OpenJDK / amber / amber
changeset 58424:7be56b2ac50d
Merge
| author | henryjen |
|---|---|
| date | Mon, 14 Oct 2019 21:01:25 +0000 |
| parents | d66bdf0e2dfe 6941d77417f4 |
| children | e0bb6b38a6f6 |
| files | src/hotspot/cpu/x86/macroAssembler_x86.cpp src/hotspot/share/runtime/compilationPolicy.cpp src/hotspot/share/runtime/compilationPolicy.hpp src/hotspot/share/runtime/tieredThresholdPolicy.cpp src/hotspot/share/runtime/tieredThresholdPolicy.hpp src/java.base/unix/native/libjava/jdk_util_md.c src/java.security.jgss/share/classes/sun/security/jgss/krb5/Krb5Context.java src/java.security.jgss/share/classes/sun/security/jgss/krb5/Krb5InitCredential.java src/jdk.internal.vm.compiler.management/share/classes/org.graalvm.compiler.hotspot.management/src/org/graalvm/compiler/hotspot/management/JMXServiceProvider.java src/jdk.internal.vm.compiler.management/share/classes/org.graalvm.compiler.hotspot.management/src/org/graalvm/compiler/hotspot/management/package-info.java test/hotspot/jtreg/compiler/graalunit/EA9Test.java test/hotspot/jtreg/compiler/graalunit/com.oracle.mxtool.junit/com/oracle/mxtool/junit/JLModule.java test/hotspot/jtreg/vmTestbase/jit/graph/CGTThread.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt0/cgt0.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt1/cgt1.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt10/cgt10.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt11/cgt11.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt2/cgt2.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt3/cgt3.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt4/cgt4.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt5/cgt5.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt6/cgt6.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt7/cgt7.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt8/cgt8.java test/hotspot/jtreg/vmTestbase/jit/graph/cgt9/cgt9.java |
| diffstat | 580 files changed, 13480 insertions(+), 11390 deletions(-) [+] |
line wrap: on
line diff
--- a/.hgtags Fri Oct 04 13:44:11 2019 -0700 +++ b/.hgtags Mon Oct 14 21:01:25 2019 +0000 @@ -590,3 +590,4 @@ 778fc2dcbdaa8981e07e929a2cacef979c72261e jdk-14+15 d29f0181ba424a95d881aba5eabf2e393abcc70f jdk-14+16 5c83830390baafb76a1fbe33443c57620bd45fb9 jdk-14+17 +e84d8379815ba0d3e50fb096d28c25894cb50b8c jdk-14+18
--- a/doc/building.html Fri Oct 04 13:44:11 2019 -0700 +++ b/doc/building.html Mon Oct 14 21:01:25 2019 +0000 @@ -281,7 +281,7 @@ <tbody> <tr class="odd"> <td style="text-align: left;">Linux</td> -<td style="text-align: left;">gcc 8.2.0</td> +<td style="text-align: left;">gcc 8.3.0</td> </tr> <tr class="even"> <td style="text-align: left;">macOS</td> @@ -293,14 +293,14 @@ </tr> <tr class="even"> <td style="text-align: left;">Windows</td> -<td style="text-align: left;">Microsoft Visual Studio 2017 update 15.9.6</td> +<td style="text-align: left;">Microsoft Visual Studio 2017 update 15.9.16</td> </tr> </tbody> </table> <p>All compilers are expected to be able to compile to the C99 language standard, as some C99 features are used in the source code. Microsoft Visual Studio doesn't fully support C99 so in practice shared code is limited to using C99 features that it does support.</p> <h3 id="gcc">gcc</h3> <p>The minimum accepted version of gcc is 4.8. Older versions will generate a warning by <code>configure</code> and are unlikely to work.</p> -<p>The JDK is currently known to be able to compile with at least version 7.4 of gcc.</p> +<p>The JDK is currently known to be able to compile with at least version 8.3 of gcc.</p> <p>In general, any version between these two should be usable.</p> <h3 id="clang">clang</h3> <p>The minimum accepted version of clang is 3.2. Older versions will not be accepted by <code>configure</code>.</p>
--- a/doc/building.md Fri Oct 04 13:44:11 2019 -0700 +++ b/doc/building.md Mon Oct 14 21:01:25 2019 +0000 @@ -323,10 +323,10 @@ Operating system Toolchain version ------------------ ------------------------------------------------------- - Linux gcc 8.2.0 + Linux gcc 8.3.0 macOS Apple Xcode 10.1 (using clang 10.0.0) Solaris Oracle Solaris Studio 12.6 (with compiler version 5.15) - Windows Microsoft Visual Studio 2017 update 15.9.6 + Windows Microsoft Visual Studio 2017 update 15.9.16 All compilers are expected to be able to compile to the C99 language standard, as some C99 features are used in the source code. Microsoft Visual Studio @@ -338,7 +338,7 @@ The minimum accepted version of gcc is 4.8. Older versions will generate a warning by `configure` and are unlikely to work. -The JDK is currently known to be able to compile with at least version 7.4 of +The JDK is currently known to be able to compile with at least version 8.3 of gcc. In general, any version between these two should be usable.
--- a/make/CreateJmods.gmk Fri Oct 04 13:44:11 2019 -0700 +++ b/make/CreateJmods.gmk Mon Oct 14 21:01:25 2019 +0000 @@ -86,16 +86,18 @@ # from there. These files were explicitly filtered or modified in <module>-copy # targets. For the rest, just pick up everything from the source legal dirs. LEGAL_NOTICES := \ - $(SUPPORT_OUTPUTDIR)/modules_legal/common \ + $(wildcard $(SUPPORT_OUTPUTDIR)/modules_legal/common) \ $(if $(wildcard $(SUPPORT_OUTPUTDIR)/modules_legal/$(MODULE)), \ $(wildcard $(SUPPORT_OUTPUTDIR)/modules_legal/$(MODULE)), \ $(call FindModuleLegalSrcDirs, $(MODULE)) \ ) -LEGAL_NOTICES_PATH := $(call PathList, $(LEGAL_NOTICES)) -DEPS += $(call FindFiles, $(LEGAL_NOTICES)) +ifneq ($(strip $(LEGAL_NOTICES)), ) + LEGAL_NOTICES_PATH := $(call PathList, $(LEGAL_NOTICES)) + DEPS += $(call FindFiles, $(LEGAL_NOTICES)) -JMOD_FLAGS += --legal-notices $(LEGAL_NOTICES_PATH) + JMOD_FLAGS += --legal-notices $(LEGAL_NOTICES_PATH) +endif ifeq ($(filter-out jdk.incubator.%, $(MODULE)), ) JMOD_FLAGS += --do-not-resolve-by-default
--- a/make/RunTestsPrebuilt.gmk Fri Oct 04 13:44:11 2019 -0700 +++ b/make/RunTestsPrebuilt.gmk Mon Oct 14 21:01:25 2019 +0000 @@ -230,7 +230,7 @@ NUM_CORES := $(shell /usr/sbin/sysctl -n hw.ncpu) MEMORY_SIZE := $(shell $(EXPR) `/usr/sbin/sysctl -n hw.memsize` / 1024 / 1024) else ifeq ($(OPENJDK_TARGET_OS), solaris) - NUM_CORES := $(shell LC_MESSAGES=C /usr/sbin/psrinfo -v | $(GREP) -c on-line) + NUM_CORES := $(shell /usr/sbin/psrinfo -v | $(GREP) -c on-line) MEMORY_SIZE := $(shell \ /usr/sbin/prtconf 2> /dev/null | $(GREP) "^Memory [Ss]ize" | $(AWK) '{print $$3}' \ )
--- a/make/RunTestsPrebuiltSpec.gmk Fri Oct 04 13:44:11 2019 -0700 +++ b/make/RunTestsPrebuiltSpec.gmk Mon Oct 14 21:01:25 2019 +0000 @@ -27,6 +27,9 @@ # Fake minimalistic spec file for RunTestsPrebuilt.gmk. ################################################################################ +# Make sure all shell commands are executed with the C locale +export LC_ALL := C + define VerifyVariable ifeq ($$($1), ) $$(info Error: Variable $1 is missing, needed by RunTestPrebuiltSpec.gmk)
--- a/make/autoconf/basics.m4 Fri Oct 04 13:44:11 2019 -0700 +++ b/make/autoconf/basics.m4 Mon Oct 14 21:01:25 2019 +0000 @@ -427,7 +427,7 @@ # Save the path variable before it gets changed ORIGINAL_PATH="$PATH" AC_SUBST(ORIGINAL_PATH) - DATE_WHEN_CONFIGURED=`LANG=C date` + DATE_WHEN_CONFIGURED=`date` AC_SUBST(DATE_WHEN_CONFIGURED) AC_MSG_NOTICE([Configuration created at $DATE_WHEN_CONFIGURED.]) ])
--- a/make/autoconf/build-performance.m4 Fri Oct 04 13:44:11 2019 -0700 +++ b/make/autoconf/build-performance.m4 Mon Oct 14 21:01:25 2019 +0000 @@ -35,7 +35,7 @@ FOUND_CORES=yes elif test -x /usr/sbin/psrinfo; then # Looks like a Solaris system - NUM_CORES=`LC_MESSAGES=C /usr/sbin/psrinfo -v | grep -c on-line` + NUM_CORES=`/usr/sbin/psrinfo -v | grep -c on-line` FOUND_CORES=yes elif test -x /usr/sbin/sysctl; then # Looks like a MacOSX system
--- a/make/autoconf/configure Fri Oct 04 13:44:11 2019 -0700 +++ b/make/autoconf/configure Mon Oct 14 21:01:25 2019 +0000 @@ -43,6 +43,9 @@ export CONFIG_SHELL=$BASH export _as_can_reexec=no +# Make sure all shell commands are executed with the C locale +export LC_ALL=C + if test "x$CUSTOM_CONFIG_DIR" != x; then custom_hook=$CUSTOM_CONFIG_DIR/custom-hook.m4 if test ! -e $custom_hook; then
--- a/make/autoconf/spec.gmk.in Fri Oct 04 13:44:11 2019 -0700 +++ b/make/autoconf/spec.gmk.in Mon Oct 14 21:01:25 2019 +0000 @@ -51,6 +51,9 @@ # What make to use for main processing, after bootstrapping top-level Makefile. MAKE := @MAKE@ +# Make sure all shell commands are executed with the C locale +export LC_ALL := C + # The default make arguments MAKE_ARGS = $(MAKE_LOG_FLAGS) -r -R -I $(TOPDIR)/make/common SPEC=$(SPEC) \ MAKE_LOG_FLAGS="$(MAKE_LOG_FLAGS)" $(MAKE_LOG_VARS)
--- a/make/autoconf/toolchain_windows.m4 Fri Oct 04 13:44:11 2019 -0700 +++ b/make/autoconf/toolchain_windows.m4 Mon Oct 14 21:01:25 2019 +0000 @@ -209,6 +209,8 @@ eval SDK_INSTALL_DIR="\${VS_SDK_INSTALLDIR_${VS_VERSION}}" eval VS_ENV_ARGS="\${VS_ENV_ARGS_${VS_VERSION}}" eval VS_TOOLSET_SUPPORTED="\${VS_TOOLSET_SUPPORTED_${VS_VERSION}}" + + VS_ENV_CMD="" # When using --with-tools-dir, assume it points to the correct and default # version of Visual Studio or that --with-toolchain-version was also set. @@ -227,8 +229,6 @@ fi fi - VS_ENV_CMD="" - if test "x$VS_COMNTOOLS" != x; then TOOLCHAIN_CHECK_POSSIBLE_VISUAL_STUDIO_ROOT([${VS_VERSION}], [$VS_COMNTOOLS/../..], [$VS_COMNTOOLS_VAR variable])
--- a/make/common/JavaCompilation.gmk Fri Oct 04 13:44:11 2019 -0700 +++ b/make/common/JavaCompilation.gmk Mon Oct 14 21:01:25 2019 +0000 @@ -122,7 +122,7 @@ $$($1_BIN)$$($1_MODULE_SUBDIR)$$($2_TARGET) : $2 $$(call LogInfo, Cleaning $$(patsubst $(OUTPUTDIR)/%,%, $$@)) $$(call MakeTargetDir) - export LC_ALL=C ; ( $(CAT) $$< && $(ECHO) "" ) \ + ( $(CAT) $$< && $(ECHO) "" ) \ | $(SED) -e 's/\([^\\]\):/\1\\:/g' -e 's/\([^\\]\)=/\1\\=/g' \ -e 's/\([^\\]\)!/\1\\!/g' -e 's/^[ ]*#.*/#/g' \ | $(SED) -f "$(TOPDIR)/make/common/support/unicode2x.sed" \
--- a/make/conf/jib-profiles.js Fri Oct 04 13:44:11 2019 -0700 +++ b/make/conf/jib-profiles.js Mon Oct 14 21:01:25 2019 +0000 @@ -944,11 +944,11 @@ var getJibProfilesDependencies = function (input, common) { var devkit_platform_revisions = { - linux_x64: "gcc8.2.0-OL6.4+1.0", + linux_x64: "gcc8.3.0-OL6.4+1.0", macosx_x64: "Xcode10.1-MacOSX10.14+1.0", solaris_x64: "SS12u4-Solaris11u1+1.0", solaris_sparcv9: "SS12u6-Solaris11u3+1.0", - windows_x64: "VS2017-15.9.6+1.0", + windows_x64: "VS2017-15.9.16+1.0", linux_aarch64: "gcc8.2.0-Fedora27+1.0", linux_arm: "gcc8.2.0-Fedora27+1.0", linux_ppc64le: "gcc8.2.0-Fedora27+1.0",
--- a/make/data/charsetmapping/SingleByte-X.java.template Fri Oct 04 13:44:11 2019 -0700 +++ b/make/data/charsetmapping/SingleByte-X.java.template Mon Oct 14 21:01:25 2019 +0000 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2013, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2008, 2019, 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 @@ -48,7 +48,7 @@ } public CharsetDecoder newDecoder() { - return new SingleByte.Decoder(this, b2c, $ASCIICOMPATIBLE$); + return new SingleByte.Decoder(this, b2c, $ASCIICOMPATIBLE$, $LATIN1DECODABLE$); } public CharsetEncoder newEncoder() {
--- a/make/data/lsrdata/language-subtag-registry.txt Fri Oct 04 13:44:11 2019 -0700 +++ b/make/data/lsrdata/language-subtag-registry.txt Mon Oct 14 21:01:25 2019 +0000 @@ -1,4 +1,4 @@ -File-Date: 2019-04-03 +File-Date: 2019-09-16 %% Type: language Subtag: aa @@ -2096,6 +2096,8 @@ Subtag: ais Description: Nataoran Amis Added: 2009-07-29 +Deprecated: 2019-04-16 +Comments: see ami, szy %% Type: language Subtag: ait @@ -2633,6 +2635,7 @@ Type: language Subtag: ant Description: Antakarinya +Description: Antikarinya Added: 2009-07-29 %% Type: language @@ -3094,6 +3097,8 @@ Subtag: asd Description: Asas Added: 2009-07-29 +Deprecated: 2019-04-16 +Preferred-Value: snz %% Type: language Subtag: ase @@ -4135,7 +4140,7 @@ %% Type: language Subtag: bck -Description: Bunaba +Description: Bunuba Added: 2009-07-29 %% Type: language @@ -6930,7 +6935,7 @@ %% Type: language Subtag: bym -Description: Bidyara +Description: Bidjara Added: 2009-07-29 %% Type: language @@ -7564,6 +7569,11 @@ Added: 2009-07-29 %% Type: language +Subtag: cey +Description: Ekai Chin +Added: 2019-04-16 +%% +Type: language Subtag: cfa Description: Dijim-Bwilim Added: 2009-07-29 @@ -9439,6 +9449,7 @@ Type: language Subtag: dif Description: Dieri +Description: Diyari Added: 2009-07-29 %% Type: language @@ -9515,6 +9526,8 @@ Subtag: dit Description: Dirari Added: 2009-07-29 +Deprecated: 2019-04-29 +Preferred-Value: dif %% Type: language Subtag: diu @@ -9560,6 +9573,7 @@ Type: language Subtag: djd Description: Djamindjung +Description: Ngaliwurru Added: 2009-07-29 %% Type: language @@ -9603,6 +9617,7 @@ %% Type: language Subtag: djn +Description: Jawoyn Description: Djauan Added: 2009-07-29 %% @@ -10191,6 +10206,8 @@ Subtag: dud Description: Hun-Saare Added: 2009-07-29 +Deprecated: 2019-04-16 +Comments: see uth, uss %% Type: language Subtag: due @@ -10382,6 +10399,7 @@ Type: language Subtag: dyn Description: Dyangadi +Description: Dhanggatti Added: 2009-07-29 %% Type: language @@ -10396,6 +10414,7 @@ %% Type: language Subtag: dyy +Description: Djabugay Description: Dyaabugay Added: 2009-07-29 %% @@ -11672,7 +11691,7 @@ %% Type: language Subtag: gbd -Description: Karadjeri +Description: Karajarri Added: 2009-07-29 %% Type: language @@ -12056,7 +12075,7 @@ %% Type: language Subtag: gge -Description: Guragone +Description: Gurr-goni Added: 2009-07-29 %% Type: language @@ -12169,7 +12188,7 @@ %% Type: language Subtag: gia -Description: Kitja +Description: Kija Added: 2009-07-29 %% Type: language @@ -12955,7 +12974,7 @@ %% Type: language Subtag: gue -Description: Gurinji +Description: Gurindji Added: 2009-07-29 %% Type: language @@ -15292,6 +15311,7 @@ Type: language Subtag: jay Description: Yan-nhangu +Description: Nhangu Added: 2009-07-29 %% Type: language @@ -15488,6 +15508,7 @@ %% Type: language Subtag: jig +Description: Jingulu Description: Djingili Added: 2009-07-29 %% @@ -17222,6 +17243,7 @@ Type: language Subtag: kkp Description: Gugubera +Description: Koko-Bera Added: 2009-07-29 %% Type: language @@ -17266,6 +17288,7 @@ %% Type: language Subtag: kky +Description: Guugu Yimidhirr Description: Guguyimidjir Added: 2009-07-29 %% @@ -18320,6 +18343,7 @@ Type: language Subtag: ktd Description: Kokata +Description: Kukatha Added: 2009-07-29 %% Type: language @@ -19341,6 +19365,7 @@ Subtag: lba Description: Lui Added: 2009-07-29 +Deprecated: 2019-04-16 %% Type: language Subtag: lbb @@ -19396,7 +19421,7 @@ %% Type: language Subtag: lbn -Description: Lamet +Description: Rmeet Added: 2009-07-29 %% Type: language @@ -19446,6 +19471,7 @@ %% Type: language Subtag: lby +Description: Lamalama Description: Lamu-Lamu Added: 2009-07-29 %% @@ -20162,6 +20188,8 @@ Subtag: llo Description: Khlor Added: 2009-07-29 +Deprecated: 2019-04-16 +Preferred-Value: ngt %% Type: language Subtag: llp @@ -20654,6 +20682,11 @@ Macrolanguage: luy %% Type: language +Subtag: lsn +Description: Tibetan Sign Language +Added: 2019-04-16 +%% +Type: language Subtag: lso Description: Laos Sign Language Added: 2009-07-29 @@ -20680,6 +20713,11 @@ Added: 2009-07-29 %% Type: language +Subtag: lsv +Description: Sivia Sign Language +Added: 2019-04-16 +%% +Type: language Subtag: lsy Description: Mauritian Sign Language Added: 2010-03-11 @@ -20848,6 +20886,11 @@ Added: 2009-07-29 %% Type: language +Subtag: lvi +Description: Lavi +Added: 2019-04-16 +%% +Type: language Subtag: lvk Description: Lavukaleve Added: 2009-07-29 @@ -21454,7 +21497,7 @@ %% Type: language Subtag: mec -Description: Mara +Description: Marra Added: 2009-07-29 %% Type: language @@ -21523,7 +21566,7 @@ %% Type: language Subtag: mep -Description: Miriwung +Description: Miriwoong Added: 2009-07-29 %% Type: language @@ -21660,7 +21703,7 @@ %% Type: language Subtag: mfr -Description: Marithiel +Description: Marrithiyel Added: 2009-07-29 %% Type: language @@ -22853,12 +22896,13 @@ %% Type: language Subtag: mpb +Description: Malak Malak Description: Mullukmulluk Added: 2009-07-29 %% Type: language Subtag: mpc -Description: Mangarayi +Description: Mangarrayi Added: 2009-07-29 %% Type: language @@ -22889,6 +22933,7 @@ Type: language Subtag: mpj Description: Martu Wangka +Description: Wangkajunga Added: 2009-07-29 %% Type: language @@ -24015,6 +24060,8 @@ Subtag: myd Description: Maramba Added: 2009-07-29 +Deprecated: 2019-04-16 +Preferred-Value: aog %% Type: language Subtag: mye @@ -24040,6 +24087,7 @@ Subtag: myi Description: Mina (India) Added: 2009-07-29 +Deprecated: 2019-04-16 %% Type: language Subtag: myj @@ -24375,7 +24423,7 @@ %% Type: language Subtag: nay -Description: Narrinyeri +Description: Ngarrindjeri Added: 2009-07-29 %% Type: language @@ -24432,7 +24480,7 @@ %% Type: language Subtag: nbj -Description: Ngarinman +Description: Ngarinyman Added: 2009-07-29 %% Type: language @@ -24467,7 +24515,7 @@ %% Type: language Subtag: nbr -Description: Numana-Nunku-Gbantu-Numbu +Description: Numana Added: 2009-07-29 %% Type: language @@ -24559,7 +24607,7 @@ %% Type: language Subtag: nck -Description: Nakara +Description: Na-kara Added: 2009-07-29 %% Type: language @@ -24931,7 +24979,7 @@ %% Type: language Subtag: ngh -Description: Nǀu +Description: Nǁng Added: 2009-07-29 %% Type: language @@ -25176,7 +25224,7 @@ %% Type: language Subtag: nig -Description: Ngalakan +Description: Ngalakgan Added: 2009-07-29 %% Type: language @@ -25798,6 +25846,8 @@ Subtag: nns Description: Ningye Added: 2009-07-29 +Deprecated: 2019-04-16 +Preferred-Value: nbr %% Type: language Subtag: nnt @@ -26658,7 +26708,7 @@ %% Type: language Subtag: nyh -Description: Nyigina +Description: Nyikina Added: 2009-07-29 %% Type: language @@ -26713,7 +26763,7 @@ %% Type: language Subtag: nys -Description: Nyunga +Description: Nyungar Added: 2009-07-29 %% Type: language @@ -28707,6 +28757,11 @@ Added: 2009-07-29 %% Type: language +Subtag: pnd +Description: Mpinda +Added: 2019-04-16 +%% +Type: language Subtag: pne Description: Western Penan Added: 2009-07-29 @@ -28794,6 +28849,7 @@ %% Type: language Subtag: pnw +Description: Banyjima Description: Panytyima Added: 2009-07-29 %% @@ -29251,7 +29307,8 @@ %% Type: language Subtag: pti -Description: Pintiini +Description: Pindiini +Description: Wangkatha Added: 2009-07-29 %% Type: language @@ -30133,6 +30190,7 @@ %% Type: language Subtag: ril +Description: Riang Lang Description: Riang (Myanmar) Added: 2009-07-29 %% @@ -30153,7 +30211,7 @@ %% Type: language Subtag: rit -Description: Ritarungo +Description: Ritharrngu Added: 2009-07-29 %% Type: language @@ -30219,7 +30277,7 @@ %% Type: language Subtag: rmb -Description: Rembarunga +Description: Rembarrnga Added: 2009-07-29 %% Type: language @@ -30641,6 +30699,7 @@ Type: language Subtag: rxw Description: Karuwali +Description: Garuwali Added: 2013-09-10 %% Type: language @@ -32206,7 +32265,7 @@ %% Type: language Subtag: snz -Description: Sinsauru +Description: Kou Added: 2009-07-29 %% Type: language @@ -32883,6 +32942,7 @@ Subtag: suj Description: Shubi Added: 2009-07-29 +Comments: see also xsj %% Type: language Subtag: suk @@ -33312,6 +33372,11 @@ Added: 2009-07-29 %% Type: language +Subtag: szy +Description: Sakizaya +Added: 2019-04-16 +%% +Type: language Subtag: taa Description: Lower Tanana Added: 2009-07-29 @@ -33465,6 +33530,7 @@ %% Type: language Subtag: tbh +Description: Dharawal Description: Thurawal Added: 2009-07-29 %% @@ -33644,6 +33710,7 @@ Type: language Subtag: tcs Description: Torres Strait Creole +Description: Yumplatok Added: 2009-07-29 %% Type: language @@ -34067,6 +34134,7 @@ %% Type: language Subtag: thd +Description: Kuuk Thaayorre Description: Thayore Added: 2009-07-29 %% @@ -34310,6 +34378,11 @@ Added: 2009-07-29 %% Type: language +Subtag: tjj +Description: Tjungundji +Added: 2019-04-16 +%% +Type: language Subtag: tjl Description: Tai Laing Added: 2012-08-12 @@ -34330,6 +34403,11 @@ Added: 2009-07-29 %% Type: language +Subtag: tjp +Description: Tjupany +Added: 2019-04-16 +%% +Type: language Subtag: tjs Description: Southern Tujia Added: 2009-07-29 @@ -35679,6 +35757,11 @@ Added: 2009-07-29 %% Type: language +Subtag: tvx +Description: Taivoan +Added: 2019-04-16 +%% +Type: language Subtag: tvy Description: Timor Pidgin Added: 2009-07-29 @@ -36230,7 +36313,7 @@ %% Type: language Subtag: ulk -Description: Meriam +Description: Meriam Mir Added: 2009-07-29 %% Type: language @@ -36280,6 +36363,7 @@ %% Type: language Subtag: umg +Description: Morrobalama Description: Umbuygamu Added: 2009-07-29 %% @@ -36550,6 +36634,11 @@ Added: 2009-07-29 %% Type: language +Subtag: uss +Description: us-Saare +Added: 2019-04-16 +%% +Type: language Subtag: usu Description: Uya Added: 2009-07-29 @@ -36565,6 +36654,11 @@ Added: 2009-07-29 %% Type: language +Subtag: uth +Description: ut-Hun +Added: 2019-04-16 +%% +Type: language Subtag: utp Description: Amba (Solomon Islands) Added: 2009-07-29 @@ -37178,7 +37272,7 @@ %% Type: language Subtag: waq -Description: Wageman +Description: Wagiman Added: 2009-07-29 %% Type: language @@ -37301,7 +37395,7 @@ %% Type: language Subtag: wbt -Description: Wanman +Description: Warnman Added: 2009-07-29 %% Type: language @@ -37448,6 +37542,7 @@ %% Type: language Subtag: wgg +Description: Wangkangurru Description: Wangganguru Added: 2009-07-29 %% @@ -37521,7 +37616,7 @@ %% Type: language Subtag: wig -Description: Wik-Ngathana +Description: Wik Ngathan Added: 2009-07-29 %% Type: language @@ -37625,6 +37720,11 @@ Added: 2009-07-29 %% Type: language +Subtag: wkr +Description: Keerray-Woorroong +Added: 2019-04-16 +%% +Type: language Subtag: wku Description: Kunduvadi Added: 2009-07-29 @@ -37857,10 +37957,12 @@ Type: language Subtag: wny Description: Wanyi +Description: Waanyi Added: 2012-08-12 %% Type: language Subtag: woa +Description: Kuwema Description: Tyaraity Added: 2009-07-29 %% @@ -37951,6 +38053,7 @@ %% Type: language Subtag: wrb +Description: Waluwarra Description: Warluwara Added: 2009-07-29 %% @@ -37962,11 +38065,12 @@ Type: language Subtag: wrg Description: Warungu +Description: Gudjal Added: 2009-07-29 %% Type: language Subtag: wrh -Description: Wiradhuri +Description: Wiradjuri Added: 2009-07-29 %% Type: language @@ -38439,6 +38543,7 @@ %% Type: language Subtag: xby +Description: Batjala Description: Batyala Added: 2013-09-10 %% @@ -38998,7 +39103,7 @@ %% Type: language Subtag: xmh -Description: Kuku-Muminh +Description: Kugu-Muminh Added: 2009-07-29 %% Type: language @@ -39423,8 +39528,7 @@ Subtag: xsj Description: Subi Added: 2009-07-29 -Deprecated: 2015-02-12 -Preferred-Value: suj +Comments: see also suj %% Type: language Subtag: xsl @@ -40258,6 +40362,7 @@ %% Type: language Subtag: yin +Description: Riang Lai Description: Yinchia Added: 2009-07-29 %% @@ -41562,12 +41667,13 @@ %% Type: language Subtag: zml -Description: Madngele +Description: Matngala Added: 2009-07-29 %% Type: language Subtag: zmm Description: Marimanindji +Description: Marramaninyshi Added: 2009-07-29 %% Type: language @@ -43019,6 +43125,13 @@ Prefix: sgn %% Type: extlang +Subtag: lsn +Description: Tibetan Sign Language +Added: 2019-04-16 +Preferred-Value: lsn +Prefix: sgn +%% +Type: extlang Subtag: lso Description: Laos Sign Language Added: 2009-07-29 @@ -43041,6 +43154,13 @@ Prefix: sgn %% Type: extlang +Subtag: lsv +Description: Sivia Sign Language +Added: 2019-04-16 +Preferred-Value: lsv +Prefix: sgn +%% +Type: extlang Subtag: lsy Description: Mauritian Sign Language Added: 2010-03-11 @@ -43966,6 +44086,11 @@ Added: 2005-10-16 %% Type: script +Subtag: Chrs +Description: Chorasmian +Added: 2019-09-11 +%% +Type: script Subtag: Cirt Description: Cirth Added: 2005-10-16 @@ -44002,6 +44127,11 @@ Added: 2005-10-16 %% Type: script +Subtag: Diak +Description: Dives Akuru +Added: 2019-09-11 +%% +Type: script Subtag: Dogr Description: Dogra Added: 2017-01-13 @@ -44839,6 +44969,11 @@ Added: 2005-10-16 %% Type: script +Subtag: Yezi +Description: Yezidi +Added: 2019-09-11 +%% +Type: script Subtag: Yiii Description: Yi Added: 2005-10-16 @@ -45683,7 +45818,7 @@ %% Type: region Subtag: MK -Description: The Former Yugoslav Republic of Macedonia +Description: North Macedonia Added: 2005-10-16 %% Type: region
--- a/make/devkit/Tools.gmk Fri Oct 04 13:44:11 2019 -0700 +++ b/make/devkit/Tools.gmk Mon Oct 14 21:01:25 2019 +0000 @@ -79,20 +79,19 @@ # Define external dependencies # Latest that could be made to work. -GCC_VER := 8.2.0 -ifeq ($(GCC_VER), 8.2.0) - gcc_ver := gcc-8.2.0 - binutils_ver := binutils-2.30 - ccache_ver := ccache-3.5.1a - CCACHE_DIRNAME := ccache-3.5.1 +GCC_VER := 8.3.0 +ifeq ($(GCC_VER), 8.3.0) + gcc_ver := gcc-8.3.0 + binutils_ver := binutils-2.32 + ccache_ver := 3.7.3 mpfr_ver := mpfr-3.1.5 gmp_ver := gmp-6.1.2 mpc_ver := mpc-1.0.3 - gdb_ver := gdb-8.2.1 + gdb_ver := gdb-8.3 else ifeq ($(GCC_VER), 7.3.0) gcc_ver := gcc-7.3.0 binutils_ver := binutils-2.30 - ccache_ver := ccache-3.3.6 + ccache_ver := 3.3.6 mpfr_ver := mpfr-3.1.5 gmp_ver := gmp-6.1.2 mpc_ver := mpc-1.0.3 @@ -100,7 +99,7 @@ else ifeq ($(GCC_VER), 4.9.2) gcc_ver := gcc-4.9.2 binutils_ver := binutils-2.25 - ccache_ver := ccache-3.2.1 + ccache_ver := 3.2.1 mpfr_ver := mpfr-3.0.1 gmp_ver := gmp-4.3.2 mpc_ver := mpc-1.0.1 @@ -111,7 +110,7 @@ GCC := http://ftp.gnu.org/pub/gnu/gcc/$(gcc_ver)/$(gcc_ver).tar.xz BINUTILS := http://ftp.gnu.org/pub/gnu/binutils/$(binutils_ver).tar.xz -CCACHE := https://samba.org/ftp/ccache/$(ccache_ver).tar.xz +CCACHE := https://github.com/ccache/ccache/releases/download/v$(ccache_ver)/ccache-$(ccache_ver).tar.xz MPFR := https://www.mpfr.org/${mpfr_ver}/${mpfr_ver}.tar.bz2 GMP := http://ftp.gnu.org/pub/gnu/gmp/${gmp_ver}.tar.bz2 MPC := http://ftp.gnu.org/pub/gnu/mpc/${mpc_ver}.tar.gz
--- a/make/devkit/createWindowsDevkit2017.sh Fri Oct 04 13:44:11 2019 -0700 +++ b/make/devkit/createWindowsDevkit2017.sh Mon Oct 14 21:01:25 2019 +0000 @@ -32,10 +32,7 @@ VS_VERSION_NUM_NODOT="150" VS_DLL_VERSION="140" SDK_VERSION="10" -SDK_FULL_VERSION="10.0.16299.0" MSVC_DIR="Microsoft.VC141.CRT" -MSVC_FULL_VERSION="14.12.25827" -REDIST_FULL_VERSION="14.12.25810" SCRIPT_DIR="$(cd "$(dirname $0)" > /dev/null && pwd)" BUILD_DIR="${SCRIPT_DIR}/../../build/devkit"
--- a/make/gensrc/Gensrc-jdk.internal.vm.compiler.management.gmk Fri Oct 04 13:44:11 2019 -0700 +++ b/make/gensrc/Gensrc-jdk.internal.vm.compiler.management.gmk Mon Oct 14 21:01:25 2019 +0000 @@ -73,7 +73,7 @@ ($(CD) $(GENSRC_DIR)/META-INF/providers && \ p=""; \ impl=""; \ - for i in $$($(GREP) '^' * | $(SORT) -t ':' -k 2 | $(SED) 's/:.*//'); do \ + for i in $$($(NAWK) '$$0=FILENAME" "$$0' * | $(SORT) -k 2 | $(SED) 's/ .*//'); do \ c=$$($(CAT) $$i | $(TR) -d '\n\r'); \ if test x$$p != x$$c; then \ if test x$$p != x; then \
--- a/make/jdk/src/classes/build/tools/charsetmapping/SBCS.java Fri Oct 04 13:44:11 2019 -0700 +++ b/make/jdk/src/classes/build/tools/charsetmapping/SBCS.java Mon Oct 14 21:01:25 2019 +0000 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2015, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2008, 2019, 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 @@ -46,6 +46,7 @@ String hisName = cs.hisName; String pkgName = cs.pkgName; boolean isASCII = cs.isASCII; + boolean isLatin1Decodable = true; StringBuilder b2cSB = new StringBuilder(); StringBuilder b2cNRSB = new StringBuilder(); @@ -69,6 +70,9 @@ c2bOff += 0x100; c2bIndex[e.cp>>8] = 1; } + if (e.cp > 0xFF) { + isLatin1Decodable = false; + } } Formatter fm = new Formatter(b2cSB); @@ -178,6 +182,9 @@ if (line.indexOf("$ASCIICOMPATIBLE$") != -1) { line = line.replace("$ASCIICOMPATIBLE$", isASCII ? "true" : "false"); } + if (line.indexOf("$LATIN1DECODABLE$") != -1) { + line = line.replace("$LATIN1DECODABLE$", isLatin1Decodable ? "true" : "false"); + } if (line.indexOf("$B2CTABLE$") != -1) { line = line.replace("$B2CTABLE$", b2c); }
