diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst --- a/llvm/docs/LangRef.rst +++ b/llvm/docs/LangRef.rst @@ -7092,14 +7092,14 @@ :: = callbr [cconv] [ret attrs] [addrspace()] | () [fn attrs] - [operand bundles] to label [other labels] + [operand bundles] to label [indirect labels] Overview: """"""""" The '``callbr``' instruction causes control to transfer to a specified function, with the possibility of control flow transfer to either the -'``normal``' label or one of the '``other``' labels. +'``fallthrough``' label or one of the '``indirect``' labels. This instruction should only be used to implement the "goto" feature of gcc style inline assembly. Any other usage is an error in the IR verifier. @@ -7126,17 +7126,17 @@ type can be omitted if the function is not varargs. #. '``fnptrval``': An LLVM value containing a pointer to a function to be called. In most cases, this is a direct function call, but - indirect ``callbr``'s are just as possible, calling an arbitrary pointer + other ``callbr``'s are just as possible, calling an arbitrary pointer to function value. #. '``function args``': argument list whose types match the function signature argument types and parameter attributes. All arguments must be of :ref:`first class ` type. If the function signature indicates the function accepts a variable number of arguments, the extra arguments can be specified. -#. '``normal label``': the label reached when the called function - executes a '``ret``' instruction. -#. '``other labels``': the labels reached when a callee transfers control - to a location other than the normal '``normal label``'. The blockaddress +#. '``fallthrough label``': the label reached when the inline assembly's + execution exits the bottom. +#. '``indirect labels``': the labels reached when a callee transfers control + to a location other than the '``fallthrough label``'. The blockaddress constant for these should also be in the list of '``function args``'. #. The optional :ref:`function attributes ` list. #. The optional :ref:`operand bundles ` list. @@ -7149,6 +7149,13 @@ establishes an association with additional labels to define where control flow goes after the call. +Outputs of a '``callbr``' instruction are valid only on the '``fallthrough``' +path. Use of outputs on the '``indirect``' path(s) results in :ref:`poison +values `. Care should be taken if the '``fallthrough``' block is +also listed in the '``indirect``' blocks. The outputs may or may not result in +:ref:`poison values ` depending on the behavior of the inline +assembly. + The only use of this today is to implement the "goto" feature of gcc inline assembly where additional labels can be provided as locations for the inline assembly to jump to. @@ -7156,10 +7163,15 @@ Example: """""""" -.. code-block:: text +.. code-block:: llvm - callbr void asm "", "r,x"(i32 %x, i8 *blockaddress(@foo, %fail)) - to label %normal [label %fail] + ; "asm goto" without output constraints. + callbr void asm "", "r,X"(i32 %x, i8 *blockaddress(@foo, %indirect)) + to label %fallthrough [label %indirect] + + ; "asm goto" with output constraints. + = callbr i32 asm "", "=r,r,X"(i32 %x, i8 *blockaddress(@foo, %indirect)) + to label %fallthrough [label %indirect] .. _i_resume: diff --git a/llvm/include/llvm/CodeGen/MachineBasicBlock.h b/llvm/include/llvm/CodeGen/MachineBasicBlock.h --- a/llvm/include/llvm/CodeGen/MachineBasicBlock.h +++ b/llvm/include/llvm/CodeGen/MachineBasicBlock.h @@ -129,6 +129,9 @@ /// Indicate that this basic block is the entry block of a cleanup funclet. bool IsCleanupFuncletEntry = false; + /// Indicate that this basic block is the indirect dest of an INLINEASM_BR. + bool IsInlineAsmBrIndirectPad = false; + /// since getSymbol is a relatively heavy-weight operation, the symbol /// is only computed once and is cached. mutable MCSymbol *CachedMCSymbol = nullptr; @@ -408,6 +411,12 @@ /// Indicates if this is the entry block of a cleanup funclet. void setIsCleanupFuncletEntry(bool V = true) { IsCleanupFuncletEntry = V; } + /// Returns true if this is the indirect dest of an INLINEASM_BR. + bool isInlineAsmBrIndirectPad() const { return IsInlineAsmBrIndirectPad; } + + /// Indicates if this is the indirect dest of an INLINEASM_BR. + void setIsInlineAsmBrIndirectPad(bool V = true) { IsInlineAsmBrIndirectPad = V; } + /// Returns true if it is legal to hoist instructions into this block. bool isLegalToHoistInto() const; diff --git a/llvm/lib/AsmParser/LLParser.cpp b/llvm/lib/AsmParser/LLParser.cpp --- a/llvm/lib/AsmParser/LLParser.cpp +++ b/llvm/lib/AsmParser/LLParser.cpp @@ -6419,9 +6419,6 @@ /*IsCall=*/true)) return true; - if (isa(Callee) && !Ty->getReturnType()->isVoidTy()) - return Error(RetTypeLoc, "asm-goto outputs not supported"); - // Set up the Attribute for the function. SmallVector Args; SmallVector ArgAttrs; diff --git a/llvm/lib/CodeGen/MachineBasicBlock.cpp b/llvm/lib/CodeGen/MachineBasicBlock.cpp --- a/llvm/lib/CodeGen/MachineBasicBlock.cpp +++ b/llvm/lib/CodeGen/MachineBasicBlock.cpp @@ -1109,6 +1109,17 @@ if (Succ->isEHPad()) return false; + // Splitting the critical edge to a callbr's indirect block isn't advised. + // Don't do it in this generic function. + if (Succ->hasAddressTaken()) + if (auto *cbr = dyn_cast(getBasicBlock()->getTerminator())) + if (auto *bb = Succ->getBasicBlock()) + if (cbr->getDefaultDest() != bb) + if (llvm::any_of(cbr->getIndirectDests(), [&](const BasicBlock *succ){ + return succ == bb; + })) + return false; + const MachineFunction *MF = getParent(); // Performance might be harmed on HW that implements branching using exec mask diff --git a/llvm/lib/CodeGen/MachineVerifier.cpp b/llvm/lib/CodeGen/MachineVerifier.cpp --- a/llvm/lib/CodeGen/MachineVerifier.cpp +++ b/llvm/lib/CodeGen/MachineVerifier.cpp @@ -633,17 +633,30 @@ } // Count the number of landing pad successors. - SmallPtrSet LandingPadSuccs; - for (MachineBasicBlock::const_succ_iterator I = MBB->succ_begin(), - E = MBB->succ_end(); I != E; ++I) { - if ((*I)->isEHPad()) - LandingPadSuccs.insert(*I); - if (!FunctionBlocks.count(*I)) + SmallPtrSet LandingPadSuccs; + for (const auto *succ : MBB->successors()) { + if (succ->isEHPad()) + LandingPadSuccs.insert(succ); + if (!FunctionBlocks.count(succ)) report("MBB has successor that isn't part of the function.", MBB); - if (!MBBInfoMap[*I].Preds.count(MBB)) { + if (!MBBInfoMap[succ].Preds.count(MBB)) { report("Inconsistent CFG", MBB); errs() << "MBB is not in the predecessor list of the successor " - << printMBBReference(*(*I)) << ".\n"; + << printMBBReference(*succ) << ".\n"; + } + } + + // Count the number of INLINEASM_BR indirect pad successors. + SmallPtrSet IndirectPadSuccs; + for (const auto *succ : MBB->successors()) { + if (succ->isInlineAsmBrIndirectPad()) + IndirectPadSuccs.insert(succ); + if (!FunctionBlocks.count(succ)) + report("MBB has successor that isn't part of the function.", MBB); + if (!MBBInfoMap[succ].Preds.count(MBB)) { + report("Inconsistent CFG", MBB); + errs() << "MBB is not in the predecessor list of the successor " + << printMBBReference(*succ) << ".