Merge

make/hotspot_version

@@ -35,7 +35,7 @@ HOTSPOT_VM_COPYRIGHT=Copyright 2014
 
 HS_MAJOR_VER=25
 HS_MINOR_VER=20
-HS_BUILD_NUMBER=14
+HS_BUILD_NUMBER=15
 
 JDK_MAJOR_VER=1
 JDK_MINOR_VER=8

src/cpu/x86/vm/macroAssembler_x86.cpp

@@ -3152,10 +3152,12 @@ void MacroAssembler::fast_pow() {
   // if fast computation is not possible, result is NaN. Requires
   // fallback from user of this macro.
   // increase precision for intermediate steps of the computation
+  BLOCK_COMMENT("fast_pow {");
   increase_precision();
   fyl2x();                 // Stack: (Y*log2(X)) ...
   pow_exp_core_encoding(); // Stack: exp(X) ...
   restore_precision();
+  BLOCK_COMMENT("} fast_pow");
 }
 
 void MacroAssembler::fast_exp() {

src/os/bsd/vm/os_bsd.cpp

@@ -127,8 +127,12 @@
 // global variables
 julong os::Bsd::_physical_memory = 0;
 
-
+#ifdef __APPLE__
+mach_timebase_info_data_t os::Bsd::_timebase_info = {0, 0};
+volatile uint64_t         os::Bsd::_max_abstime   = 0;
+#else
 int (*os::Bsd::_clock_gettime)(clockid_t, struct timespec *) = NULL;
+#endif
 pthread_t os::Bsd::_main_thread;
 int os::Bsd::_page_size = -1;
 
@@ -986,13 +990,15 @@ jlong os::javaTimeMillis() {
   return jlong(time.tv_sec) * 1000  +  jlong(time.tv_usec / 1000);
 }
 
+#ifndef __APPLE__
 #ifndef CLOCK_MONOTONIC
 #define CLOCK_MONOTONIC (1)
 #endif
+#endif
 
 #ifdef __APPLE__
 void os::Bsd::clock_init() {
-        // XXXDARWIN: Investigate replacement monotonic clock
+  mach_timebase_info(&_timebase_info);
 }
 #else
 void os::Bsd::clock_init() {
@@ -1007,10 +1013,38 @@ void os::Bsd::clock_init() {
 #endif
 
 
+#ifdef __APPLE__
+
+jlong os::javaTimeNanos() {
+    const uint64_t tm = mach_absolute_time();
+    const uint64_t now = (tm * Bsd::_timebase_info.numer) / Bsd::_timebase_info.denom;
+    const uint64_t prev = Bsd::_max_abstime;
+    if (now <= prev) {
+      return prev;   // same or retrograde time;
+    }
+    const uint64_t obsv = Atomic::cmpxchg(now, (volatile jlong*)&Bsd::_max_abstime, prev);
+    assert(obsv >= prev, "invariant");   // Monotonicity
+    // If the CAS succeeded then we're done and return "now".
+    // If the CAS failed and the observed value "obsv" is >= now then
+    // we should return "obsv".  If the CAS failed and now > obsv > prv then
+    // some other thread raced this thread and installed a new value, in which case
+    // we could either (a) retry the entire operation, (b) retry trying to install now
+    // or (c) just return obsv.  We use (c).   No loop is required although in some cases
+    // we might discard a higher "now" value in deference to a slightly lower but freshly
+    // installed obsv value.   That's entirely benign -- it admits no new orderings compared
+    // to (a) or (b) -- and greatly reduces coherence traffic.
+    // We might also condition (c) on the magnitude of the delta between obsv and now.
+    // Avoiding excessive CAS operations to hot RW locations is critical.
+    // See https://blogs.oracle.com/dave/entry/cas_and_cache_trivia_invalidate
+    return (prev == obsv) ? now : obsv;
+}
+
+#else // __APPLE__
+
 jlong os::javaTimeNanos() {
   if (Bsd::supports_monotonic_clock()) {
     struct timespec tp;
-    int status = Bsd::clock_gettime(CLOCK_MONOTONIC, &tp);
+    int status = Bsd::_clock_gettime(CLOCK_MONOTONIC, &tp);
     assert(status == 0, "gettime error");
     jlong result = jlong(tp.tv_sec) * (1000 * 1000 * 1000) + jlong(tp.tv_nsec);
     return result;
@@ -1023,6 +1057,8 @@ jlong os::javaTimeNanos() {
   }
 }
 
