[CIR][CodeGen] Emit globals with constructor initializer #197

htyu · 2023-07-31T23:34:50Z

This change does the CIR generation for globals initialized by a constructor call. It currently only covers C++ to CIR generation. The corresponding LLVM lowering will be in a follow-up commit.

A motivating example is

class Init {
  friend class ios_base;

public:
  Init(bool);
  ~Init();

private:
  static bool _S_synced_with_stdio;
};

static Init ioinit(true);

Unlike what the default Clang codegen generates LLVM that detaches the initialization code from the global var definition (like below), we are taking a different approach that keeps them together, which we think will make the later dataflow analysis/transform easier.

@_ZL8ioinit = internal global %class.Init zeroinitializer, align 1, !dbg !0

define internal void @cxx_global_var_init() #0 section ".text.startup" !dbg !23 {
entry:
  call void @_ZN4InitC2Ev(ptr noundef nonnull align 1 dereferenceable(1) @_ZL8ioinit), !dbg !27
  %0 = call i32 @cxa_atexit(ptr @_ZN4InitD1Ev, ptr @_ZL8ioinit, ptr @dso_handle) #3, !dbg !29
  ret void, !dbg !27
}

So on CIR, we have something like:

  cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 {
    %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8)
    %1 = cir.const(#true) : !cir.bool loc(#loc5)
    cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6)
}

The destructor support will also be in a separate change.

bcardosolopes

I'm very excited for this, thanks for putting the work and starting this in incremental way. Overall the approach looks great.

Some missing bits:

Please add a testcase for roundtriping parsing/printing CIR ctor inits (like others in clang/test/CIR/IR.
I know you mentioned dtor support coming next, but if all the CIR parts for dtors are already part of this patch, please include simple parsing/printing support for it too (codegen could naturally come in later patches, not a problem).

@sitio-couto did some initial support for simple global inits in #176, probably a good idea to rebase this PR on top of his work sooner (#176 is already approved and should land soon).

clang/include/clang/CIR/Dialect/IR/CIROps.td

clang/lib/CIR/CodeGen/CIRGenDeclCXX.cpp

clang/lib/CIR/CodeGen/CIRGenFunction.cpp

htyu · 2023-08-03T00:18:30Z

Please add a testcase for roundtriping parsing/printing CIR ctor inits (like others in clang/test/CIR/IR.

There is a roundtrip test in static.cpp, i.e, // RUN: cir-opt %t.cir -o - | FileCheck %s -check-prefix=CIR. Would you like to move it to clang/test/CIR/IR?

As for the dtor parsing/printing support, I can add it here but I don't have a good way to test it. Would a separate change make more sense?

bcardosolopes · 2023-08-04T15:36:03Z

There is a roundtrip test in static.cpp, i.e, // RUN: cir-opt %t.cir -o - | FileCheck %s -check-prefix=CIR. Would you like to move it to clang/test/CIR/IR?

You can leave the roundtrip test there (definitely doesn't hurt!), but we need cir-to-cir ones that are independent of codegen, doesn't need to be anything fancy, just one that proves parsing/printing, globals.cir could be a good place for it.

As for the dtor parsing/printing support, I can add it here but I don't have a good way to test it. Would a separate change make more sense?

I believe so! You can also add them incrementally before codegen support if you prefer.

bcardosolopes

Thanks for applying updates! Can you mark the conversations as resolved as you fix the requested bits? It makes follow up reviews easier.

clang/lib/CIR/CodeGen/CIRGenExpr.cpp

bcardosolopes

This is pretty cool Hongtao, LGTM with a few remaining minor nitpicks!

clang/lib/CIR/CodeGen/CIRGenCXX.cpp

bcardosolopes · 2023-08-07T17:10:13Z

clang/lib/CIR/Dialect/IR/CIRDialect.cpp

+void GlobalOp::getSuccessorRegions(std::optional<unsigned> index,
+                                   ArrayRef<Attribute> operands,
+                                   SmallVectorImpl<RegionSuccessor> &regions) {
+  // The `then` and the `else` region branch back to the parent operation.