\n"; } } @@ -684,11 +697,13 @@ // It's possible that the block legitimately ends with a noreturn // call or an unreachable, in which case it won't actually fall // out the bottom of the function. - } else if (MBB->succ_size() == LandingPadSuccs.size()) { + } else if (MBB->succ_size() == LandingPadSuccs.size() || + MBB->succ_size() == IndirectPadSuccs.size()) { // It's possible that the block legitimately ends with a noreturn // call or an unreachable, in which case it won't actually fall // out of the block. - } else if (MBB->succ_size() != 1+LandingPadSuccs.size()) { + } else if (MBB->succ_size() != 1 + LandingPadSuccs.size() && + MBB->succ_size() != 1 + IndirectPadSuccs.size()) { report("MBB exits via unconditional fall-through but doesn't have " "exactly one CFG successor!", MBB); } else if (!MBB->isSuccessor(&*MBBI)) { @@ -710,7 +725,10 @@ // landingpad, accept it as valid control flow. if (MBB->succ_size() != 1+LandingPadSuccs.size() && (MBB->succ_size() != 1 || LandingPadSuccs.size() != 1 || - *MBB->succ_begin() != *LandingPadSuccs.begin())) { + *MBB->succ_begin() != *LandingPadSuccs.begin()) && + MBB->succ_size() != 1 + IndirectPadSuccs.size() && + (MBB->succ_size() != 1 || IndirectPadSuccs.size() != 1 || + *MBB->succ_begin() != *IndirectPadSuccs.begin())) { report("MBB exits via unconditional branch but doesn't have " "exactly one CFG successor!", MBB); } else if (!MBB->isSuccessor(TBB)) { @@ -840,8 +858,14 @@ if (!FirstTerminator) FirstTerminator = MI; } else if (FirstTerminator && !MI->isDebugEntryValue()) { - report("Non-terminator instruction after the first terminator", MI); - errs() << "First terminator was:\t" << *FirstTerminator; + // An "INLINEASM_BR" will fallthrough to the successor block executing any + // "COPY" instructions that exist so that they can be assigned in the + // fallthrough block.. + if (FirstTerminator->getOpcode() != TargetOpcode::INLINEASM_BR || + MI->getOpcode() != TargetOpcode::COPY) { + report("Non-terminator instruction after the first terminator", MI); + errs() << "First terminator was:\t" << *FirstTerminator; + } } } diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp --- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp @@ -2847,6 +2847,7 @@ assert(isa(I.getCalledValue()) && "Only know how to handle inlineasm callbr"); visitInlineAsm(&I); + CopyToExportRegsIfNeeded(&I); // Retrieve successors. MachineBasicBlock *Return = FuncInfo.MBBMap[I.getDefaultDest()]; @@ -2856,6 +2857,7 @@ for (unsigned i = 0, e = I.getNumIndirectDests(); i < e; ++i) { MachineBasicBlock *Target = FuncInfo.MBBMap[I.getIndirectDest(i)]; addSuccessorWithProb(CallBrMBB, Target); + Target->setIsInlineAsmBrIndirectPad(); } CallBrMBB->normalizeSuccProbs(); diff --git a/llvm/lib/IR/Verifier.cpp b/llvm/lib/IR/Verifier.cpp --- a/llvm/lib/IR/Verifier.cpp +++ b/llvm/lib/IR/Verifier.cpp @@ -2501,8 +2501,6 @@ void Verifier::visitCallBrInst(CallBrInst &CBI) { Assert(CBI.isInlineAsm(), "Callbr is currently only used for asm-goto!", &CBI); - Assert(CBI.getType()->isVoidTy(), "Callbr return value is not supported!", - &CBI); for (unsigned i = 0, e = CBI.getNumSuccessors(); i != e; ++i) Assert(CBI.getSuccessor(i)->getType()->isLabelTy(), "Callbr successors must all have pointer type!", &CBI); diff --git