OpenJDK / bsd-port / bsd-port / hotspot
changeset 5228:49f4a81eabd5
Merge
| author | asaha |
|---|---|
| date | Mon, 03 Mar 2014 11:56:00 -0800 |
| parents | beeaee2df810 22cae361773d |
| children | ae99b9328cf3 |
| files | |
| diffstat | 23 files changed, 510 insertions(+), 87 deletions(-) [+] |
line wrap: on
line diff
--- a/make/bsd/makefiles/buildtree.make Thu Feb 27 08:30:16 2014 -0800 +++ b/make/bsd/makefiles/buildtree.make Mon Mar 03 11:56:00 2014 -0800 @@ -1,5 +1,6 @@ # # Copyright (c) 2005, 2011, Oracle and/or its affiliates. All rights reserved. +# Copyright (c) 2013 Red Hat, Inc. # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. # # This code is free software; you can redistribute it and/or modify it @@ -59,7 +60,7 @@ # needs to be set here since this Makefile doesn't include defs.make OS_VENDOR:=$(shell uname -s) --include $(SPEC) +include $(GAMMADIR)/make/defs.make include $(GAMMADIR)/make/scm.make include $(GAMMADIR)/make/altsrc.make @@ -215,6 +216,7 @@ echo "HOTSPOT_BUILD_USER = $(HOTSPOT_BUILD_USER)"; \ echo "HOTSPOT_VM_DISTRO = $(HOTSPOT_VM_DISTRO)"; \ echo "OPENJDK = $(OPENJDK)"; \ + echo "ZERO_BUILD = $(ZERO_BUILD)"; \ echo; \ echo "# Used for platform dispatching"; \ echo "TARGET_DEFINES = -DTARGET_OS_FAMILY_\$$(Platform_os_family)"; \
--- a/make/bsd/makefiles/gcc.make Thu Feb 27 08:30:16 2014 -0800 +++ b/make/bsd/makefiles/gcc.make Mon Mar 03 11:56:00 2014 -0800 @@ -1,5 +1,6 @@ # # Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved. +# Copyright (c) 2013 Red Hat, Inc. # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. # # This code is free software; you can redistribute it and/or modify it @@ -105,10 +106,10 @@ VM_PICFLAG/AOUT = VM_PICFLAG = $(VM_PICFLAG/$(LINK_INTO)) -ifeq ($(JVM_VARIANT_ZERO), true) +ifeq ($(TYPE),ZERO) CFLAGS += $(LIBFFI_CFLAGS) endif -ifeq ($(JVM_VARIANT_ZEROSHARK), true) +ifeq ($(TYPE),SHARK) CFLAGS += $(LIBFFI_CFLAGS) CFLAGS += $(LLVM_CFLAGS) endif
--- a/make/bsd/makefiles/vm.make Thu Feb 27 08:30:16 2014 -0800 +++ b/make/bsd/makefiles/vm.make Mon Mar 03 11:56:00 2014 -0800 @@ -1,5 +1,6 @@ # # Copyright (c) 1999, 2012, Oracle and/or its affiliates. All rights reserved. +# Copyright (c) 2013 Red Hat, Inc. # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. # # This code is free software; you can redistribute it and/or modify it @@ -42,7 +43,9 @@ -include $(DEP_DIR)/*.d # read machine-specific adjustments (%%% should do this via buildtree.make?) -ifeq ($(findstring true, $(JVM_VARIANT_ZERO) $(JVM_VARIANT_ZEROSHARK)), true) +ifeq ($(TYPE),ZERO) + include $(MAKEFILES_DIR)/zeroshark.make +else ifeq ($(TYPE),SHARK) include $(MAKEFILES_DIR)/zeroshark.make else include $(MAKEFILES_DIR)/$(BUILDARCH).make @@ -271,10 +274,10 @@ LIBS_VM += $(LIBS) endif -ifeq ($(JVM_VARIANT_ZERO), true) +ifeq ($(TYPE),ZERO) LIBS_VM += $(LIBFFI_LIBS) endif -ifeq ($(JVM_VARIANT_ZEROSHARK), true) +ifeq ($(TYPE),SHARK) LIBS_VM += $(LIBFFI_LIBS) $(LLVM_LIBS) LFLAGS_VM += $(LLVM_LDFLAGS) endif
--- a/make/linux/makefiles/buildtree.make Thu Feb 27 08:30:16 2014 -0800 +++ b/make/linux/makefiles/buildtree.make Mon Mar 03 11:56:00 2014 -0800 @@ -1,5 +1,6 @@ # # Copyright (c) 2005, 2012, Oracle and/or its affiliates. All rights reserved. +# Copyright (c) 2013 Red Hat, Inc. # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. # # This code is free software; you can redistribute it and/or modify it @@ -56,7 +57,7 @@ # The makefiles are split this way so that "make foo" will run faster by not # having to read the dependency files for the vm. --include $(SPEC) +include $(GAMMADIR)/make/defs.make include $(GAMMADIR)/make/scm.make include $(GAMMADIR)/make/altsrc.make @@ -208,6 +209,7 @@ echo "HOTSPOT_BUILD_USER = $(HOTSPOT_BUILD_USER)"; \ echo "HOTSPOT_VM_DISTRO = $(HOTSPOT_VM_DISTRO)"; \ echo "OPENJDK = $(OPENJDK)"; \ + echo "ZERO_BUILD = $(ZERO_BUILD)"; \ echo; \ echo "# Used for platform dispatching"; \ echo "TARGET_DEFINES = -DTARGET_OS_FAMILY_\$$(Platform_os_family)"; \
--- a/make/linux/makefiles/gcc.make Thu Feb 27 08:30:16 2014 -0800 +++ b/make/linux/makefiles/gcc.make Mon Mar 03 11:56:00 2014 -0800 @@ -1,5 +1,6 @@ # # Copyright (c) 1999, 2012, Oracle and/or its affiliates. All rights reserved. +# Copyright (c) 2013 Red Hat, Inc. # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. # # This code is free software; you can redistribute it and/or modify it @@ -72,10 +73,10 @@ VM_PICFLAG/AOUT = VM_PICFLAG = $(VM_PICFLAG/$(LINK_INTO)) -ifeq ($(JVM_VARIANT_ZERO), true) +ifeq ($(TYPE),ZERO) CFLAGS += $(LIBFFI_CFLAGS) endif -ifeq ($(JVM_VARIANT_ZEROSHARK), true) +ifeq ($(TYPE),SHARK) CFLAGS += $(LIBFFI_CFLAGS) CFLAGS += $(LLVM_CFLAGS) endif
