Dereference(blocks[len(blocks)-1-i].Root(), common.Hash{})
-		chain.stateCache.TrieDB().Dereference(forks[len(blocks)-1-i].Root(), common.Hash{})
+		chain.stateCache.TrieDB().Dereference(blocks[len(blocks)-1-i].Root())
+		chain.stateCache.TrieDB().Dereference(forks[len(blocks)-1-i].Root())
 	}
 	if len(chain.stateCache.TrieDB().Nodes()) > 0 {
 		t.Fatalf("stale tries still alive after garbase collection")

core/state/statedb.go

@@ -596,7 +596,7 @@ func (s *StateDB) Commit(deleteEmptyObjects bool) (root common.Hash, err error)
 		case isDirty:
 			// Write any contract code associated with the state object
 			if stateObject.code != nil && stateObject.dirtyCode {
-				s.db.TrieDB().Insert(common.BytesToHash(stateObject.CodeHash()), stateObject.code)
+				s.db.TrieDB().InsertBlob(common.BytesToHash(stateObject.CodeHash()), stateObject.code)
 				stateObject.dirtyCode = false
 			}
 			// Write any storage changes in the state object to its storage trie.

eth/api_tracer.go

@@ -297,7 +297,7 @@ func (api *PrivateDebugAPI) traceChain(ctx context.Context, start, end *types.Bl
 				database.TrieDB().Reference(root, common.Hash{})
 			}
 			// Dereference all past tries we ourselves are done working with
-			database.TrieDB().Dereference(proot, common.Hash{})
+			database.TrieDB().Dereference(proot)
 			proot = root
 
 			// TODO(karalabe): Do we need the preimages? Won't they accumulate too much?
@@ -320,7 +320,7 @@ func (api *PrivateDebugAPI) traceChain(ctx context.Context, start, end *types.Bl
 			done[uint64(result.Block)] = result
 
 			// Dereference any paret tries held in memory by this task
-			database.TrieDB().Dereference(res.rootref, common.Hash{})
+			database.TrieDB().Dereference(res.rootref)
 
 			// Stream completed traces to the user, aborting on the first error
 			for result, ok := done[next]; ok; result, ok = done[next] {
@@ -526,7 +526,7 @@ func (api *PrivateDebugAPI) computeStateDB(block *types.Block, reexec uint64) (*
 			return nil, err
 		}
 		database.TrieDB().Reference(root, common.Hash{})
-		database.TrieDB().Dereference(proot, common.Hash{})
+		database.TrieDB().Dereference(proot)
 		proot = root
 	}
 	nodes, imgs := database.TrieDB().Size()