Move EntryExitInstrumentation pass location

This seems to be more of a Clang thing rather than a generic LLVM thing,
so this moves it out of LLVM pipelines and as Clang extension hooks into
LLVM pipelines.

Move the post-inline EEInstrumentation out of the backend pipeline and
into a late pass, similar to other sanitizer passes. It doesn't fit
into the codegen pipeline.

Also fix up EntryExitInstrumentation not running at -O0 under the new
PM. PR49143

Reviewed By: hans

Differential Revision: https://reviews.llvm.org/D97608

include/llvm/Transforms/Utils/EntryExitInstrumenter.h

@@ -28,6 +28,8 @@ struct EntryExitInstrumenterPass
   PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM);
 
   bool PostInlining;
+
+  static bool isRequired() { return true; }
 };
 
 } // namespace llvm

lib/CodeGen/TargetPassConfig.cpp

@@ -864,9 +864,6 @@ void TargetPassConfig::addIRPasses() {
   if (getOptLevel() != CodeGenOpt::None && !DisablePartialLibcallInlining)
     addPass(createPartiallyInlineLibCallsPass());
 
-  // Instrument function entry and exit, e.g. with calls to mcount().
-  addPass(createPostInlineEntryExitInstrumenterPass());
-
   // Add scalarization of target's unsupported masked memory intrinsics pass.
   // the unsupported intrinsic will be replaced with a chain of basic blocks,
   // that stores/loads element one-by-one if the appropriate mask bit is set.

lib/Transforms/IPO/PassManagerBuilder.cpp

@@ -300,7 +300,6 @@ void PassManagerBuilder::addInitialAliasAnalysisPasses(
 void PassManagerBuilder::populateFunctionPassManager(
     legacy::FunctionPassManager &FPM) {
   addExtensionsToPM(EP_EarlyAsPossible, FPM);
-  FPM.add(createEntryExitInstrumenterPass());
 
   // Add LibraryInfo if we have some.
   if (LibraryInfo)

test/CodeGen/AArch64/O0-pipeline.ll

@@ -21,7 +21,6 @@
 ; CHECK-NEXT:       Shadow Stack GC Lowering
 ; CHECK-NEXT:       Lower constant intrinsics
 ; CHECK-NEXT:       Remove unreachable blocks from the CFG
-; CHECK-NEXT:       Instrument function entry/exit with calls to e.g. mcount() (post inlining)
 ; CHECK-NEXT:       Scalarize Masked Memory Intrinsics
 ; CHECK-NEXT:       Expand reduction intrinsics
 ; CHECK-NEXT:       AArch64 Stack Tagging

test/CodeGen/AArch64/O3-pipeline.ll

@@ -56,7 +56,6 @@
 ; CHECK-NEXT:       Constant Hoisting
 ; CHECK-NEXT:       Replace intrinsics with calls to vector library
 ; CHECK-NEXT:       Partially inline calls to library functions
-; CHECK-NEXT:       Instrument function entry/exit with calls to e.g. mcount() (post inlining)
 ; CHECK-NEXT:       Scalarize Masked Memory Intrinsics
 ; CHECK-NEXT:       Expand reduction intrinsics
 ; CHECK-NEXT:     Stack Safety Analysis

test/CodeGen/AMDGPU/opt-pipeline.ll

@@ -10,7 +10,6 @@
 ; GCN-O0-NEXT:   FunctionPass Manager
 ; GCN-O0-NEXT:     Early propagate attributes from kernels to functions
 ; GCN-O0-NEXT:     Replace builtin math calls with that native versions.
-; GCN-O0-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 
 ; GCN-O0-NEXT: Pass Arguments:
 ; GCN-O0-NEXT: Target Library Information
@@ -42,7 +41,6 @@
 ; GCN-O1-NEXT:     Basic Alias Analysis (stateless AA impl)
 ; GCN-O1-NEXT:     Function Alias Analysis Results
 ; GCN-O1-NEXT:     Simplify well-known AMD library calls
-; GCN-O1-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; GCN-O1-NEXT:     Simplify the CFG
 ; GCN-O1-NEXT:     Dominator Tree Construction
 ; GCN-O1-NEXT:     SROA
@@ -352,7 +350,6 @@
 ; GCN-O2-NEXT:     Basic Alias Analysis (stateless AA impl)
 ; GCN-O2-NEXT:     Function Alias Analysis Results
 ; GCN-O2-NEXT:     Simplify well-known AMD library calls
-; GCN-O2-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; GCN-O2-NEXT:     Simplify the CFG
 ; GCN-O2-NEXT:     Dominator Tree Construction
 ; GCN-O2-NEXT:     SROA
@@ -707,7 +704,6 @@
 ; GCN-O3-NEXT:     Basic Alias Analysis (stateless AA impl)
 ; GCN-O3-NEXT:     Function Alias Analysis Results
 ; GCN-O3-NEXT:     Simplify well-known AMD library calls
-; GCN-O3-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; GCN-O3-NEXT:     Simplify the CFG
 ; GCN-O3-NEXT:     Dominator Tree Construction
 ; GCN-O3-NEXT:     SROA