+#endif // __APPLE__
+
 void os::javaTimeNanos_info(jvmtiTimerInfo *info_ptr) {
   if (Bsd::supports_monotonic_clock()) {
     info_ptr->max_value = ALL_64_BITS;

src/os/bsd/vm/os_bsd.hpp

@@ -58,7 +58,13 @@ class Bsd {
   // For signal flags diagnostics
   static int sigflags[MAXSIGNUM];
 
+#ifdef __APPLE__
+  // mach_absolute_time
+  static mach_timebase_info_data_t _timebase_info;
+  static volatile uint64_t         _max_abstime;
+#else
   static int (*_clock_gettime)(clockid_t, struct timespec *);
+#endif
 
   static GrowableArray<int>* _cpu_to_node;
 
@@ -135,11 +141,11 @@ class Bsd {
   static void clock_init(void);
 
   static inline bool supports_monotonic_clock() {
+#ifdef __APPLE__
+    return true;
+#else
     return _clock_gettime != NULL;
-  }
-
-  static int clock_gettime(clockid_t clock_id, struct timespec *tp) {
-    return _clock_gettime ? _clock_gettime(clock_id, tp) : -1;
+#endif
   }
 
   // Stack repair handling

src/os/solaris/vm/os_solaris.cpp

@@ -415,11 +415,7 @@ julong os::physical_memory() {
 
 static hrtime_t first_hrtime = 0;
 static const hrtime_t hrtime_hz = 1000*1000*1000;
-const int LOCK_BUSY = 1;
-const int LOCK_FREE = 0;
-const int LOCK_INVALID = -1;
 static volatile hrtime_t max_hrtime = 0;
-static volatile int max_hrtime_lock = LOCK_FREE;     // Update counter with LSB as lock-in-progress
 
 
 void os::Solaris::initialize_system_info() {
@@ -1534,58 +1530,31 @@ void* os::thread_local_storage_at(int index) {
 }
 
