Lower contract creation using the new operator using zkevm intrinsics

This commit enables the generation of $zk_create[2],
$zk_data{size|offset|copy} zkevm intrinsics when lowering the new
operator

libevmasm/Assembly.cpp

@@ -579,12 +579,14 @@ LinkerObject const& Assembly::assemble() const
 			dataRef.insert(make_pair(h256(i.data()), ret.bytecode.size()));
 			ret.bytecode.resize(ret.bytecode.size() + bytesPerDataRef);
 			break;
+		case ZKEVMPushSub:
 		case PushSub:
 			assertThrow(i.data() <= numeric_limits<size_t>::max(), AssemblyException, "");
 			ret.bytecode.push_back(dataRefPush);
 			subRef.insert(make_pair(static_cast<size_t>(i.data()), ret.bytecode.size()));
 			ret.bytecode.resize(ret.bytecode.size() + bytesPerDataRef);
 			break;
+		case ZKEVMPushSubSize:
 		case PushSubSize:
 		{
 			assertThrow(i.data() <= numeric_limits<size_t>::max(), AssemblyException, "");

libevmasm/Assembly.h

@@ -91,6 +91,15 @@ class Assembly
 	/// Adds a subroutine to the code (in the data section) and pushes its size (via a tag)
 	/// on the stack. @returns the pushsub assembly item.
 	AssemblyItem appendSubroutine(AssemblyPointer const& _assembly) { auto sub = newSub(_assembly); append(newPushSubSize(size_t(sub.data()))); return sub; }
+	/// zkevm version of appendSubroutine():
+	AssemblyItem zkevmAppendSubroutine(AssemblyPointer const& _assembly)
+	{
+		m_subs.push_back(_assembly);
+		auto sub = AssemblyItem(ZKEVMPushSub, m_subs.size() - 1);
+		append(AssemblyItem(ZKEVMPushSubSize, size_t(sub.data())));
+		return sub;
+	}
+
 	void pushSubroutineSize(size_t _subRoutine) { append(newPushSubSize(_subRoutine)); }
 	/// Pushes the offset of the subroutine.
 	void pushSubroutineOffset(size_t _subRoutine) { append(AssemblyItem(PushSub, _subRoutine)); }

libevmasm/AssemblyItem.cpp

@@ -86,6 +86,10 @@ pair<string, string> AssemblyItem::nameAndData() const
 		return {"PUSH [$]", toString(util::h256(data()))};
 	case PushSubSize:
 		return {"PUSH #[$]", toString(util::h256(data()))};
+	case ZKEVMPushSub:
+		return {"$ZK_PUSH [$]", toString(util::h256(data()))};
+	case ZKEVMPushSubSize:
+		return {"$ZK_PUSH #[$]", toString(util::h256(data()))};
 	case PushProgramSize:
 		return {"PUSHSIZE", ""};
 	case PushLibraryAddress:
@@ -126,11 +130,13 @@ size_t AssemblyItem::bytesRequired(size_t _addressLength, Precision _precision)
 	case Push:
 		return 1 + max<size_t>(1, numberEncodingSize(data()));
 	case PushSubSize:
+	case ZKEVMPushSubSize:
 	case PushProgramSize:
 		return 1 + 4;		// worst case: a 16MB program
 	case PushTag:
 	case PushData:
 	case PushSub:
+	case ZKEVMPushSub:
 		return 1 + _addressLength;
 	case PushLibraryAddress:
 	case PushDeployTimeAddress:
@@ -188,6 +194,8 @@ size_t AssemblyItem::returnValues() const
 	case PushData:
 	case PushSub:
 	case PushSubSize:
+	case ZKEVMPushSub:
+	case ZKEVMPushSubSize:
 	case PushProgramSize:
 	case PushLibraryAddress:
 	case PushImmutable:
@@ -216,6 +224,8 @@ bool AssemblyItem::canBeFunctional() const
 	case PushData:
 	case PushSub:
 	case PushSubSize:
+	case ZKEVMPushSub:
+	case ZKEVMPushSubSize:
 	case PushProgramSize:
 	case PushLibraryAddress:
 	case PushDeployTimeAddress:
@@ -277,6 +287,8 @@ string AssemblyItem::toAssemblyText(Assembly const& _assembly) const
 		break;
 	case PushSub:
 	case PushSubSize:
+	case ZKEVMPushSub:
+	case ZKEVMPushSubSize:
 	{
 		vector<string> subPathComponents;
 		for (size_t subPathComponentId: _assembly.decodeSubPath(static_cast<size_t>(data())))
@@ -353,9 +365,15 @@ ostream& solidity::evmasm::operator<<(ostream& _out, AssemblyItem const& _item)
 	case PushSub:
 		_out << " PushSub " << hex << static_cast<size_t>(_item.data()) << dec;
 		break;
+	case ZKEVMPushSub:
+		_out << " ZKEVMPushSub " << hex << static_cast<size_t>(_item.data()) << dec;
+		break;
 	case PushSubSize:
 		_out << " PushSubSize " << hex << static_cast<size_t>(_item.data()) << dec;
 		break;
+	case ZKEVMPushSubSize:
+		_out << " ZKEVMPushSubSize " << hex << static_cast<size_t>(_item.data()) << dec;
+		break;
 	case PushProgramSize:
 		_out << " PushProgramSize";
 		break;