--- a/make/scripts/compare.sh Fri Oct 04 13:44:11 2019 -0700 +++ b/make/scripts/compare.sh Mon Oct 14 21:01:25 2019 +0000 @@ -34,6 +34,9 @@ exit 1 fi +# Make sure all shell commands are executed with the C locale +export LC_ALL=C + if [ "$OPENJDK_TARGET_OS" = "macosx" ]; then FULLDUMP_CMD="$OTOOL -v -V -h -X -d" LDD_CMD="$OTOOL -L" @@ -110,7 +113,7 @@ " fi elif [ "$OPENJDK_TARGET_OS" = "macosx" ]; then - DIS_DIFF_FILTER="LANG=C $SED \ + DIS_DIFF_FILTER="$SED \ -e 's/0x[0-9a-f]\{3,16\}/<HEXSTR>/g' -e 's/^[0-9a-f]\{12,20\}/<ADDR>/' \ -e 's/-20[0-9][0-9]-[0-1][0-9]-[0-3][0-9]-[0-2][0-9]\{5\}/<DATE>/g' \ -e 's/), built on .*/), <DATE>/' \ @@ -134,7 +137,7 @@ if [[ "$THIS_FILE" = *"META-INF/MANIFEST.MF" ]]; then # Filter out date string, ant version and java version differences. - TMP=$(LC_ALL=C $DIFF $OTHER_FILE $THIS_FILE | \ + TMP=$($DIFF $OTHER_FILE $THIS_FILE | \ $GREP '^[<>]' | \ $SED -e '/[<>] Ant-Version: Apache Ant .*/d' \ -e '/[<>] Created-By: .* (Oracle [Corpatin)]*/d' \ @@ -142,7 +145,7 @@ -e '/[<>].*[0-9]\{4\}_[0-9]\{2\}_[0-9]\{2\}_[0-9]\{2\}_[0-9]\{2\}-b[0-9]\{2\}.*/d') fi if test "x$SUFFIX" = "xjava"; then - TMP=$(LC_ALL=C $DIFF $OTHER_FILE $THIS_FILE | \ + TMP=$($DIFF $OTHER_FILE $THIS_FILE | \ $GREP '^[<>]' | \ $SED -e '/[<>] \* from.*\.idl/d' \ -e '/[<>] .*[0-9]\{4\}_[0-9]\{2\}_[0-9]\{2\}_[0-9]\{2\}_[0-9]\{2\}-b[0-9]\{2\}.*/d' \ @@ -197,7 +200,7 @@ fi if test "x$SUFFIX" = "xproperties"; then # Filter out date string differences. - TMP=$(LC_ALL=C $DIFF $OTHER_FILE $THIS_FILE | \ + TMP=$($DIFF $OTHER_FILE $THIS_FILE | \ $GREP '^[<>]' | \ $SED -e '/[<>].*[0-9]\{4\}-[0-9]\{2\}-[0-9]\{2\}-[0-9]\{6\}.*/d') fi @@ -207,7 +210,7 @@ -e 's/<font size=-1>/<font size=\"-1\">/g'" $CAT $THIS_FILE | eval "$HTML_FILTER" > $THIS_FILE.filtered $CAT $OTHER_FILE | eval "$HTML_FILTER" > $OTHER_FILE.filtered - TMP=$(LC_ALL=C $DIFF $OTHER_FILE.filtered $THIS_FILE.filtered | \ + TMP=$($DIFF $OTHER_FILE.filtered $THIS_FILE.filtered | \ $GREP '^[<>]' | \ $SED -e '/[<>] <!-- Generated by javadoc .* on .* -->/d' \ -e '/[<>] <meta name="date" content=".*">/d' ) @@ -554,11 +557,11 @@ CONTENTS_DIFF_FILE=$WORK_DIR/$ZIP_FILE.diff # On solaris, there is no -q option. if [ "$OPENJDK_TARGET_OS" = "solaris" ]; then - LC_ALL=C $DIFF -r $OTHER_UNZIPDIR $THIS_UNZIPDIR \ + $DIFF -r $OTHER_UNZIPDIR $THIS_UNZIPDIR \ | $GREP -v -e "^<" -e "^>" -e "^Common subdirectories:" \ > $CONTENTS_DIFF_FILE else - LC_ALL=C $DIFF -rq $OTHER_UNZIPDIR $THIS_UNZIPDIR > $CONTENTS_DIFF_FILE + $DIFF -rq $OTHER_UNZIPDIR $THIS_UNZIPDIR > $CONTENTS_DIFF_FILE fi ONLY_OTHER=$($GREP "^Only in $OTHER_UNZIPDIR" $CONTENTS_DIFF_FILE) @@ -605,11 +608,11 @@ if [ -n "$SHOW_DIFFS" ]; then for i in $(cat $WORK_DIR/$ZIP_FILE.difflist) ; do if [ -f "${OTHER_UNZIPDIR}/$i.javap" ]; then - LC_ALL=C $DIFF ${OTHER_UNZIPDIR}/$i.javap ${THIS_UNZIPDIR}/$i.javap + $DIFF ${OTHER_UNZIPDIR}/$i.javap ${THIS_UNZIPDIR}/$i.javap elif [ -f "${OTHER_UNZIPDIR}/$i.cleaned" ]; then - LC_ALL=C $DIFF ${OTHER_UNZIPDIR}/$i.cleaned ${THIS_UNZIPDIR}/$i + $DIFF ${OTHER_UNZIPDIR}/$i.cleaned ${THIS_UNZIPDIR}/$i else - LC_ALL=C $DIFF ${OTHER_UNZIPDIR}/$i ${THIS_UNZIPDIR}/$i + $DIFF ${OTHER_UNZIPDIR}/$i ${THIS_UNZIPDIR}/$i fi done fi @@ -642,7 +645,7 @@ $JMOD list $THIS_JMOD | sort > $THIS_JMOD_LIST $JMOD list $OTHER_JMOD | sort > $OTHER_JMOD_LIST JMOD_LIST_DIFF_FILE=$WORK_DIR/$JMOD_FILE.list.diff - LC_ALL=C $DIFF $THIS_JMOD_LIST $OTHER_JMOD_LIST > $JMOD_LIST_DIFF_FILE + $DIFF $THIS_JMOD_LIST $OTHER_JMOD_LIST > $JMOD_LIST_DIFF_FILE ONLY_THIS=$($GREP "^<" $JMOD_LIST_DIFF_FILE) ONLY_OTHER=$($GREP "^>" $JMOD_LIST_DIFF_FILE) @@ -924,7 +927,7 @@ > $WORK_FILE_BASE.symbols.this fi - LC_ALL=C $DIFF $WORK_FILE_BASE.symbols.other $WORK_FILE_BASE.symbols.this > $WORK_FILE_BASE.symbols.diff + $DIFF $WORK_FILE_BASE.symbols.other $WORK_FILE_BASE.symbols.this > $WORK_FILE_BASE.symbols.diff if [ -s $WORK_FILE_BASE.symbols.diff ]; then SYM_MSG=" diff " if [[ "$ACCEPTED_SYM_DIFF" != *"$BIN_FILE"* ]]; then @@ -964,9 +967,9 @@ | $UNIQ > $WORK_FILE_BASE.deps.this.uniq) (cd $FILE_WORK_DIR && $RM -f $NAME) - LC_ALL=C $DIFF $WORK_FILE_BASE.deps.other $WORK_FILE_BASE.deps.this \ + $DIFF $WORK_FILE_BASE.deps.other $WORK_FILE_BASE.deps.this \ > $WORK_FILE_BASE.deps.diff - LC_ALL=C $DIFF $WORK_FILE_BASE.deps.other.uniq $WORK_FILE_BASE.deps.this.uniq \ + $DIFF $WORK_FILE_BASE.deps.other.uniq $WORK_FILE_BASE.deps.this.uniq \ > $WORK_FILE_BASE.deps.diff.uniq if [ -s $WORK_FILE_BASE.deps.diff ]; then @@ -1016,7 +1019,7 @@ > $WORK_FILE_BASE.fulldump.this 2>&1 & wait - LC_ALL=C $DIFF $WORK_FILE_BASE.fulldump.other $WORK_FILE_BASE.fulldump.this \ + $DIFF $WORK_FILE_BASE.fulldump.other $WORK_FILE_BASE.fulldump.this \ > $WORK_FILE_BASE.fulldump.diff if [ -s $WORK_FILE_BASE.fulldump.diff ]; then @@ -1063,7 +1066,7 @@ | eval "$this_DIS_DIFF_FILTER" > $WORK_FILE_BASE.dis.this 2>&1 & wait - LC_ALL=C $DIFF $WORK_FILE_BASE.dis.other $WORK_FILE_BASE.dis.this > $WORK_FILE_BASE.dis.diff + $DIFF $WORK_FILE_BASE.dis.other $WORK_FILE_BASE.dis.this > $WORK_FILE_BASE.dis.diff if [ -s $WORK_FILE_BASE.dis.diff ]; then DIS_DIFF_SIZE=$(ls -n $WORK_FILE_BASE.dis.diff | awk '{print $5}')
--- a/src/hotspot/cpu/aarch64/aarch64.ad Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/aarch64.ad Mon Oct 14 21:01:25 2019 +0000 @@ -2513,17 +2513,8 @@ __ INSN(REG, as_Register(BASE)); \ } -typedef void (MacroAssembler::* mem_insn)(Register Rt, const Address &adr); -typedef void (MacroAssembler::* mem_float_insn)(FloatRegister Rt, const Address &adr); -typedef void (MacroAssembler::* mem_vector_insn)(FloatRegister Rt, - MacroAssembler::SIMD_RegVariant T, const Address &adr); - - // Used for all non-volatile memory accesses. The use of - // $mem->opcode() to discover whether this pattern uses sign-extended - // offsets is something of a kludge. - static void loadStore(MacroAssembler masm, mem_insn insn, - Register reg, int opcode, - Register base, int index, int size, int disp) + +static Address mem2address(int opcode, Register base, int index, int size, int disp) { Address::extend scale; @@ -2542,16 +2533,34 @@ } if (index == -1) { - (masm.*insn)(reg, Address(base, disp)); + return Address(base, disp); } else { assert(disp == 0, "unsupported address mode: disp = %d", disp); - (masm.*insn)(reg, Address(base, as_Register(index), scale)); + return Address(base, as_Register(index), scale); } } + +typedef void (MacroAssembler::* mem_insn)(Register Rt, const Address &adr); +typedef void (MacroAssembler::* mem_insn2)(Register Rt, Register adr); +typedef void (MacroAssembler::* mem_float_insn)(FloatRegister Rt, const Address &adr); +typedef void (MacroAssembler::* mem_vector_insn)(FloatRegister Rt, + MacroAssembler::SIMD_RegVariant T, const Address &adr); + + // Used for all non-volatile memory accesses. The use of + // $mem->opcode() to discover whether this pattern uses sign-extended + // offsets is something of a kludge. + static void loadStore(MacroAssembler masm, mem_insn insn, + Register reg, int opcode, + Register base, int index, int size, int disp) + { + Address addr = mem2address(opcode, base, index, size, disp); + (masm.*insn)(reg, addr); + } + static void loadStore(MacroAssembler masm, mem_float_insn insn, - FloatRegister reg, int opcode, - Register base, int index, int size, int disp) + FloatRegister reg, int opcode, + Register base, int index, int size, int disp) { Address::extend scale; @@ -2573,8 +2582,8 @@ } static void loadStore(MacroAssembler masm, mem_vector_insn insn, - FloatRegister reg, MacroAssembler::SIMD_RegVariant T, - int opcode, Register base, int index, int size, int disp) + FloatRegister reg, MacroAssembler::SIMD_RegVariant T, + int opcode, Register base, int index, int size, int disp) { if (index == -1) { (masm.*insn)(reg, T, Address(base, disp)); @@ -3791,7 +3800,7 @@ static const int hi[Op_RegL + 1] = { // enum name 0, // Op_Node 0, // Op_Set - OptoReg::Bad, // Op_RegN + OptoReg::Bad, // Op_RegN OptoReg::Bad, // Op_RegI R0_H_num, // Op_RegP OptoReg::Bad, // Op_RegF @@ -6923,7 +6932,7 @@ instruct loadP(iRegPNoSp dst, memory mem) %{ match(Set dst (LoadP mem)); - predicate(!needs_acquiring_load(n)); + predicate(!needs_acquiring_load(n) && (n->as_Load()->barrier_data() == 0)); ins_cost(4 * INSN_COST); format %{ "ldr $dst, $mem\t# ptr" %} @@ -7616,6 +7625,7 @@ instruct loadP_volatile(iRegPNoSp dst, /* sync_memory*/indirect mem) %{ match(Set dst (LoadP mem)); + predicate(n->as_Load()->barrier_data() == 0); ins_cost(VOLATILE_REF_COST); format %{ "ldar $dst, $mem\t# ptr" %} @@ -8552,6 +8562,7 @@ instruct compareAndSwapP(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ match(Set res (CompareAndSwapP mem (Binary oldval newval))); + predicate(n->as_LoadStore()->barrier_data() == 0); ins_cost(2 * VOLATILE_REF_COST); effect(KILL cr); @@ -8665,7 +8676,7 @@ instruct compareAndSwapPAcq(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ - predicate(needs_acquiring_load_exclusive(n)); + predicate(needs_acquiring_load_exclusive(n) && (n->as_LoadStore()->barrier_data() == 0)); match(Set res (CompareAndSwapP mem (Binary oldval newval))); ins_cost(VOLATILE_REF_COST); @@ -8796,6 +8807,7 @@ %} instruct compareAndExchangeP(iRegPNoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ + predicate(n->as_LoadStore()->barrier_data() == 0); match(Set res (CompareAndExchangeP mem (Binary oldval newval))); ins_cost(2 * VOLATILE_REF_COST); effect(TEMP_DEF res, KILL cr); @@ -8895,7 +8907,7 @@ %} instruct compareAndExchangePAcq(iRegPNoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ - predicate(needs_acquiring_load_exclusive(n)); + predicate(needs_acquiring_load_exclusive(n) && (n->as_LoadStore()->barrier_data() == 0)); match(Set res (CompareAndExchangeP mem (Binary oldval newval))); ins_cost(VOLATILE_REF_COST); effect(TEMP_DEF res, KILL cr); @@ -8996,6 +9008,7 @@ %} instruct weakCompareAndSwapP(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ + predicate(n->as_LoadStore()->barrier_data() == 0); match(Set res (WeakCompareAndSwapP mem (Binary oldval newval))); ins_cost(2 * VOLATILE_REF_COST); effect(KILL cr); @@ -9103,8 +9116,8 @@ %} instruct weakCompareAndSwapPAcq(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ - predicate(needs_acquiring_load_exclusive(n)); match(Set res (WeakCompareAndSwapP mem (Binary oldval newval))); + predicate(needs_acquiring_load_exclusive(n) && (n->as_LoadStore()->barrier_data() == 0)); ins_cost(VOLATILE_REF_COST); effect(KILL cr); format %{ @@ -9154,6 +9167,7 @@ %} instruct get_and_setP(indirect mem, iRegP newv, iRegPNoSp prev) %{ + predicate(n->as_LoadStore()->barrier_data() == 0); match(Set prev (GetAndSetP mem newv)); ins_cost(2 * VOLATILE_REF_COST); format %{ "atomic_xchg $prev, $newv, [$mem]" %} @@ -9197,7 +9211,7 @@ %} instruct get_and_setPAcq(indirect mem, iRegP newv, iRegPNoSp prev) %{ - predicate(needs_acquiring_load_exclusive(n)); + predicate(needs_acquiring_load_exclusive(n) && (n->as_LoadStore()->barrier_data() == 0)); match(Set prev (GetAndSetP mem newv)); ins_cost(VOLATILE_REF_COST); format %{ "atomic_xchg_acq $prev, $newv, [$mem]" %}
--- a/src/hotspot/cpu/aarch64/abstractInterpreter_aarch64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/abstractInterpreter_aarch64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -26,6 +26,7 @@ #include "precompiled.hpp" #include "interpreter/interpreter.hpp" #include "oops/constMethod.hpp" +#include "oops/klass.inline.hpp" #include "oops/method.hpp" #include "runtime/frame.inline.hpp" #include "utilities/align.hpp"
--- a/src/hotspot/cpu/aarch64/compiledIC_aarch64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/compiledIC_aarch64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -1,5 +1,5 @@ /* - * Copyright (c) 1997, 2018, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 1997, 2019, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2014, 2018, Red Hat Inc. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * @@ -162,16 +162,12 @@ // Creation also verifies the object. NativeMovConstReg* method_holder = nativeMovConstReg_at(stub + NativeInstruction::instruction_size); -#ifndef PRODUCT - NativeGeneralJump* jump = nativeGeneralJump_at(method_holder->next_instruction_address()); - // read the value once - volatile intptr_t data = method_holder->data(); - assert(data == 0 || data == (intptr_t)callee(), - "a) MT-unsafe modification of inline cache"); - assert(data == 0 || jump->jump_destination() == entry, - "b) MT-unsafe modification of inline cache"); +#ifdef ASSERT + NativeGeneralJump* jump = nativeGeneralJump_at(method_holder->next_instruction_address()); + verify_mt_safe(callee, entry, method_holder, jump); #endif + // Update stub. method_holder->set_data((intptr_t)callee()); NativeGeneralJump::insert_unconditional(method_holder->next_instruction_address(), entry); @@ -189,6 +185,10 @@ NativeMovConstReg* method_holder = nativeMovConstReg_at(stub + NativeInstruction::instruction_size); method_holder->set_data(0); + if (!static_stub->is_aot()) { + NativeJump* jump = nativeJump_at(method_holder->next_instruction_address()); + jump->set_jump_destination((address)-1); + } } //-----------------------------------------------------------------------------
--- a/src/hotspot/cpu/aarch64/gc/z/zBarrierSetAssembler_aarch64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/gc/z/zBarrierSetAssembler_aarch64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -24,22 +24,23 @@ #include "precompiled.hpp" #include "asm/macroAssembler.inline.hpp" #include "code/codeBlob.hpp" +#include "code/vmreg.inline.hpp" #include "gc/z/zBarrier.inline.hpp" #include "gc/z/zBarrierSet.hpp" #include "gc/z/zBarrierSetAssembler.hpp" #include "gc/z/zBarrierSetRuntime.hpp" +#include "gc/z/zThreadLocalData.hpp" #include "memory/resourceArea.hpp" +#include "runtime/sharedRuntime.hpp" +#include "utilities/macros.hpp" #ifdef COMPILER1 #include "c1/c1_LIRAssembler.hpp" #include "c1/c1_MacroAssembler.hpp" #include "gc/z/c1/zBarrierSetC1.hpp" #endif // COMPILER1 - -#include "gc/z/zThreadLocalData.hpp" - -ZBarrierSetAssembler::ZBarrierSetAssembler() : - _load_barrier_slow_stub(), - _load_barrier_weak_slow_stub() {} +#ifdef COMPILER2 +#include "gc/z/c2/zBarrierSetC2.hpp" +#endif // COMPILER2 #ifdef PRODUCT #define BLOCK_COMMENT(str) /* nothing */ @@ -66,7 +67,7 @@ assert_different_registers(rscratch1, rscratch2, src.base()); assert_different_registers(rscratch1, rscratch2, dst); - RegSet savedRegs = RegSet::range(r0,r28) - RegSet::of(dst, rscratch1, rscratch2); + RegSet savedRegs = RegSet::range(r0, r28) - RegSet::of(dst, rscratch1, rscratch2); Label done; @@ -206,7 +207,8 @@ // The Address offset is too large to direct load - -784. Our range is +127, -128. __ mov(tmp, (long int)(in_bytes(ZThreadLocalData::address_bad_mask_offset()) - - in_bytes(JavaThread::jni_environment_offset()))); + in_bytes(JavaThread::jni_environment_offset()))); + // Load address bad mask __ add(tmp, jni_env, tmp); __ ldr(tmp, Address(tmp)); @@ -294,12 +296,12 @@ __ prologue("zgc_load_barrier stub", false); // We don't use push/pop_clobbered_registers() - we need to pull out the result from r0. - for (int i = 0; i < 32; i +=2) { - __ stpd(as_FloatRegister(i), as_FloatRegister(i+1), Address(__ pre(sp,-16))); + for (int i = 0; i < 32; i += 2) { + __ stpd(as_FloatRegister(i), as_FloatRegister(i + 1), Address(__ pre(sp,-16))); } - RegSet saveRegs = RegSet::range(r0,r28) - RegSet::of(r0); - __ push(saveRegs, sp); + const RegSet save_regs = RegSet::range(r1, r28); + __ push(save_regs, sp); // Setup arguments __ load_parameter(0, c_rarg0); @@ -307,98 +309,161 @@ __ call_VM_leaf(ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded_addr(decorators), 2); - __ pop(saveRegs, sp); + __ pop(save_regs, sp); - for (int i = 30; i >0; i -=2) { - __ ldpd(as_FloatRegister(i), as_FloatRegister(i+1), Address(__ post(sp, 16))); - } + for (int i = 30; i >= 0; i -= 2) { + __ ldpd(as_FloatRegister(i), as_FloatRegister(i + 1), Address(__ post(sp, 16))); + } __ epilogue(); } #endif // COMPILER1 +#ifdef COMPILER2 + +OptoReg::Name ZBarrierSetAssembler::refine_register(const Node* node, OptoReg::Name opto_reg) { + if (!OptoReg::is_reg(opto_reg)) { + return OptoReg::Bad; + } + + const VMReg vm_reg = OptoReg::as_VMReg(opto_reg); + if (vm_reg->is_FloatRegister()) { + return opto_reg & ~1; + } + + return opto_reg; +} + #undef __ -#define __ cgen->assembler()-> +#define __ _masm-> + +class ZSaveLiveRegisters { +private: + MacroAssembler* const _masm; + RegSet _gp_regs; + RegSet _fp_regs; + +public: + void initialize(ZLoadBarrierStubC2* stub) { + // Create mask of live registers + RegMask live = stub->live(); -// Generates a register specific stub for calling -// ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded() or -// ZBarrierSetRuntime::load_barrier_on_weak_oop_field_preloaded(). -// -// The raddr register serves as both input and output for this stub. When the stub is -// called the raddr register contains the object field address (oop*) where the bad oop -// was loaded from, which caused the slow path to be taken. On return from the stub the -// raddr register contains the good/healed oop returned from -// ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded() or -// ZBarrierSetRuntime::load_barrier_on_weak_oop_field_preloaded(). -static address generate_load_barrier_stub(StubCodeGenerator* cgen, Register raddr, DecoratorSet decorators) { - // Don't generate stub for invalid registers - if (raddr == zr || raddr == r29 || raddr == r30) { - return NULL; + // Record registers that needs to be saved/restored + while (live.is_NotEmpty()) { + const OptoReg::Name opto_reg = live.find_first_elem(); + live.Remove(opto_reg); + if (OptoReg::is_reg(opto_reg)) { + const VMReg vm_reg = OptoReg::as_VMReg(opto_reg); + if (vm_reg->is_Register()) { + _gp_regs += RegSet::of(vm_reg->as_Register()); + } else if (vm_reg->is_FloatRegister()) { + _fp_regs += RegSet::of((Register)vm_reg->as_FloatRegister()); + } else { + fatal("Unknown register type"); + } + } + } + + // Remove C-ABI SOE registers, scratch regs and _ref register that will be updated + _gp_regs -= RegSet::range(r19, r30) + RegSet::of(r8, r9, stub->ref()); + } + + ZSaveLiveRegisters(MacroAssembler* masm, ZLoadBarrierStubC2* stub) : + _masm(masm), + _gp_regs(), + _fp_regs() { + + // Figure out what registers to save/restore + initialize(stub); + + // Save registers + __ push(_gp_regs, sp); + __ push_fp(_fp_regs, sp); } - // Create stub name - char name[64]; - const bool weak = (decorators & ON_WEAK_OOP_REF) != 0; - os::snprintf(name, sizeof(name), "zgc_load_barrier%s_stub_%s", weak ? "_weak" : "", raddr->name()); + ~ZSaveLiveRegisters() { + // Restore registers + __ pop_fp(_fp_regs, sp); + __ pop(_gp_regs, sp); + } +}; + +#undef __ +#define __ _masm-> - __ align(CodeEntryAlignment); - StubCodeMark mark(cgen, "StubRoutines", os::strdup(name, mtCode)); - address start = __ pc(); +class ZSetupArguments { +private: + MacroAssembler* const _masm; + const Register _ref; + const Address _ref_addr; + +public: + ZSetupArguments(MacroAssembler* masm, ZLoadBarrierStubC2* stub) : + _masm(masm), + _ref(stub->ref()), + _ref_addr(stub->ref_addr()) { - // Save live registers - RegSet savedRegs = RegSet::range(r0,r18) - RegSet::of(raddr); - - __ enter(); - __ push(savedRegs, sp); - - // Setup arguments - if (raddr != c_rarg1) { - __ mov(c_rarg1, raddr); + // Setup arguments + if (_ref_addr.base() == noreg) { + // No self healing + if (_ref != c_rarg0) { + __ mov(c_rarg0, _ref); + } + __ mov(c_rarg1, 0); + } else { + // Self healing + if (_ref == c_rarg0) { + // _ref is already at correct place + __ lea(c_rarg1, _ref_addr); + } else if (_ref != c_rarg1) { + // _ref is in wrong place, but not in c_rarg1, so fix it first + __ lea(c_rarg1, _ref_addr); + __ mov(c_rarg0, _ref); + } else if (_ref_addr.base() != c_rarg0 && _ref_addr.index() != c_rarg0) { + assert(_ref == c_rarg1, "Mov ref first, vacating c_rarg0"); + __ mov(c_rarg0, _ref); + __ lea(c_rarg1, _ref_addr); + } else { + assert(_ref == c_rarg1, "Need to vacate c_rarg1 and _ref_addr is using c_rarg0"); + if (_ref_addr.base() == c_rarg0 || _ref_addr.index() == c_rarg0) { + __ mov(rscratch2, c_rarg1); + __ lea(c_rarg1, _ref_addr); + __ mov(c_rarg0, rscratch2); + } else { + ShouldNotReachHere(); + } + } + } } - __ ldr(c_rarg0, Address(raddr)); + ~ZSetupArguments() { + // Transfer result + if (_ref != r0) { + __ mov(_ref, r0); + } + } +}; + +#undef __ +#define __ masm-> - // Call barrier function - __ call_VM_leaf(ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded_addr(decorators), c_rarg0, c_rarg1); +void ZBarrierSetAssembler::generate_c2_load_barrier_stub(MacroAssembler* masm, ZLoadBarrierStubC2* stub) const { + BLOCK_COMMENT("ZLoadBarrierStubC2"); + + // Stub entry + __ bind(*stub->entry()); - // Move result returned in r0 to raddr, if needed - if (raddr != r0) { - __ mov(raddr, r0); + { + ZSaveLiveRegisters save_live_registers(masm, stub); + ZSetupArguments setup_arguments(masm, stub); + __ mov(rscratch1, stub->slow_path()); + __ blr(rscratch1); } - __ pop(savedRegs, sp); - __ leave(); - __ ret(lr); - - return start; + // Stub exit + __ b(*stub->continuation()); } #undef __ -static void barrier_stubs_init_inner(const char* label, const DecoratorSet decorators, address* stub) { - const int nregs = 28; // Exclude FP, XZR, SP from calculation. - const int code_size = nregs * 254; // Rough estimate of code size - - ResourceMark rm; - - CodeBuffer buf(BufferBlob::create(label, code_size)); - StubCodeGenerator cgen(&buf); - - for (int i = 0; i < nregs; i++) { - const Register reg = as_Register(i); - stub[i] = generate_load_barrier_stub(&cgen, reg, decorators); - } -} - -void ZBarrierSetAssembler::barrier_stubs_init() { - barrier_stubs_init_inner("zgc_load_barrier_stubs", ON_STRONG_OOP_REF, _load_barrier_slow_stub); - barrier_stubs_init_inner("zgc_load_barrier_weak_stubs", ON_WEAK_OOP_REF, _load_barrier_weak_slow_stub); -} - -address ZBarrierSetAssembler::load_barrier_slow_stub(Register reg) { - return _load_barrier_slow_stub[reg->encoding()]; -} - -address ZBarrierSetAssembler::load_barrier_weak_slow_stub(Register reg) { - return _load_barrier_weak_slow_stub[reg->encoding()]; -} +#endif // COMPILER2
--- a/src/hotspot/cpu/aarch64/gc/z/zBarrierSetAssembler_aarch64.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/gc/z/zBarrierSetAssembler_aarch64.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -24,6 +24,12 @@ #ifndef CPU_AARCH64_GC_Z_ZBARRIERSETASSEMBLER_AARCH64_HPP #define CPU_AARCH64_GC_Z_ZBARRIERSETASSEMBLER_AARCH64_HPP +#include "code/vmreg.hpp" +#include "oops/accessDecorators.hpp" +#ifdef COMPILER2 +#include "opto/optoreg.hpp" +#endif // COMPILER2 + #ifdef COMPILER1 class LIR_Assembler; class LIR_OprDesc; @@ -32,14 +38,13 @@ class ZLoadBarrierStubC1; #endif // COMPILER1 +#ifdef COMPILER2 +class Node; +class ZLoadBarrierStubC2; +#endif // COMPILER2 + class ZBarrierSetAssembler : public ZBarrierSetAssemblerBase { -private: - address _load_barrier_slow_stub[RegisterImpl::number_of_registers]; - address _load_barrier_weak_slow_stub[RegisterImpl::number_of_registers]; - public: - ZBarrierSetAssembler(); - virtual void load_at(MacroAssembler* masm, DecoratorSet decorators, BasicType type, @@ -83,10 +88,13 @@ DecoratorSet decorators) const; #endif // COMPILER1 - virtual void barrier_stubs_init(); +#ifdef COMPILER2 + OptoReg::Name refine_register(const Node* node, + OptoReg::Name opto_reg); - address load_barrier_slow_stub(Register reg); - address load_barrier_weak_slow_stub(Register reg); + void generate_c2_load_barrier_stub(MacroAssembler* masm, + ZLoadBarrierStubC2* stub) const; +#endif // COMPILER2 }; #endif // CPU_AARCH64_GC_Z_ZBARRIERSETASSEMBLER_AARCH64_HPP
--- a/src/hotspot/cpu/aarch64/gc/z/z_aarch64.ad Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/gc/z/z_aarch64.ad Mon Oct 14 21:01:25 2019 +0000 @@ -24,155 +24,244 @@ source_hpp %{ #include "gc/z/c2/zBarrierSetC2.hpp" +#include "gc/z/zThreadLocalData.hpp" %} source %{ -#include "gc/z/zBarrierSetAssembler.hpp" - -static void z_load_barrier_slow_reg(MacroAssembler& _masm, Register dst, - Register base, int index, int scale, - int disp, bool weak) { - const address stub = weak ? ZBarrierSet::assembler()->load_barrier_weak_slow_stub(dst) - : ZBarrierSet::assembler()->load_barrier_slow_stub(dst); +static void z_load_barrier(MacroAssembler& _masm, const MachNode* node, Address ref_addr, Register ref, Register tmp, bool weak) { + ZLoadBarrierStubC2* const stub = ZLoadBarrierStubC2::create(node, ref_addr, ref, tmp, weak); + __ ldr(tmp, Address(rthread, ZThreadLocalData::address_bad_mask_offset())); + __ andr(tmp, tmp, ref); + __ cbnz(tmp, *stub->entry()); + __ bind(*stub->continuation()); +} - if (index == -1) { - if (disp != 0) { - __ lea(dst, Address(base, disp)); - } else { - __ mov(dst, base); - } - } else { - Register index_reg = as_Register(index); - if (disp == 0) { - __ lea(dst, Address(base, index_reg, Address::lsl(scale))); - } else { - __ lea(dst, Address(base, disp)); - __ lea(dst, Address(dst, index_reg, Address::lsl(scale))); - } - } - - __ far_call(RuntimeAddress(stub)); +static void z_load_barrier_slow_path(MacroAssembler& _masm, const MachNode* node, Address ref_addr, Register ref, Register tmp) { + ZLoadBarrierStubC2* const stub = ZLoadBarrierStubC2::create(node, ref_addr, ref, tmp, false /* weak */); + __ b(*stub->entry()); + __ bind(*stub->continuation()); } %} -// -// Execute ZGC load barrier (strong) slow path -// -instruct loadBarrierSlowReg(iRegP dst, memory src, rFlagsReg cr, - vRegD_V0 v0, vRegD_V1 v1, vRegD_V2 v2, vRegD_V3 v3, vRegD_V4 v4, - vRegD_V5 v5, vRegD_V6 v6, vRegD_V7 v7, vRegD_V8 v8, vRegD_V9 v9, - vRegD_V10 v10, vRegD_V11 v11, vRegD_V12 v12, vRegD_V13 v13, vRegD_V14 v14, - vRegD_V15 v15, vRegD_V16 v16, vRegD_V17 v17, vRegD_V18 v18, vRegD_V19 v19, - vRegD_V20 v20, vRegD_V21 v21, vRegD_V22 v22, vRegD_V23 v23, vRegD_V24 v24, - vRegD_V25 v25, vRegD_V26 v26, vRegD_V27 v27, vRegD_V28 v28, vRegD_V29 v29, - vRegD_V30 v30, vRegD_V31 v31) %{ - match(Set dst (LoadBarrierSlowReg src dst)); - predicate(!n->as_LoadBarrierSlowReg()->is_weak()); +// Load Pointer +instruct zLoadP(iRegPNoSp dst, memory mem, rFlagsReg cr) +%{ + match(Set dst (LoadP mem)); + predicate(UseZGC && !needs_acquiring_load(n) && (n->as_Load()->barrier_data() == ZLoadBarrierStrong)); + effect(TEMP dst, KILL cr); - effect(KILL cr, - KILL v0, KILL v1, KILL v2, KILL v3, KILL v4, KILL v5, KILL v6, KILL v7, - KILL v8, KILL v9, KILL v10, KILL v11, KILL v12, KILL v13, KILL v14, - KILL v15, KILL v16, KILL v17, KILL v18, KILL v19, KILL v20, KILL v21, - KILL v22, KILL v23, KILL v24, KILL v25, KILL v26, KILL v27, KILL v28, - KILL v29, KILL v30, KILL v31); + ins_cost(4 * INSN_COST); - format %{ "lea $dst, $src\n\t" - "call #ZLoadBarrierSlowPath" %} + format %{ "ldr $dst, $mem" %} ins_encode %{ - z_load_barrier_slow_reg(_masm, $dst$$Register, $src$$base$$Register, - $src$$index, $src$$scale, $src$$disp, false); + const Address ref_addr = mem2address($mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp); + __ ldr($dst$$Register, ref_addr); + if (barrier_data() != ZLoadBarrierElided) { + z_load_barrier(_masm, this, ref_addr, $dst$$Register, rscratch2 /* tmp */, false /* weak */); + } %} - ins_pipe(pipe_slow); + + ins_pipe(iload_reg_mem); %} -// -// Execute ZGC load barrier (weak) slow path -// -instruct loadBarrierWeakSlowReg(iRegP dst, memory src, rFlagsReg cr, - vRegD_V0 v0, vRegD_V1 v1, vRegD_V2 v2, vRegD_V3 v3, vRegD_V4 v4, - vRegD_V5 v5, vRegD_V6 v6, vRegD_V7 v7, vRegD_V8 v8, vRegD_V9 v9, - vRegD_V10 v10, vRegD_V11 v11, vRegD_V12 v12, vRegD_V13 v13, vRegD_V14 v14, - vRegD_V15 v15, vRegD_V16 v16, vRegD_V17 v17, vRegD_V18 v18, vRegD_V19 v19, - vRegD_V20 v20, vRegD_V21 v21, vRegD_V22 v22, vRegD_V23 v23, vRegD_V24 v24, - vRegD_V25 v25, vRegD_V26 v26, vRegD_V27 v27, vRegD_V28 v28, vRegD_V29 v29, - vRegD_V30 v30, vRegD_V31 v31) %{ - match(Set dst (LoadBarrierSlowReg src dst)); - predicate(n->as_LoadBarrierSlowReg()->is_weak()); +// Load Weak Pointer +instruct zLoadWeakP(iRegPNoSp dst, memory mem, rFlagsReg cr) +%{ + match(Set dst (LoadP mem)); + predicate(UseZGC && !needs_acquiring_load(n) && (n->as_Load()->barrier_data() == ZLoadBarrierWeak)); + effect(TEMP dst, KILL cr); - effect(KILL cr, - KILL v0, KILL v1, KILL v2, KILL v3, KILL v4, KILL v5, KILL v6, KILL v7, - KILL v8, KILL v9, KILL v10, KILL v11, KILL v12, KILL v13, KILL v14, - KILL v15, KILL v16, KILL v17, KILL v18, KILL v19, KILL v20, KILL v21, - KILL v22, KILL v23, KILL v24, KILL v25, KILL v26, KILL v27, KILL v28, - KILL v29, KILL v30, KILL v31); + ins_cost(4 * INSN_COST); - format %{ "lea $dst, $src\n\t" - "call #ZLoadBarrierSlowPath" %} + format %{ "ldr $dst, $mem" %} ins_encode %{ - z_load_barrier_slow_reg(_masm, $dst$$Register, $src$$base$$Register, - $src$$index, $src$$scale, $src$$disp, true); + const Address ref_addr = mem2address($mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp); + __ ldr($dst$$Register, ref_addr); + z_load_barrier(_masm, this, ref_addr, $dst$$Register, rscratch2 /* tmp */, true /* weak */); %} - ins_pipe(pipe_slow); + + ins_pipe(iload_reg_mem); %} +// Load Pointer Volatile +instruct zLoadPVolatile(iRegPNoSp dst, indirect mem /* sync_memory */, rFlagsReg cr) +%{ + match(Set dst (LoadP mem)); + predicate(UseZGC && needs_acquiring_load(n) && n->as_Load()->barrier_data() == ZLoadBarrierStrong); + effect(TEMP dst, KILL cr); -// Specialized versions of compareAndExchangeP that adds a keepalive that is consumed -// but doesn't affect output. + ins_cost(VOLATILE_REF_COST); -instruct z_compareAndExchangeP(iRegPNoSp res, indirect mem, - iRegP oldval, iRegP newval, iRegP keepalive, - rFlagsReg cr) %{ - match(Set res (ZCompareAndExchangeP (Binary mem keepalive) (Binary