 
-// gethrtime can move backwards if read from one cpu and then a different cpu
-// getTimeNanos is guaranteed to not move backward on Solaris
-// local spinloop created as faster for a CAS on an int than
-// a CAS on a 64bit jlong. Also Atomic::cmpxchg for jlong is not
-// supported on sparc v8 or pre supports_cx8 intel boxes.
-// oldgetTimeNanos for systems which do not support CAS on 64bit jlong
-// i.e. sparc v8 and pre supports_cx8 (i486) intel boxes
-inline hrtime_t oldgetTimeNanos() {
-  int gotlock = LOCK_INVALID;
-  hrtime_t newtime = gethrtime();
-
-  for (;;) {
-// grab lock for max_hrtime
-    int curlock = max_hrtime_lock;
-    if (curlock & LOCK_BUSY)  continue;
-    if (gotlock = Atomic::cmpxchg(LOCK_BUSY, &max_hrtime_lock, LOCK_FREE) != LOCK_FREE) continue;
-    if (newtime > max_hrtime) {
-      max_hrtime = newtime;
-    } else {
-      newtime = max_hrtime;
-    }
-    // release lock
-    max_hrtime_lock = LOCK_FREE;
-    return newtime;
-  }
-}
-// gethrtime can move backwards if read from one cpu and then a different cpu
-// getTimeNanos is guaranteed to not move backward on Solaris
+// gethrtime() should be monotonic according to the documentation,
+// but some virtualized platforms are known to break this guarantee.
+// getTimeNanos() must be guaranteed not to move backwards, so we
+// are forced to add a check here.
 inline hrtime_t getTimeNanos() {
-  if (VM_Version::supports_cx8()) {
-    const hrtime_t now = gethrtime();
-    // Use atomic long load since 32-bit x86 uses 2 registers to keep long.
-    const hrtime_t prev = Atomic::load((volatile jlong*)&max_hrtime);
-    if (now <= prev)  return prev;   // same or retrograde time;
-    const hrtime_t obsv = Atomic::cmpxchg(now, (volatile jlong*)&max_hrtime, prev);
-    assert(obsv >= prev, "invariant");   // Monotonicity
-    // If the CAS succeeded then we're done and return "now".
-    // If the CAS failed and the observed value "obs" is >= now then
-    // we should return "obs".  If the CAS failed and now > obs > prv then
-    // some other thread raced this thread and installed a new value, in which case
-    // we could either (a) retry the entire operation, (b) retry trying to install now
-    // or (c) just return obs.  We use (c).   No loop is required although in some cases
-    // we might discard a higher "now" value in deference to a slightly lower but freshly
-    // installed obs value.   That's entirely benign -- it admits no new orderings compared
-    // to (a) or (b) -- and greatly reduces coherence traffic.
-    // We might also condition (c) on the magnitude of the delta between obs and now.
-    // Avoiding excessive CAS operations to hot RW locations is critical.
-    // See http://blogs.sun.com/dave/entry/cas_and_cache_trivia_invalidate
-    return (prev == obsv) ? now : obsv ;
-  } else {
-    return oldgetTimeNanos();
-  }
+  const hrtime_t now = gethrtime();
+  const hrtime_t prev = max_hrtime;
+  if (now <= prev) {
+    return prev;   // same or retrograde time;
+  }
+  const hrtime_t obsv = Atomic::cmpxchg(now, (volatile jlong*)&max_hrtime, prev);
+  assert(obsv >= prev, "invariant");   // Monotonicity
+  // If the CAS succeeded then we're done and return "now".
+  // If the CAS failed and the observed value "obsv" is >= now then
+  // we should return "obsv".  If the CAS failed and now > obsv > prv then
+  // some other thread raced this thread and installed a new value, in which case
+  // we could either (a) retry the entire operation, (b) retry trying to install now
+  // or (c) just return obsv.  We use (c).   No loop is required although in some cases
+  // we might discard a higher "now" value in deference to a slightly lower but freshly
+  // installed obsv value.   That's entirely benign -- it admits no new orderings compared
+  // to (a) or (b) -- and greatly reduces coherence traffic.
+  // We might also condition (c) on the magnitude of the delta between obsv and now.
+  // Avoiding excessive CAS operations to hot RW locations is critical.
+  // See https://blogs.oracle.com/dave/entry/cas_and_cache_trivia_invalidate
+  return (prev == obsv) ? now : obsv;
 }
 
 // Time since start-up in seconds to a fine granularity.

src/share/vm/c1/c1_GraphBuilder.cpp

@@ -1697,6 +1697,15 @@ Values* GraphBuilder::args_list_for_profiling(ciMethod* target, int& start, bool
   return NULL;
 }
 
+void GraphBuilder::check_args_for_profiling(Values* obj_args, int expected) {
+#ifdef ASSERT
+  bool ignored_will_link;
+  ciSignature* declared_signature = NULL;
+  ciMethod* real_target = method()->get_method_at_bci(bci(), ignored_will_link, &declared_signature);
+  assert(expected == obj_args->length() || real_target->is_method_handle_intrinsic(), "missed on arg?");
+#endif
+}
+
 // Collect arguments that we want to profile in a list
 Values* GraphBuilder::collect_args_for_profiling(Values* args, ciMethod* target, bool may_have_receiver) {
   int start = 0;
@@ -1705,13 +1714,14 @@ Values* GraphBuilder::collect_args_for_profiling(Values* args, ciMethod* target,
     return NULL;
   }
   int s = obj_args->size();
-  for (int i = start, j = 0; j < s; i++) {
+  // if called through method handle invoke, some arguments may have been popped
+  for (int i = start, j = 0; j < s && i < args->length(); i++) {
     if (args->at(i)->type()->is_object_kind()) {
       obj_args->push(args->at(i));
       j++;
     }
   }
-  assert(s == obj_args->length(), "missed on arg?");
+  check_args_for_profiling(obj_args, s);
   return obj_args;
 }
 