Please update the comment with what's relevant for this specific operation

yugr · 2023-08-09T13:18:29Z

Are there plans to extend this with arbitrary global initializers e.g.

int foo();
int c = foo();

?

htyu · 2023-08-09T15:50:42Z

Are there plans to extend this with arbitrary global initializers e.g.
int foo();
int c = foo();
?

Yes. It'll be handled in the same way, i.e, with initialization code placed in the ctor block.

I'm currently working on the LLVM lowering which should support arbitrary initialization code.

…zers (#235) As a follow up to #197, this change pre-lowers high-level CIR for global initializers. High-level CIR: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` After pre-lowering: ``` cir.global "private" internal @_ZL8__ioinit = #cir.zero : !ty_22class2EInit22 {ast = #cir.vardecl.ast} cir.func internal private @__cxx_global_var_init() { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> %1 = cir.const(#true) : !cir.bool cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () cir.return } cir.func private @_GLOBAL__sub_I_static.cpp() { cir.call @__cxx_global_var_init() : () -> () cir.return } ``` There is still work to be done to fully lower to LLVM. E.g, add `llvm.global_ctors` global to list all module initializers like `_GLOBAL__sub_I_static`. This will be handled in a separate change.

This change does the CIR generation for globals initialized by a constructor call. It currently only covers C++ to CIR generation. The corresponding LLVM lowering will be in a follow-up commit. A motivating example is ``` class Init { friend class ios_base; public: Init(bool); ~Init(); private: static bool _S_synced_with_stdio; }; static Init ioinit(true); ``` Unlike what the default Clang codegen generates LLVM that detaches the initialization code from the global var definition (like below), we are taking a different approach that keeps them together, which we think will make the later dataflow analysis/transform easier. ``` @_ZL8ioinit = internal global %class.Init zeroinitializer, align 1, !dbg !0 define internal void @cxx_global_var_init() #0 section ".text.startup" !dbg !23 { entry: call void @_ZN4InitC2Ev(ptr noundef nonnull align 1 dereferenceable(1) @_ZL8ioinit), !dbg !27 %0 = call i32 @cxa_atexit(ptr @_ZN4InitD1Ev, ptr @_ZL8ioinit, ptr @dso_handle) #3, !dbg !29 ret void, !dbg !27 } ``` So on CIR, we have something like: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` The destructor support will also be in a separate change.

…zers (#235) As a follow up to #197, this change pre-lowers high-level CIR for global initializers. High-level CIR: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` After pre-lowering: ``` cir.global "private" internal @_ZL8__ioinit = #cir.zero : !ty_22class2EInit22 {ast = #cir.vardecl.ast} cir.func internal private @__cxx_global_var_init() { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> %1 = cir.const(#true) : !cir.bool cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () cir.return } cir.func private @_GLOBAL__sub_I_static.cpp() { cir.call @__cxx_global_var_init() : () -> () cir.return } ``` There is still work to be done to fully lower to LLVM. E.g, add `llvm.global_ctors` global to list all module initializers like `_GLOBAL__sub_I_static`. This will be handled in a separate change.

This change does the CIR generation for globals initialized by a constructor call. It currently only covers C++ to CIR generation. The corresponding LLVM lowering will be in a follow-up commit. A motivating example is ``` class Init { friend class ios_base; public: Init(bool); ~Init(); private: static bool _S_synced_with_stdio; }; static Init ioinit(true); ``` Unlike what the default Clang codegen generates LLVM that detaches the initialization code from the global var definition (like below), we are taking a different approach that keeps them together, which we think will make the later dataflow analysis/transform easier. ``` @_ZL8ioinit = internal global %class.Init zeroinitializer, align 1, !dbg !0 define internal void @cxx_global_var_init() #0 section ".text.startup" !dbg !23 { entry: call void @_ZN4InitC2Ev(ptr noundef nonnull align 1 dereferenceable(1) @_ZL8ioinit), !dbg !27 %0 = call i32 @cxa_atexit(ptr @_ZN4InitD1Ev, ptr @_ZL8ioinit, ptr @dso_handle) #3, !dbg !29 ret void, !dbg !27 } ``` So on CIR, we have something like: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` The destructor support will also be in a separate change.