oldval newval))); - ins_cost(2 * VOLATILE_REF_COST); - effect(TEMP_DEF res, KILL cr); - format %{ - "cmpxchg $res = $mem, $oldval, $newval\t# (ptr, weak) if $mem == $oldval then $mem <-- $newval" + format %{ "ldar $dst, $mem\t" %} + + ins_encode %{ + __ ldar($dst$$Register, $mem$$Register); + if (barrier_data() != ZLoadBarrierElided) { + z_load_barrier(_masm, this, Address($mem$$Register), $dst$$Register, rscratch2 /* tmp */, false /* weak */); + } %} - ins_encode %{ - __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, - Assembler::xword, /*acquire*/ false, /*release*/ true, - /*weak*/ false, $res$$Register); - %} - ins_pipe(pipe_slow); + + ins_pipe(pipe_serial); %} -instruct z_compareAndSwapP(iRegINoSp res, - indirect mem, - iRegP oldval, iRegP newval, iRegP keepalive, - rFlagsReg cr) %{ - - match(Set res (ZCompareAndSwapP (Binary mem keepalive) (Binary oldval newval))); - match(Set res (ZWeakCompareAndSwapP (Binary mem keepalive) (Binary oldval newval))); +instruct zCompareAndSwapP(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ + match(Set res (CompareAndSwapP mem (Binary oldval newval))); + match(Set res (WeakCompareAndSwapP mem (Binary oldval newval))); + predicate(UseZGC && !needs_acquiring_load_exclusive(n) && n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong); + effect(KILL cr, TEMP_DEF res); ins_cost(2 * VOLATILE_REF_COST); - effect(KILL cr); + format %{ "cmpxchg $mem, $oldval, $newval\n\t" + "cset $res, EQ" %} - format %{ - "cmpxchg $mem, $oldval, $newval\t# (ptr) if $mem == $oldval then $mem <-- $newval" - "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)" - %} - - ins_encode(aarch64_enc_cmpxchg(mem, oldval, newval), - aarch64_enc_cset_eq(res)); + ins_encode %{ + guarantee($mem$$index == -1 && $mem$$disp == 0, "impossible encoding"); + __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, Assembler::xword, + false /* acquire */, true /* release */, false /* weak */, rscratch2); + __ cset($res$$Register, Assembler::EQ); + if (barrier_data() != ZLoadBarrierElided) { + Label good; + __ ldr(rscratch1, Address(rthread, ZThreadLocalData::address_bad_mask_offset())); + __ andr(rscratch1, rscratch1, rscratch2); + __ cbz(rscratch1, good); + z_load_barrier_slow_path(_masm, this, Address($mem$$Register), rscratch2 /* ref */, rscratch1 /* tmp */); + __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, Assembler::xword, + false /* acquire */, true /* release */, false /* weak */, rscratch2); + __ cset($res$$Register, Assembler::EQ); + __ bind(good); + } + %} ins_pipe(pipe_slow); %} +instruct zCompareAndSwapPAcq(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ + match(Set res (CompareAndSwapP mem (Binary oldval newval))); + match(Set res (WeakCompareAndSwapP mem (Binary oldval newval))); + predicate(UseZGC && needs_acquiring_load_exclusive(n) && (n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong)); + effect(KILL cr, TEMP_DEF res); -instruct z_get_and_setP(indirect mem, iRegP newv, iRegPNoSp prev, - iRegP keepalive) %{ - match(Set prev (ZGetAndSetP mem (Binary newv keepalive))); + ins_cost(2 * VOLATILE_REF_COST); + + format %{ "cmpxchg $mem, $oldval, $newval\n\t" + "cset $res, EQ" %} + + ins_encode %{ + guarantee($mem$$index == -1 && $mem$$disp == 0, "impossible encoding"); + __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, Assembler::xword, + true /* acquire */, true /* release */, false /* weak */, rscratch2); + __ cset($res$$Register, Assembler::EQ); + if (barrier_data() != ZLoadBarrierElided) { + Label good; + __ ldr(rscratch1, Address(rthread, ZThreadLocalData::address_bad_mask_offset())); + __ andr(rscratch1, rscratch1, rscratch2); + __ cbz(rscratch1, good); + z_load_barrier_slow_path(_masm, this, Address($mem$$Register), rscratch2 /* ref */, rscratch1 /* tmp */ ); + __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, Assembler::xword, + true /* acquire */, true /* release */, false /* weak */, rscratch2); + __ cset($res$$Register, Assembler::EQ); + __ bind(good); + } + %} + + ins_pipe(pipe_slow); +%} + +instruct zCompareAndExchangeP(iRegPNoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ + match(Set res (CompareAndExchangeP mem (Binary oldval newval))); + predicate(UseZGC && !needs_acquiring_load_exclusive(n) && n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong); + effect(TEMP_DEF res, KILL cr); ins_cost(2 * VOLATILE_REF_COST); - format %{ "atomic_xchg $prev, $newv, [$mem]" %} + + format %{ "cmpxchg $res = $mem, $oldval, $newval" %} + + ins_encode %{ + guarantee($mem$$index == -1 && $mem$$disp == 0, "impossible encoding"); + __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, Assembler::xword, + false /* acquire */, true /* release */, false /* weak */, $res$$Register); + if (barrier_data() != ZLoadBarrierElided) { + Label good; + __ ldr(rscratch1, Address(rthread, ZThreadLocalData::address_bad_mask_offset())); + __ andr(rscratch1, rscratch1, $res$$Register); + __ cbz(rscratch1, good); + z_load_barrier_slow_path(_masm, this, Address($mem$$Register), $res$$Register /* ref */, rscratch1 /* tmp */); + __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, Assembler::xword, + false /* acquire */, true /* release */, false /* weak */, $res$$Register); + __ bind(good); + } + %} + + ins_pipe(pipe_slow); +%} + +instruct zCompareAndExchangePAcq(iRegPNoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{ + match(Set res (CompareAndExchangeP mem (Binary oldval newval))); + predicate(UseZGC && needs_acquiring_load_exclusive(n) && n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong); + effect(TEMP_DEF res, KILL cr); + + ins_cost(2 * VOLATILE_REF_COST); + + format %{ "cmpxchg $res = $mem, $oldval, $newval" %} + ins_encode %{ - __ atomic_xchg($prev$$Register, $newv$$Register, as_Register($mem$$base)); + guarantee($mem$$index == -1 && $mem$$disp == 0, "impossible encoding"); + __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, Assembler::xword, + true /* acquire */, true /* release */, false /* weak */, $res$$Register); + if (barrier_data() != ZLoadBarrierElided) { + Label good; + __ ldr(rscratch1, Address(rthread, ZThreadLocalData::address_bad_mask_offset())); + __ andr(rscratch1, rscratch1, $res$$Register); + __ cbz(rscratch1, good); + z_load_barrier_slow_path(_masm, this, Address($mem$$Register), $res$$Register /* ref */, rscratch1 /* tmp */); + __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register, Assembler::xword, + true /* acquire */, true /* release */, false /* weak */, $res$$Register); + __ bind(good); + } + %} + + ins_pipe(pipe_slow); +%} + +instruct zGetAndSetP(indirect mem, iRegP newv, iRegPNoSp prev, rFlagsReg cr) %{ + match(Set prev (GetAndSetP mem newv)); + predicate(UseZGC && !needs_acquiring_load_exclusive(n) && n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong); + effect(TEMP_DEF prev, KILL cr); + + ins_cost(2 * VOLATILE_REF_COST); + + format %{ "atomic_xchg $prev, $newv, [$mem]" %} + + ins_encode %{ + __ atomic_xchg($prev$$Register, $newv$$Register, $mem$$Register); + if (barrier_data() != ZLoadBarrierElided) { + z_load_barrier(_masm, this, Address(noreg, 0), $prev$$Register, rscratch2 /* tmp */, false /* weak */); + } + %} + + ins_pipe(pipe_serial); +%} + +instruct zGetAndSetPAcq(indirect mem, iRegP newv, iRegPNoSp prev, rFlagsReg cr) %{ + match(Set prev (GetAndSetP mem newv)); + predicate(UseZGC && needs_acquiring_load_exclusive(n) && (n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong)); + effect(TEMP_DEF prev, KILL cr); + + ins_cost(VOLATILE_REF_COST); + + format %{ "atomic_xchg_acq $prev, $newv, [$mem]" %} + + ins_encode %{ + __ atomic_xchgal($prev$$Register, $newv$$Register, $mem$$Register); + if (barrier_data() != ZLoadBarrierElided) { + z_load_barrier(_masm, this, Address(noreg, 0), $prev$$Register, rscratch2 /* tmp */, false /* weak */); + } %} ins_pipe(pipe_serial); %}
--- a/src/hotspot/cpu/aarch64/macroAssembler_aarch64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/macroAssembler_aarch64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -2132,6 +2132,65 @@ return count; } + +// Push lots of registers in the bit set supplied. Don't push sp. +// Return the number of words pushed +int MacroAssembler::push_fp(unsigned int bitset, Register stack) { + int words_pushed = 0; + + // Scan bitset to accumulate register pairs + unsigned char regs[32]; + int count = 0; + for (int reg = 0; reg <= 31; reg++) { + if (1 & bitset) + regs[count++] = reg; + bitset >>= 1; + } + regs[count++] = zr->encoding_nocheck(); + count &= ~1; // Only push an even number of regs + + // Always pushing full 128 bit registers. + if (count) { + stpq(as_FloatRegister(regs[0]), as_FloatRegister(regs[1]), Address(pre(stack, -count * wordSize * 2))); + words_pushed += 2; + } + for (int i = 2; i < count; i += 2) { + stpq(as_FloatRegister(regs[i]), as_FloatRegister(regs[i+1]), Address(stack, i * wordSize * 2)); + words_pushed += 2; + } + + assert(words_pushed == count, "oops, pushed != count"); + return count; +} + +int MacroAssembler::pop_fp(unsigned int bitset, Register stack) { + int words_pushed = 0; + + // Scan bitset to accumulate register pairs + unsigned char regs[32]; + int count = 0; + for (int reg = 0; reg <= 31; reg++) { + if (1 & bitset) + regs[count++] = reg; + bitset >>= 1; + } + regs[count++] = zr->encoding_nocheck(); + count &= ~1; + + for (int i = 2; i < count; i += 2) { + ldpq(as_FloatRegister(regs[i]), as_FloatRegister(regs[i+1]), Address(stack, i * wordSize * 2)); + words_pushed += 2; + } + if (count) { + ldpq(as_FloatRegister(regs[0]), as_FloatRegister(regs[1]), Address(post(stack, count * wordSize * 2))); + words_pushed += 2; + } + + assert(words_pushed == count, "oops, pushed != count"); + + return count; +} + #ifdef ASSERT void MacroAssembler::verify_heapbase(const char* msg) { #if 0
--- a/src/hotspot/cpu/aarch64/macroAssembler_aarch64.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/macroAssembler_aarch64.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -442,12 +442,18 @@ int push(unsigned int bitset, Register stack); int pop(unsigned int bitset, Register stack); + int push_fp(unsigned int bitset, Register stack); + int pop_fp(unsigned int bitset, Register stack); + void mov(Register dst, Address a); public: void push(RegSet regs, Register stack) { if (regs.bits()) push(regs.bits(), stack); } void pop(RegSet regs, Register stack) { if (regs.bits()) pop(regs.bits(), stack); } + void push_fp(RegSet regs, Register stack) { if (regs.bits()) push_fp(regs.bits(), stack); } + void pop_fp(RegSet regs, Register stack) { if (regs.bits()) pop_fp(regs.bits(), stack); } + // Push and pop everything that might be clobbered by a native // runtime call except rscratch1 and rscratch2. (They are always // scratch, so we don't have to protect them.) Only save the lower
--- a/src/hotspot/cpu/aarch64/nativeInst_aarch64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/nativeInst_aarch64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -332,9 +332,14 @@ // We use jump to self as the unresolved address which the inline // cache code (and relocs) know about + // As a special case we also use sequence movptr(r,0); br(r); + // i.e. jump to 0 when we need leave space for a wide immediate + // load - // return -1 if jump to self - dest = (dest == (address) this) ? (address) -1 : dest; + // return -1 if jump to self or to 0 + if ((dest == (address)this) || dest == 0) { + dest = (address) -1; + } return dest; } @@ -356,9 +361,13 @@ // We use jump to self as the unresolved address which the inline // cache code (and relocs) know about + // As a special case we also use jump to 0 when first generating + // a general jump - // return -1 if jump to self - dest = (dest == (address) this) ? (address) -1 : dest; + // return -1 if jump to self or to 0 + if ((dest == (address)this) || dest == 0) { + dest = (address) -1; + } return dest; }
--- a/src/hotspot/cpu/aarch64/register_aarch64.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/register_aarch64.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -230,6 +230,11 @@ return *this; } + RegSet &operator-=(const RegSet aSet) { + *this = *this - aSet; + return *this; + } + static RegSet of(Register r1) { return RegSet(r1); }
--- a/src/hotspot/cpu/aarch64/sharedRuntime_aarch64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/aarch64/sharedRuntime_aarch64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -34,6 +34,7 @@ #include "logging/log.hpp" #include "memory/resourceArea.hpp" #include "oops/compiledICHolder.hpp" +#include "oops/klass.inline.hpp" #include "runtime/safepointMechanism.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/vframeArray.hpp"
--- a/src/hotspot/cpu/arm/abstractInterpreter_arm.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/arm/abstractInterpreter_arm.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -27,6 +27,7 @@ #include "interpreter/bytecode.hpp" #include "interpreter/interpreter.hpp" #include "oops/constMethod.hpp" +#include "oops/klass.inline.hpp" #include "oops/method.hpp" #include "prims/methodHandles.hpp" #include "runtime/handles.inline.hpp"
--- a/src/hotspot/cpu/arm/compiledIC_arm.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/arm/compiledIC_arm.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -1,5 +1,5 @@ /* - * Copyright (c) 1997, 2018, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 1997, 2019, 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 @@ -115,16 +115,7 @@ // Creation also verifies the object. NativeMovConstReg* method_holder = nativeMovConstReg_at(stub); NativeJump* jump = nativeJump_at(method_holder->next_instruction_address()); - -#ifdef ASSERT - // read the value once - volatile intptr_t data = method_holder->data(); - volatile address destination = jump->jump_destination(); - assert(data == 0 || data == (intptr_t)callee(), - "a) MT-unsafe modification of inline cache"); - assert(destination == (address)-1 || destination == entry, - "b) MT-unsafe modification of inline cache"); -#endif + verify_mt_safe(callee, entry, method_holder, jump); // Update stub. method_holder->set_data((intptr_t)callee());
--- a/src/hotspot/cpu/arm/sharedRuntime_arm.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/arm/sharedRuntime_arm.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -32,6 +32,7 @@ #include "logging/log.hpp" #include "memory/resourceArea.hpp" #include "oops/compiledICHolder.hpp" +#include "oops/klass.inline.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/vframeArray.hpp" #include "utilities/align.hpp"
