In previous post, WTSC 7: Semantic Analysis, we have get the Type-annotated AST for our helloworld example show as following:
(source_file "Helloworld.swift"
(top_level_code_decl range=[Helloworld.swift:1:1 - line:1:21]
(brace_stmt implicit range=[Helloworld.swift:1:1 - line:1:21]
(call_expr type='()' location=Helloworld.swift:1:1 range=[Helloworld.swift:1:1 - line:1:21] nothrow arg_labels=_:
(declref_expr type='(Any..., String, String) -> ()' location=Helloworld.swift:1:1 range=[Helloworld.swift:1:1 - line:1:1] decl=Swift.(file).print(_:separator:terminator:) function_ref=single)
(tuple_expr type='(Any..., separator: String, terminator: String)' location=Helloworld.swift:1:6 range=[Helloworld.swift:1:6 - line:1:21] names='',separator,terminator
(vararg_expansion_expr implicit type='[Any]' location=Helloworld.swift:1:7 range=[Helloworld.swift:1:7 - line:1:7]
(array_expr implicit type='[Any]' location=Helloworld.swift:1:7 range=[Helloworld.swift:1:7 - line:1:7] initializer=**NULL**
(erasure_expr implicit type='Any' location=Helloworld.swift:1:7 range=[Helloworld.swift:1:7 - line:1:7]
(string_literal_expr type='String' location=Helloworld.swift:1:7 range=[Helloworld.swift:1:7 - line:1:7] encoding=utf8 value="Hello world!" builtin_initializer=Swift.(file).String extension.init(_builtinStringLiteral:utf8CodeUnitCount:isASCII:) initializer=**NULL**))))
(default_argument_expr implicit type='String' location=Helloworld.swift:1:6 range=[Helloworld.swift:1:6 - line:1:6] default_args_owner=Swift.(file).print(_:separator:terminator:) param=1)
(default_argument_expr implicit type='String' location=Helloworld.swift:1:6 range=[Helloworld.swift:1:6 - line:1:6] default_args_owner=Swift.(file).print(_:separator:terminator:) param=2))))))
And in this post, we gonna get a raw SIL, Swift Intermediate Language, from above AST as following:
swiftc -o hello -emit-silgen hello.swift
cat hellosil_stage raw
import Builtin
import Swift
import SwiftShims
sil [ossa] @main : $@convention(c) (Int32, UnsafeMutablePointer<Optional<UnsafeMutablePointer<Int8>>>) -> Int32 {
bb0(%0 : $Int32, %1 : $UnsafeMutablePointer<Optional<UnsafeMutablePointer<Int8>>>):
%2 = integer_literal $Builtin.Word, 1 // user: %4
// function_ref _allocateUninitializedArray<A>(_:)
%3 = function_ref @$ss27_allocateUninitializedArrayySayxG_BptBwlF : $@convention(thin) <τ_0_0> (Builtin.Word) -> (@owned Array<τ_0_0>, Builtin.RawPointer) // user: %4
%4 = apply %3<Any>(%2) : $@convention(thin) <τ_0_0> (Builtin.Word) -> (@owned Array<τ_0_0>, Builtin.RawPointer) // user: %5
(%5, %6) = destructure_tuple %4 : $(Array<Any>, Builtin.RawPointer) // users: %24, %21, %7
%7 = pointer_to_address %6 : $Builtin.RawPointer to [strict] $*Any // user: %14
%8 = string_literal utf8 "hello world!" // user: %13
%9 = integer_literal $Builtin.Word, 12 // user: %13
%10 = integer_literal $Builtin.Int1, -1 // user: %13
%11 = metatype $@thin String.Type // user: %13
// function_ref String.init(_builtinStringLiteral:utf8CodeUnitCount:isASCII:)
%12 = function_ref @$sSS21_builtinStringLiteral17utf8CodeUnitCount7isASCIISSBp_BwBi1_tcfC : $@convention(method) (Builtin.RawPointer, Builtin.Word, Builtin.Int1, @thin String.Type) -> @owned String // user: %13
%13 = apply %12(%8, %9, %10, %11) : $@convention(method) (Builtin.RawPointer, Builtin.Word, Builtin.Int1, @thin String.Type) -> @owned String // user: %15
%14 = init_existential_addr %7 : $*Any, $String // user: %15
store %13 to [init] %14 : $*String // id: %15
// function_ref default argument 1 of print(_:separator:terminator:)
%16 = function_ref @$ss5print_9separator10terminatoryypd_S2StFfA0_ : $@convention(thin) () -> @owned String // user: %17
%17 = apply %16() : $@convention(thin) () -> @owned String // users: %23, %21
// function_ref default argument 2 of print(_:separator:terminator:)
%18 = function_ref @$ss5print_9separator10terminatoryypd_S2StFfA1_ : $@convention(thin) () -> @owned String // user: %19
%19 = apply %18() : $@convention(thin) () -> @owned String // users: %22, %21
// function_ref print(_:separator:terminator:)
%20 = function_ref @$ss5print_9separator10terminatoryypd_S2StF : $@convention(thin) (@guaranteed Array<Any>, @guaranteed String, @guaranteed String) -> () // user: %21
%21 = apply %20(%5, %17, %19) : $@convention(thin) (@guaranteed Array<Any>, @guaranteed String, @guaranteed String) -> ()
destroy_value %19 : $String // id: %22
destroy_value %17 : $String // id: %23
destroy_value %5 : $Array<Any> // id: %24
%25 = integer_literal $Builtin.Int32, 0 // user: %26
%26 = struct $Int32 (%25 : $Builtin.Int32) // user: %27
return %26 : $Int32 // id: %27