a/llvm/test/CodeGen/X86/callbr-asm-outputs.ll b/llvm/test/CodeGen/X86/callbr-asm-outputs.ll --- a/llvm/test/CodeGen/X86/callbr-asm-outputs.ll +++ b/llvm/test/CodeGen/X86/callbr-asm-outputs.ll @@ -1,18 +1,118 @@ -; RUN: not llc -mtriple=i686-- < %s 2> %t -; RUN: FileCheck %s < %t +; RUN: llc -mtriple=i686-- -verify-machineinstrs < %s | FileCheck %s -; CHECK: error: asm-goto outputs not supported +; A test for asm-goto output -; A test for asm-goto output prohibition - -define i32 @test(i32 %a) { +; CHECK-LABEL: test1: +; CHECK: movl 4(%esp), %eax +; CHECK-NEXT: addl $4, %eax +; CHECK-NEXT: #APP +; CHECK-NEXT: xorl %eax, %eax +; CHECK-NEXT: jmp .Ltmp0 +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: .LBB0_1: +; CHECK-NEXT: retl +; CHECK-NEXT: .Ltmp0: +define i32 @test1(i32 %x) { entry: - %0 = add i32 %a, 4 - %1 = callbr i32 asm "xorl $1, $1; jmp ${1:l}", "=&r,r,X,~{dirflag},~{fpsr},~{flags}"(i32 %0, i8* blockaddress(@test, %fail)) to label %normal [label %fail] + %add = add nsw i32 %x, 4 + %ret = callbr i32 asm "xorl $1, $0; jmp ${2:l}", "=r,r,X,~{dirflag},~{fpsr},~{flags}"(i32 %add, i8* blockaddress(@test1, %abnormal)) + to label %normal [label %abnormal] normal: - ret i32 %1 + ret i32 %ret -fail: +abnormal: ret i32 1 } + +; CHECK-LABEL: test2: +; CHECK: # %bb.1: # %if.then +; CHECK-NEXT: #APP +; CHECK-NEXT: testl %esi, %esi +; CHECK-NEXT: testl %edi, %esi +; CHECK-NEXT: jne .Ltmp1 +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: .LBB1_2: +; CHECK-NEXT: jmp .LBB1_4 +; CHECK-NEXT: .LBB1_3: # %if.else +; CHECK-NEXT: #APP +; CHECK-NEXT: testl %esi, %edi +; CHECK-NEXT: testl %esi, %edi +; CHECK-NEXT: jne .Ltmp2 +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: .LBB1_4: +; CHECK-NEXT: movl %esi, %eax +; CHECK-NEXT: addl %edi, %eax +; CHECK-NEXT: jmp .LBB1_6 +; CHECK-NEXT: .Ltmp1: +; CHECK-NEXT: .LBB1_5: # %label_true +; CHECK-NEXT: movl $-2, %eax +; CHECK-NEXT: .Ltmp2: +; CHECK-NEXT: .LBB1_6: # %return +define i32 @test2(i32 %out1, i32 %out2) { +entry: + %cmp = icmp slt i32 %out1, %out2 + br i1 %cmp, label %if.then, label %if.else + +if.then: ; preds = %entry + %0 = callbr { i32, i32 } asm sideeffect "testl $0, $0; testl $1, $2; jne ${5:l}", "={si},={di},r,0,1,X,X,~{dirflag},~{fpsr},~{flags}"(i32 %out1, i32 %out1, i32 %out2, i8* blockaddress(@test2, %label_true), i8* blockaddress(@test2, %return)) + to label %if.end [label %label_true, label %return] + +if.else: ; preds = %entry + %1 = callbr { i32, i32 } asm sideeffect "testl $0, $1; testl $2, $3; jne ${7:l}", "={si},={di},r,r,0,1,X,X,~{dirflag},~{fpsr},~{flags}"(i32 %out1, i32 %out2, i32 %out1, i32 %out2, i8* blockaddress(@test2, %label_true), i8* blockaddress(@test2, %return)) + to label %if.end [label %label_true, label %return] + +if.end: ; preds = %if.else, %if.then + %.sink11 = phi { i32, i32 } [ %0, %if.then ], [ %1, %if.else ] + %asmresult3 = extractvalue { i32, i32 } %.sink11, 0 + %asmresult4 = extractvalue { i32, i32 } %.sink11, 1 + %add = add nsw i32 %asmresult4, %asmresult3 + br label %return + +label_true: ; preds = %if.else, %if.then + br label %return + +return: ; preds = %if.then, %if.else, %label_true, %if.end + %retval.0 = phi i32 [ %add, %if.end ], [ -2, %label_true ], [ -1, %if.else ], [ -1, %if.then ] + ret i32 %retval.0 +} + +; CHECK-LABEL: test3: +; CHECK: # %bb.1: # %true +; CHECK-NEXT: #APP +; CHECK-NEXT: .short %esi +; CHECK-NEXT: .short %edi +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: .LBB2_2: +; CHECK-NEXT: movl %edi, %eax +; CHECK-NEXT: jmp .LBB2_5 +; CHECK-NEXT: .LBB2_3: # %false +; CHECK-NEXT: #APP +; CHECK-NEXT: .short %eax +; CHECK-NEXT: .short %edx +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: .LBB2_4: +; CHECK-NEXT: movl %edx, %eax +; CHECK-NEXT: jmp .LBB2_5 +; CHECK-NEXT: .Ltmp3: # Block address taken +; CHECK-NEXT: .LBB2_6: # %indirect +; CHECK-NEXT: movl $42, %eax +; CHECK-NEXT: .LBB2_5: # %asm.fallthrough +define i32 @test3(i1 %cmp) { +entry: + br i1 %cmp, label %true, label %false + +true: + %0 = callbr { i32, i32 } asm sideeffect ".word $0, $1", "={si},={di},X" (i8* blockaddress(@test3, %indirect)) to label %asm.fallthrough [label %indirect] + +false: + %1 = callbr { i32, i32 } asm sideeffect ".word $0, $1", "={ax},={dx},X" (i8* blockaddress(@test3, %indirect)) to label %asm.fallthrough [label %indirect] + +asm.fallthrough: + %vals = phi { i32, i32 } [ %0, %true ], [ %1, %false ] + %v = extractvalue { i32, i32 } %vals, 1 + ret i32 %v + +indirect: + ret i32 42 +} diff --git a/llvm/test/CodeGen/X86/callbr-asm.ll b/llvm/test/CodeGen/X86/callbr-asm.ll --- a/llvm/test/CodeGen/X86/callbr-asm.ll +++ b/llvm/test/CodeGen/X86/callbr-asm.ll @@ -1,5 +1,5 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=i686-- -O3 | FileCheck %s +; RUN: llc < %s -mtriple=i686-- -O3 -verify-machineinstrs | FileCheck %s ; Tests for using callbr as an asm-goto wrapper @@ -159,3 +159,54 @@ cleanup: ; preds = %asm.fallthrough, %quux ret void } + +define i32 @test5(i32 %out1, i32 %out2) { +; Test 5 - asm-goto with output constraints. +; CHECK-LABEL: test5: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: movl $-1, %eax +; CHECK-NEXT: movl 4(%esp), %ecx +; CHECK-NEXT: #APP +; CHECK-NEXT: testl %ecx, %ecx +; CHECK-NEXT: testl %edx, %ecx +; CHECK-NEXT: jne .Ltmp6 +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: .LBB4_1: +; CHECK-NEXT: #APP +; CHECK-NEXT: testl %ecx, %edx +; CHECK-NEXT: testl %ecx, %edx +; CHECK-NEXT: jne .Ltmp7 +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: .LBB4_2: +; CHECK-NEXT: addl %edx, %ecx +; CHECK-NEXT: movl %ecx, %eax +; CHECK-NEXT: retl +; CHECK-NEXT: .Ltmp6: +; CHECK-NEXT: .LBB4_3: +; CHECK-NEXT: movl $-2, %eax +; CHECK-NEXT: .Ltmp7: +; CHECK-NEXT: .LBB4_4: +; CHECK-NEXT: retl +entry: + %0 = callbr { i32, i32 } asm sideeffect "testl $0, $0; testl $1, $2; jne ${3:l}", "=r,=r,r,X,X,~{dirflag},~{fpsr},~{flags}"(i32 %out1, i8* blockaddress(@test5, %label_true), i8* blockaddress(@test5, %return)) + to label %asm.fallthrough [label %label_true, label %return] + +asm.fallthrough: ; preds = %entry + %asmresult = extractvalue { i32, i32 } %0, 0 + %asmresult1 = extractvalue { i32, i32 } %0, 1 + %1 = callbr { i32, i32 } asm sideeffect "testl $0, $1; testl $2, $3; jne ${5:l}", "=r,=r,r,r,X,X,~{dirflag},~{fpsr},~{flags}"(i32 %asmresult, i32 %asmresult1, i8* blockaddress(@test5, %label_true), i8* blockaddress(@test5, %return)) + to label %asm.fallthrough2 [label %label_true, label %return] + +asm.fallthrough2: ; preds = %asm.fallthrough + %asmresult3 = extractvalue { i32, i32 } %1, 0 + %asmresult4 = extractvalue { i32, i32 } %1, 1 + %add = add nsw i32 %asmresult3, %asmresult4 + br label %return + +label_true: ; preds = %asm.fallthrough, %entry + br label %return + +return: ; preds = %entry, %asm.fallthrough, %label_true, %asm.fallthrough2 + %retval.0 = phi i32 [ %add, %asm.fallthrough2 ], [ -2, %label_true ], [ -1, %asm.fallthrough ], [ -1, %entry ] + ret i32 %retval.0 +}