libevmasm/AssemblyItem.h

@@ -43,6 +43,8 @@ enum AssemblyItemType
 	PushTag,
 	PushSub,
 	PushSubSize,
+	ZKEVMPushSub,
+	ZKEVMPushSubSize,
 	PushProgramSize,
 	Tag,
 	PushData,

libevmasm/GasMeter.cpp

@@ -49,6 +49,8 @@ GasMeter::GasConsumption GasMeter::estimateMax(AssemblyItem const& _item, bool _
 	case PushData:
 	case PushSub:
 	case PushSubSize:
+	case ZKEVMPushSub:
+	case ZKEVMPushSubSize:
 	case PushProgramSize:
 	case PushLibraryAddress:
 	case PushDeployTimeAddress:

libevmasm/Instruction.cpp

@@ -162,6 +162,9 @@ std::map<std::string, Instruction> const solidity::evmasm::c_instructions =
 	{ "LOG2", Instruction::LOG2 },
 	{ "LOG3", Instruction::LOG3 },
 	{ "LOG4", Instruction::LOG4 },
+	{ "$ZK_CREATE", Instruction::ZK_CREATE },
+	{ "$ZK_CREATE2", Instruction::ZK_CREATE2 },
+	{ "$ZK_DATACOPY", Instruction::ZK_DATACOPY },
 	{ "CREATE", Instruction::CREATE },
 	{ "CALL", Instruction::CALL },
 	{ "CALLCODE", Instruction::CALLCODE },
@@ -309,6 +312,9 @@ static std::map<Instruction, InstructionInfo> const c_instructionInfo =
 	{ Instruction::LOG2,		{ "LOG2",			0, 4, 0, true, Tier::Special } },
 	{ Instruction::LOG3,		{ "LOG3",			0, 5, 0, true, Tier::Special } },
 	{ Instruction::LOG4,		{ "LOG4",			0, 6, 0, true, Tier::Special } },
+	{ Instruction::ZK_CREATE,	{"$ZK_CREATE",		0, 3, 1, true, Tier::Special } },
+	{ Instruction::ZK_CREATE2,	{"$ZK_CREATE2",		0, 4, 1, true, Tier::Special } },
+	{ Instruction::ZK_DATACOPY,	{"$ZK_DATACOPY",	0, 3, 0, true, Tier::Special } },
 	{ Instruction::CREATE,		{ "CREATE",			0, 3, 1, true, Tier::Special } },
 	{ Instruction::CALL,		{ "CALL",			0, 7, 1, true, Tier::Special } },
 	{ Instruction::CALLCODE,	{ "CALLCODE",		0, 7, 1, true, Tier::Special } },

libevmasm/Instruction.h

@@ -175,6 +175,10 @@ enum class Instruction: uint8_t
 	LOG3,				///< Makes a log entry; 3 topics.
 	LOG4,				///< Makes a log entry; 4 topics.
 