--- a/make/linux/makefiles/vm.make Thu Feb 27 08:30:16 2014 -0800 +++ b/make/linux/makefiles/vm.make Mon Mar 03 11:56:00 2014 -0800 @@ -1,5 +1,6 @@ # # Copyright (c) 1999, 2012, Oracle and/or its affiliates. All rights reserved. +# Copyright (c) 2013 Red Hat, Inc. # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. # # This code is free software; you can redistribute it and/or modify it @@ -42,7 +43,9 @@ -include $(DEP_DIR)/*.d # read machine-specific adjustments (%%% should do this via buildtree.make?) -ifeq ($(findstring true, $(JVM_VARIANT_ZERO) $(JVM_VARIANT_ZEROSHARK)), true) +ifeq ($(TYPE),ZERO) + include $(MAKEFILES_DIR)/zeroshark.make +else ifeq ($(TYPE),SHARK) include $(MAKEFILES_DIR)/zeroshark.make else include $(MAKEFILES_DIR)/$(BUILDARCH).make @@ -244,7 +247,7 @@ vm.def: $(Res_Files) $(Obj_Files) sh $(GAMMADIR)/make/linux/makefiles/build_vm_def.sh *.o > $@ -ifeq ($(JVM_VARIANT_ZEROSHARK), true) +ifeq ($(TYPE),SHARK) STATIC_CXX = false else ifeq ($(ZERO_LIBARCH), ppc64) @@ -276,10 +279,10 @@ LIBS_VM += $(LIBS) endif -ifeq ($(JVM_VARIANT_ZERO), true) +ifeq ($(TYPE),ZERO) LIBS_VM += $(LIBFFI_LIBS) endif -ifeq ($(JVM_VARIANT_ZEROSHARK), true) +ifeq ($(TYPE),SHARK) LIBS_VM += $(LIBFFI_LIBS) $(LLVM_LIBS) LFLAGS_VM += $(LLVM_LDFLAGS) endif
--- a/src/cpu/sparc/vm/sparc.ad Thu Feb 27 08:30:16 2014 -0800 +++ b/src/cpu/sparc/vm/sparc.ad Mon Mar 03 11:56:00 2014 -0800 @@ -3399,8 +3399,8 @@ interface(CONST_INTER); %} -// Unsigned (positive) Integer Immediate: 13-bit -operand immU13() %{ +// Unsigned Integer Immediate: 12-bit (non-negative that fits in simm13) +operand immU12() %{ predicate((0 <= n->get_int()) && Assembler::is_simm13(n->get_int())); match(ConI); op_cost(0); @@ -3436,6 +3436,17 @@ interface(CONST_INTER); %} +// Int Immediate non-negative +operand immU31() +%{ + predicate(n->get_int() >= 0); + match(ConI); + + op_cost(0); + format %{ %} + interface(CONST_INTER); +%} + // Integer Immediate: 0-bit operand immI0() %{ predicate(n->get_int() == 0); @@ -5732,7 +5743,6 @@ effect(TEMP dst, TEMP tmp); ins_cost(MEMORY_REF_COST + 2*DEFAULT_COST); - size((3+1)*4); // set may use two instructions. format %{ "LDUH $mem,$dst\t! ushort/char & 16-bit mask -> long\n\t" "SET $mask,$tmp\n\t" "AND $dst,$tmp,$dst" %} @@ -5854,13 +5864,13 @@ ins_pipe(iload_mem); %} -// Load Integer with a 13-bit mask into a Long Register -instruct loadI2L_immI13(iRegL dst, memory mem, immI13 mask) %{ +// Load Integer with a 12-bit mask into a Long Register +instruct loadI2L_immU12(iRegL dst, memory mem, immU12 mask) %{ match(Set dst (ConvI2L (AndI (LoadI mem) mask))); ins_cost(MEMORY_REF_COST + DEFAULT_COST); size(2*4); - format %{ "LDUW $mem,$dst\t! int & 13-bit mask -> long\n\t" + format %{ "LDUW $mem,$dst\t! int & 12-bit mask -> long\n\t" "AND $dst,$mask,$dst" %} ins_encode %{ Register Rdst = $dst$$Register; @@ -5870,14 +5880,13 @@ ins_pipe(iload_mem); %} -// Load Integer with a 32-bit mask into a Long Register -instruct loadI2L_immI(iRegL dst, memory mem, immI mask, iRegL tmp) %{ +// Load Integer with a 31-bit mask into a Long Register +instruct loadI2L_immU31(iRegL dst, memory mem, immU31 mask, iRegL tmp) %{ match(Set dst (ConvI2L (AndI (LoadI mem) mask))); effect(TEMP dst, TEMP tmp); ins_cost(MEMORY_REF_COST + 2*DEFAULT_COST); - size((3+1)*4); // set may use two instructions. - format %{ "LDUW $mem,$dst\t! int & 32-bit mask -> long\n\t" + format %{ "LDUW $mem,$dst\t! int & 31-bit mask -> long\n\t" "SET $mask,$tmp\n\t" "AND $dst,$tmp,$dst" %} ins_encode %{ @@ -8919,7 +8928,7 @@ ins_pipe(ialu_cconly_reg_reg); %} -instruct compU_iReg_imm13(flagsRegU icc, iRegI op1, immU13 op2 ) %{ +instruct compU_iReg_imm13(flagsRegU icc, iRegI op1, immU12 op2 ) %{ match(Set icc (CmpU op1 op2)); size(4);
--- a/src/cpu/x86/vm/x86_32.ad Thu Feb 27 08:30:16 2014 -0800 +++ b/src/cpu/x86/vm/x86_32.ad Mon Mar 03 11:56:00 2014 -0800 @@ -3919,6 +3919,17 @@ interface(CONST_INTER); %} +// Int Immediate non-negative +operand immU31() +%{ + predicate(n->get_int() >= 0); + match(ConI); + + op_cost(0); + format %{ %} + interface(CONST_INTER); +%} + // Constant for long shifts operand immI_32() %{ predicate( n->get_int() == 32 ); @@ -6135,12 +6146,12 @@ ins_pipe(ialu_reg_mem); %} -// Load Integer with 32-bit mask into Long Register -instruct loadI2L_immI(eRegL dst, memory mem, immI mask, eFlagsReg cr) %{ +// Load Integer with 31-bit mask into Long Register +instruct loadI2L_immU31(eRegL dst, memory mem, immU31 mask, eFlagsReg cr) %{ match(Set dst (ConvI2L (AndI (LoadI mem) mask))); effect(KILL cr); - format %{ "MOV $dst.lo,$mem\t# int & 32-bit mask -> long\n\t" + format %{ "MOV $dst.lo,$mem\t# int & 31-bit mask -> long\n\t" "XOR $dst.hi,$dst.hi\n\t" "AND $dst.lo,$mask" %} ins_encode %{
