Merge pull request #12074 from ethereum/output-selection-in-assembler-mode

CLI output selection in assembler mode

Changelog.md

@@ -8,6 +8,7 @@ Compiler Features:
  * Code Generator: Skip existence check for external contract if return data is expected. In this case, the ABI decoder will revert if the contract does not exist.
  * Commandline Interface: Accept nested brackets in step sequences passed to ``--yul-optimizations``.
  * Commandline Interface: Add ``--debug-info`` option for selecting how much extra debug information should be included in the produced EVM assembly and Yul code.
+ * Commandline Interface: Support ``--asm``, ``--bin``, ``--ir-optimized``, ``--ewasm`` and ``--ewasm-ir`` output selection options in assembler mode.
  * Commandline Interface: Use different colors when printing errors, warnings and infos.
  * SMTChecker: Output values for ``block.*``, ``msg.*`` and ``tx.*`` variables that are present in the called functions.
  * SMTChecker: Report contract invariants and reentrancy properties. This can be enabled via the CLI option ``--model-checker-invariants`` or the Standard JSON option ``settings.modelChecker.invariants``.

solc/CommandLineInterface.cpp

@@ -1042,34 +1042,52 @@ bool CommandLineInterface::assemble(yul::AssemblyStack::Language _language, yul:
 
 		yul::AssemblyStack& stack = assemblyStacks[src.first];
 
-		sout() << endl << "Pretty printed source:" << endl;
+		if (m_options.compiler.outputs.irOptimized)
-		sout() << stack.print() << endl;
+		{
+			// NOTE: This actually outputs unoptimized code when the optimizer is disabled but
+			// 'ir' output in StandardCompiler works the same way.
+			sout() << endl << "Pretty printed source:" << endl;
+			sout() << stack.print() << endl;
+		}
 
 		if (_language != yul::AssemblyStack::Language::Ewasm && _targetMachine == yul::AssemblyStack::Machine::Ewasm)
 		{
 			stack.translate(yul::AssemblyStack::Language::Ewasm);
 			stack.optimize();
 
-			sout() << endl << "==========================" << endl;
+			if (m_options.compiler.outputs.ewasmIR)
-			sout() << endl << "Translated source:" << endl;
+			{
-			sout() << stack.print() << endl;
+				sout() << endl << "==========================" << endl;
+				sout() << endl << "Translated source:" << endl;
+				sout() << stack.print() << endl;
+			}
 		}
 
 		yul::MachineAssemblyObject object;
 		object = stack.assemble(_targetMachine);
 		object.bytecode->link(m_options.linker.libraries);
 
-		sout() << endl << "Binary representation:" << endl;
+		if (m_options.compiler.outputs.binary)
-		if (object.bytecode)
+		{
-			sout() << object.bytecode->toHex() << endl;
+			sout() << endl << "Binary representation:" << endl;
-		else
+			if (object.bytecode)
-			serr() << "No binary representation found." << endl;
+				sout() << object.bytecode->toHex() << endl;
+			else
+				serr() << "No binary representation found." << endl;
+		}
 
-		sout() << endl << "Text representation:" << endl;
+		solAssert(_targetMachine == yul::AssemblyStack::Machine::Ewasm || _targetMachine == yul::AssemblyStack::Machine::EVM, "");
-		if (!object.assembly.empty())
+		if (
-			sout() << object.assembly << endl;
+			(_targetMachine == yul::AssemblyStack::Machine::EVM && m_options.compiler.outputs.asm_) ||
-		else
+			(_targetMachine == yul::AssemblyStack::Machine::Ewasm && m_options.compiler.outputs.ewasm)
-			serr() << "No text representation found." << endl;
+		)
+		{
+			sout() << endl << "Text representation:" << endl;
+			if (!object.assembly.empty())
+				sout() << object.assembly << endl;
+			else
+				serr() << "No text representation found." << endl;
+		}
 	}
 
 	return true;