+	ZK_CREATE = 0xb0,	///< ZKEVM version of CREATE
+	ZK_CREATE2,			///< ZKEVM version of CREATE2
+	ZK_DATACOPY,		///< ZKEVM version of yul's datacopy instruction
+
 	CREATE = 0xf0,		///< create a new account with associated code
 	CALL,				///< message-call into an account
 	CALLCODE,			///< message-call with another account's code only

libsolidity/codegen/CompilerContext.h

@@ -225,6 +225,11 @@ class CompilerContext
 	/// Adds a subroutine to the code (in the data section) and pushes its size (via a tag)
 	/// on the stack. @returns the pushsub assembly item.
 	evmasm::AssemblyItem addSubroutine(evmasm::AssemblyPointer const& _assembly) { return m_asm->appendSubroutine(_assembly); }
+	/// zkevm version of addSubroutine():
+	evmasm::AssemblyItem zkevmAddSubroutine(evmasm::AssemblyPointer const& _assembly)
+	{
+		return m_asm->zkevmAppendSubroutine(_assembly);
+	}
 	/// Pushes the size of the subroutine.
 	void pushSubroutineSize(size_t _subRoutine) { m_asm->pushSubroutineSize(_subRoutine); }
 	/// Pushes the offset of the subroutine.

libsolidity/codegen/CompilerUtils.cpp

@@ -1507,24 +1507,26 @@ void CompilerUtils::computeHashStatic()
 	m_context << u256(32) << u256(0) << Instruction::KECCAK256;
 }
 
-void CompilerUtils::copyContractCodeToMemory(ContractDefinition const& contract, bool _creation)
+void CompilerUtils::copyContractCodeToMemory(ContractDefinition const& contract, bool _creation, bool _zkevm)
 {
 	string which = _creation ? "Creation" : "Runtime";
 	m_context.callLowLevelFunction(
 		"$copyContract" + which + "CodeToMemory_" + contract.type()->identifier(),
 		1,
 		1,
-		[&contract, _creation](CompilerContext& _context)
+		[&contract, _creation, _zkevm](CompilerContext& _context)
 		{
 			// copy the contract's code into memory
 			shared_ptr<evmasm::Assembly> assembly =
 				_creation ?
 				_context.compiledContract(contract) :
 				_context.compiledContractRuntime(contract);
 			// pushes size
-			auto subroutine = _context.addSubroutine(assembly);
-			_context << Instruction::DUP1 << subroutine;
-			_context << Instruction::DUP4 << Instruction::CODECOPY;
+			if (_zkevm)
+				_context << Instruction::DUP1 << _context.zkevmAddSubroutine(assembly);
+			else
+				_context << Instruction::DUP1 << _context.addSubroutine(assembly);
+			_context << Instruction::DUP4 << (_zkevm ? Instruction::ZK_DATACOPY : Instruction::CODECOPY);
 			_context << Instruction::ADD;
 		}
 	);

libsolidity/codegen/CompilerUtils.h

@@ -300,8 +300,9 @@ class CompilerUtils
 	/// Stack pre: Memory position
 	/// Stack post: Updated memory position
 	/// @param creation if true, copies creation code, if false copies runtime code.
+	/// @param _zkevem enables zkevm specific lowering
 	/// @note the contract has to be compiled already, so beware of cyclic dependencies!
-	void copyContractCodeToMemory(ContractDefinition const& contract, bool _creationCode);
+	void copyContractCodeToMemory(ContractDefinition const& contract, bool _creationCode, bool _zkevm = false);
 
 	/// Bytes we need to the start of call data.
 	///  - The size in bytes of the function (hash) identifier.

libsolidity/codegen/ExpressionCompiler.cpp

@@ -708,7 +708,7 @@ bool ExpressionCompiler::visit(FunctionCall const& _functionCall)
 			ContractDefinition const* contract =
 				&dynamic_cast<ContractType const&>(*function.returnParameterTypes().front()).contractDefinition();
 			utils().fetchFreeMemoryPointer();
-			utils().copyContractCodeToMemory(*contract, true);
+			utils().copyContractCodeToMemory(*contract, true, /*zkevm=*/true);
 			utils().abiEncode(argumentTypes, function.parameterTypes());
 			// now on stack: [salt], [value], memory_end_ptr
 			// need: [salt], size, offset, value
@@ -730,9 +730,9 @@ bool ExpressionCompiler::visit(FunctionCall const& _functionCall)
 