--- a/src/cpu/x86/vm/x86_64.ad Thu Feb 27 08:30:16 2014 -0800 +++ b/src/cpu/x86/vm/x86_64.ad Mon Mar 03 11:56:00 2014 -0800 @@ -3142,6 +3142,17 @@ interface(CONST_INTER); %} +// Int Immediate non-negative +operand immU31() +%{ + predicate(n->get_int() >= 0); + match(ConI); + + op_cost(0); + format %{ %} + interface(CONST_INTER); +%} + // Constant for long shifts operand immI_32() %{ @@ -5083,12 +5094,12 @@ ins_pipe(ialu_reg_mem); %} -// Load Integer with a 32-bit mask into Long Register -instruct loadI2L_immI(rRegL dst, memory mem, immI mask, rFlagsReg cr) %{ +// Load Integer with a 31-bit mask into Long Register +instruct loadI2L_immU31(rRegL dst, memory mem, immU31 mask, rFlagsReg cr) %{ match(Set dst (ConvI2L (AndI (LoadI mem) mask))); effect(KILL cr); - format %{ "movl $dst, $mem\t# int & 32-bit mask -> long\n\t" + format %{ "movl $dst, $mem\t# int & 31-bit mask -> long\n\t" "andl $dst, $mask" %} ins_encode %{ Register Rdst = $dst$$Register;
--- a/src/cpu/zero/vm/entryFrame_zero.hpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/cpu/zero/vm/entryFrame_zero.hpp Mon Mar 03 11:56:00 2014 -0800 @@ -1,6 +1,6 @@ /* * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright 2008, 2010 Red Hat, Inc. + * Copyright (c) 2013 Red Hat, Inc. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -58,8 +58,8 @@ JavaCallWrapper* call_wrapper, TRAPS); public: - JavaCallWrapper *call_wrapper() const { - return (JavaCallWrapper *) value_of_word(call_wrapper_off); + JavaCallWrapper **call_wrapper() const { + return (JavaCallWrapper **) addr_of_word(call_wrapper_off); } public:
--- a/src/cpu/zero/vm/frame_zero.inline.hpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/cpu/zero/vm/frame_zero.inline.hpp Mon Mar 03 11:56:00 2014 -0800 @@ -1,6 +1,6 @@ /* * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2007, 2013, Red Hat, Inc. + * Copyright (c) 2013 Red Hat, Inc. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -132,7 +132,7 @@ return fp(); } -inline JavaCallWrapper* frame::entry_frame_call_wrapper() const { +inline JavaCallWrapper** frame::entry_frame_call_wrapper_addr() const { return zero_entryframe()->call_wrapper(); }
--- a/src/os/linux/vm/os_linux.cpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/os/linux/vm/os_linux.cpp Mon Mar 03 11:56:00 2014 -0800 @@ -4827,6 +4827,10 @@ Linux::capture_initial_stack(JavaThread::stack_size_at_create()); +#if defined(IA32) + workaround_expand_exec_shield_cs_limit(); +#endif + Linux::libpthread_init(); if (PrintMiscellaneous && (Verbose || WizardMode)) { tty->print_cr("[HotSpot is running with %s, %s(%s)]\n",
--- a/src/os_cpu/linux_x86/vm/os_linux_x86.cpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/os_cpu/linux_x86/vm/os_linux_x86.cpp Mon Mar 03 11:56:00 2014 -0800 @@ -873,3 +873,46 @@ #endif } #endif + + +/* + * IA32 only: execute code at a high address in case buggy NX emulation is present. I.e. avoid CS limit + * updates (JDK-8023956). + */ +void os::workaround_expand_exec_shield_cs_limit() { +#if defined(IA32) + size_t page_size = os::vm_page_size(); + /* + * Take the highest VA the OS will give us and exec + * + * Although using -(pagesz) as mmap hint works on newer kernel as you would + * think, older variants affected by this work-around don't (search forward only). + * + * On the affected distributions, we understand the memory layout to be: + * + * TASK_LIMIT= 3G, main stack base close to TASK_LIMT. + * + * A few pages south main stack will do it. + * + * If we are embedded in an app other than launcher (initial != main stack), + * we don't have much control or understanding of the address space, just let it slide. + */ + char* hint = (char*) (Linux::initial_thread_stack_bottom() - + ((StackYellowPages + StackRedPages + 1) * page_size)); + char* codebuf = os::reserve_memory(page_size, hint); + if ( (codebuf == NULL) || (!os::commit_memory(codebuf, page_size, true)) ) { + return; // No matter, we tried, best effort. + } + if (PrintMiscellaneous && (Verbose || WizardMode)) { + tty->print_cr("[CS limit NX emulation work-around, exec code at: %p]", codebuf); + } + + // Some code to exec: the 'ret' instruction + codebuf[0] = 0xC3; + + // Call the code in the codebuf + __asm__ volatile("call *%0" : : "r"(codebuf)); + + // keep the page mapped so CS limit isn't reduced. +#endif +}
