7096366: PPC: corruption of floating-point values with DeoptimizeALot
Summary: fix for a deoptimization found on PPC, which could impact other big endian platforms
Reviewed-by: roland, dholmes
--- a/hotspot/src/share/vm/c1/c1_LinearScan.cpp Tue Oct 04 14:30:04 2011 -0700
+++ b/hotspot/src/share/vm/c1/c1_LinearScan.cpp Fri Oct 07 13:28:44 2011 +0200
@@ -2619,6 +2619,24 @@
Location::Type loc_type = float_saved_as_double ? Location::float_in_dbl : Location::normal;
VMReg rname = frame_map()->fpu_regname(opr->fpu_regnr());
+#ifndef __SOFTFP__
+#ifndef VM_LITTLE_ENDIAN
+ if (! float_saved_as_double) {
+ // On big endian system, we may have an issue if float registers use only
+ // the low half of the (same) double registers.
+ // Both the float and the double could have the same regnr but would correspond
+ // to two different addresses once saved.
+
+ // get next safely (no assertion checks)
+ VMReg next = VMRegImpl::as_VMReg(1+rname->value());
+ if (next->is_reg() &&
+ (next->as_FloatRegister() == rname->as_FloatRegister())) {
+ // the back-end does use the same numbering for the double and the float
+ rname = next; // VMReg for the low bits, e.g. the real VMReg for the float
+ }
+ }
+#endif
+#endif
LocationValue* sv = new LocationValue(Location::new_reg_loc(loc_type, rname));
scope_values->append(sv);