solc/CommandLineParser.cpp

@@ -446,10 +446,18 @@ bool CommandLineParser::parseOutputSelection()
 {
 	static auto outputSupported = [](InputMode _mode, string_view _outputName)
 	{
-		static set<string> const compilerModeOutputs =
+		static set<string> const compilerModeOutputs = (
 			CompilerOutputs::componentMap() |
 			ranges::views::keys |
-			ranges::to<set>();
+			ranges::to<set>()
+		) - set<string>{CompilerOutputs::componentName(&CompilerOutputs::ewasmIR)};
+		static set<string> const assemblerModeOutputs = {
+			CompilerOutputs::componentName(&CompilerOutputs::asm_),
+			CompilerOutputs::componentName(&CompilerOutputs::binary),
+			CompilerOutputs::componentName(&CompilerOutputs::irOptimized),
+			CompilerOutputs::componentName(&CompilerOutputs::ewasm),
+			CompilerOutputs::componentName(&CompilerOutputs::ewasmIR),
+		};
 
 		switch (_mode)
 		{
@@ -461,6 +469,7 @@ bool CommandLineParser::parseOutputSelection()
 		case InputMode::CompilerWithASTImport:
 			return contains(compilerModeOutputs, _outputName);
 		case InputMode::Assembler:
+			return contains(assemblerModeOutputs, _outputName);
 		case InputMode::StandardJson:
 		case InputMode::Linker:
 			return false;
@@ -472,6 +481,17 @@ bool CommandLineParser::parseOutputSelection()
 	for (auto&& [optionName, outputComponent]: CompilerOutputs::componentMap())
 		m_options.compiler.outputs.*outputComponent = (m_args.count(optionName) > 0);
 
+	if (m_options.input.mode == InputMode::Assembler && m_options.compiler.outputs == CompilerOutputs{})
+	{
+		// In assembly mode keep the default outputs enabled for backwards-compatibility.
+		// TODO: Remove this (must be done in a breaking release).
+		m_options.compiler.outputs.asm_ = true;
+		m_options.compiler.outputs.binary = true;
+		m_options.compiler.outputs.irOptimized = true;
+		m_options.compiler.outputs.ewasm = true;
+		m_options.compiler.outputs.ewasmIR = true;
+	}
+
 	vector<string> unsupportedOutputs;
 	for (auto&& [optionName, outputComponent]: CompilerOutputs::componentMap())
 		if (m_options.compiler.outputs.*outputComponent && !outputSupported(m_options.input.mode, optionName))
@@ -692,6 +712,7 @@ General Information)").c_str(),
 		(CompilerOutputs::componentName(&CompilerOutputs::ir).c_str(), "Intermediate Representation (IR) of all contracts (EXPERIMENTAL).")
 		(CompilerOutputs::componentName(&CompilerOutputs::irOptimized).c_str(), "Optimized intermediate Representation (IR) of all contracts (EXPERIMENTAL).")
 		(CompilerOutputs::componentName(&CompilerOutputs::ewasm).c_str(), "Ewasm text representation of all contracts (EXPERIMENTAL).")
+		(CompilerOutputs::componentName(&CompilerOutputs::ewasmIR).c_str(), "Intermediate representation (IR) converted to a form that can be translated directly into Ewasm text representation (EXPERIMENTAL).")
 		(CompilerOutputs::componentName(&CompilerOutputs::signatureHashes).c_str(), "Function signature hashes of the contracts.")
 		(CompilerOutputs::componentName(&CompilerOutputs::natspecUser).c_str(), "Natspec user documentation of all contracts.")
 		(CompilerOutputs::componentName(&CompilerOutputs::natspecDev).c_str(), "Natspec developer documentation of all contracts.")
@@ -906,11 +927,12 @@ bool CommandLineParser::processArgs()
 	if (!checkMutuallyExclusive({g_strColor, g_strNoColor}))
 		return false;
 