 			// now: [salt], [value], [salt], size, offset, value
 			if (function.saltSet())
-				m_context << Instruction::CREATE2;
+				m_context << Instruction::ZK_CREATE2;
 			else
-				m_context << Instruction::CREATE;
+				m_context << Instruction::ZK_CREATE;
 
 			// now: [salt], [value], address
 

libsolidity/codegen/ir/IRGeneratorForStatements.cpp

@@ -1483,14 +1483,14 @@ void IRGeneratorForStatements::endVisit(FunctionCall const& _functionCall)
 		m_context.subObjectsCreated().insert(contract);
 
 		Whiskers t(R"(let <memPos> := <allocateUnbounded>()
-			let <memEnd> := add(<memPos>, datasize("<object>"))
+			let <memEnd> := add(<memPos>, $zk_datasize("<object>"))
 			if or(gt(<memEnd>, 0xffffffffffffffff), lt(<memEnd>, <memPos>)) { <panic>() }
-			datacopy(<memPos>, dataoffset("<object>"), datasize("<object>"))
+			$zk_datacopy(<memPos>, $zk_dataoffset("<object>"), $zk_datasize("<object>"))
 			<memEnd> := <abiEncode>(<memEnd><constructorParams>)
 			<?saltSet>
-				let <address> := create2(<value>, <memPos>, sub(<memEnd>, <memPos>), <salt>)
+				let <address> := $zk_create2(<value>, <memPos>, sub(<memEnd>, <memPos>), <salt>)
 			<!saltSet>
-				let <address> := create(<value>, <memPos>, sub(<memEnd>, <memPos>))
+				let <address> := $zk_create(<value>, <memPos>, sub(<memEnd>, <memPos>))
 			</saltSet>
 			<?isTryCall>
 				let <success> := iszero(iszero(<address>))

libyul/backends/evm/AbstractAssembly.h

@@ -23,6 +23,7 @@
 
 #pragma once
 
+#include "liblangutil/Exceptions.h"
 #include <libyul/ASTForward.h>
 
 #include <libsolutil/Common.h>
@@ -116,6 +117,20 @@ class AbstractAssembly
 
 	/// Mark this assembly as invalid. Any attempt to request bytecode from it should throw.
 	virtual void markAsInvalid() = 0;
+
+	/// zkevm version of appendDataOffset()
+	virtual void appendZKEVMDataOffset(std::vector<SubID> const& _subPath)
+	{
+		(void) _subPath;
+		solUnimplemented("ZKEVM dataoffset lowering not implemented");
+	}
+
+	/// zkevm version of appendDataSize()
+	virtual void appendZKEVMDataSize(std::vector<SubID> const& _subPath)
+	{
+		(void) _subPath;
+		solUnimplemented("ZKEVM datasize lowering not implemented");
+	}
 };
 
 enum class IdentifierContext { LValue, RValue, VariableDeclaration, NonExternal };

libyul/backends/evm/EVMDialect.cpp

@@ -249,6 +249,49 @@ map<YulString, BuiltinFunctionForEVM> createBuiltins(langutil::EVMVersion _evmVe
 			})
 		);
 