} // end sil function 'main'
// _allocateUninitializedArray<A>(_:)
sil [serialized] [always_inline] [_semantics "array.uninitialized_intrinsic"] @$ss27_allocateUninitializedArrayySayxG_BptBwlF : $@convention(thin) <τ_0_0> (Builtin.Word) -> (@owned Array<τ_0_0>, Builtin.RawPointer)
// String.init(_builtinStringLiteral:utf8CodeUnitCount:isASCII:)
sil [serialized] [always_inline] [readonly] [_semantics "string.makeUTF8"] @$sSS21_builtinStringLiteral17utf8CodeUnitCount7isASCIISSBp_BwBi1_tcfC : $@convention(method) (Builtin.RawPointer, Builtin.Word, Builtin.Int1, @thin String.Type) -> @owned String
// default argument 1 of print(_:separator:terminator:)
sil hidden_external [serialized] @$ss5print_9separator10terminatoryypd_S2StFfA0_ : $@convention(thin) () -> @owned String
// default argument 2 of print(_:separator:terminator:)
sil hidden_external [serialized] @$ss5print_9separator10terminatoryypd_S2StFfA1_ : $@convention(thin) () -> @owned String
// print(_:separator:terminator:)
sil @$ss5print_9separator10terminatoryypd_S2StF : $@convention(thin) (@guaranteed Array<Any>, @guaranteed String, @guaranteed String) -> ()
Let us find out how this transformation perfromed.
We start from following code snippet, that will run performCompileStepsPostSema() right after finishing SemanticAnalysis.
/* /wtsc/swift/lib/FrontendTool/FrontendTool.cpp */
static bool performAction(CompilerInstance &Instance,
int &ReturnValue,
FrontendObserver *observer) {
...
case FrontendOptions::ActionType::EmitObject:
...
return withSemanticAnalysis(
Instance, observer, [&](CompilerInstance &Instance) {
assert(FrontendOptions::doesActionGenerateSIL(opts.RequestedAction) &&
"All actions not requiring SILGen must have been handled!");
return performCompileStepsPostSema(Instance, ReturnValue, observer);
});
...
}
Then we use our lovely lldb to step through what would be performed to generate SIL. I will point out some important break points to show how this above raw sil is generated.
Firstly, set a breakpoint at just right after an empty SIL Module is created.
(lldb) b SILGen.cpp : 2037
auto silMod = SILModule::createEmptyModule(desc.context, desc.conv, desc.opts);
Then in lldb, we can show what the sil module looks like at that point.
(lldb) expr silMod->dump(true);
sil_stage raw
import Builtin
import Swift
import SwiftShims
Second, set a breakpoint at creation of SourceFileScope that emits the main function in SIL Module.
(lldb) b SILGen.cpp : 1979
SourceFileScope scope(SGM, sf);
(lldb) expr SGM.M.dump(true)
sil_stage raw
import Builtin
import Swift
import SwiftShims
// main
sil [ossa] @main : $@convention(c) (Int32, UnsafeMutablePointer<Optional<UnsafeMutablePointer<Int8>>>) -> Int32 {
bb0(%0 : $Int32, %1 : $UnsafeMutablePointer<Optional<UnsafeMutablePointer<Int8>>>):
bb1:
bb2(%2 : @owned $Error):
} // end sil function 'main'
The actual action to create those new stuff is inside the constructor of SourceFileScope.
RegularLocation TopLevelLoc = RegularLocation::getModuleLocation();
SILFunction *toplevel = sgm.emitTopLevelFunction(TopLevelLoc);
// Assign a debug scope pointing into the void to the top level function.
toplevel->setDebugScope(new (sgm.M) SILDebugScope(TopLevelLoc, toplevel));
sgm.TopLevelSGF = new SILGenFunction(sgm, *toplevel, sf);
sgm.TopLevelSGF->MagicFunctionName = sgm.SwiftModule->getName();
auto moduleCleanupLoc = CleanupLocation::getModuleCleanupLocation();
sgm.TopLevelSGF->prepareEpilog(false, true, moduleCleanupLoc);
// Create the argc and argv arguments.
auto prologueLoc = RegularLocation::getModuleLocation();
prologueLoc.markAsPrologue();
auto entry = sgm.TopLevelSGF->B.getInsertionBB();
auto context = sgm.TopLevelSGF->getTypeExpansionContext();
auto paramTypeIter = sgm.TopLevelSGF->F.getConventions()
.getParameterSILTypes(context)
.begin();
entry->createFunctionArgument(*paramTypeIter);
entry->createFunctionArgument(*std::next(paramTypeIter));
scope.emplace(sgm.TopLevelSGF->Cleanups, moduleCleanupLoc);
We can use the same trick to go over the function calls to see how the SIL is generated.
One more thing, the SILModule::dump() method is worthwhile to look over. Because it iterates over a sil module to print all the content inside it. Therefore we can get some inside about the structre of module. Simply saying, a module contains some global variables, functions and some more information about this module. And a function contains basic blocks, in turn, basic block containts instructions.
This post shallowly shows the sil generation process. I will give a deeper look inside the whole process after the end of this series posts.
Stay tune.