@@ -3843,14 +3853,7 @@ bool GraphBuilder::try_inline_full(ciMethod* callee, bool holder_known, Bytecode
             j++;
           }
         }
-#ifdef ASSERT
-        {
-          bool ignored_will_link;
-          ciSignature* declared_signature = NULL;
-          ciMethod* real_target = method()->get_method_at_bci(bci(), ignored_will_link, &declared_signature);
-          assert(s == obj_args->length() || real_target->is_method_handle_intrinsic(), "missed on arg?");
-        }
-#endif
+        check_args_for_profiling(obj_args, s);
       }
       profile_call(callee, recv, holder_known ? callee->holder() : NULL, obj_args, true);
     }

src/share/vm/c1/c1_GraphBuilder.hpp

@@ -392,6 +392,7 @@ class GraphBuilder VALUE_OBJ_CLASS_SPEC {
 
   Values* args_list_for_profiling(ciMethod* target, int& start, bool may_have_receiver);
   Values* collect_args_for_profiling(Values* args, ciMethod* target, bool may_have_receiver);
+  void check_args_for_profiling(Values* obj_args, int expected);
 
  public:
   NOT_PRODUCT(void print_stats();)

src/share/vm/c1/c1_LIRGenerator.cpp

@@ -2634,8 +2634,10 @@ ciKlass* LIRGenerator::profile_type(ciMethodData* md, int md_base_offset, int md
       // LIR_Assembler::emit_profile_type() from emitting useless code
       profiled_k = ciTypeEntries::with_status(result, profiled_k);
     }
-    if (exact_signature_k != NULL && exact_klass != exact_signature_k) {
-      assert(exact_klass == NULL, "obj and signature disagree?");
+    // exact_klass and exact_signature_k can be both non NULL but
+    // different if exact_klass is loaded after the ciObject for
+    // exact_signature_k is created.
+    if (exact_klass == NULL && exact_signature_k != NULL && exact_klass != exact_signature_k) {
       // sometimes the type of the signature is better than the best type
       // the compiler has
       exact_klass = exact_signature_k;
@@ -2646,8 +2648,7 @@ ciKlass* LIRGenerator::profile_type(ciMethodData* md, int md_base_offset, int md
       if (improved_klass == NULL) {
         improved_klass = comp->cha_exact_type(callee_signature_k);
       }
-      if (improved_klass != NULL && exact_klass != improved_klass) {
-        assert(exact_klass == NULL, "obj and signature disagree?");
+      if (exact_klass == NULL && improved_klass != NULL && exact_klass != improved_klass) {
         exact_klass = exact_signature_k;
       }
     }

src/share/vm/opto/library_call.cpp

@@ -216,7 +216,7 @@ class LibraryCallKit : public GraphKit {
   bool inline_math_subtractExactL(bool is_decrement);
   bool inline_exp();
   bool inline_pow();
-  void finish_pow_exp(Node* result, Node* x, Node* y, const TypeFunc* call_type, address funcAddr, const char* funcName);
+  Node* finish_pow_exp(Node* result, Node* x, Node* y, const TypeFunc* call_type, address funcAddr, const char* funcName);
   bool inline_min_max(vmIntrinsics::ID id);
   Node* generate_min_max(vmIntrinsics::ID id, Node* x, Node* y);
   // This returns Type::AnyPtr, RawPtr, or OopPtr.
@@ -1678,7 +1678,7 @@ bool LibraryCallKit::inline_trig(vmIntrinsics::ID id) {
   return true;
 }
 
