ConcurrentChain: Optimize Load method

NBitcoin.Bench/README.md

@@ -4,9 +4,9 @@ This project use BenchmarkDotnet to measure performance of parts of NBitcoin imp
 
 You can generate flamegraph to view with Perfview on Windows with the following command:
 ```powershell
-dotnet run -c Release -- --runtimes netcoreapp2.1 --filter *GolombRiceFilters* --profiler ETW
+dotnet run -c Release -- --runtimes net8.0 --filter *GolombRiceFilters* --profiler ETW
 ```
 
 Where `GolombRiceFilters` is the name of the benchmark class you are insterested in.
 
-For generating flamegraph on linux, see [this comment](https://github.com/MetacoSA/NBitcoin/pull/656#issuecomment-462927805).
+For generating flamegraph on linux, see [this comment](https://github.com/MetacoSA/NBitcoin/pull/656#issuecomment-462927805).

NBitcoin/ConcurrentChain.cs

@@ -289,12 +289,21 @@ private ChainedBlock SetTipNoLock(ChainedBlock block)
 				height--;
 			}
 			var fork = GetBlockNoLock(height);
-			foreach (var newBlock in block.EnumerateToGenesis()
-				.TakeWhile(c => c != fork))
+
+			// The while loop was manually changed from more idiomatic but slower:
+			//   foreach (var newBlock in block.EnumerateToGenesis().TakeWhile(c => c != fork))
+			var newBlock = block;
+			while (newBlock != fork)
 			{
-				_BlocksById.AddOrReplace(newBlock.HashBlock, newBlock);
+				_BlocksById[newBlock.HashBlock] = newBlock;
 				AddOrReplaceBlocksByHeight(newBlock.Height, newBlock);
+
+				newBlock = newBlock.Previous;
+
+				if (newBlock is null)
+					break;
 			}
+
 			_Tip = block;
 			return fork;
 		}