[PATCH] [X86] Fix a bug with fetch_add(INT32_MIN)

[JF Bastien]

================
Comment at: lib/Target/X86/X86ISelDAGToDAG.cpp:1717
@@ +1716,3 @@
+    // Quit if INT32_MIN: it would be negated as it is negative and overflow,
+    // producing an immediate too big to be merged into the sub.
+    if (CNVal == std::numeric_limits<int32_t>::min())
----------------
The key thing to understand is that negating `INT32_MIN` makes it not fit in 32-bits. I'm not sure the comment makes that clear.

================
Comment at: lib/Target/X86/X86ISelDAGToDAG.cpp:1718
@@ +1717,3 @@
+    // producing an immediate too big to be merged into the sub.
+    if (CNVal == std::numeric_limits<int32_t>::min())
+      return CurDAG->getTargetConstant(CNVal, NVT);
----------------
It looks like the prevalent thing is to use `INT32_MIN` from `<stdint.h>` instead of `std::numeric_limits`. I know I pointed you at `std::numeric_limits`, but I think we should follow what other parts of LLVM use.

http://reviews.llvm.org/D5655