--- a/src/os_cpu/linux_x86/vm/os_linux_x86.hpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/os_cpu/linux_x86/vm/os_linux_x86.hpp Mon Mar 03 11:56:00 2014 -0800 @@ -36,4 +36,17 @@ // Note: Currently only used in 64 bit Windows implementations static bool register_code_area(char *low, char *high) { return true; } + /* + * Work-around for broken NX emulation using CS limit, Red Hat patch "Exec-Shield" + * (IA32 only). + * + * Map and execute at a high VA to prevent CS lazy updates race with SMP MM + * invalidation.Further code generation by the JVM will no longer cause CS limit + * updates. + * + * Affects IA32: RHEL 5 & 6, Ubuntu 10.04 (LTS), 10.10, 11.04, 11.10, 12.04. + * @see JDK-8023956 + */ + static void workaround_expand_exec_shield_cs_limit(); + #endif // OS_CPU_LINUX_X86_VM_OS_LINUX_X86_HPP
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/cmsOopClosures.hpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/share/vm/gc_implementation/concurrentMarkSweep/cmsOopClosures.hpp Mon Mar 03 11:56:00 2014 -0800 @@ -58,8 +58,22 @@ MarkRefsIntoClosure(MemRegion span, CMSBitMap* bitMap); virtual void do_oop(oop* p); virtual void do_oop(narrowOop* p); - inline void do_oop_nv(oop* p) { MarkRefsIntoClosure::do_oop_work(p); } - inline void do_oop_nv(narrowOop* p) { MarkRefsIntoClosure::do_oop_work(p); } + + Prefetch::style prefetch_style() { + return Prefetch::do_read; + } +}; + +class Par_MarkRefsIntoClosure: public OopsInGenClosure { + private: + const MemRegion _span; + CMSBitMap* _bitMap; + protected: + DO_OOP_WORK_DEFN + public: + Par_MarkRefsIntoClosure(MemRegion span, CMSBitMap* bitMap); + virtual void do_oop(oop* p); + virtual void do_oop(narrowOop* p); bool do_header() { return true; } Prefetch::style prefetch_style() { return Prefetch::do_read;
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp Mon Mar 03 11:56:00 2014 -0800 @@ -575,6 +575,7 @@ _restart_addr(NULL), _overflow_list(NULL), _stats(cmsGen), + _eden_chunk_lock(new Mutex(Mutex::leaf + 1, "CMS_eden_chunk_lock", true)), _eden_chunk_array(NULL), // may be set in ctor body _eden_chunk_capacity(0), // -- ditto -- _eden_chunk_index(0), // -- ditto -- @@ -754,7 +755,7 @@ assert(_eden_chunk_array != NULL || _eden_chunk_capacity == 0, "Error"); // Support for parallelizing survivor space rescan - if (CMSParallelRemarkEnabled && CMSParallelSurvivorRemarkEnabled) { + if ((CMSParallelRemarkEnabled && CMSParallelSurvivorRemarkEnabled) || CMSParallelInitialMarkEnabled) { const size_t max_plab_samples = ((DefNewGeneration*)_young_gen)->max_survivor_size()/MinTLABSize; @@ -2136,6 +2137,39 @@ } +void CMSCollector::print_eden_and_survivor_chunk_arrays() { + DefNewGeneration* dng = _young_gen->as_DefNewGeneration(); + EdenSpace* eden_space = dng->eden(); + ContiguousSpace* from_space = dng->from(); + ContiguousSpace* to_space = dng->to(); + // Eden + if (_eden_chunk_array != NULL) { + gclog_or_tty->print_cr("eden " PTR_FORMAT "-" PTR_FORMAT "-" PTR_FORMAT "(" SIZE_FORMAT ")", + eden_space->bottom(), eden_space->top(), + eden_space->end(), eden_space->capacity()); + gclog_or_tty->print_cr("_eden_chunk_index=" SIZE_FORMAT ", " + "_eden_chunk_capacity=" SIZE_FORMAT, + _eden_chunk_index, _eden_chunk_capacity); + for (size_t i = 0; i < _eden_chunk_index; i++) { + gclog_or_tty->print_cr("_eden_chunk_array[" SIZE_FORMAT "]=" PTR_FORMAT, + i, _eden_chunk_array[i]); + } + } + // Survivor + if (_survivor_chunk_array != NULL) { + gclog_or_tty->print_cr("survivor " PTR_FORMAT "-" PTR_FORMAT "-" PTR_FORMAT "(" SIZE_FORMAT ")", + from_space->bottom(), from_space->top(), + from_space->end(), from_space->capacity()); + gclog_or_tty->print_cr("_survivor_chunk_index=" SIZE_FORMAT ", " + "_survivor_chunk_capacity=" SIZE_FORMAT, + _survivor_chunk_index, _survivor_chunk_capacity); + for (size_t i = 0; i < _survivor_chunk_index; i++) { + gclog_or_tty->print_cr("_survivor_chunk_array[" SIZE_FORMAT "]=" PTR_FORMAT, + i, _survivor_chunk_array[i]); + } + } +} + void CMSCollector::getFreelistLocks() const { // Get locks for all free lists in all generations that this // collector is responsible for @@ -3526,6 +3560,31 @@ // CMS work +// The common parts of CMSParInitialMarkTask and CMSParRemarkTask. +class CMSParMarkTask : public AbstractGangTask { + protected: + CMSCollector* _collector; + int _n_workers; + CMSParMarkTask(const char* name, CMSCollector* collector, int n_workers) : + AbstractGangTask(name), + _collector(collector), + _n_workers(n_workers) {} + // Work