diff --git a/lib/Transforms/IPO/GlobalOpt.cpp b/lib/Transforms/IPO/GlobalOpt.cpp index 7bed9e22d11..fd71dfb8261 100644 --- a/lib/Transforms/IPO/GlobalOpt.cpp +++ b/lib/Transforms/IPO/GlobalOpt.cpp @@ -21,6 +21,7 @@ #include "llvm/Instructions.h" #include "llvm/IntrinsicInst.h" #include "llvm/Module.h" +#include "llvm/ParameterAttributes.h" #include "llvm/Pass.h" #include "llvm/Analysis/ConstantFolding.h" #include "llvm/Target/TargetData.h" @@ -46,6 +47,7 @@ STATISTIC(NumLocalized , "Number of globals localized"); STATISTIC(NumShrunkToBool , "Number of global vars shrunk to booleans"); STATISTIC(NumFastCallFns , "Number of functions converted to fastcc"); STATISTIC(NumCtorsEvaluated, "Number of static ctors evaluated"); +STATISTIC(NumNestRemoved , "Number of nest attributes removed"); namespace { struct VISIBILITY_HIDDEN GlobalOpt : public ModulePass { @@ -1590,6 +1592,35 @@ static void ChangeCalleesToFastCall(Function *F) { } } +static const ParamAttrsList *StripNest(const ParamAttrsList *Attrs) { + if (Attrs) { + for (unsigned i = 0, e = Attrs->size(); i != e; ++i) { + uint16_t A = Attrs->getParamAttrsAtIndex(i); + if (A & ParamAttr::Nest) { + Attrs = ParamAttrsList::excludeAttrs(Attrs, Attrs->getParamIndex(i), + ParamAttr::Nest); + // There can be only one. + break; + } + } + } + + return Attrs; +} + +static void RemoveNestAttribute(Function *F) { + F->setParamAttrs(StripNest(F->getParamAttrs())); + for (Value::use_iterator UI = F->use_begin(), E = F->use_end(); UI != E;++UI){ + Instruction *User = cast(*UI); + if (CallInst *CI = dyn_cast(User)) { + CI->setParamAttrs(StripNest(CI->getParamAttrs())); + } else { + InvokeInst *II = cast(User); + II->setParamAttrs(StripNest(II->getParamAttrs())); + } + } +} + bool GlobalOpt::OptimizeFunctions(Module &M) { bool Changed = false; // Optimize functions. @@ -1601,16 +1632,27 @@ bool GlobalOpt::OptimizeFunctions(Module &M) { M.getFunctionList().erase(F); Changed = true; ++NumFnDeleted; - } else if (F->hasInternalLinkage() && - F->getCallingConv() == CallingConv::C && !F->isVarArg() && - OnlyCalledDirectly(F)) { - // If this function has C calling conventions, is not a varargs - // function, and is only called directly, promote it to use the Fast - // calling convention. - F->setCallingConv(CallingConv::Fast); - ChangeCalleesToFastCall(F); - ++NumFastCallFns; - Changed = true; + } else if (F->hasInternalLinkage()) { + if (F->getCallingConv() == CallingConv::C && !F->isVarArg() && + OnlyCalledDirectly(F)) { + // If this function has C calling conventions, is not a varargs + // function, and is only called directly, promote it to use the Fast + // calling convention. + F->setCallingConv(CallingConv::Fast); + ChangeCalleesToFastCall(F); + ++NumFastCallFns; + Changed = true; + } + + if (F->getParamAttrs() && + F->getParamAttrs()->hasAttrSomewhere(ParamAttr::Nest) && + OnlyCalledDirectly(F)) { + // The function