-	array<string, 8> const conflictingWithStopAfter{
+	array<string, 9> const conflictingWithStopAfter{
 		CompilerOutputs::componentName(&CompilerOutputs::binary),
 		CompilerOutputs::componentName(&CompilerOutputs::ir),
 		CompilerOutputs::componentName(&CompilerOutputs::irOptimized),
 		CompilerOutputs::componentName(&CompilerOutputs::ewasm),
+		CompilerOutputs::componentName(&CompilerOutputs::ewasmIR),
 		g_strGas,
 		CompilerOutputs::componentName(&CompilerOutputs::asm_),
 		CompilerOutputs::componentName(&CompilerOutputs::asmJson),

solc/CommandLineParser.h

@@ -78,6 +78,7 @@ struct CompilerOutputs
 			{"ir", &CompilerOutputs::ir},
 			{"ir-optimized", &CompilerOutputs::irOptimized},
 			{"ewasm", &CompilerOutputs::ewasm},
+			{"ewasm-ir", &CompilerOutputs::ewasmIR},
 			{"hashes", &CompilerOutputs::signatureHashes},
 			{"userdoc", &CompilerOutputs::natspecUser},
 			{"devdoc", &CompilerOutputs::natspecDev},
@@ -97,6 +98,7 @@ struct CompilerOutputs
 	bool ir = false;
 	bool irOptimized = false;
 	bool ewasm = false;
+	bool ewasmIR = false;
 	bool signatureHashes = false;
 	bool natspecUser = false;
 	bool natspecDev = false;

test/cmdlineTests/evm_to_wasm_output_selection_asm_only/args

@@ -0,0 +1 @@
+--assemble --optimize --yul-dialect evm --machine ewasm --asm

test/cmdlineTests/evm_to_wasm_output_selection_asm_only/err

@@ -0,0 +1 @@
+Warning: Yul is still experimental. Please use the output with care.

test/cmdlineTests/evm_to_wasm_output_selection_asm_only/input.yul

@@ -0,0 +1,4 @@
+{
+    let x := 42
+    sstore(0, x)
+}

test/cmdlineTests/evm_to_wasm_output_selection_asm_only/output

@@ -0,0 +1,2 @@
+
+======= evm_to_wasm_output_selection_asm_only/input.yul (Ewasm) =======

test/cmdlineTests/evm_to_wasm_output_selection_ewasm_ir_only/args

@@ -0,0 +1 @@
+--assemble --optimize --yul-dialect evm --machine ewasm --ewasm-ir

test/cmdlineTests/evm_to_wasm_output_selection_ewasm_ir_only/err

@@ -0,0 +1 @@
+Warning: Yul is still experimental. Please use the output with care.

test/cmdlineTests/evm_to_wasm_output_selection_ewasm_ir_only/input.yul

@@ -0,0 +1,4 @@
+{
+    let x := 42
+    sstore(0, x)
+}

test/cmdlineTests/evm_to_wasm_output_selection_ewasm_ir_only/output

@@ -0,0 +1,34 @@
+
+======= evm_to_wasm_output_selection_ewasm_ir_only/input.yul (Ewasm) =======
+
+==========================
+
+Translated source:
+object "object" {
+    code {
+        function main()
+        {
+            let hi := i64.shl(i64.extend_i32_u(bswap32(i32.wrap_i64(0))), 32)
+            let y := i64.or(hi, i64.extend_i32_u(bswap32(i32.wrap_i64(i64.shr_u(0, 32)))))
+            i64.store(0:i32, y)
+            i64.store(i32.add(0:i32, 8:i32), y)
+            i64.store(i32.add(0:i32, 16:i32), y)
+            i64.store(i32.add(0:i32, 24:i32), y)
+            i64.store(32:i32, y)
+            i64.store(i32.add(32:i32, 8:i32), y)
+            i64.store(i32.add(32:i32, 16:i32), y)
+            let hi_1 := i64.shl(i64.extend_i32_u(bswap32(i32.wrap_i64(42))), 32)
+            i64.store(i32.add(32:i32, 24:i32), i64.or(hi_1, i64.extend_i32_u(bswap32(i32.wrap_i64(i64.shr_u(42, 32))))))
+            eth.storageStore(0:i32, 32:i32)
+        }
+        function bswap16(x:i32) -> y:i32
+        {
+            y := i32.or(i32.and(i32.shl(x, 8:i32), 0xff00:i32), i32.and(i32.shr_u(x, 8:i32), 0xff:i32))
+        }
+        function bswap32(x:i32) -> y:i32
+        {
+            let hi:i32 := i32.shl(bswap16(x), 16:i32)
+            y := i32.or(hi, bswap16(i32.shr_u(x, 16:i32)))
+        }
+    }
+}