+		builtins.emplace(createFunction("$zk_datasize", 1, 1, SideEffects{}, {LiteralKind::String}, [](
+			FunctionCall const& _call,
+			AbstractAssembly& _assembly,
+			BuiltinContext& _context
+		) {
+			yulAssert(_context.currentObject, "No object available.");
+			yulAssert(_call.arguments.size() == 1, "");
+			Expression const& arg = _call.arguments.front();
+			YulString dataName = std::get<Literal>(arg).value;
+			if (_context.currentObject->name == dataName)
+				_assembly.appendAssemblySize();
+			else
+			{
+				vector<size_t> subIdPath =
+					_context.subIDs.count(dataName) == 0 ?
+						_context.currentObject->pathToSubObject(dataName) :
+						vector<size_t>{_context.subIDs.at(dataName)};
+				yulAssert(!subIdPath.empty(), "Could not find assembly object <" + dataName.str() + ">.");
+				_assembly.appendZKEVMDataSize(subIdPath);
+			}
+		}));
+		builtins.emplace(createFunction("$zk_dataoffset", 1, 1, SideEffects{}, {LiteralKind::String}, [](
+			FunctionCall const& _call,
+			AbstractAssembly& _assembly,
+			BuiltinContext& _context
+		) {
+			yulAssert(_context.currentObject, "No object available.");
+			yulAssert(_call.arguments.size() == 1, "");
+			Expression const& arg = _call.arguments.front();
+			YulString dataName = std::get<Literal>(arg).value;
+			if (_context.currentObject->name == dataName)
+				_assembly.appendConstant(0);
+			else
+			{
+				vector<size_t> subIdPath =
+					_context.subIDs.count(dataName) == 0 ?
+						_context.currentObject->pathToSubObject(dataName) :
+						vector<size_t>{_context.subIDs.at(dataName)};
+				yulAssert(!subIdPath.empty(), "Could not find assembly object <" + dataName.str() + ">.");
+				_assembly.appendZKEVMDataOffset(subIdPath);
+			}
+		}));
+
 		builtins.emplace(createFunction("datasize", 1, 1, SideEffects{}, {LiteralKind::String}, [](
 			FunctionCall const& _call,
 			AbstractAssembly& _assembly,

libyul/backends/evm/EthAssemblyAdapter.cpp

@@ -141,6 +141,18 @@ void EthAssemblyAdapter::appendDataOffset(vector<AbstractAssembly::SubID> const&
 	m_assembly.pushSubroutineOffset(m_assembly.encodeSubPath(_subPath));
 }
 
+void EthAssemblyAdapter::appendZKEVMDataOffset(vector<AbstractAssembly::SubID> const& _subPath)
+{
+	if (auto it = m_dataHashBySubId.find(_subPath[0]); it != m_dataHashBySubId.end())
+	{
+		yulAssert(_subPath.size() == 1, "");
+		m_assembly << evmasm::AssemblyItem(evmasm::PushData, it->second);
+		return;
+	}
+
+	m_assembly.append(evmasm::AssemblyItem(evmasm::ZKEVMPushSub, m_assembly.encodeSubPath(_subPath)));
+}
+
 void EthAssemblyAdapter::appendDataSize(vector<AbstractAssembly::SubID> const& _subPath)
 {
 	if (auto it = m_dataHashBySubId.find(_subPath[0]); it != m_dataHashBySubId.end())
@@ -153,6 +165,18 @@ void EthAssemblyAdapter::appendDataSize(vector<AbstractAssembly::SubID> const& _
 	m_assembly.pushSubroutineSize(m_assembly.encodeSubPath(_subPath));
 }
 
+void EthAssemblyAdapter::appendZKEVMDataSize(vector<AbstractAssembly::SubID> const& _subPath)
+{
+	if (auto it = m_dataHashBySubId.find(_subPath[0]); it != m_dataHashBySubId.end())
+	{
+		yulAssert(_subPath.size() == 1, "");
+		m_assembly << u256(m_assembly.data(h256(it->second)).size());
+		return;
+	}
+
+	m_assembly.append(evmasm::AssemblyItem(evmasm::ZKEVMPushSubSize, m_assembly.encodeSubPath(_subPath)));
+}
+
 AbstractAssembly::SubID EthAssemblyAdapter::appendData(bytes const& _data)
 {
 	evmasm::AssemblyItem pushData = m_assembly.newData(_data);

libyul/backends/evm/EthAssemblyAdapter.h

@@ -58,6 +58,8 @@ class EthAssemblyAdapter: public AbstractAssembly
 	std::pair<std::shared_ptr<AbstractAssembly>, SubID> createSubAssembly(bool _creation, std::string _name = {}) override;
 	void appendDataOffset(std::vector<SubID> const& _subPath) override;
 	void appendDataSize(std::vector<SubID> const& _subPath) override;
+	void appendZKEVMDataOffset(std::vector<SubID> const& _subPath) override;
+	void appendZKEVMDataSize(std::vector<SubID> const& _subPath) override;
 	SubID appendData(bytes const& _data) override;
 
 	void appendToAuxiliaryData(bytes const& _data) override;