is not used by a trampoline intrinsic, so it is safe + // to remove the 'nest' attribute. + RemoveNestAttribute(F); + ++NumNestRemoved; + Changed = true; + } } } return Changed; diff --git a/test/Transforms/GlobalOpt/2008-02-16-NestAttr.ll b/test/Transforms/GlobalOpt/2008-02-16-NestAttr.ll new file mode 100644 index 00000000000..59996c5a98b --- /dev/null +++ b/test/Transforms/GlobalOpt/2008-02-16-NestAttr.ll @@ -0,0 +1,57 @@ +; RUN: llvm-as < %s | opt -globalopt | llvm-dis | grep { nest } | count 1 + %struct.FRAME.nest = type { i32, i32 (i32)* } + %struct.__builtin_trampoline = type { [10 x i8] } +@.str = internal constant [7 x i8] c"%d %d\0A\00" ; <[7 x i8]*> [#uses=1] + +define i32 @process(i32 (i32)* %func) nounwind { +entry: + %tmp2 = tail call i32 %func( i32 1 ) nounwind ; [#uses=1] + ret i32 %tmp2 +} + +define internal fastcc i32 @g.1478(%struct.FRAME.nest* nest %CHAIN.1, i32 %m) nounwind { +entry: + %tmp3 = getelementptr %struct.FRAME.nest* %CHAIN.1, i32 0, i32 0 ; [#uses=1] + %tmp4 = load i32* %tmp3, align 4 ; [#uses=1] + %tmp7 = icmp eq i32 %tmp4, %m ; [#uses=1] + %tmp78 = zext i1 %tmp7 to i32 ; [#uses=1] + ret i32 %tmp78 +} + +define internal i32 @f.1481(%struct.FRAME.nest* nest %CHAIN.2, i32 %m) nounwind { +entry: + %tmp4 = tail call fastcc i32 @g.1478( %struct.FRAME.nest* nest %CHAIN.2, i32 %m ) nounwind ; [#uses=1] + %tmp6 = getelementptr %struct.FRAME.nest* %CHAIN.2, i32 0, i32 0 ; [#uses=1] + %tmp7 = load i32* %tmp6, align 4 ; [#uses=1] + %tmp9 = icmp eq i32 %tmp4, %tmp7 ; [#uses=1] + %tmp910 = zext i1 %tmp9 to i32 ; [#uses=1] + ret i32 %tmp910 +} + +define i32 @nest(i32 %n) nounwind { +entry: + %TRAMP.316 = alloca [10 x i8] ; <[10 x i8]*> [#uses=1] + %FRAME.0 = alloca %struct.FRAME.nest ; <%struct.FRAME.nest*> [#uses=3] + %TRAMP.316.sub = getelementptr [10 x i8]* %TRAMP.316, i32 0, i32 0 ; [#uses=1] + %tmp3 = getelementptr %struct.FRAME.nest* %FRAME.0, i32 0, i32 0 ; [#uses=1] + store i32 %n, i32* %tmp3, align 8 + %FRAME.06 = bitcast %struct.FRAME.nest* %FRAME.0 to i8* ; [#uses=1] + %tramp = call i8* @llvm.init.trampoline( i8* %TRAMP.316.sub, i8* bitcast (i32 (%struct.FRAME.nest*, i32)* @f.1481 to i8*), i8* %FRAME.06 ) ; [#uses=1] + %tmp7 = getelementptr %struct.FRAME.nest* %FRAME.0, i32 0, i32 1 ; [#uses=1] + %tmp89 = bitcast i8* %tramp to i32 (i32)* ; [#uses=2] + store i32 (i32)* %tmp89, i32 (i32)** %tmp7, align 4 + %tmp13 = call i32 @process( i32 (i32)* %tmp89 ) nounwind ; [#uses=1] + ret i32 %tmp13 +} + +declare i8* @llvm.init.trampoline(i8*, i8*, i8*) nounwind + +define i32 @main() nounwind { +entry: + %tmp = tail call i32 @nest( i32 2 ) nounwind ; [#uses=1] + %tmp1 = tail call i32 @nest( i32 1 ) nounwind ; [#uses=1] + %tmp3 = tail call i32 (i8*, ...)* @printf( i8* noalias getelementptr ([7 x i8]* @.str, i32 0, i32 0), i32 %tmp1, i32 %tmp ) nounwind ; [#uses=0] + ret i32 undef +} + +declare i32 @printf(i8*, ...) nounwind