…zers (#235) As a follow up to #197, this change pre-lowers high-level CIR for global initializers. High-level CIR: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` After pre-lowering: ``` cir.global "private" internal @_ZL8__ioinit = #cir.zero : !ty_22class2EInit22 {ast = #cir.vardecl.ast} cir.func internal private @__cxx_global_var_init() { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> %1 = cir.const(#true) : !cir.bool cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () cir.return } cir.func private @_GLOBAL__sub_I_static.cpp() { cir.call @__cxx_global_var_init() : () -> () cir.return } ``` There is still work to be done to fully lower to LLVM. E.g, add `llvm.global_ctors` global to list all module initializers like `_GLOBAL__sub_I_static`. This will be handled in a separate change.

This change does the CIR generation for globals initialized by a constructor call. It currently only covers C++ to CIR generation. The corresponding LLVM lowering will be in a follow-up commit. A motivating example is ``` class Init { friend class ios_base; public: Init(bool); ~Init(); private: static bool _S_synced_with_stdio; }; static Init ioinit(true); ``` Unlike what the default Clang codegen generates LLVM that detaches the initialization code from the global var definition (like below), we are taking a different approach that keeps them together, which we think will make the later dataflow analysis/transform easier. ``` @_ZL8ioinit = internal global %class.Init zeroinitializer, align 1, !dbg !0 define internal void @cxx_global_var_init() #0 section ".text.startup" !dbg !23 { entry: call void @_ZN4InitC2Ev(ptr noundef nonnull align 1 dereferenceable(1) @_ZL8ioinit), !dbg !27 %0 = call i32 @cxa_atexit(ptr @_ZN4InitD1Ev, ptr @_ZL8ioinit, ptr @dso_handle) #3, !dbg !29 ret void, !dbg !27 } ``` So on CIR, we have something like: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` The destructor support will also be in a separate change.

…zers (#235) As a follow up to #197, this change pre-lowers high-level CIR for global initializers. High-level CIR: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` After pre-lowering: ``` cir.global "private" internal @_ZL8__ioinit = #cir.zero : !ty_22class2EInit22 {ast = #cir.vardecl.ast} cir.func internal private @__cxx_global_var_init() { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> %1 = cir.const(#true) : !cir.bool cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () cir.return } cir.func private @_GLOBAL__sub_I_static.cpp() { cir.call @__cxx_global_var_init() : () -> () cir.return } ``` There is still work to be done to fully lower to LLVM. E.g, add `llvm.global_ctors` global to list all module initializers like `_GLOBAL__sub_I_static`. This will be handled in a separate change.

…zers (llvm#235) As a follow up to llvm/clangir#197, this change pre-lowers high-level CIR for global initializers. High-level CIR: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` After pre-lowering: ``` cir.global "private" internal @_ZL8__ioinit = #cir.zero : !ty_22class2EInit22 {ast = #cir.vardecl.ast} cir.func internal private @__cxx_global_var_init() { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> %1 = cir.const(#true) : !cir.bool cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () cir.return } cir.func private @_GLOBAL__sub_I_static.cpp() { cir.call @__cxx_global_var_init() : () -> () cir.return } ``` There is still work to be done to fully lower to LLVM. E.g, add `llvm.global_ctors` global to list all module initializers like `_GLOBAL__sub_I_static`. This will be handled in a separate change.