test/cmdlineTests/evm_to_wasm_output_selection_ewasm_only/args

@@ -0,0 +1 @@
+--assemble --optimize --yul-dialect evm --machine ewasm --ewasm

test/cmdlineTests/evm_to_wasm_output_selection_ewasm_only/err

@@ -0,0 +1 @@
+Warning: Yul is still experimental. Please use the output with care.

test/cmdlineTests/evm_to_wasm_output_selection_ewasm_only/input.yul

@@ -0,0 +1,4 @@
+{
+    let x := 42
+    sstore(0, x)
+}

test/cmdlineTests/evm_to_wasm_output_selection_ewasm_only/output

@@ -0,0 +1,54 @@
+
+======= evm_to_wasm_output_selection_ewasm_only/input.yul (Ewasm) =======
+
+Text representation:
+(module
+    (import "ethereum" "storageStore" (func $eth.storageStore (param i32 i32)))
+    (memory $memory (export "memory") 1)
+    (export "main" (func $main))
+
+(func $main
+    (local $hi i64)
+    (local $y i64)
+    (local $hi_1 i64)
+    (block $label_
+        (local.set $hi (i64.shl (i64.extend_i32_u (call $bswap32 (i32.wrap_i64 (i64.const 0)))) (i64.const 32)))
+        (local.set $y (i64.or (local.get $hi) (i64.extend_i32_u (call $bswap32 (i32.wrap_i64 (i64.shr_u (i64.const 0) (i64.const 32)))))))
+        (i64.store (i32.const 0) (local.get $y))
+        (i64.store (i32.add (i32.const 0) (i32.const 8)) (local.get $y))
+        (i64.store (i32.add (i32.const 0) (i32.const 16)) (local.get $y))
+        (i64.store (i32.add (i32.const 0) (i32.const 24)) (local.get $y))
+        (i64.store (i32.const 32) (local.get $y))
+        (i64.store (i32.add (i32.const 32) (i32.const 8)) (local.get $y))
+        (i64.store (i32.add (i32.const 32) (i32.const 16)) (local.get $y))
+        (local.set $hi_1 (i64.shl (i64.extend_i32_u (call $bswap32 (i32.wrap_i64 (i64.const 42)))) (i64.const 32)))
+        (i64.store (i32.add (i32.const 32) (i32.const 24)) (i64.or (local.get $hi_1) (i64.extend_i32_u (call $bswap32 (i32.wrap_i64 (i64.shr_u (i64.const 42) (i64.const 32)))))))
+        (call $eth.storageStore (i32.const 0) (i32.const 32))
+    )
+)
+
+(func $bswap16
+    (param $x i32)
+    (result i32)
+    (local $y i32)
+    (block $label__1
+        (local.set $y (i32.or (i32.and (i32.shl (local.get $x) (i32.const 8)) (i32.const 65280)) (i32.and (i32.shr_u (local.get $x) (i32.const 8)) (i32.const 255))))
+
+    )
+    (local.get $y)
+)
+
+(func $bswap32
+    (param $x i32)
+    (result i32)
+    (local $y i32)
+    (local $hi i32)
+    (block $label__2
+        (local.set $hi (i32.shl (call $bswap16 (local.get $x)) (i32.const 16)))
+        (local.set $y (i32.or (local.get $hi) (call $bswap16 (i32.shr_u (local.get $x) (i32.const 16)))))
+
+    )
+    (local.get $y)
+)
+
+)

test/cmdlineTests/output_selection_ewasm_ir_only/args

@@ -0,0 +1 @@
+--optimize --ewasm-ir

test/cmdlineTests/output_selection_ewasm_ir_only/err

@@ -0,0 +1 @@
+The following outputs are not supported in compiler mode: --ewasm-ir.