method in support of parallel rescan ... of young gen spaces + void do_young_space_rescan(uint worker_id, OopsInGenClosure* cl, + ContiguousSpace* space, + HeapWord** chunk_array, size_t chunk_top); + void work_on_young_gen_roots(uint worker_id, OopsInGenClosure* cl); +}; + +// Parallel initial mark task +class CMSParInitialMarkTask: public CMSParMarkTask { + public: + CMSParInitialMarkTask(CMSCollector* collector, int n_workers) : + CMSParMarkTask("Scan roots and young gen for initial mark in parallel", + collector, n_workers) {} + void work(uint worker_id); +}; + // Checkpoint the roots into this generation from outside // this generation. [Note this initial checkpoint need only // be approximate -- we'll do a catch up phase subsequently.] @@ -3615,20 +3674,42 @@ // weak reference processing has not started yet. ref_processor()->set_enqueuing_is_done(false); + if (CMSPrintEdenSurvivorChunks) { + print_eden_and_survivor_chunk_arrays(); + } + { // This is not needed. DEBUG_ONLY(RememberKlassesChecker imx(true);) COMPILER2_PRESENT(DerivedPointerTableDeactivate dpt_deact;) - gch->rem_set()->prepare_for_younger_refs_iterate(false); // Not parallel. - gch->gen_process_strong_roots(_cmsGen->level(), - true, // younger gens are roots - true, // activate StrongRootsScope - true, // collecting perm gen - SharedHeap::ScanningOption(roots_scanning_options()), - ¬Older, - true, // walk all of code cache if (so & SO_CodeCache) - NULL); - } - + if (CMSParallelInitialMarkEnabled && CollectedHeap::use_parallel_gc_threads()) { + // The parallel version. + FlexibleWorkGang* workers = gch->workers(); + assert(workers != NULL, "Need parallel worker threads."); + int n_workers = workers->active_workers(); + CMSParInitialMarkTask tsk(this, n_workers); + gch->set_par_threads(n_workers); + initialize_sequential_subtasks_for_young_gen_rescan(n_workers); + if (n_workers > 1) { + GenCollectedHeap::StrongRootsScope srs(gch); + workers->run_task(&tsk); + } else { + GenCollectedHeap::StrongRootsScope srs(gch); + tsk.work(0); + } + gch->set_par_threads(0); + } else { + // The serial version. + gch->rem_set()->prepare_for_younger_refs_iterate(false); // Not parallel. + gch->gen_process_strong_roots(_cmsGen->level(), + true, // younger gens are roots + true, // activate StrongRootsScope + true, // collecting perm gen + SharedHeap::ScanningOption(roots_scanning_options()), + ¬Older, + true, // walk all of code cache if (so & SO_CodeCache) + NULL); + } + } // Clear mod-union table; it will be dirtied in the prologue of // CMS generation per each younger generation collection. @@ -4410,7 +4491,9 @@ verify_overflow_empty(); _abort_preclean = false; if (CMSPrecleaningEnabled) { - _eden_chunk_index = 0; + if (!CMSEdenChunksRecordAlways) { + _eden_chunk_index = 0; + } size_t used = get_eden_used(); size_t capacity = get_eden_capacity(); // Don't start sampling unless we will get sufficiently @@ -4519,7 +4602,9 @@ if (!_start_sampling) { return; } - if (_eden_chunk_array) { + // When CMSEdenChunksRecordAlways is true, the eden chunk array + // is populated by the young generation. + if (_eden_chunk_array != NULL && !CMSEdenChunksRecordAlways) { if (_eden_chunk_index < _eden_chunk_capacity) { _eden_chunk_array[_eden_chunk_index] = *_top_addr; // take sample assert(_eden_chunk_array[_eden_chunk_index] <= *_end_addr, @@ -4996,6 +5081,10 @@ // Update the saved marks which may affect the root scans. gch->save_marks(); + if (CMSPrintEdenSurvivorChunks) { + print_eden_and_survivor_chunk_arrays(); + } + { COMPILER2_PRESENT(DerivedPointerTableDeactivate dpt_deact;) @@ -5103,10 +5192,51 @@ } } +void CMSParInitialMarkTask::work(uint worker_id) { + elapsedTimer _timer; + ResourceMark rm; + HandleMark hm; + + // ---------- scan from roots -------------- + _timer.start(); + GenCollectedHeap* gch = GenCollectedHeap::heap(); + Par_MarkRefsIntoClosure par_mri_cl(_collector->_span, &(_collector->_markBitMap)); + + // ---------- young gen roots -------------- + { + work_on_young_gen_roots(worker_id, &par_mri_cl); + _timer.stop(); + if (PrintCMSStatistics != 0) { + gclog_or_tty->print_cr( + "Finished young gen initial mark scan work in %dth thread: %3.3f sec", + worker_id, _timer.seconds()); + } + } + + // ---------- remaining roots -------------- + _timer.reset(); + _timer.start(); + gch->gen_process_strong_roots(_collector->_cmsGen->level(), + false, // yg was scanned above + false, // this is parallel code + true, // collecting perm gen + SharedHeap::ScanningOption(_collector->CMSCollector::roots_scanning_options()), + &par_mri_cl, + true, // walk