Fix x86 aget-wide compilation to use loadBaseIndexedDisp.

loadPair doesn't work properly when the result register is floating
point. Using loadBaseIndexedDisp fixes the run-tests again.

Change-Id: I06cc455abe230068eea862846b91c09681139c55
diff --git a/src/compiler/codegen/ b/src/compiler/codegen/
index e9c348a..0d2cf00 100644
--- a/src/compiler/codegen/
+++ b/src/compiler/codegen/
@@ -1506,7 +1506,8 @@
     oatFreeTemp(cUnit, rlArray.lowReg);
     oatFreeTemp(cUnit, rlIndex.lowReg);
     rlResult = oatEvalLoc(cUnit, rlDest, regClass, true);
-    loadPair(cUnit, regAddr, rlResult.lowReg, rlResult.highReg);
+    loadBaseIndexedDisp(cUnit, NULL, regAddr, INVALID_REG, 0, 0, rlResult.lowReg,
+                        rlResult.highReg, size, INVALID_SREG);
     storeValueWide(cUnit, rlDest, rlResult);
   } else {
     rlResult = oatEvalLoc(cUnit, rlDest, regClass, true);