reimplement __sync_* builtins to be variadic and to follow the same
semantic rules that gcc and icc use.  This implements the variadic
and concrete versions as builtins and has sema do the 
disambiguation.  There are probably a bunch of details to finish up
but this seems like a large monotonic step forward :)


git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@71212 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/test/CodeGen/atomic.c b/test/CodeGen/atomic.c
index d1b3cbb..c224cbd 100644
--- a/test/CodeGen/atomic.c
+++ b/test/CodeGen/atomic.c
@@ -1,15 +1,15 @@
 // RUN: clang-cc %s -emit-llvm -o - > %t1 &&
 // RUN: grep @llvm.atomic.load.add.i32 %t1 | count 3 &&
-// RUN: grep @llvm.atomic.load.sub.i32 %t1 | count 3 &&
+// RUN: grep @llvm.atomic.load.sub.i8 %t1 | count 2 &&
 // RUN: grep @llvm.atomic.load.min.i32 %t1 &&
 // RUN: grep @llvm.atomic.load.max.i32 %t1 &&
 // RUN: grep @llvm.atomic.load.umin.i32 %t1 &&
 // RUN: grep @llvm.atomic.load.umax.i32 %t1 &&
 // RUN: grep @llvm.atomic.swap.i32 %t1 &&
 // RUN: grep @llvm.atomic.cmp.swap.i32 %t1 | count 3 &&
-// RUN: grep @llvm.atomic.load.and.i32 %t1 | count 3 &&
-// RUN: grep @llvm.atomic.load.or.i32 %t1 | count 3 &&
-// RUN: grep @llvm.atomic.load.xor.i32 %t1 | count 3
+// RUN: grep @llvm.atomic.load.and.i32 %t1 | count 2 &&
+// RUN: grep @llvm.atomic.load.or.i8 %t1  &&
+// RUN: grep @llvm.atomic.load.xor.i8 %t1
 
 
 int atomic(void)
@@ -17,11 +17,12 @@
   // nonsenical test for sync functions
   int old;
   int val = 1;
+  char valc = 1;
   unsigned int uval = 1;
   int cmp = 0;
 
   old = __sync_fetch_and_add(&val, 1);
-  old = __sync_fetch_and_sub(&val, 2);
+  old = __sync_fetch_and_sub(&valc, 2);
   old = __sync_fetch_and_min(&val, 3);
   old = __sync_fetch_and_max(&val, 4);
   old = __sync_fetch_and_umin(&uval, 5u);
@@ -35,9 +36,9 @@
 
   old = __sync_add_and_fetch(&val, 1);
   old = __sync_sub_and_fetch(&val, 2);
-  old = __sync_and_and_fetch(&val, 3);
-  old = __sync_or_and_fetch(&val, 4);
-  old = __sync_xor_and_fetch(&val, 5);
+  old = __sync_and_and_fetch(&valc, 3);
+  old = __sync_or_and_fetch(&valc, 4);
+  old = __sync_xor_and_fetch(&valc, 5);
 
   return old;
 }