test/cmdlineTests/output_selection_ewasm_ir_only/exit

@@ -0,0 +1 @@
+1

test/cmdlineTests/output_selection_ewasm_ir_only/input.sol

@@ -0,0 +1,4 @@
+// SPDX-License-Identifier: GPL-2.0
+pragma solidity *;
+
+contract C {}

test/cmdlineTests/standard_cli_output_selection_invalid/args

@@ -1 +1 @@
---ast-compact-json --asm --asm-json --opcodes --bin --bin-runtime --abi --ir --ir-optimized --ewasm --hashes --userdoc --devdoc --metadata --storage-layout
+--ast-compact-json --asm --asm-json --opcodes --bin --bin-runtime --abi --ir --ir-optimized --ewasm --ewasm-ir --hashes --userdoc --devdoc --metadata --storage-layout

test/cmdlineTests/standard_cli_output_selection_invalid/err

@@ -1 +1 @@
-The following outputs are not supported in standard JSON mode: --abi, --asm, --asm-json, --ast-compact-json, --bin, --bin-runtime, --devdoc, --ewasm, --hashes, --ir, --ir-optimized, --metadata, --opcodes, --storage-layout, --userdoc.
+The following outputs are not supported in standard JSON mode: --abi, --asm, --asm-json, --ast-compact-json, --bin, --bin-runtime, --devdoc, --ewasm, --ewasm-ir, --hashes, --ir, --ir-optimized, --metadata, --opcodes, --storage-layout, --userdoc.

test/cmdlineTests/strict_asm_output_selection_asm_only/args

@@ -0,0 +1 @@
+--strict-assembly --optimize --asm

test/cmdlineTests/strict_asm_output_selection_asm_only/err

@@ -0,0 +1 @@
+Warning: Yul is still experimental. Please use the output with care.

test/cmdlineTests/strict_asm_output_selection_asm_only/input.yul

@@ -0,0 +1,4 @@
+{
+    let x := 42
+    sstore(0, x)
+}

test/cmdlineTests/strict_asm_output_selection_asm_only/output

@@ -0,0 +1,12 @@
+
+======= strict_asm_output_selection_asm_only/input.yul (EVM) =======
+
+Text representation:
+    /* "strict_asm_output_selection_asm_only/input.yul":15:17   */
+  0x2a
+    /* "strict_asm_output_selection_asm_only/input.yul":29:30   */
+  0x00
+    /* "strict_asm_output_selection_asm_only/input.yul":22:34   */
+  sstore
+    /* "strict_asm_output_selection_asm_only/input.yul":0:36   */
+  stop

test/cmdlineTests/strict_asm_output_selection_bin_only/args

@@ -0,0 +1 @@
+--strict-assembly --optimize --bin

test/cmdlineTests/strict_asm_output_selection_bin_only/err

@@ -0,0 +1 @@
+Warning: Yul is still experimental. Please use the output with care.

test/cmdlineTests/strict_asm_output_selection_bin_only/input.yul

@@ -0,0 +1,4 @@
+{
+    let x := 42
+    sstore(0, x)
+}

test/cmdlineTests/strict_asm_output_selection_bin_only/output

@@ -0,0 +1,5 @@
+
+======= strict_asm_output_selection_bin_only/input.yul (EVM) =======
+
+Binary representation:
+602a60005500

test/cmdlineTests/strict_asm_output_selection_ewasm_ir_only/args

@@ -0,0 +1 @@
+--strict-assembly --optimize --ewasm-ir

test/cmdlineTests/strict_asm_output_selection_ewasm_ir_only/err

@@ -0,0 +1 @@
+Warning: Yul is still experimental. Please use the output with care.