test/CodeGen/ARM/O3-pipeline.ll

@@ -36,7 +36,6 @@
 ; CHECK-NEXT:      Constant Hoisting
 ; CHECK-NEXT:      Replace intrinsics with calls to vector library
 ; CHECK-NEXT:      Partially inline calls to library functions
-; CHECK-NEXT:      Instrument function entry/exit with calls to e.g. mcount() (post inlining)
 ; CHECK-NEXT:      Scalarize Masked Memory Intrinsics
 ; CHECK-NEXT:      Expand reduction intrinsics
 ; CHECK-NEXT:      Natural Loop Information

test/CodeGen/ARM/gnu_mcount_nc.ll

@@ -18,6 +18,7 @@ define dso_local void @callee() #0 {
 ; CHECK-THUMB-FAST-ISEL-NEXT:   bl      __gnu_mcount_nc
 ; CHECK-THUMB-GLOBAL-ISEL:      push    {lr}
 ; CHECK-THUMB-GLOBAL-ISEL-NEXT: bl      __gnu_mcount_nc
+  call void @llvm.arm.gnu.eabi.mcount()
   ret void
 }
 
@@ -34,8 +35,12 @@ define dso_local void @caller() #0 {
 ; CHECK-THUMB-FAST-ISEL-NEXT:   bl      __gnu_mcount_nc
 ; CHECK-THUMB-GLOBAL-ISEL:      push    {lr}
 ; CHECK-THUMB-GLOBAL-ISEL-NEXT: bl      __gnu_mcount_nc
+  call void @llvm.arm.gnu.eabi.mcount()
   call void @callee()
   ret void
 }
 
-attributes #0 = { nofree nounwind "instrument-function-entry-inlined"="llvm.arm.gnu.eabi.mcount" }
+declare void @llvm.arm.gnu.eabi.mcount() #1
+
+attributes #0 = { nofree nounwind }
+attributes #1 = { nounwind }

test/CodeGen/Mips/long-call-mcount.ll

@@ -4,9 +4,9 @@
 ; RUN: llc -march=mips -target-abi o32 --mattr=-long-calls,+noabicalls < %s \
 ; RUN:   -mips-jalr-reloc=false | FileCheck -check-prefixes=CHECK,SHORT %s
 
-; Function Attrs: noinline nounwind optnone
+define void @foo() {
-define void @foo() #0 {
 entry:
+  call void @_mcount()
   ret void
 
 ; CHECK-LABEL: foo
@@ -16,4 +16,4 @@ entry:
 ; SHORT:         jal     _mcount
 }
 
-attributes #0 = { "instrument-function-entry-inlined"="_mcount" }
+declare void @_mcount()

test/CodeGen/Mips/mcount.ll

@@ -15,8 +15,7 @@
 
 ; Test that checks ABI for _mcount calls.
 
-; Function Attrs: noinline nounwind optnone
+define void @foo() {
-define void @foo() #0 {
 ; MIPS32-LABEL: foo:
 ; MIPS32:       # %bb.0: # %entry
 ; MIPS32-NEXT:    addiu $sp, $sp, -24
@@ -117,7 +116,8 @@ define void @foo() #0 {
 ; MIPS32-MM-PIC-NEXT:    jr $ra
 ; MIPS32-MM-PIC-NEXT:    addiu $sp, $sp, 24
 entry:
+  call void @_mcount()
   ret void
 }
 
-attributes #0 = { "instrument-function-entry-inlined"="_mcount" }
+declare void @_mcount()

test/CodeGen/PowerPC/mcount-insertion.ll

@@ -1,4 +1,4 @@
-; RUN: opt -ee-instrument < %s | opt -inline | llc -mtriple=powerpc64-unknown-linux-gnu | FileCheck %s
+; RUN: opt -ee-instrument < %s | opt -inline | opt -post-inline-ee-instrument | llc -mtriple=powerpc64-unknown-linux-gnu | FileCheck %s
 
 ; The run-line mimics how Clang might run the instrumentation passes.
 

test/CodeGen/X86/O0-pipeline.ll

@@ -24,7 +24,6 @@
 ; CHECK-NEXT:       Shadow Stack GC Lowering
 ; CHECK-NEXT:       Lower constant intrinsics
 ; CHECK-NEXT:       Remove unreachable blocks from the CFG
-; CHECK-NEXT:       Instrument function entry/exit with calls to e.g. mcount() (post inlining)
 ; CHECK-NEXT:       Scalarize Masked Memory Intrinsics
 ; CHECK-NEXT:       Expand reduction intrinsics
 ; CHECK-NEXT:       Expand indirectbr instructions

test/CodeGen/X86/musttail-inalloca.ll

@@ -13,6 +13,7 @@ target triple = "i386-pc-windows-msvc19.16.0"
 
 declare dso_local x86_thiscallcc void @methodWithVtorDisp(i8* nocapture readonly, <{ %struct.Args }>* inalloca)
 
