[llvm-commits] [llvm] r45853 - in /llvm/trunk: lib/Transforms/Utils/InlineFunction.cpp test/Transforms/Inline/byval.ll

[Chris Lattner]

Author: lattner
Date: Fri Jan 11 00:09:30 2008
New Revision: 45853

URL: http://llvm.org/viewvc/llvm-project?rev=45853&view=rev
Log:
When inlining a functino with a byval argument, make an explicit 
copy of it in case the callee modifies the struct.

Added:
    llvm/trunk/test/Transforms/Inline/byval.ll
Modified:
    llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp

Modified: llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp?rev=45853&r1=45852&r2=45853&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp Fri Jan 11 00:09:30 2008
@@ -18,7 +18,9 @@
 #include "llvm/Module.h"
 #include "llvm/Instructions.h"
 #include "llvm/Intrinsics.h"
+#include "llvm/ParameterAttributes.h"
 #include "llvm/Analysis/CallGraph.h"
+#include "llvm/Target/TargetData.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/Support/CallSite.h"
 using namespace llvm;
@@ -201,7 +203,6 @@
   BasicBlock *OrigBB = TheCall->getParent();
   Function *Caller = OrigBB->getParent();
 
-  
   // GC poses two hazards to inlining, which only occur when the callee has GC:
   //  1. If the caller has no GC, then the callee's GC must be propagated to the
   //     caller.
@@ -213,7 +214,6 @@
       return false;
   }
   
-  
   // Get an iterator to the last basic block in the function, which will have
   // the new function inlined after it.
   //
@@ -228,15 +228,66 @@
   { // Scope to destroy ValueMap after cloning.
     DenseMap<const Value*, Value*> ValueMap;
 
-    // Calculate the vector of arguments to pass into the function cloner, which
-    // matches up the formal to the actual argument values.
     assert(std::distance(CalledFunc->arg_begin(), CalledFunc->arg_end()) ==
            std::distance(CS.arg_begin(), CS.arg_end()) &&
            "No varargs calls can be inlined!");
+    
+    // Calculate the vector of arguments to pass into the function cloner, which
+    // matches up the formal to the actual argument values.
     CallSite::arg_iterator AI = CS.arg_begin();
+    unsigned ArgNo = 0;
     for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
-           E = CalledFunc->arg_end(); I != E; ++I, ++AI)
-      ValueMap[I] = *AI;
+         E = CalledFunc->arg_end(); I != E; ++I, ++AI, ++ArgNo) {
+      Value *ActualArg = *AI;
+      
+      // When byval arguments actually inlined, we need to make the copy implied
+      // by them actually explicit.
+      // TODO: If we know that the callee never modifies the struct, we can
+      // remove this copy.
+      if (CalledFunc->paramHasAttr(ArgNo+1, ParamAttr::ByVal)) {
+        const Type *AggTy = cast<PointerType>(I->getType())->getElementType();
+        const Type *VoidPtrTy = PointerType::getUnqual(Type::Int8Ty);
+        
+        // Create the alloca.  If we have TargetData, use nice alignment.
+        unsigned Align = 1;
+        if (TD) Align = TD->getPrefTypeAlignment(AggTy);
+        Value *NewAlloca = new AllocaInst(AggTy, 0, Align, I->getName(), 
+                                          Caller->begin()->begin());
+        // Emit a memcpy.
+        Function *MemCpyFn = Intrinsic::getDeclaration(Caller->getParent(),
+                                                       Intrinsic::memcpy_i64);
+        Value *DestCast = new BitCastInst(NewAlloca, VoidPtrTy, "tmp", TheCall);
+        Value *SrcCast = new BitCastInst(*AI, VoidPtrTy, "tmp", TheCall);
+        
+        Value *Size;
+        if (TD == 0)
+          Size = ConstantExpr::getSizeOf(AggTy);
+        else
+          Size = ConstantInt::get(Type::Int64Ty, TD->getTypeStoreSize(AggTy));
+        
+        // Always generate a memcpy of alignment 1 here because we don't know
+        // the alignment of the src pointer.  Other optimizations can infer
+        // better alignment.
+        Value *CallArgs[] = {
+          DestCast, SrcCast, Size, ConstantInt::get(Type::Int32Ty, 1)
+        };
+        CallInst *TheMemCpy =
+          new CallInst(MemCpyFn, CallArgs, CallArgs+4, "", TheCall);
+        
+        // If we have a call graph, update it.
+        if (CG) {
+          CallGraphNode *MemCpyCGN = CG->getOrInsertFunction(MemCpyFn);
+          CallGraphNode *CallerNode = (*CG)[Caller];
+          CallerNode->addCalledFunction(TheMemCpy, MemCpyCGN);
+        }
+        
+        // Uses of the argument in the function should use our new alloca
+        // instead.
+        ActualArg = NewAlloca;
+      }
+      
+      ValueMap[I] = ActualArg;
+    }
 
     // We want the inliner to prune the code as it copies.  We would LOVE to
     // have no dead or constant instructions leftover after inlining occurs

Added: llvm/trunk/test/Transforms/Inline/byval.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/Inline/byval.ll?rev=45853&view=auto

==============================================================================
--- llvm/trunk/test/Transforms/Inline/byval.ll (added)
+++ llvm/trunk/test/Transforms/Inline/byval.ll Fri Jan 11 00:09:30 2008
@@ -0,0 +1,28 @@
+; RUN: llvm-as < %s | opt -inline | llvm-dis | grep {llvm.memcpy}
+
+; Inlining a byval struct should cause an explicit copy into an alloca.
+
+	%struct.ss = type { i32, i64 }
+ at .str = internal constant [10 x i8] c"%d, %lld\0A\00"		; <[10 x i8]*> [#uses=1]
+
+define internal void @f(%struct.ss* byval  %b) nounwind  {
+entry:
+	%tmp = getelementptr %struct.ss* %b, i32 0, i32 0		; <i32*> [#uses=2]
+	%tmp1 = load i32* %tmp, align 4		; <i32> [#uses=1]
+	%tmp2 = add i32 %tmp1, 1		; <i32> [#uses=1]
+	store i32 %tmp2, i32* %tmp, align 4
+	ret void
+}
+
+declare i32 @printf(i8*, ...) nounwind 
+
+define i32 @main() nounwind  {
+entry:
+	%S = alloca %struct.ss		; <%struct.ss*> [#uses=4]
+	%tmp1 = getelementptr %struct.ss* %S, i32 0, i32 0		; <i32*> [#uses=1]
+	store i32 1, i32* %tmp1, align 8
+	%tmp4 = getelementptr %struct.ss* %S, i32 0, i32 1		; <i64*> [#uses=1]
+	store i64 2, i64* %tmp4, align 4
+	call void @f( %struct.ss* byval  %S ) nounwind 
+	ret i32 0
+}