This change does the CIR generation for globals initialized by a constructor call. It currently only covers C++ to CIR generation. The corresponding LLVM lowering will be in a follow-up commit. A motivating example is ``` class Init { friend class ios_base; public: Init(bool); ~Init(); private: static bool _S_synced_with_stdio; }; static Init ioinit(true); ``` Unlike what the default Clang codegen generates LLVM that detaches the initialization code from the global var definition (like below), we are taking a different approach that keeps them together, which we think will make the later dataflow analysis/transform easier. ``` @_ZL8ioinit = internal global %class.Init zeroinitializer, align 1, !dbg !0 define internal void @cxx_global_var_init() #0 section ".text.startup" !dbg !23 { entry: call void @_ZN4InitC2Ev(ptr noundef nonnull align 1 dereferenceable(1) @_ZL8ioinit), !dbg !27 %0 = call i32 @cxa_atexit(ptr @_ZN4InitD1Ev, ptr @_ZL8ioinit, ptr @dso_handle) #3, !dbg !29 ret void, !dbg !27 } ``` So on CIR, we have something like: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` The destructor support will also be in a separate change.

…zers (#235) As a follow up to #197, this change pre-lowers high-level CIR for global initializers. High-level CIR: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` After pre-lowering: ``` cir.global "private" internal @_ZL8__ioinit = #cir.zero : !ty_22class2EInit22 {ast = #cir.vardecl.ast} cir.func internal private @__cxx_global_var_init() { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> %1 = cir.const(#true) : !cir.bool cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () cir.return } cir.func private @_GLOBAL__sub_I_static.cpp() { cir.call @__cxx_global_var_init() : () -> () cir.return } ``` There is still work to be done to fully lower to LLVM. E.g, add `llvm.global_ctors` global to list all module initializers like `_GLOBAL__sub_I_static`. This will be handled in a separate change.

This change does the CIR generation for globals initialized by a constructor call. It currently only covers C++ to CIR generation. The corresponding LLVM lowering will be in a follow-up commit. A motivating example is ``` class Init { friend class ios_base; public: Init(bool); ~Init(); private: static bool _S_synced_with_stdio; }; static Init ioinit(true); ``` Unlike what the default Clang codegen generates LLVM that detaches the initialization code from the global var definition (like below), we are taking a different approach that keeps them together, which we think will make the later dataflow analysis/transform easier. ``` @_ZL8ioinit = internal global %class.Init zeroinitializer, align 1, !dbg !0 define internal void @cxx_global_var_init() #0 section ".text.startup" !dbg !23 { entry: call void @_ZN4InitC2Ev(ptr noundef nonnull align 1 dereferenceable(1) @_ZL8ioinit), !dbg !27 %0 = call i32 @cxa_atexit(ptr @_ZN4InitD1Ev, ptr @_ZL8ioinit, ptr @dso_handle) llvm#3, !dbg !29 ret void, !dbg !27 } ``` So on CIR, we have something like: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` The destructor support will also be in a separate change.

…zers (llvm#235) As a follow up to llvm#197, this change pre-lowers high-level CIR for global initializers. High-level CIR: ``` cir.global "private" internal @_ZL8__ioinit = ctor : !ty_22class2EInit22 { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> loc(#loc8) %1 = cir.const(#true) : !cir.bool loc(#loc5) cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () loc(#loc6) } ``` After pre-lowering: ``` cir.global "private" internal @_ZL8__ioinit = #cir.zero : !ty_22class2EInit22 {ast = #cir.vardecl.ast} cir.func internal private @__cxx_global_var_init() { %0 = cir.get_global @_ZL8__ioinit : cir.ptr <!ty_22class2EInit22> %1 = cir.const(#true) : !cir.bool cir.call @_ZN4InitC1Eb(%0, %1) : (!cir.ptr<!ty_22class2EInit22>, !cir.bool) -> () cir.return } cir.func private @_GLOBAL__sub_I_static.cpp() { cir.call @__cxx_global_var_init() : () -> () cir.return } ``` There is still work to be done to fully lower to LLVM. E.g, add `llvm.global_ctors` global to list all module initializers like `_GLOBAL__sub_I_static`. This will be handled in a separate change.