Fix byval arguments in the fastcc calling convention. The fastcc convention

delegates to the regular x86-32 convention which handles byval, but only
after it handles a few cases, and it's necessary to handle byval before
handling those cases. This fixes PR3122 (and rdar://6400815), llvm-gcc
miscompiling LLVM.

llvm-svn: 60453

lib/Target/X86/X86CallingConv.td

@@ -321,6 +321,11 @@ def CC_X86_32_FastCall : CallingConv<[
 ]>;
 
 def CC_X86_32_FastCC : CallingConv<[
+  // Handles byval parameters.  Note that we can't rely on the delegation
+  // to CC_X86_32_Common for this because that happens after code that
+  // handles i32 arguments.
+  CCIfByVal<CCPassByVal<4, 4>>,
+
   // Promote i8/i16 arguments to i32.
   CCIfType<[i8, i16], CCPromoteToType<i32>>,
 

test/CodeGen/X86/fastcc-byval.ll

@@ -0,0 +1,20 @@
+; RUN: llvm-as < %s | llc | grep {movl\[\[:space:\]\]*8(%esp), %eax} | count 2
+; PR3122
+; rdar://6400815
+
+; byval requires a copy, even with fastcc.
+
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
+target triple = "i386-apple-darwin9.5"
+	%struct.MVT = type { i32 }
+
+define fastcc i32 @bar() nounwind {
+	%V = alloca %struct.MVT
+	%a = getelementptr %struct.MVT* %V, i32 0, i32 0
+	store i32 1, i32* %a
+	call fastcc void @foo(%struct.MVT* byval %V) nounwind
+	%t = load i32* %a
+	ret i32 %t
+}
+
+declare fastcc void @foo(%struct.MVT* byval)

test/CodeGen/X86/tailcallbyval.ll

@@ -1,6 +1,5 @@
 ; RUN: llvm-as < %s | llc -march=x86 -tailcallopt | grep TAILCALL
-; check for the 2 byval moves
-; RUN: llvm-as < %s | llc -march=x86 -tailcallopt | grep movl | grep ecx | grep eax | wc -l | grep 1
+; RUN: llvm-as < %s | llc -march=x86 -tailcallopt | grep {movl\[\[:space:\]\]*4(%esp), %eax} | count 1
 %struct.s = type {i32, i32, i32, i32, i32, i32, i32, i32,
                   i32, i32, i32, i32, i32, i32, i32, i32,
                   i32, i32, i32, i32, i32, i32, i32, i32 }