[x32] Fix function indirect calls

[pavel.v.chupin]

Summary

Zero-extend register to 64-bit for callq/jmpq.

[nadav]

LGTM. :)

[pavel.v.chupin]

Sorry, but this patch should be better than previous version. Modified 1 more test to catch ineffectiveness I didn't noticed before.
We need zero-extend only if callee address is in register.
Please review.

[dschuff]

thanks, still LGTM

Closed by commit rL217942 (authored by pvchupin).

lib/Target/X86/X86ISelLowering.cpp

X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   bool IsWin64        = Subtarget->isCallingConvWin64(CallConv);
   StructReturnType SR = callIsStructReturn(Outs);
   bool IsSibcall      = false;
   X86MachineFunctionInfo *X86Info = MF.getInfo<X86MachineFunctionInfo>();
 
   if (MF.getTarget().Options.DisableTailCalls)
     isTailCall = false;
 
+  if (Subtarget->isTarget64BitILP32() && Callee->getValueType(0) == MVT::i32) {
+    // Zero-extend the 32-bit Callee address into a 64-bit according to x32 ABI
+    Callee = DAG.getNode(ISD::ZERO_EXTEND, dl, MVT::i64, Callee);
+  }
+
   bool IsMustTail = CLI.CS && CLI.CS->isMustTailCall();
   if (IsMustTail) {
     // Force this to be a tail call.  The verifier rules are enough to ensure
     // that we can lower this successfully without moving the return address
     // around.
     isTailCall = true;
   } else if (isTailCall) {
     // Check if it's really possible to do a tail call.

test/CodeGen/X86/x32-function_pointer-1.ll

+; RUN: llc < %s -mtriple=x86_64-linux-gnux32  | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-linux-gnux32 -fast-isel | FileCheck %s
+
+; Test for x32 function pointer tail call
+
+@foo1 = external global void (i8*)*
+@foo2 = external global void (i8*)*
+
+define void @bar(i8* %h) nounwind uwtable {
+entry:
+  %0 = load void (i8*)** @foo1, align 4
+; CHECK: movl foo1(%rip), %e{{[^,]*}}
+  tail call void %0(i8* %h) nounwind
+; CHECK: callq  *%r{{[^,]*}}
+  %1 = load void (i8*)** @foo2, align 4
+; CHECK: movl foo2(%rip), %e{{[^,]*}}
+  tail call void %1(i8* %h) nounwind
+; CHECK: jmpq *%r{{[^,]*}}
+  ret void
+}

test/CodeGen/X86/x32-function_pointer-2.ll

+; RUN: llc < %s -mtriple=x86_64-linux-gnux32 | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-linux-gnux32 -fast-isel | FileCheck %s
+
+; Test call function pointer with function argument
+;
+; void bar (void * h, void (*foo) (void *))
+;    {
+;      foo (h);
+;      foo (h);
+;    }
+
+
+define void @bar(i8* %h, void (i8*)* nocapture %foo) nounwind {
+entry:
+  tail call void %foo(i8* %h) nounwind
+; CHECK: mov{{l|q}} %{{e|r}}si, %{{e|r}}[[REG:.*]]{{d?}}
+; CHECK: callq  *%r[[REG]]
+  tail call void %foo(i8* %h) nounwind
+; CHECK: jmpq *%r{{[^,]*}}
+  ret void
+}

test/CodeGen/X86/x32-function_pointer-3.ll

+; RUN: llc < %s -mtriple=x86_64-linux-gnux32 | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-linux-gnux32 -fast-isel | FileCheck %s
+
+; Test calling function pointer passed in struct
+
+;    The fuction argument `h' in
+
+;    struct foo {
+;      void (*f) (void);
+;      int i;
+;    };
+;    void
+;    bar (struct foo h)
+;    {
+;      h.f ();
+;    }
+
+;    is passed in the 64-bit %rdi register.  The `f' field is in the lower 32
+;    bits of %rdi register and the `i' field is in the upper 32 bits of %rdi
+;    register.  We need to zero-extend %edi to %rdi before branching via %rdi.
+
+define void @bar(i64 %h.coerce) nounwind {
+entry:
+  %h.sroa.0.0.extract.trunc = trunc i64 %h.coerce to i32
+  %0 = inttoptr i32 %h.sroa.0.0.extract.trunc to void ()*
+; CHECK: movl %edi, %e[[REG:.*]]
+  tail call void %0() nounwind
+; CHECK: jmpq *%r[[REG]]
+  ret void
+}