Implement expansion in type legalization for add/sub with overflow.  The
expansion is the same as that used by LegalizeDAG.

The resulting code sucks in terms of performance/codesize on x86-32 for a
64-bit operation; I haven't looked into whether different expansions might be
better in general.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@105378 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/test/CodeGen/Generic/add-with-overflow-128.ll b/test/CodeGen/Generic/add-with-overflow-128.ll
new file mode 100644
index 0000000..c46c820
--- /dev/null
+++ b/test/CodeGen/Generic/add-with-overflow-128.ll
@@ -0,0 +1,42 @@
+; RUN: llc < %s
+
+@ok = internal constant [4 x i8] c"%d\0A\00"
+@no = internal constant [4 x i8] c"no\0A\00"
+
+define i1 @func1(i128 signext %v1, i128 signext %v2) nounwind {
+entry:
+  %t = call {i128, i1} @llvm.sadd.with.overflow.i128(i128 %v1, i128 %v2)
+  %sum = extractvalue {i128, i1} %t, 0
+  %sum32 = trunc i128 %sum to i32
+  %obit = extractvalue {i128, i1} %t, 1
+  br i1 %obit, label %overflow, label %normal
+
+normal:
+  %t1 = tail call i32 (i8*, ...)* @printf( i8* getelementptr ([4 x i8]* @ok, i32 0, i32 0), i32 %sum32 ) nounwind
+  ret i1 true
+
+overflow:
+  %t2 = tail call i32 (i8*, ...)* @printf( i8* getelementptr ([4 x i8]* @no, i32 0, i32 0) ) nounwind
+  ret i1 false
+}
+
+define i1 @func2(i128 zeroext %v1, i128 zeroext %v2) nounwind {
+entry:
+  %t = call {i128, i1} @llvm.uadd.with.overflow.i128(i128 %v1, i128 %v2)
+  %sum = extractvalue {i128, i1} %t, 0
+  %sum32 = trunc i128 %sum to i32
+  %obit = extractvalue {i128, i1} %t, 1
+  br i1 %obit, label %carry, label %normal
+
+normal:
+  %t1 = tail call i32 (i8*, ...)* @printf( i8* getelementptr ([4 x i8]* @ok, i32 0, i32 0), i32 %sum32 ) nounwind
+  ret i1 true
+
+carry:
+  %t2 = tail call i32 (i8*, ...)* @printf( i8* getelementptr ([4 x i8]* @no, i32 0, i32 0) ) nounwind
+  ret i1 false
+}
+
+declare i32 @printf(i8*, ...) nounwind
+declare {i128, i1} @llvm.sadd.with.overflow.i128(i128, i128)
+declare {i128, i1} @llvm.uadd.with.overflow.i128(i128, i128)