all of code cache if (so & SO_CodeCache) + NULL); + assert(_collector->should_unload_classes() + || (_collector->CMSCollector::roots_scanning_options() & SharedHeap::SO_CodeCache), + "if we didn't scan the code cache, we have to be ready to drop nmethods with expired weak oops"); + _timer.stop(); + if (PrintCMSStatistics != 0) { + gclog_or_tty->print_cr( + "Finished remaining root initial mark scan work in %dth thread: %3.3f sec", + worker_id, _timer.seconds()); + } +} + // Parallel remark task -class CMSParRemarkTask: public AbstractGangTask { - CMSCollector* _collector; - int _n_workers; +class CMSParRemarkTask: public CMSParMarkTask { CompactibleFreeListSpace* _cms_space; CompactibleFreeListSpace* _perm_space; @@ -5122,10 +5252,9 @@ CompactibleFreeListSpace* perm_space, int n_workers, FlexibleWorkGang* workers, OopTaskQueueSet* task_queues): - AbstractGangTask("Rescan roots and grey objects in parallel"), - _collector(collector), + CMSParMarkTask("Rescan roots and grey objects in parallel", + collector, n_workers), _cms_space(cms_space), _perm_space(perm_space), - _n_workers(n_workers), _task_queues(task_queues), _term(n_workers, task_queues) { } @@ -5139,11 +5268,6 @@ void work(uint worker_id); private: - // Work method in support of parallel rescan ... of young gen spaces - void do_young_space_rescan(int i, Par_MarkRefsIntoAndScanClosure* cl, - ContiguousSpace* space, - HeapWord** chunk_array, size_t chunk_top); - // ... of dirty cards in old space void do_dirty_card_rescan_tasks(CompactibleFreeListSpace* sp, int i, Par_MarkRefsIntoAndScanClosure* cl); @@ -5152,6 +5276,25 @@ void do_work_steal(int i, Par_MarkRefsIntoAndScanClosure* cl, int* seed); }; +void CMSParMarkTask::work_on_young_gen_roots(uint worker_id, OopsInGenClosure* cl) { + DefNewGeneration* dng = _collector->_young_gen->as_DefNewGeneration(); + EdenSpace* eden_space = dng->eden(); + ContiguousSpace* from_space = dng->from(); + ContiguousSpace* to_space = dng->to(); + + HeapWord** eca = _collector->_eden_chunk_array; + size_t ect = _collector->_eden_chunk_index; + HeapWord** sca = _collector->_survivor_chunk_array; + size_t sct = _collector->_survivor_chunk_index; + + assert(ect <= _collector->_eden_chunk_capacity, "out of bounds"); + assert(sct <= _collector->_survivor_chunk_capacity, "out of bounds"); + + do_young_space_rescan(worker_id, cl, to_space, NULL, 0); + do_young_space_rescan(worker_id, cl, from_space, sca, sct); + do_young_space_rescan(worker_id, cl, eden_space, eca, ect); +} + // work_queue(i) is passed to the closure // Par_MarkRefsIntoAndScanClosure. The "i" parameter // also is passed to do_dirty_card_rescan_tasks() and to @@ -5176,23 +5319,7 @@ // work first. // ---------- young gen roots -------------- { - DefNewGeneration* dng = _collector->_young_gen->as_DefNewGeneration(); - EdenSpace* eden_space = dng->eden(); - ContiguousSpace* from_space = dng->from(); - ContiguousSpace* to_space = dng->to(); - - HeapWord** eca = _collector->_eden_chunk_array; - size_t ect = _collector->_eden_chunk_index; - HeapWord** sca = _collector->_survivor_chunk_array; - size_t sct = _collector->_survivor_chunk_index; - - assert(ect <= _collector->_eden_chunk_capacity, "out of bounds"); - assert(sct <= _collector->_survivor_chunk_capacity, "out of bounds"); - - do_young_space_rescan(worker_id, &par_mrias_cl, to_space, NULL, 0); - do_young_space_rescan(worker_id, &par_mrias_cl, from_space, sca, sct); - do_young_space_rescan(worker_id, &par_mrias_cl, eden_space, eca, ect); - + work_on_young_gen_roots(worker_id, &par_mrias_cl); _timer.stop(); if (PrintCMSStatistics != 0) { gclog_or_tty->print_cr( @@ -5253,8 +5380,8 @@ // Note that parameter "i" is not used. void -CMSParRemarkTask::do_young_space_rescan(int i, - Par_MarkRefsIntoAndScanClosure* cl, ContiguousSpace* space, +CMSParMarkTask::do_young_space_rescan(uint worker_id, + OopsInGenClosure* cl, ContiguousSpace* space, HeapWord** chunk_array, size_t chunk_top) { // Until all tasks completed: // . claim an unclaimed task @@ -5450,6 +5577,32 @@ "Else our work is not yet done"); } +// Record object boundaries in _eden_chunk_array by sampling the eden +// top in the slow-path eden object allocation code path and record +// the boundaries, if CMSEdenChunksRecordAlways is true. If +// CMSEdenChunksRecordAlways is false, we use the other asynchronous +// sampling in sample_eden() that activates during the part of the +// preclean phase. +void CMSCollector::sample_eden_chunk() { + if (CMSEdenChunksRecordAlways && _eden_chunk_array != NULL) { + if (_eden_chunk_lock->try_lock()) { + // Record a sample. This is the critical section. The contents + // of the _eden_chunk_array have to be non-decreasing in the + // address order. + _eden_chunk_array[_eden_chunk_index] = *_top_addr; + assert(_eden_chunk_array[_eden_chunk_index] <= *_end_addr, + "Unexpected state of Eden"); + if (_eden_chunk_index == 0 || + ((_eden_chunk_array[_eden_chunk_index] > _eden_chunk_array[_eden_chunk_index-1]) && + (pointer_delta(_eden_chunk_array[_eden_chunk_index], + _eden_chunk_array[_eden_chunk_index-1]) >= CMSSamplingGrain))) { + _eden_chunk_index++; // commit sample + } + _eden_chunk_lock->unlock(); + } + } +} + // Return a thread-local PLAB recording array, as appropriate. void* CMSCollector::get_data_recorder(int thr_num) { if (_survivor_plab_array != NULL && @@ -5473,12 +5626,13 @@ // Merge the per-thread plab arrays into the global survivor chunk // array which will provide the partitioning of the survivor space -// for CMS rescan. +// for CMS initial scan and rescan. void CMSCollector::merge_survivor_plab_arrays(ContiguousSpace* surv, int no_of_gc_threads) { assert(_survivor_plab_array != NULL, "Error"); assert(_survivor_chunk_array != NULL, "Error"); - assert(_collectorState == FinalMarking, "Error"); + assert(_collectorState == FinalMarking || + (CMSParallelInitialMarkEnabled && _collectorState == InitialMarking), "Error"); for (int j = 0; j < no_of_gc_threads; j++) { _cursor[j] = 0; } @@ -5541,7 +5695,7 @@ } // Set up the space's par_seq_tasks structure for work claiming -// for parallel rescan of young gen. +// for parallel initial scan and rescan of young gen. // See ParRescanTask where this is currently used. void CMSCollector:: @@ -6691,6 +6845,28 @@ void MarkRefsIntoClosure::do_oop(oop* p) { MarkRefsIntoClosure::do_oop_work(p); } void MarkRefsIntoClosure::do_oop(narrowOop* p) { MarkRefsIntoClosure::do_oop_work(p); } +Par_MarkRefsIntoClosure::Par_MarkRefsIntoClosure( + MemRegion span, CMSBitMap* bitMap): + _span(span), + _bitMap(bitMap) +{ + assert(_ref_processor == NULL, "deliberately left NULL"); + assert(_bitMap->covers(_span), "_bitMap/_span mismatch"); +} + +void Par_MarkRefsIntoClosure::do_oop(oop obj) { + // if p points into _span, then mark corresponding bit in _markBitMap + assert(obj->is_oop(), "expected an oop"); + HeapWord* addr = (HeapWord*)obj; + if (_span.contains(addr)) { + // this should be made more efficient + _bitMap->par_mark(addr); + } +} + +void Par_MarkRefsIntoClosure::do_oop(oop* p) { Par_MarkRefsIntoClosure::do_oop_work(p); } +void Par_MarkRefsIntoClosure::do_oop(narrowOop* p) { Par_MarkRefsIntoClosure::do_oop_work(p); } + // A variant of the above, used for CMS marking verification. MarkRefsIntoVerifyClosure::MarkRefsIntoVerifyClosure( MemRegion span, CMSBitMap* verification_bm, CMSBitMap* cms_bm): @@ -9356,7 +9532,6 @@ return; } } - // Transfer some number of overflown objects to usual marking // stack. Return true if some objects were transferred. bool MarkRefsIntoAndScanClosure::take_from_overflow_list() {
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.hpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.hpp Mon Mar 03 11:56:00 2014 -0800 @@ -516,6 +516,8 @@ friend class ConcurrentMarkSweepThread; friend class ConcurrentMarkSweepGeneration; friend class CompactibleFreeListSpace; + friend class CMSParMarkTask; + friend class CMSParInitialMarkTask; friend class CMSParRemarkTask; friend class CMSConcMarkingTask; friend class CMSRefProcTaskProxy; @@ -752,6 +754,7 @@ Generation* _young_gen; // the younger gen HeapWord** _top_addr; // ... Top of Eden HeapWord** _end_addr; // ... End of Eden + Mutex* _eden_chunk_lock; HeapWord** _eden_chunk_array; // ... Eden partitioning array size_t _eden_chunk_index; // ... top (exclusive) of array size_t _eden_chunk_capacity; // ... max entries in array @@ -953,6 +956,7 @@ // Support for parallel remark of survivor space void* get_data_recorder(int thr_num); + void sample_eden_chunk(); CMSBitMap* markBitMap() { return &_markBitMap; } void directAllocated(HeapWord* start, size_t size); @@ -1031,6 +1035,8 @@ // Initialization errors bool completed_initialization() { return _completed_initialization; } + + void print_eden_and_survivor_chunk_arrays(); }; class CMSExpansionCause : public AllStatic { @@ -1317,6 +1323,10 @@ //Delegate to collector return collector()->get_data_recorder(thr_num); } + void sample_eden_chunk() { + //Delegate to collector + return collector()->sample_eden_chunk(); + } // Printing const char* name() const;