-void LibraryCallKit::finish_pow_exp(Node* result, Node* x, Node* y, const TypeFunc* call_type, address funcAddr, const char* funcName) {
+Node* LibraryCallKit::finish_pow_exp(Node* result, Node* x, Node* y, const TypeFunc* call_type, address funcAddr, const char* funcName) {
   //-------------------
   //result=(result.isNaN())? funcAddr():result;
   // Check: If isNaN() by checking result!=result? then either trap
@@ -1694,7 +1694,7 @@ void LibraryCallKit::finish_pow_exp(Node* result, Node* x, Node* y, const TypeFu
       uncommon_trap(Deoptimization::Reason_intrinsic,
                     Deoptimization::Action_make_not_entrant);
     }
-    set_result(result);
+    return result;
   } else {
     // If this inlining ever returned NaN in the past, we compile a call
     // to the runtime to properly handle corner cases
@@ -1724,9 +1724,10 @@ void LibraryCallKit::finish_pow_exp(Node* result, Node* x, Node* y, const TypeFu
 
       result_region->init_req(2, control());
       result_val->init_req(2, value);
-      set_result(result_region, result_val);
+      set_control(_gvn.transform(result_region));
+      return _gvn.transform(result_val);
     } else {
-      set_result(result);
+      return result;
     }
   }
 }
@@ -1738,7 +1739,8 @@ bool LibraryCallKit::inline_exp() {
   Node* arg = round_double_node(argument(0));
   Node* n   = _gvn.transform(new (C) ExpDNode(C, control(), arg));
 
-  finish_pow_exp(n, arg, NULL, OptoRuntime::Math_D_D_Type(), CAST_FROM_FN_PTR(address, SharedRuntime::dexp), "EXP");
+  n = finish_pow_exp(n, arg, NULL, OptoRuntime::Math_D_D_Type(), CAST_FROM_FN_PTR(address, SharedRuntime::dexp), "EXP");
+  set_result(n);
 
   C->set_has_split_ifs(true); // Has chance for split-if optimization
   return true;
@@ -1748,27 +1750,48 @@ bool LibraryCallKit::inline_exp() {
 // Inline power instructions, if possible.
 bool LibraryCallKit::inline_pow() {
   // Pseudocode for pow
-  // if (x <= 0.0) {
-  //   long longy = (long)y;
-  //   if ((double)longy == y) { // if y is long
-  //     if (y + 1 == y) longy = 0; // huge number: even
-  //     result = ((1&longy) == 0)?-DPow(abs(x), y):DPow(abs(x), y);
+  // if (y == 2) {
+  //   return x * x;
+  // } else {
+  //   if (x <= 0.0) {
+  //     long longy = (long)y;
+  //     if ((double)longy == y) { // if y is long
+  //       if (y + 1 == y) longy = 0; // huge number: even
+  //       result = ((1&longy) == 0)?-DPow(abs(x), y):DPow(abs(x), y);
+  //     } else {
+  //       result = NaN;
+  //     }
   //   } else {
-  //     result = NaN;
+  //     result = DPow(x,y);
   //   }
-  // } else {
-  //   result = DPow(x,y);
-  // }
-  // if (result != result)?  {
-  //   result = uncommon_trap() or runtime_call();
+  //   if (result != result)?  {
+  //     result = uncommon_trap() or runtime_call();
+  //   }
+  //   return result;
   // }
-  // return result;
 
   Node* x = round_double_node(argument(0));
   Node* y = round_double_node(argument(2));
 
   Node* result = NULL;
 
+  Node*   const_two_node = makecon(TypeD::make(2.0));
+  Node*   cmp_node       = _gvn.transform(new (C) CmpDNode(y, const_two_node));
+  Node*   bool_node      = _gvn.transform(new (C) BoolNode(cmp_node, BoolTest::eq));
+  IfNode* if_node        = create_and_xform_if(control(), bool_node, PROB_STATIC_INFREQUENT, COUNT_UNKNOWN);
+  Node*   if_true        = _gvn.transform(new (C) IfTrueNode(if_node));
+  Node*   if_false       = _gvn.transform(new (C) IfFalseNode(if_node));
+
+  RegionNode* region_node = new (C) RegionNode(3);
+  region_node->init_req(1, if_true);
+
+  Node* phi_node = new (C) PhiNode(region_node, Type::DOUBLE);
+  // special case for x^y where y == 2, we can convert it to x * x
+  phi_node->init_req(1, _gvn.transform(new (C) MulDNode(x, x)));
+
+  // set control to if_false since we will now process the false branch
+  set_control(if_false);
+
   if (!too_many_traps(Deoptimization::Reason_intrinsic)) {
     // Short form: skip the fancy tests and just check for NaN result.
     result = _gvn.transform(new (C) PowDNode(C, control(), x, y));
@@ -1892,7 +1915,15 @@ bool LibraryCallKit::inline_pow() {
     result = _gvn.transform(phi);
   }
 