+; Function Attrs: nounwind optsize
 define dso_local x86_thiscallcc void @methodWithVtorDisp_thunk(i8* %0, <{ %struct.Args }>* inalloca %1) #0 {
 ; CHECK-LABEL: methodWithVtorDisp_thunk:
 ; CHECK:       # %bb.0:
@@ -31,8 +32,16 @@ define dso_local x86_thiscallcc void @methodWithVtorDisp_thunk(i8* %0, <{ %struc
   %5 = load i32, i32* %4, align 4
   %6 = sub i32 0, %5
   %7 = getelementptr i8, i8* %0, i32 %6
+  %8 = call i8* @llvm.returnaddress(i32 0)
+  call void @__cyg_profile_func_exit(i8* bitcast (void (i8*, <{ %struct.Args }>*)* @methodWithVtorDisp_thunk to i8*), i8* %8)
   musttail call x86_thiscallcc void @methodWithVtorDisp(i8* %7, <{ %struct.Args }>* inalloca nonnull %1)
   ret void
 }
 
-attributes #0 = { nounwind optsize "instrument-function-exit-inlined"="__cyg_profile_func_exit"  }
+declare void @__cyg_profile_func_exit(i8*, i8*)
+
+; Function Attrs: nofree nosync nounwind readnone willreturn
+declare i8* @llvm.returnaddress(i32 immarg) #1
+
+attributes #0 = { nounwind optsize }
+attributes #1 = { nofree nosync nounwind readnone willreturn }

test/CodeGen/X86/opt-pipeline.ll

@@ -53,7 +53,6 @@
 ; CHECK-NEXT:       Constant Hoisting
 ; CHECK-NEXT:       Replace intrinsics with calls to vector library
 ; CHECK-NEXT:       Partially inline calls to library functions
-; CHECK-NEXT:       Instrument function entry/exit with calls to e.g. mcount() (post inlining)
 ; CHECK-NEXT:       Scalarize Masked Memory Intrinsics
 ; CHECK-NEXT:       Expand reduction intrinsics
 ; CHECK-NEXT:       Interleaved Access Pass

test/Other/opt-O0-pipeline-enable-matrix.ll

@@ -6,7 +6,6 @@
 ; CHECK-NEXT: Target Transform Information
 ; CHECK-NEXT:   FunctionPass Manager
 ; CHECK-NEXT:     Module Verifier
-; CHECK-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; CHECK-NEXT:     Lower the matrix intrinsics (minimal)
 
 

test/Other/opt-O0-pipeline.ll

@@ -10,7 +10,6 @@
 ; CHECK-NEXT:   FunctionPass Manager
 ; CHECK-NEXT:     Module Verifier
 ; CHECK-EXT:     Good Bye World Pass
-; CHECK-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; CHECK-NEXT: Pass Arguments:
 ; CHECK-NEXT: Target Library Information
 ; CHECK-NEXT: Target Transform Information

test/Other/opt-O2-pipeline.ll

@@ -12,7 +12,6 @@
 ; CHECK-NEXT:     Module Verifier
 ; CHECK-EXT:      Good Bye World Pass
 ; CHECK-NOEXT-NOT:      Good Bye World Pass
-; CHECK-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; CHECK-NEXT:     Simplify the CFG
 ; CHECK-NEXT:     Dominator Tree Construction
 ; CHECK-NEXT:     SROA

test/Other/opt-O3-pipeline-enable-matrix.ll

@@ -12,7 +12,6 @@
 ; CHECK-NEXT:     Module Verifier
 ; CHECK-EXT:      Good Bye World Pass
 ; CHECK-NOEXT-NOT:      Good Bye World Pass
-; CHECK-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; CHECK-NEXT:     Simplify the CFG
 ; CHECK-NEXT:     Dominator Tree Construction
 ; CHECK-NEXT:     SROA

test/Other/opt-O3-pipeline.ll

@@ -12,7 +12,6 @@
 ; CHECK-NEXT:     Module Verifier
 ; CHECK-EXT:      Good Bye World Pass
 ; CHECK-NOEXT-NOT:      Good Bye World Pass
-; CHECK-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; CHECK-NEXT:     Simplify the CFG
 ; CHECK-NEXT:     Dominator Tree Construction
 ; CHECK-NEXT:     SROA

test/Other/opt-Os-pipeline.ll

@@ -12,7 +12,6 @@
 ; CHECK-NEXT:     Module Verifier
 ; CHECK-EXT:      Good Bye World Pass
 ; CHECK-NOEXT-NOT:      Good Bye World Pass
-; CHECK-NEXT:     Instrument function entry/exit with calls to e.g. mcount() (pre inlining)
 ; CHECK-NEXT:     Simplify the CFG
 ; CHECK-NEXT:     Dominator Tree Construction
 ; CHECK-NEXT:     SROA