Update Wasm benchmarks (#2957)

In https://github.com/paritytech/polkadot-sdk/pull/2941 we found out
that the new Wasmi (register) is very effective at optimizing away
certain benchmark bytecode constructs in a way that created an unfair
advantage over Wasmi (stack) which yielded our former benchmarks to be
ineffective at properly measuring the performance impact.

This PR adjusts both affected benchmarks to fix the stated problems.
Affected are
- `instr_i64const` -> `instr_i64add`: Renamed since it now measures the
performance impact of the Wasm `i64.add` instruction with locals as
inputs and outputs. This makes it impossible for Wasmi (register) to
aggressively optimize away the entire function body (as it previously
did) but still provides a way for Wasmi (register) to shine with its
register based execution model.
- `call_with_code_per_byte`: Now uses `local.get` instead of `i32.const`
for the `if` condition which prevents Wasmi (register) to aggressively
optimizing away whole parts of the `if` creating an unfair advantage.

cc @athei

---------

Co-authored-by: command-bot <>
Co-authored-by: Alexander Theißen <alex.theissen@me.com>
Co-authored-by: Ignacio Palacios <ignacio.palacios.santos@gmail.com>

substrate/frame/contracts/src/benchmarking/mod.rs

@@ -48,7 +48,7 @@ use pallet_balances;
 use pallet_contracts_uapi::CallFlags;
 use sp_runtime::traits::{Bounded, Hash};
 use sp_std::prelude::*;
-use wasm_instrument::parity_wasm::elements::{BlockType, Instruction, ValueType};
+use wasm_instrument::parity_wasm::elements::{BlockType, Instruction, Local, ValueType};
 
 /// How many runs we do per API benchmark.
 ///
@@ -2582,19 +2582,45 @@ benchmarks! {
 		let origin = RawOrigin::Signed(instance.caller.clone());
 	}: call(origin, instance.addr, 0u32.into(), Weight::MAX, None, vec![])
 
-	// We make the assumption that pushing a constant and dropping a value takes roughly
-	// the same amount of time. We call this weight `w_base`.
-	// The weight that would result from the respective benchmark we call: `w_bench`.
+	// We load `i64` values from random linear memory locations and store the loaded
+	// values back into yet another random linear memory location.
+	// The random addresses are uniformely distributed across the entire span of the linear memory.
+	// We do this to enforce random memory accesses which are particularly expensive.
 	//
-	// w_base = w_i{32,64}const = w_drop = w_bench / 2
+	// The combination of this computation is our weight base `w_base`.
 	#[pov_mode = Ignored]
-	instr_i64const {
+	instr_i64_load_store {
 		let r in 0 .. INSTR_BENCHMARK_RUNS;
+
+		use rand::prelude::*;
+
+		// We do not need to be secure here. Fixed seed allows for determinstic results.
+		let mut rng = rand_pcg::Pcg32::seed_from_u64(8446744073709551615);
+
+		let memory = ImportedMemory::max::<T>();
+		let bytes_per_page = 65536;
+		let bytes_per_memory = memory.max_pages * bytes_per_page;
 		let mut sbox = Sandbox::from(&WasmModule::<T>::from(ModuleDefinition {
-			call_body: Some(body::repeated_dyn(r, vec![
-				RandomI64Repeated(1),
-				Regular(Instruction::Drop),
-			])),
+			memory: Some(memory),
+			call_body: Some(body::repeated_with_locals_using(
+				&[Local::new(1, ValueType::I64)],
+				r,
+				|| {
+					// Instruction sequence to load a `i64` from linear memory
+					// at a random memory location and store it back into another
+					// location of the linear memory.
+					let c0: i32 = rng.gen_range(0..bytes_per_memory as i32);
+					let c1: i32 = rng.gen_range(0..bytes_per_memory as i32);
+					[
+						Instruction::I32Const(c0), // address for `i64.load_8s`
+						Instruction::I64Load8S(0, 0),
+						Instruction::SetLocal(0),  // temporarily store value loaded in `i64.load_8s`
+						Instruction::I32Const(c1), // address for `i64.store8`
+						Instruction::GetLocal(0),  // value to be stores in `i64.store8`
+						Instruction::I64Store8(0, 0),
+					]
+				}
+			)),
 			.. Default::default()
 		}));
 	}: {