-  finish_pow_exp(result, x, y, OptoRuntime::Math_DD_D_Type(), CAST_FROM_FN_PTR(address, SharedRuntime::dpow), "POW");
+  result = finish_pow_exp(result, x, y, OptoRuntime::Math_DD_D_Type(), CAST_FROM_FN_PTR(address, SharedRuntime::dpow), "POW");
+
+  // control from finish_pow_exp is now input to the region node
+  region_node->set_req(2, control());
+  // the result from finish_pow_exp is now input to the phi node
+  phi_node->init_req(2, result);
+  set_control(_gvn.transform(region_node));
+  record_for_igvn(region_node);
+  set_result(_gvn.transform(phi_node));
 
   C->set_has_split_ifs(true); // Has chance for split-if optimization
   return true;

src/share/vm/runtime/os.hpp

@@ -48,6 +48,9 @@
 #ifdef TARGET_OS_FAMILY_bsd
 # include "jvm_bsd.h"
 # include <setjmp.h>
+# ifdef __APPLE__
+#  include <mach/mach_time.h>
+# endif
 #endif
 
 class AgentLibrary;

test/compiler/profiling/TestMethodHandleInvokesIntrinsic.java

+/*
+ * Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test
+ * @bug 8041458
+ * @summary profiling of arguments in C1 at MethodHandle invoke of intrinsic tries to profile popped argument.
+ * @run main/othervm -XX:-BackgroundCompilation -XX:-UseOnStackReplacement -XX:TieredStopAtLevel=3 TestMethodHandleInvokesIntrinsic
+ *
+ */
+
+import java.lang.invoke.*;
+
+public class TestMethodHandleInvokesIntrinsic {
+
+    static final MethodHandle mh_nanoTime;
+    static final MethodHandle mh_getClass;
+    static {
+        MethodHandles.Lookup lookup = MethodHandles.lookup();
+        MethodType mt = MethodType.methodType(long.class);
+        MethodHandle MH = null;
+        try {
+            MH = lookup.findStatic(System.class, "nanoTime", mt);
+        } catch(NoSuchMethodException nsme) {
+            nsme.printStackTrace();
+            throw new RuntimeException("TEST FAILED", nsme);
+        } catch(IllegalAccessException iae) {
+            iae.printStackTrace();
+            throw new RuntimeException("TEST FAILED", iae);
+        }
+        mh_nanoTime = MH;
+
+        mt = MethodType.methodType(Class.class);
+        MH = null;
+        try {
+            MH = lookup.findVirtual(Object.class, "getClass", mt);
+        } catch(NoSuchMethodException nsme) {
+            nsme.printStackTrace();
+            throw new RuntimeException("TEST FAILED", nsme);
+        } catch(IllegalAccessException iae) {
+            iae.printStackTrace();
+            throw new RuntimeException("TEST FAILED", iae);
+        }
+        mh_getClass = MH;
+    }
+
+    static long m1() throws Throwable {
+        return (long)mh_nanoTime.invokeExact();
+    }
+
+    static Class m2(Object o) throws Throwable {
+        return (Class)mh_getClass.invokeExact(o);
+    }
+
+    static public void main(String[] args) {
+        try {
+            for (int i = 0; i < 20000; i++) {
+                m1();
+            }
+            TestMethodHandleInvokesIntrinsic o = new TestMethodHandleInvokesIntrinsic();
+            for (int i = 0; i < 20000; i++) {
+                m2(o);
+            }
+        } catch(Throwable t) {
+            System.out.println("Unexpected exception");
+            t.printStackTrace();
+            throw new RuntimeException("TEST FAILED", t);
+        }
+
+        System.out.println("TEST PASSED");
+    }
+}