--- a/src/share/vm/memory/defNewGeneration.cpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/share/vm/memory/defNewGeneration.cpp Mon Mar 03 11:56:00 2014 -0800 @@ -1004,6 +1004,9 @@ // have to use it here, as well. HeapWord* result = eden()->par_allocate(word_size); if (result != NULL) { + if (CMSEdenChunksRecordAlways && _next_gen != NULL) { + _next_gen->sample_eden_chunk(); + } return result; } do { @@ -1034,13 +1037,19 @@ // circular dependency at compile time. if (result == NULL) { result = allocate_from_space(word_size); + } else if (CMSEdenChunksRecordAlways && _next_gen != NULL) { + _next_gen->sample_eden_chunk(); } return result; } HeapWord* DefNewGeneration::par_allocate(size_t word_size, bool is_tlab) { - return eden()->par_allocate(word_size); + HeapWord* res = eden()->par_allocate(word_size); + if (CMSEdenChunksRecordAlways && _next_gen != NULL) { + _next_gen->sample_eden_chunk(); + } + return res; } void DefNewGeneration::gc_prologue(bool full) {
--- a/src/share/vm/memory/generation.hpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/share/vm/memory/generation.hpp Mon Mar 03 11:56:00 2014 -0800 @@ -457,6 +457,7 @@ // expected to be GC worker thread-local, with the worker index // indicated by "thr_num". virtual void* get_data_recorder(int thr_num) { return NULL; } + virtual void sample_eden_chunk() {} // Some generations may require some cleanup actions before allowing // a verification.
--- a/src/share/vm/memory/sharedHeap.cpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/share/vm/memory/sharedHeap.cpp Mon Mar 03 11:56:00 2014 -0800 @@ -64,7 +64,8 @@ } _sh = this; // ch is static, should be set only once. if ((UseParNewGC || - (UseConcMarkSweepGC && CMSParallelRemarkEnabled) || + (UseConcMarkSweepGC && (CMSParallelInitialMarkEnabled || + CMSParallelRemarkEnabled)) || UseG1GC) && ParallelGCThreads > 0) { _workers = new FlexibleWorkGang("Parallel GC Threads", ParallelGCThreads,
--- a/src/share/vm/runtime/globals.hpp Thu Feb 27 08:30:16 2014 -0800 +++ b/src/share/vm/runtime/globals.hpp Mon Mar 03 11:56:00 2014 -0800 @@ -1697,6 +1697,9 @@ product(bool, CMSAbortSemantics, false, \ "Whether abort-on-overflow semantics is implemented") \ \ + product(bool, CMSParallelInitialMarkEnabled, false, \ + "Use the parallel initial mark.") \ + \ product(bool, CMSParallelRemarkEnabled, true, \ "Whether parallel remark enabled (only if ParNewGC)") \ \ @@ -1708,6 +1711,14 @@ "Whether to always record survivor space PLAB bdries" \ " (effective only if CMSParallelSurvivorRemarkEnabled)") \ \ + product(bool, CMSEdenChunksRecordAlways, false, \ + "Whether to always record eden chunks used for " \ + "the parallel initial mark or remark of eden" ) \ + \ + product(bool, CMSPrintEdenSurvivorChunks, false, \ + "Print the eden and the survivor chunks used for the parallel " \ + "initial mark or remark of the eden/survivor spaces") \ + \ product(bool, CMSConcurrentMTEnabled, true, \ "Whether multi-threaded concurrent work enabled (if ParNewGC)") \ \
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/test/compiler/codegen/LoadWithMask.java Mon Mar 03 11:56:00 2014 -0800 @@ -0,0 +1,44 @@ +/* + * 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 8032207 + * @summary Invalid node sizing for loadUS2L_immI16 and loadI2L_immI + * @run main/othervm -server -Xbatch -XX:-TieredCompilation -XX:CompileCommand=compileonly,LoadWithMask.foo LoadWithMask + * + */ +public class LoadWithMask { + static int x[] = new int[1]; + static long foo() { + return x[0] & 0xfff0ffff; + } + + public static void main(String[] args) { + x[0] = -1; + long l = 0; + for (int i = 0; i < 100000; ++i) { + l = foo(); + } + } +}
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/test/compiler/codegen/LoadWithMask2.java Mon Mar 03 11:56:00 2014 -0800 @@ -0,0 +1,55 @@ +/* + * 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 8031743 + * @summary loadI2L_immI broken for negative memory values + * @run main/othervm -server -Xbatch -XX:-TieredCompilation -XX:CompileCommand=compileonly,*.foo* LoadWithMask2 + * + */ +public class LoadWithMask2 { + static int x; + static long foo1() { + return x & 0xfffffffe; + } + static long foo2() { + return x & 0xff000000; + } + static long foo3() { + return x & 0x8abcdef1; + } + + public static void main(String[] args) { + x = -1; + long l = 0; + for (int i = 0; i < 100000; ++i) { + l = foo1() & foo2() & foo3(); + } + if (l > 0) { + System.out.println("FAILED"); + System.exit(97); + } + System.out.println("PASSED"); + } +}