test/cmdlineTests/strict_asm_output_selection_ewasm_ir_only/input.yul

@@ -0,0 +1,4 @@
+{
+    let x := 42
+    sstore(0, x)
+}

test/cmdlineTests/strict_asm_output_selection_ewasm_ir_only/output

@@ -0,0 +1,2 @@
+
+======= strict_asm_output_selection_ewasm_ir_only/input.yul (EVM) =======

test/cmdlineTests/strict_asm_output_selection_ewasm_only/args

@@ -0,0 +1 @@
+--strict-assembly --optimize --ewasm

test/cmdlineTests/strict_asm_output_selection_ewasm_only/err

@@ -0,0 +1 @@
+Warning: Yul is still experimental. Please use the output with care.

test/cmdlineTests/strict_asm_output_selection_ewasm_only/input.yul

@@ -0,0 +1,4 @@
+{
+    let x := 42
+    sstore(0, x)
+}

test/cmdlineTests/strict_asm_output_selection_ewasm_only/output

@@ -0,0 +1,2 @@
+
+======= strict_asm_output_selection_ewasm_only/input.yul (EVM) =======

test/cmdlineTests/strict_asm_output_selection_invalid/err

@@ -1 +1 @@
-The following outputs are not supported in assembler mode: --abi, --asm, --asm-json, --bin, --bin-runtime, --devdoc, --ewasm, --hashes, --ir, --ir-optimized, --metadata, --opcodes, --storage-layout, --userdoc.
+The following outputs are not supported in assembler mode: --abi, --asm-json, --bin-runtime, --devdoc, --hashes, --ir, --metadata, --opcodes, --storage-layout, --userdoc.

test/cmdlineTests/strict_asm_output_selection_ir_optimized_only/args

@@ -0,0 +1 @@
+--strict-assembly --optimize --ir-optimized

test/cmdlineTests/strict_asm_output_selection_ir_optimized_only/err

@@ -0,0 +1 @@
+Warning: Yul is still experimental. Please use the output with care.

test/cmdlineTests/strict_asm_output_selection_ir_optimized_only/input.yul

@@ -0,0 +1,4 @@
+{
+    let x := 42
+    sstore(0, x)
+}

test/cmdlineTests/strict_asm_output_selection_ir_optimized_only/output

@@ -0,0 +1,7 @@
+
+======= strict_asm_output_selection_ir_optimized_only/input.yul (EVM) =======
+
+Pretty printed source:
+object "object" {
+    code { { sstore(0, 42) } }
+}

test/solc/CommandLineParser.cpp

@@ -198,7 +198,9 @@ BOOST_AUTO_TEST_CASE(cli_mode_options)
 			true, true, true, true, true,
 			true, true, true, true, true,
 			true, true, true, true, true,
+			true,
 		};
+		expectedOptions.compiler.outputs.ewasmIR = false;
 		expectedOptions.compiler.estimateGas = true;
 		expectedOptions.compiler.combinedJsonRequests = {
 			true, true, true, true, true,
@@ -297,6 +299,11 @@ BOOST_AUTO_TEST_CASE(assembly_mode_options)
 				"underflow,"
 				"divByZero",
 			"--model-checker-timeout=5",   // Ignored in assembly mode
+			"--asm",
+			"--bin",
+			"--ir-optimized",
+			"--ewasm",
+			"--ewasm-ir",
 		};
 		commandLine += assemblyOptions;
 		if (expectedLanguage == AssemblyStack::Language::StrictAssembly || expectedLanguage == AssemblyStack::Language::Ewasm)
@@ -333,6 +340,11 @@ BOOST_AUTO_TEST_CASE(assembly_mode_options)
 		};
 		expectedOptions.formatting.coloredOutput = false;
 		expectedOptions.formatting.withErrorIds = true;
+		expectedOptions.compiler.outputs.asm_ = true;
+		expectedOptions.compiler.outputs.binary = true;
+		expectedOptions.compiler.outputs.irOptimized = true;
+		expectedOptions.compiler.outputs.ewasm = true;
+		expectedOptions.compiler.outputs.ewasmIR = true;
 		if (expectedLanguage == AssemblyStack::Language::StrictAssembly || expectedLanguage == AssemblyStack::Language::Ewasm)
 		{
 			expectedOptions.optimizer.enabled = true;