--- a/src/hotspot/cpu/ppc/abstractInterpreter_ppc.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/ppc/abstractInterpreter_ppc.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -26,6 +26,7 @@ #include "precompiled.hpp" #include "interpreter/interpreter.hpp" #include "oops/constMethod.hpp" +#include "oops/klass.inline.hpp" #include "oops/method.hpp" #include "runtime/frame.inline.hpp" #include "utilities/debug.hpp"
--- a/src/hotspot/cpu/ppc/compiledIC_ppc.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/ppc/compiledIC_ppc.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -1,5 +1,5 @@ /* - * Copyright (c) 1997, 2018, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 1997, 2019, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2015 SAP SE. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * @@ -178,15 +178,7 @@ NativeMovConstReg* method_holder = nativeMovConstReg_at(stub + IC_pos_in_java_to_interp_stub); NativeJump* jump = nativeJump_at(method_holder->next_instruction_address()); -#ifdef ASSERT - // read the value once - volatile intptr_t data = method_holder->data(); - volatile address destination = jump->jump_destination(); - assert(data == 0 || data == (intptr_t)callee(), - "a) MT-unsafe modification of inline cache"); - assert(destination == (address)-1 || destination == entry, - "b) MT-unsafe modification of inline cache"); -#endif + verify_mt_safe(callee, entry, method_holder, jump); // Update stub. method_holder->set_data((intptr_t)callee());
--- a/src/hotspot/cpu/ppc/macroAssembler_ppc.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/ppc/macroAssembler_ppc.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -32,6 +32,7 @@ #include "interpreter/interpreter.hpp" #include "memory/resourceArea.hpp" #include "nativeInst_ppc.hpp" +#include "oops/klass.inline.hpp" #include "prims/methodHandles.hpp" #include "runtime/biasedLocking.hpp" #include "runtime/icache.hpp"
--- a/src/hotspot/cpu/ppc/sharedRuntime_ppc.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/ppc/sharedRuntime_ppc.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -34,6 +34,7 @@ #include "interpreter/interp_masm.hpp" #include "memory/resourceArea.hpp" #include "oops/compiledICHolder.hpp" +#include "oops/klass.inline.hpp" #include "runtime/safepointMechanism.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/vframeArray.hpp"
--- a/src/hotspot/cpu/ppc/templateTable_ppc_64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/ppc/templateTable_ppc_64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -32,6 +32,7 @@ #include "interpreter/templateInterpreter.hpp" #include "interpreter/templateTable.hpp" #include "memory/universe.hpp" +#include "oops/klass.inline.hpp" #include "oops/objArrayKlass.hpp" #include "oops/oop.inline.hpp" #include "prims/methodHandles.hpp"
--- a/src/hotspot/cpu/ppc/vtableStubs_ppc_64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/ppc/vtableStubs_ppc_64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -30,6 +30,7 @@ #include "memory/resourceArea.hpp" #include "oops/compiledICHolder.hpp" #include "oops/instanceKlass.hpp" +#include "oops/klass.inline.hpp" #include "oops/klassVtable.hpp" #include "runtime/sharedRuntime.hpp" #include "vmreg_ppc.inline.hpp"
--- a/src/hotspot/cpu/s390/abstractInterpreter_s390.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/s390/abstractInterpreter_s390.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -26,6 +26,7 @@ #include "precompiled.hpp" #include "interpreter/interpreter.hpp" #include "oops/constMethod.hpp" +#include "oops/klass.inline.hpp" #include "oops/method.hpp" #include "runtime/frame.inline.hpp" #include "utilities/debug.hpp"
--- a/src/hotspot/cpu/s390/compiledIC_s390.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/s390/compiledIC_s390.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2018, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2016, 2019, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2016 SAP SE. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * @@ -104,19 +104,7 @@ // Creation also verifies the object. NativeMovConstReg* method_holder = nativeMovConstReg_at(stub + NativeCall::get_IC_pos_in_java_to_interp_stub()); NativeJump* jump = nativeJump_at(method_holder->next_instruction_address()); - -#ifdef ASSERT - // A generated lambda form might be deleted from the Lambdaform - // cache in MethodTypeForm. If a jit compiled lambdaform method - // becomes not entrant and the cache access returns null, the new - // resolve will lead to a new generated LambdaForm. - volatile intptr_t data = method_holder->data(); - volatile address destination = jump->jump_destination(); - assert(data == 0 || data == (intptr_t)callee() || callee->is_compiled_lambda_form(), - "a) MT-unsafe modification of inline cache"); - assert(destination == (address)-1 || destination == entry, - "b) MT-unsafe modification of inline cache"); -#endif + verify_mt_safe(callee, entry, method_holder, jump); // Update stub. method_holder->set_data((intptr_t)callee(), relocInfo::metadata_type);
--- a/src/hotspot/cpu/s390/sharedRuntime_s390.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/s390/sharedRuntime_s390.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -33,6 +33,7 @@ #include "interpreter/interp_masm.hpp" #include "memory/resourceArea.hpp" #include "oops/compiledICHolder.hpp" +#include "oops/klass.inline.hpp" #include "registerSaver_s390.hpp" #include "runtime/safepointMechanism.hpp" #include "runtime/sharedRuntime.hpp"
--- a/src/hotspot/cpu/s390/vtableStubs_s390.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/s390/vtableStubs_s390.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -30,6 +30,7 @@ #include "memory/resourceArea.hpp" #include "oops/compiledICHolder.hpp" #include "oops/instanceKlass.hpp" +#include "oops/klass.inline.hpp" #include "oops/klassVtable.hpp" #include "runtime/sharedRuntime.hpp" #include "vmreg_s390.inline.hpp"
--- a/src/hotspot/cpu/sparc/abstractInterpreter_sparc.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/sparc/abstractInterpreter_sparc.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -25,6 +25,7 @@ #include "precompiled.hpp" #include "interpreter/interpreter.hpp" #include "oops/constMethod.hpp" +#include "oops/klass.inline.hpp" #include "oops/method.hpp" #include "runtime/arguments.hpp" #include "runtime/frame.inline.hpp"
--- a/src/hotspot/cpu/sparc/compiledIC_sparc.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/sparc/compiledIC_sparc.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -1,5 +1,5 @@ /* - * Copyright (c) 1997, 2018, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 1997, 2019, 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 @@ -104,16 +104,7 @@ // Creation also verifies the object. NativeMovConstReg* method_holder = nativeMovConstReg_at(stub); NativeJump* jump = nativeJump_at(method_holder->next_instruction_address()); - -#ifdef ASSERT - // read the value once - volatile intptr_t data = method_holder->data(); - volatile address destination = jump->jump_destination(); - assert(data == 0 || data == (intptr_t)callee(), - "a) MT-unsafe modification of inline cache"); - assert(destination == (address)-1 || destination == entry, - "b) MT-unsafe modification of inline cache"); -#endif + verify_mt_safe(callee, entry, method_holder, jump); // Update stub. method_holder->set_data((intptr_t)callee());
--- a/src/hotspot/cpu/sparc/sharedRuntime_sparc.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/sparc/sharedRuntime_sparc.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -32,6 +32,7 @@ #include "logging/log.hpp" #include "memory/resourceArea.hpp" #include "oops/compiledICHolder.hpp" +#include "oops/klass.inline.hpp" #include "runtime/safepointMechanism.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/vframeArray.hpp"
--- a/src/hotspot/cpu/x86/abstractInterpreter_x86.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/abstractInterpreter_x86.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -25,6 +25,7 @@ #include "precompiled.hpp" #include "ci/ciMethod.hpp" #include "interpreter/interpreter.hpp" +#include "oops/klass.inline.hpp" #include "runtime/frame.inline.hpp"
--- a/src/hotspot/cpu/x86/compiledIC_x86.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/compiledIC_x86.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -1,5 +1,5 @@ /* - * Copyright (c) 1997, 2018, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 1997, 2019, 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 @@ -157,16 +157,7 @@ // Creation also verifies the object. NativeMovConstReg* method_holder = nativeMovConstReg_at(stub); NativeJump* jump = nativeJump_at(method_holder->next_instruction_address()); - -#ifdef ASSERT - Method* old_method = reinterpret_cast<Method*>(method_holder->data()); - address destination = jump->jump_destination(); - assert(old_method == NULL || old_method == callee() || - !old_method->method_holder()->is_loader_alive(), - "a) MT-unsafe modification of inline cache"); - assert(destination == (address)-1 || destination == entry, - "b) MT-unsafe modification of inline cache"); -#endif + verify_mt_safe(callee, entry, method_holder, jump); // Update stub. method_holder->set_data((intptr_t)callee());
--- a/src/hotspot/cpu/x86/gc/z/zArguments_x86.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/gc/z/zArguments_x86.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -23,20 +23,7 @@ #include "precompiled.hpp" #include "gc/z/zArguments.hpp" -#include "runtime/globals.hpp" -#include "runtime/globals_extension.hpp" -#include "utilities/debug.hpp" void ZArguments::initialize_platform() { -#ifdef COMPILER2 - // The C2 barrier slow path expects vector registers to be least - // 16 bytes wide, which is the minimum width available on all - // x86-64 systems. However, the user could have speficied a lower - // number on the command-line, in which case we print a warning - // and raise it to 16. - if (MaxVectorSize < 16) { - warning("ZGC requires MaxVectorSize to be at least 16"); - FLAG_SET_DEFAULT(MaxVectorSize, 16); - } -#endif + // Does nothing }
--- a/src/hotspot/cpu/x86/gc/z/zBarrierSetAssembler_x86.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/gc/z/zBarrierSetAssembler_x86.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -24,22 +24,22 @@ #include "precompiled.hpp" #include "asm/macroAssembler.inline.hpp" #include "code/codeBlob.hpp" +#include "code/vmreg.inline.hpp" #include "gc/z/zBarrier.inline.hpp" #include "gc/z/zBarrierSet.hpp" #include "gc/z/zBarrierSetAssembler.hpp" #include "gc/z/zBarrierSetRuntime.hpp" #include "memory/resourceArea.hpp" -#include "runtime/stubCodeGenerator.hpp" +#include "runtime/sharedRuntime.hpp" #include "utilities/macros.hpp" #ifdef COMPILER1 #include "c1/c1_LIRAssembler.hpp" #include "c1/c1_MacroAssembler.hpp" #include "gc/z/c1/zBarrierSetC1.hpp" #endif // COMPILER1 - -ZBarrierSetAssembler::ZBarrierSetAssembler() : - _load_barrier_slow_stub(), - _load_barrier_weak_slow_stub() {} +#ifdef COMPILER2 +#include "gc/z/c2/zBarrierSetC2.hpp" +#endif // COMPILER2 #ifdef PRODUCT #define BLOCK_COMMENT(str) /* nothing */ @@ -344,137 +344,327 @@ #endif // COMPILER1 +#ifdef COMPILER2 + +OptoReg::Name ZBarrierSetAssembler::refine_register(const Node* node, OptoReg::Name opto_reg) { + if (!OptoReg::is_reg(opto_reg)) { + return OptoReg::Bad; + } + + const VMReg vm_reg = OptoReg::as_VMReg(opto_reg); + if (vm_reg->is_XMMRegister()) { + opto_reg &= ~15; + switch (node->ideal_reg()) { + case Op_VecX: + opto_reg |= 2; + break; + case Op_VecY: + opto_reg |= 4; + break; + case Op_VecZ: + opto_reg |= 8; + break; + default: + opto_reg |= 1; + break; + } + } + + return opto_reg; +} + +// We use the vec_spill_helper from the x86.ad file to avoid reinventing this wheel +extern int vec_spill_helper(CodeBuffer *cbuf, bool do_size, bool is_load, + int stack_offset, int reg, uint ireg, outputStream* st); + #undef __ -#define __ cgen->assembler()-> +#define __ _masm-> + +class ZSaveLiveRegisters { +private: + struct XMMRegisterData { + XMMRegister _reg; + int _size; + + // Used by GrowableArray::find() + bool operator == (const XMMRegisterData& other) { + return _reg == other._reg; + } + }; + + MacroAssembler* const _masm; + GrowableArray<Register> _gp_registers; + GrowableArray<XMMRegisterData> _xmm_registers; + int _spill_size; + int _spill_offset; -// Generates a register specific stub for calling -// ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded() or -// ZBarrierSetRuntime::load_barrier_on_weak_oop_field_preloaded(). -// -// The raddr register serves as both input and output for this stub. When the stub is -// called the raddr register contains the object field address (oop*) where the bad oop -// was loaded from, which caused the slow path to be taken. On return from the stub the -// raddr register contains the good/healed oop returned from -// ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded() or -// ZBarrierSetRuntime::load_barrier_on_weak_oop_field_preloaded(). -static address generate_load_barrier_stub(StubCodeGenerator* cgen, Register raddr, DecoratorSet decorators) { - // Don't generate stub for invalid registers - if (raddr == rsp || raddr == r15) { - return NULL; + static int xmm_compare_register_size(XMMRegisterData* left, XMMRegisterData* right) { + if (left->_size == right->_size) { + return 0; + } + + return (left->_size < right->_size) ? -1 : 1; + } + + static int xmm_slot_size(OptoReg::Name opto_reg) { + // The low order 4 bytes denote what size of the XMM register is live + return (opto_reg & 15) << 3; + } + + static uint xmm_ideal_reg_for_size(int reg_size) { + switch (reg_size) { + case 8: + return Op_VecD; + case 16: + return Op_VecX; + case 32: + return Op_VecY; + case 64: + return Op_VecZ; + default: + fatal("Invalid register size %d", reg_size); + return 0; + } + } + + bool xmm_needs_vzeroupper() const { + return _xmm_registers.is_nonempty() && _xmm_registers.at(0)._size > 16; + } + + void xmm_register_save(const XMMRegisterData& reg_data) { + const OptoReg::Name opto_reg = OptoReg::as_OptoReg(reg_data._reg->as_VMReg()); + const uint ideal_reg = xmm_ideal_reg_for_size(reg_data._size); + _spill_offset -= reg_data._size; + vec_spill_helper(__ code(), false /* do_size */, false /* is_load */, _spill_offset, opto_reg, ideal_reg, tty); + } + + void xmm_register_restore(const XMMRegisterData& reg_data) { + const OptoReg::Name opto_reg = OptoReg::as_OptoReg(reg_data._reg->as_VMReg()); + const uint ideal_reg = xmm_ideal_reg_for_size(reg_data._size); + vec_spill_helper(__ code(), false /* do_size */, true /* is_load */, _spill_offset, opto_reg, ideal_reg, tty); + _spill_offset += reg_data._size; + } + + void gp_register_save(Register reg) { + _spill_offset -= 8; + __ movq(Address(rsp, _spill_offset), reg); + } + + void gp_register_restore(Register reg) { + __ movq(reg, Address(rsp, _spill_offset)); + _spill_offset += 8; } - // Create stub name - char name[64]; - const bool weak = (decorators & ON_WEAK_OOP_REF) != 0; - os::snprintf(name, sizeof(name), "zgc_load_barrier%s_stub_%s", weak ? "_weak" : "", raddr->name()); + void initialize(ZLoadBarrierStubC2* stub) { + // Create mask of caller saved registers that need to + // be saved/restored if live + RegMask caller_saved; + caller_saved.Insert(OptoReg::as_OptoReg(rax->as_VMReg())); + caller_saved.Insert(OptoReg::as_OptoReg(rcx->as_VMReg())); + caller_saved.Insert(OptoReg::as_OptoReg(rdx->as_VMReg())); + caller_saved.Insert(OptoReg::as_OptoReg(rsi->as_VMReg())); + caller_saved.Insert(OptoReg::as_OptoReg(rdi->as_VMReg())); + caller_saved.Insert(OptoReg::as_OptoReg(r8->as_VMReg())); + caller_saved.Insert(OptoReg::as_OptoReg(r9->as_VMReg())); + caller_saved.Insert(OptoReg::as_OptoReg(r10->as_VMReg())); + caller_saved.Insert(OptoReg::as_OptoReg(r11->as_VMReg())); + caller_saved.Remove(OptoReg::as_OptoReg(stub->ref()->as_VMReg())); - __ align(CodeEntryAlignment); - StubCodeMark mark(cgen, "StubRoutines", os::strdup(name, mtCode)); - address start = __ pc(); + // Create mask of live registers + RegMask live = stub->live(); + if (stub->tmp() != noreg) { + live.Insert(OptoReg::as_OptoReg(stub->tmp()->as_VMReg())); + } + + int gp_spill_size = 0; + int xmm_spill_size = 0; + + // Record registers that needs to be saved/restored + while (live.is_NotEmpty()) { + const OptoReg::Name opto_reg = live.find_first_elem(); + const VMReg vm_reg = OptoReg::as_VMReg(opto_reg); + + live.Remove(opto_reg); - // Save live registers - if (raddr != rax) { - __ push(rax); - } - if (raddr != rcx) { - __ push(rcx); - } - if (raddr != rdx) { - __ push(rdx); - } - if (raddr != rsi) { - __ push(rsi); - } - if (raddr != rdi) { - __ push(rdi); - } - if (raddr != r8) { - __ push(r8); - } - if (raddr != r9) { - __ push(r9); - } - if (raddr != r10) { - __ push(r10); - } - if (raddr != r11) { - __ push(r11); + if (vm_reg->is_Register()) { + if (caller_saved.Member(opto_reg)) { + _gp_registers.append(vm_reg->as_Register()); + gp_spill_size += 8; + } + } else if (vm_reg->is_XMMRegister()) { + // We encode in the low order 4 bits of the opto_reg, how large part of the register is live + const VMReg vm_reg_base = OptoReg::as_VMReg(opto_reg & ~15); + const int reg_size = xmm_slot_size(opto_reg); + const XMMRegisterData reg_data = { vm_reg_base->as_XMMRegister(), reg_size }; + const int reg_index = _xmm_registers.find(reg_data); + if (reg_index == -1) { + // Not previously appended + _xmm_registers.append(reg_data); + xmm_spill_size += reg_size; + } else { + // Previously appended, update size + const int reg_size_prev = _xmm_registers.at(reg_index)._size; + if (reg_size > reg_size_prev) { + _xmm_registers.at_put(reg_index, reg_data); + xmm_spill_size += reg_size - reg_size_prev; + } + } + } else { + fatal("Unexpected register type"); + } + } + + // Sort by size, largest first + _xmm_registers.sort(xmm_compare_register_size); + + // Stack pointer must be 16 bytes aligned for the call + _spill_offset = _spill_size = align_up(xmm_spill_size + gp_spill_size, 16); } - // Setup arguments - if (raddr != c_rarg1) { - __ movq(c_rarg1, raddr); - } - __ movq(c_rarg0, Address(raddr, 0)); +public: + ZSaveLiveRegisters(MacroAssembler* masm, ZLoadBarrierStubC2* stub) : + _masm(masm), + _gp_registers(), + _xmm_registers(), + _spill_size(0), + _spill_offset(0) { - // Call barrier function - __ call_VM_leaf(ZBarrierSetRuntime::load_barrier_on_oop_field_preloaded_addr(decorators), c_rarg0, c_rarg1); + // + // Stack layout after registers have been spilled: + // + // | ... | original rsp, 16 bytes aligned + // ------------------ + // | zmm0 high | + // | ... | + // | zmm0 low | 16 bytes aligned + // | ... | + // | ymm1 high | + // | ... | + // | ymm1 low | 16 bytes aligned + // | ... | + // | xmmN high | + // | ... | + // | xmmN low | 8 bytes aligned + // | reg0 | 8 bytes aligned + // | reg1 | + // | ... | + // | regN | new rsp, if 16 bytes aligned + // | <padding> | else new rsp, 16 bytes aligned + // ------------------ + // - // Move result returned in rax to raddr, if needed - if (raddr != rax) { - __ movq(raddr, rax); + // Figure out what registers to save/restore + initialize(stub); + + // Allocate stack space + if (_spill_size > 0) { + __ subptr(rsp, _spill_size); + } + + // Save XMM/YMM/ZMM registers + for (int i = 0; i < _xmm_registers.length(); i++) { + xmm_register_save(_xmm_registers.at(i)); + } + + if (xmm_needs_vzeroupper()) { + __ vzeroupper(); + } + + // Save general purpose registers + for (int i = 0; i < _gp_registers.length(); i++) { + gp_register_save(_gp_registers.at(i)); + } } - // Restore saved registers - if (raddr != r11) { - __ pop(r11); - } - if (raddr != r10) { - __ pop(r10); - } - if (raddr != r9) { - __ pop(r9); - } - if (raddr != r8) { - __ pop(r8); + ~ZSaveLiveRegisters() { + // Restore general purpose registers + for (int i = _gp_registers.length() - 1; i >= 0; i--) { + gp_register_restore(_gp_registers.at(i)); + } + + __ vzeroupper(); + + // Restore XMM/YMM/ZMM registers + for (int i = _xmm_registers.length() - 1; i >= 0; i--) { + xmm_register_restore(_xmm_registers.at(i)); + } + + // Free stack space + if (_spill_size > 0) { + __ addptr(rsp, _spill_size); + } } - if (raddr != rdi) { - __ pop(rdi); - } - if (raddr != rsi) { - __ pop(rsi); - } - if (raddr != rdx) { - __ pop(rdx); - } - if (raddr != rcx) { - __ pop(rcx); - } - if (raddr != rax) { - __ pop(rax); +}; + +class ZSetupArguments { +private: + MacroAssembler* const _masm; + const Register _ref; + const Address _ref_addr; + +public: + ZSetupArguments(MacroAssembler* masm, ZLoadBarrierStubC2* stub) : + _masm(masm), + _ref(stub->ref()), + _ref_addr(stub->ref_addr()) { + + // Setup arguments + if (_ref_addr.base() == noreg) { + // No self healing + if (_ref != c_rarg0) { + __ movq(c_rarg0, _ref); + } + __ xorq(c_rarg1, c_rarg1); + } else { + // Self healing + if (_ref == c_rarg0) { + __ lea(c_rarg1, _ref_addr); + } else if (_ref != c_rarg1) { + __ lea(c_rarg1, _ref_addr); + __ movq(c_rarg0, _ref); + } else if (_ref_addr.base() != c_rarg0 && _ref_addr.index() != c_rarg0) { + __ movq(c_rarg0, _ref); + __ lea(c_rarg1, _ref_addr); + } else { + __ xchgq(c_rarg0, c_rarg1); + if (_ref_addr.base() == c_rarg0) { + __ lea(c_rarg1, Address(c_rarg1, _ref_addr.index(), _ref_addr.scale(), _ref_addr.disp())); + } else if (_ref_addr.index() == c_rarg0) { + __ lea(c_rarg1, Address(_ref_addr.base(), c_rarg1, _ref_addr.scale(), _ref_addr.disp())); + } else { + ShouldNotReachHere(); + } + } + } } - __ ret(0); + ~ZSetupArguments() { + // Transfer result + if (_ref != rax) { + __ movq(_ref, rax); + } + } +}; + +#undef __ +#define __ masm-> - return start; +void ZBarrierSetAssembler::generate_c2_load_barrier_stub(MacroAssembler* masm, ZLoadBarrierStubC2* stub) const { + BLOCK_COMMENT("ZLoadBarrierStubC2"); + + // Stub entry + __ bind(*stub->entry()); + + { + ZSaveLiveRegisters save_live_registers(masm, stub); + ZSetupArguments setup_arguments(masm, stub); + __ call(RuntimeAddress(stub->slow_path())); + } + + // Stub exit + __ jmp(*stub->continuation()); } #undef __ -static void barrier_stubs_init_inner(const char* label, const DecoratorSet decorators, address* stub) { - const int nregs = RegisterImpl::number_of_registers; - const int code_size = nregs * 128; // Rough estimate of code size - - ResourceMark rm; - - CodeBuffer buf(BufferBlob::create(label, code_size)); - StubCodeGenerator cgen(&buf); - - for (int i = 0; i < nregs; i++) { - const Register reg = as_Register(i); - stub[i] = generate_load_barrier_stub(&cgen, reg, decorators); - } -} - -void ZBarrierSetAssembler::barrier_stubs_init() { - barrier_stubs_init_inner("zgc_load_barrier_stubs", ON_STRONG_OOP_REF, _load_barrier_slow_stub); - barrier_stubs_init_inner("zgc_load_barrier_weak_stubs", ON_WEAK_OOP_REF, _load_barrier_weak_slow_stub); -} - -address ZBarrierSetAssembler::load_barrier_slow_stub(Register reg) { - return _load_barrier_slow_stub[reg->encoding()]; -} - -address ZBarrierSetAssembler::load_barrier_weak_slow_stub(Register reg) { - return _load_barrier_weak_slow_stub[reg->encoding()]; -} +#endif // COMPILER2
--- a/src/hotspot/cpu/x86/gc/z/zBarrierSetAssembler_x86.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/gc/z/zBarrierSetAssembler_x86.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -24,6 +24,14 @@ #ifndef CPU_X86_GC_Z_ZBARRIERSETASSEMBLER_X86_HPP #define CPU_X86_GC_Z_ZBARRIERSETASSEMBLER_X86_HPP +#include "code/vmreg.hpp" +#include "oops/accessDecorators.hpp" +#ifdef COMPILER2 +#include "opto/optoreg.hpp" +#endif // COMPILER2 + +class MacroAssembler; + #ifdef COMPILER1 class LIR_Assembler; class LIR_OprDesc; @@ -32,14 +40,13 @@ class ZLoadBarrierStubC1; #endif // COMPILER1 +#ifdef COMPILER2 +class Node; +class ZLoadBarrierStubC2; +#endif // COMPILER2 + class ZBarrierSetAssembler : public ZBarrierSetAssemblerBase { -private: - address _load_barrier_slow_stub[RegisterImpl::number_of_registers]; - address _load_barrier_weak_slow_stub[RegisterImpl::number_of_registers]; - public: - ZBarrierSetAssembler(); - virtual void load_at(MacroAssembler* masm, DecoratorSet decorators, BasicType type, @@ -82,10 +89,13 @@ DecoratorSet decorators) const; #endif // COMPILER1 - virtual void barrier_stubs_init(); +#ifdef COMPILER2 + OptoReg::Name refine_register(const Node* node, + OptoReg::Name opto_reg); - address load_barrier_slow_stub(Register reg); - address load_barrier_weak_slow_stub(Register reg); + void generate_c2_load_barrier_stub(MacroAssembler* masm, + ZLoadBarrierStubC2* stub) const; +#endif // COMPILER2 }; #endif // CPU_X86_GC_Z_ZBARRIERSETASSEMBLER_X86_HPP
--- a/src/hotspot/cpu/x86/gc/z/z_x86_64.ad Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/gc/z/z_x86_64.ad Mon Oct 14 21:01:25 2019 +0000 @@ -24,190 +24,144 @@ source_hpp %{ #include "gc/z/c2/zBarrierSetC2.hpp" +#include "gc/z/zThreadLocalData.hpp" %} source %{ -#include "gc/z/zBarrierSetAssembler.hpp" +static void z_load_barrier(MacroAssembler& _masm, const MachNode* node, Address ref_addr, Register ref, Register tmp, bool weak) { + ZLoadBarrierStubC2* const stub = ZLoadBarrierStubC2::create(node, ref_addr, ref, tmp, weak); + __ testptr(ref, Address(r15_thread, ZThreadLocalData::address_bad_mask_offset())); + __ jcc(Assembler::notZero, *stub->entry()); + __ bind(*stub->continuation()); +} -static void z_load_barrier_slow_reg(MacroAssembler& _masm, Register dst, Address src, bool weak) { - assert(dst != rsp, "Invalid register"); - assert(dst != r15, "Invalid register"); - - const address stub = weak ? ZBarrierSet::assembler()->load_barrier_weak_slow_stub(dst) - : ZBarrierSet::assembler()->load_barrier_slow_stub(dst); - __ lea(dst, src); - __ call(RuntimeAddress(stub)); +static void z_load_barrier_slow_path(MacroAssembler& _masm, const MachNode* node, Address ref_addr, Register ref, Register tmp) { + ZLoadBarrierStubC2* const stub = ZLoadBarrierStubC2::create(node, ref_addr, ref, tmp, false /* weak */); + __ jmp(*stub->entry()); + __ bind(*stub->continuation()); } %} -// For XMM and YMM enabled processors -instruct zLoadBarrierSlowRegXmmAndYmm(rRegP dst, memory src, rFlagsReg cr, - rxmm0 x0, rxmm1 x1, rxmm2 x2, rxmm3 x3, - rxmm4 x4, rxmm5 x5, rxmm6 x6, rxmm7 x7, - rxmm8 x8, rxmm9 x9, rxmm10 x10, rxmm11 x11, - rxmm12 x12, rxmm13 x13, rxmm14 x14, rxmm15 x15) %{ - match(Set dst (LoadBarrierSlowReg src dst)); - predicate(UseAVX <= 2 && !n->as_LoadBarrierSlowReg()->is_weak()); +// Load Pointer +instruct zLoadP(rRegP dst, memory mem, rFlagsReg cr) +%{ + predicate(UseZGC && n->as_Load()->barrier_data() == ZLoadBarrierStrong); + match(Set dst (LoadP mem)); + effect(KILL cr, TEMP dst); - effect(KILL cr, - KILL x0, KILL x1, KILL x2, KILL x3, - KILL x4, KILL x5, KILL x6, KILL x7, - KILL x8, KILL x9, KILL x10, KILL x11, - KILL x12, KILL x13, KILL x14, KILL x15); + ins_cost(125); - format %{ "lea $dst, $src\n\t" - "call #ZLoadBarrierSlowPath" %} + format %{ "movq $dst, $mem" %} ins_encode %{ - z_load_barrier_slow_reg(_masm, $dst$$Register, $src$$Address, false /* weak */); + __ movptr($dst$$Register, $mem$$Address); + if (barrier_data() != ZLoadBarrierElided) { + z_load_barrier(_masm, this, $mem$$Address, $dst$$Register, noreg /* tmp */, false /* weak */); + } %} - ins_pipe(pipe_slow); + + ins_pipe(ialu_reg_mem); %} -// For ZMM enabled processors -instruct zLoadBarrierSlowRegZmm(rRegP dst, memory src, rFlagsReg cr, - rxmm0 x0, rxmm1 x1, rxmm2 x2, rxmm3 x3, - rxmm4 x4, rxmm5 x5, rxmm6 x6, rxmm7 x7, - rxmm8 x8, rxmm9 x9, rxmm10 x10, rxmm11 x11, - rxmm12 x12, rxmm13 x13, rxmm14 x14, rxmm15 x15, - rxmm16 x16, rxmm17 x17, rxmm18 x18, rxmm19 x19, - rxmm20 x20, rxmm21 x21, rxmm22 x22, rxmm23 x23, - rxmm24 x24, rxmm25 x25, rxmm26 x26, rxmm27 x27, - rxmm28 x28, rxmm29 x29, rxmm30 x30, rxmm31 x31) %{ +// Load Weak Pointer +instruct zLoadWeakP(rRegP dst, memory mem, rFlagsReg cr) +%{ + predicate(UseZGC && n->as_Load()->barrier_data() == ZLoadBarrierWeak); + match(Set dst (LoadP mem)); + effect(KILL cr, TEMP dst); - match(Set dst (LoadBarrierSlowReg src dst)); - predicate(UseAVX == 3 && !n->as_LoadBarrierSlowReg()->is_weak()); + ins_cost(125); - effect(KILL cr, - KILL x0, KILL x1, KILL x2, KILL x3, - KILL x4, KILL x5, KILL x6, KILL x7, - KILL x8, KILL x9, KILL x10, KILL x11, - KILL x12, KILL x13, KILL x14, KILL x15, - KILL x16, KILL x17, KILL x18, KILL x19, - KILL x20, KILL x21, KILL x22, KILL x23, - KILL x24, KILL x25, KILL x26, KILL x27, - KILL x28, KILL x29, KILL x30, KILL x31); - - format %{ "lea $dst, $src\n\t" - "call #ZLoadBarrierSlowPath" %} + format %{ "movq $dst, $mem" %} ins_encode %{ - z_load_barrier_slow_reg(_masm, $dst$$Register, $src$$Address, false /* weak */); + __ movptr($dst$$Register, $mem$$Address); + z_load_barrier(_masm, this, $mem$$Address, $dst$$Register, noreg /* tmp */, true /* weak */); %} - ins_pipe(pipe_slow); -%} - -// For XMM and YMM enabled processors -instruct zLoadBarrierWeakSlowRegXmmAndYmm(rRegP dst, memory src, rFlagsReg cr, - rxmm0 x0, rxmm1 x1, rxmm2 x2, rxmm3 x3, - rxmm4 x4, rxmm5 x5, rxmm6 x6, rxmm7 x7, - rxmm8 x8, rxmm9 x9, rxmm10 x10, rxmm11 x11, - rxmm12 x12, rxmm13 x13, rxmm14 x14, rxmm15 x15) %{ - match(Set dst (LoadBarrierSlowReg src dst)); - predicate(UseAVX <= 2 && n->as_LoadBarrierSlowReg()->is_weak()); - effect(KILL cr, - KILL x0, KILL x1, KILL x2, KILL x3, - KILL x4, KILL x5, KILL x6, KILL x7, - KILL x8, KILL x9, KILL x10, KILL x11, - KILL x12, KILL x13, KILL x14, KILL x15); - - format %{ "lea $dst, $src\n\t" - "call #ZLoadBarrierSlowPath" %} - - ins_encode %{ - z_load_barrier_slow_reg(_masm, $dst$$Register, $src$$Address, true /* weak */); - %} - ins_pipe(pipe_slow); + ins_pipe(ialu_reg_mem); %} -// For ZMM enabled processors -instruct zLoadBarrierWeakSlowRegZmm(rRegP dst, memory src, rFlagsReg cr, - rxmm0 x0, rxmm1 x1, rxmm2 x2, rxmm3 x3, - rxmm4 x4, rxmm5 x5, rxmm6 x6, rxmm7 x7, - rxmm8 x8, rxmm9 x9, rxmm10 x10, rxmm11 x11, - rxmm12 x12, rxmm13 x13, rxmm14 x14, rxmm15 x15, - rxmm16 x16, rxmm17 x17, rxmm18 x18, rxmm19 x19, - rxmm20 x20, rxmm21 x21, rxmm22 x22, rxmm23 x23, - rxmm24 x24, rxmm25 x25, rxmm26 x26, rxmm27 x27, - rxmm28 x28, rxmm29 x29, rxmm30 x30, rxmm31 x31) %{ +instruct zCompareAndExchangeP(memory mem, rax_RegP oldval, rRegP newval, rRegP tmp, rFlagsReg cr) %{ + match(Set oldval (CompareAndExchangeP mem (Binary oldval newval))); + predicate(UseZGC && n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong); + effect(KILL cr, TEMP tmp); - match(Set dst (LoadBarrierSlowReg src dst)); - predicate(UseAVX == 3 && n->as_LoadBarrierSlowReg()->is_weak()); - - effect(KILL cr, - KILL x0, KILL x1, KILL x2, KILL x3, - KILL x4, KILL x5, KILL x6, KILL x7, - KILL x8, KILL x9, KILL x10, KILL x11, - KILL x12, KILL x13, KILL x14, KILL x15, - KILL x16, KILL x17, KILL x18, KILL x19, - KILL x20, KILL x21, KILL x22, KILL x23, - KILL x24, KILL x25, KILL x26, KILL x27, - KILL x28, KILL x29, KILL x30, KILL x31); - - format %{ "lea $dst, $src\n\t" - "call #ZLoadBarrierSlowPath" %} + format %{ "lock\n\t" + "cmpxchgq $newval, $mem" %} ins_encode %{ - z_load_barrier_slow_reg(_masm, $dst$$Register, $src$$Address, true /* weak */); + if (barrier_data() != ZLoadBarrierElided) { + __ movptr($tmp$$Register, $oldval$$Register); + } + __ lock(); + __ cmpxchgptr($newval$$Register, $mem$$Address); + if (barrier_data() != ZLoadBarrierElided) { + Label good; + __ testptr($oldval$$Register, Address(r15_thread, ZThreadLocalData::address_bad_mask_offset())); + __ jcc(Assembler::zero, good); + z_load_barrier_slow_path(_masm, this, $mem$$Address, $oldval$$Register, $tmp$$Register); + __ movptr($oldval$$Register, $tmp$$Register); + __ lock(); + __ cmpxchgptr($newval$$Register, $mem$$Address); + __ bind(good); + } %} - ins_pipe(pipe_slow); + + ins_pipe(pipe_cmpxchg); %} -// Specialized versions of compareAndExchangeP that adds a keepalive that is consumed -// but doesn't affect output. +instruct zCompareAndSwapP(rRegI res, memory mem, rRegP newval, rRegP tmp, rFlagsReg cr, rax_RegP oldval) %{ + match(Set res (CompareAndSwapP mem (Binary oldval newval))); + match(Set res (WeakCompareAndSwapP mem (Binary oldval newval))); + predicate(UseZGC && n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong); + effect(KILL cr, KILL oldval, TEMP tmp); + + format %{ "lock\n\t" + "cmpxchgq $newval, $mem\n\t" + "sete $res\n\t" + "movzbl $res, $res" %} -instruct z_compareAndExchangeP( - memory mem_ptr, - rax_RegP oldval, rRegP newval, rRegP keepalive, - rFlagsReg cr) %{ - predicate(VM_Version::supports_cx8()); - match(Set oldval (ZCompareAndExchangeP (Binary mem_ptr keepalive) (Binary oldval newval))); - effect(KILL cr); + ins_encode %{ + if (barrier_data() != ZLoadBarrierElided) { + __ movptr($tmp$$Register, $oldval$$Register); + } + __ lock(); + __ cmpxchgptr($newval$$Register, $mem$$Address); + if (barrier_data() != ZLoadBarrierElided) { + Label good; + __ testptr($oldval$$Register, Address(r15_thread, ZThreadLocalData::address_bad_mask_offset())); + __ jcc(Assembler::zero, good); + z_load_barrier_slow_path(_masm, this, $mem$$Address, $oldval$$Register, $tmp$$Register); + __ movptr($oldval$$Register, $tmp$$Register); + __ lock(); + __ cmpxchgptr($newval$$Register, $mem$$Address); + __ bind(good); + __ cmpptr($tmp$$Register, $oldval$$Register); + } + __ setb(Assembler::equal, $res$$Register); + __ movzbl($res$$Register, $res$$Register); + %} - format %{ "cmpxchgq $mem_ptr,$newval\t# " - "If rax == $mem_ptr then store $newval into $mem_ptr\n\t" %} - opcode(0x0F, 0xB1); - ins_encode(lock_prefix, - REX_reg_mem_wide(newval, mem_ptr), - OpcP, OpcS, - reg_mem(newval, mem_ptr) // lock cmpxchg - ); - ins_pipe( pipe_cmpxchg ); + ins_pipe(pipe_cmpxchg); %} -instruct z_compareAndSwapP(rRegI res, - memory mem_ptr, - rax_RegP oldval, rRegP newval, rRegP keepalive, - rFlagsReg cr) %{ - predicate(VM_Version::supports_cx8()); - match(Set res (ZCompareAndSwapP (Binary mem_ptr keepalive) (Binary oldval newval))); - match(Set res (ZWeakCompareAndSwapP (Binary mem_ptr keepalive) (Binary oldval newval))); - effect(KILL cr, KILL oldval); +instruct zXChgP(memory mem, rRegP newval, rFlagsReg cr) %{ + match(Set newval (GetAndSetP mem newval)); + predicate(UseZGC && n->as_LoadStore()->barrier_data() == ZLoadBarrierStrong); + effect(KILL cr); + + format %{ "xchgq $newval, $mem" %} - format %{ "cmpxchgq $mem_ptr,$newval\t# " - "If rax == $mem_ptr then store $newval into $mem_ptr\n\t" - "sete $res\n\t" - "movzbl $res, $res" %} - opcode(0x0F, 0xB1); - ins_encode(lock_prefix, - REX_reg_mem_wide(newval, mem_ptr), - OpcP, OpcS, - reg_mem(newval, mem_ptr), - REX_breg(res), Opcode(0x0F), Opcode(0x94), reg(res), // sete - REX_reg_breg(res, res), // movzbl - Opcode(0xF), Opcode(0xB6), reg_reg(res, res)); - ins_pipe( pipe_cmpxchg ); + ins_encode %{ + __ xchgptr($newval$$Register, $mem$$Address); + if (barrier_data() != ZLoadBarrierElided) { + z_load_barrier(_masm, this, Address(noreg, 0), $newval$$Register, noreg /* tmp */, false /* weak */); + } + %} + + ins_pipe(pipe_cmpxchg); %} - -instruct z_xchgP( memory mem, rRegP newval, rRegP keepalive) %{ - match(Set newval (ZGetAndSetP mem (Binary newval keepalive))); - format %{ "XCHGQ $newval,[$mem]" %} - ins_encode %{ - __ xchgq($newval$$Register, $mem$$Address); - %} - ins_pipe( pipe_cmpxchg ); -%}
--- a/src/hotspot/cpu/x86/macroAssembler_x86.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/macroAssembler_x86.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -824,11 +824,13 @@ } void MacroAssembler::stop(const char* msg) { - address rip = pc(); - pusha(); // get regs on stack + if (ShowMessageBoxOnError) { + address rip = pc(); + pusha(); // get regs on stack + lea(c_rarg1, InternalAddress(rip)); + movq(c_rarg2, rsp); // pass pointer to regs array + } lea(c_rarg0, ExternalAddress((address) msg)); - lea(c_rarg1, InternalAddress(rip)); - movq(c_rarg2, rsp); // pass pointer to regs array andq(rsp, -16); // align stack as required by ABI call(RuntimeAddress(CAST_FROM_FN_PTR(address, MacroAssembler::debug64))); hlt();
--- a/src/hotspot/cpu/x86/sharedRuntime_x86_32.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/sharedRuntime_x86_32.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -34,6 +34,7 @@ #include "logging/log.hpp" #include "memory/resourceArea.hpp" #include "oops/compiledICHolder.hpp" +#include "oops/klass.inline.hpp" #include "runtime/safepointMechanism.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/vframeArray.hpp"
--- a/src/hotspot/cpu/x86/sharedRuntime_x86_64.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/sharedRuntime_x86_64.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -41,6 +41,7 @@ #include "memory/resourceArea.hpp" #include "memory/universe.hpp" #include "oops/compiledICHolder.hpp" +#include "oops/klass.inline.hpp" #include "runtime/safepointMechanism.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/vframeArray.hpp"
--- a/src/hotspot/cpu/x86/x86.ad Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/x86.ad Mon Oct 14 21:01:25 2019 +0000 @@ -1097,138 +1097,6 @@ reg_class_dynamic vectorz_reg(vectorz_reg_evex, vectorz_reg_legacy, %{ VM_Version::supports_evex() %} ); reg_class_dynamic vectorz_reg_vl(vectorz_reg_evex, vectorz_reg_legacy, %{ VM_Version::supports_evex() && VM_Version::supports_avx512vl() %} ); -reg_class xmm0_reg(XMM0, XMM0b, XMM0c, XMM0d); -reg_class ymm0_reg(XMM0, XMM0b, XMM0c, XMM0d, XMM0e, XMM0f, XMM0g, XMM0h); -reg_class zmm0_reg(XMM0, XMM0b, XMM0c, XMM0d, XMM0e, XMM0f, XMM0g, XMM0h, XMM0i, XMM0j, XMM0k, XMM0l, XMM0m, XMM0n, XMM0o, XMM0p); - -reg_class xmm1_reg(XMM1, XMM1b, XMM1c, XMM1d); -reg_class ymm1_reg(XMM1, XMM1b, XMM1c, XMM1d, XMM1e, XMM1f, XMM1g, XMM1h); -reg_class zmm1_reg(XMM1, XMM1b, XMM1c, XMM1d, XMM1e, XMM1f, XMM1g, XMM1h, XMM1i, XMM1j, XMM1k, XMM1l, XMM1m, XMM1n, XMM1o, XMM1p); - -reg_class xmm2_reg(XMM2, XMM2b, XMM2c, XMM2d); -reg_class ymm2_reg(XMM2, XMM2b, XMM2c, XMM2d, XMM2e, XMM2f, XMM2g, XMM2h); -reg_class zmm2_reg(XMM2, XMM2b, XMM2c, XMM2d, XMM2e, XMM2f, XMM2g, XMM2h, XMM2i, XMM2j, XMM2k, XMM2l, XMM2m, XMM2n, XMM2o, XMM2p); - -reg_class xmm3_reg(XMM3, XMM3b, XMM3c, XMM3d); -reg_class ymm3_reg(XMM3, XMM3b, XMM3c, XMM3d, XMM3e, XMM3f, XMM3g, XMM3h); -reg_class zmm3_reg(XMM3, XMM3b, XMM3c, XMM3d, XMM3e, XMM3f, XMM3g, XMM3h, XMM3i, XMM3j, XMM3k, XMM3l, XMM3m, XMM3n, XMM3o, XMM3p); - -reg_class xmm4_reg(XMM4, XMM4b, XMM4c, XMM4d); -reg_class ymm4_reg(XMM4, XMM4b, XMM4c, XMM4d, XMM4e, XMM4f, XMM4g, XMM4h); -reg_class zmm4_reg(XMM4, XMM4b, XMM4c, XMM4d, XMM4e, XMM4f, XMM4g, XMM4h, XMM4i, XMM4j, XMM4k, XMM4l, XMM4m, XMM4n, XMM4o, XMM4p); - -reg_class xmm5_reg(XMM5, XMM5b, XMM5c, XMM5d); -reg_class ymm5_reg(XMM5, XMM5b, XMM5c, XMM5d, XMM5e, XMM5f, XMM5g, XMM5h); -reg_class zmm5_reg(XMM5, XMM5b, XMM5c, XMM5d, XMM5e, XMM5f, XMM5g, XMM5h, XMM5i, XMM5j, XMM5k, XMM5l, XMM5m, XMM5n, XMM5o, XMM5p); - -reg_class xmm6_reg(XMM6, XMM6b, XMM6c, XMM6d); -reg_class ymm6_reg(XMM6, XMM6b, XMM6c, XMM6d, XMM6e, XMM6f, XMM6g, XMM6h); -reg_class zmm6_reg(XMM6, XMM6b, XMM6c, XMM6d, XMM6e, XMM6f, XMM6g, XMM6h, XMM6i, XMM6j, XMM6k, XMM6l, XMM6m, XMM6n, XMM6o, XMM6p); - -reg_class xmm7_reg(XMM7, XMM7b, XMM7c, XMM7d); -reg_class ymm7_reg(XMM7, XMM7b, XMM7c, XMM7d, XMM7e, XMM7f, XMM7g, XMM7h); -reg_class zmm7_reg(XMM7, XMM7b, XMM7c, XMM7d, XMM7e, XMM7f, XMM7g, XMM7h, XMM7i, XMM7j, XMM7k, XMM7l, XMM7m, XMM7n, XMM7o, XMM7p); - -#ifdef _LP64 - -reg_class xmm8_reg(XMM8, XMM8b, XMM8c, XMM8d); -reg_class ymm8_reg(XMM8, XMM8b, XMM8c, XMM8d, XMM8e, XMM8f, XMM8g, XMM8h); -reg_class zmm8_reg(XMM8, XMM8b, XMM8c, XMM8d, XMM8e, XMM8f, XMM8g, XMM8h, XMM8i, XMM8j, XMM8k, XMM8l, XMM8m, XMM8n, XMM8o, XMM8p); - -reg_class xmm9_reg(XMM9, XMM9b, XMM9c, XMM9d); -reg_class ymm9_reg(XMM9, XMM9b, XMM9c, XMM9d, XMM9e, XMM9f, XMM9g, XMM9h); -reg_class zmm9_reg(XMM9, XMM9b, XMM9c, XMM9d, XMM9e, XMM9f, XMM9g, XMM9h, XMM9i, XMM9j, XMM9k, XMM9l, XMM9m, XMM9n, XMM9o, XMM9p); - -reg_class xmm10_reg(XMM10, XMM10b, XMM10c, XMM10d); -reg_class ymm10_reg(XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h); -reg_class zmm10_reg(XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h, XMM10i, XMM10j, XMM10k, XMM10l, XMM10m, XMM10n, XMM10o, XMM10p); - -reg_class xmm11_reg(XMM11, XMM11b, XMM11c, XMM11d); -reg_class ymm11_reg(XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h); -reg_class zmm11_reg(XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h, XMM11i, XMM11j, XMM11k, XMM11l, XMM11m, XMM11n, XMM11o, XMM11p); - -reg_class xmm12_reg(XMM12, XMM12b, XMM12c, XMM12d); -reg_class ymm12_reg(XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h); -reg_class zmm12_reg(XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h, XMM12i, XMM12j, XMM12k, XMM12l, XMM12m, XMM12n, XMM12o, XMM12p); - -reg_class xmm13_reg(XMM13, XMM13b, XMM13c, XMM13d); -reg_class ymm13_reg(XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h); -reg_class zmm13_reg(XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h, XMM13i, XMM13j, XMM13k, XMM13l, XMM13m, XMM13n, XMM13o, XMM13p); - -reg_class xmm14_reg(XMM14, XMM14b, XMM14c, XMM14d); -reg_class ymm14_reg(XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h); -reg_class zmm14_reg(XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h, XMM14i, XMM14j, XMM14k, XMM14l, XMM14m, XMM14n, XMM14o, XMM14p); - -reg_class xmm15_reg(XMM15, XMM15b, XMM15c, XMM15d); -reg_class ymm15_reg(XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h); -reg_class zmm15_reg(XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h, XMM15i, XMM15j, XMM15k, XMM15l, XMM15m, XMM15n, XMM15o, XMM15p); - -reg_class xmm16_reg(XMM16, XMM16b, XMM16c, XMM16d); -reg_class ymm16_reg(XMM16, XMM16b, XMM16c, XMM16d, XMM16e, XMM16f, XMM16g, XMM16h); -reg_class zmm16_reg(XMM16, XMM16b, XMM16c, XMM16d, XMM16e, XMM16f, XMM16g, XMM16h, XMM16i, XMM16j, XMM16k, XMM16l, XMM16m, XMM16n, XMM16o, XMM16p); - -reg_class xmm17_reg(XMM17, XMM17b, XMM17c, XMM17d); -reg_class ymm17_reg(XMM17, XMM17b, XMM17c, XMM17d, XMM17e, XMM17f, XMM17g, XMM17h); -reg_class zmm17_reg(XMM17, XMM17b, XMM17c, XMM17d, XMM17e, XMM17f, XMM17g, XMM17h, XMM17i, XMM17j, XMM17k, XMM17l, XMM17m, XMM17n, XMM17o, XMM17p); - -reg_class xmm18_reg(XMM18, XMM18b, XMM18c, XMM18d); -reg_class ymm18_reg(XMM18, XMM18b, XMM18c, XMM18d, XMM18e, XMM18f, XMM18g, XMM18h); -reg_class zmm18_reg(XMM18, XMM18b, XMM18c, XMM18d, XMM18e, XMM18f, XMM18g, XMM18h, XMM18i, XMM18j, XMM18k, XMM18l, XMM18m, XMM18n, XMM18o, XMM18p); - -reg_class xmm19_reg(XMM19, XMM19b, XMM19c, XMM19d); -reg_class ymm19_reg(XMM19, XMM19b, XMM19c, XMM19d, XMM19e, XMM19f, XMM19g, XMM19h); -reg_class zmm19_reg(XMM19, XMM19b, XMM19c, XMM19d, XMM19e, XMM19f, XMM19g, XMM19h, XMM19i, XMM19j, XMM19k, XMM19l, XMM19m, XMM19n, XMM19o, XMM19p); - -reg_class xmm20_reg(XMM20, XMM20b, XMM20c, XMM20d); -reg_class ymm20_reg(XMM20, XMM20b, XMM20c, XMM20d, XMM20e, XMM20f, XMM20g, XMM20h); -reg_class zmm20_reg(XMM20, XMM20b, XMM20c, XMM20d, XMM20e, XMM20f, XMM20g, XMM20h, XMM20i, XMM20j, XMM20k, XMM20l, XMM20m, XMM20n, XMM20o, XMM20p); - -reg_class xmm21_reg(XMM21, XMM21b, XMM21c, XMM21d); -reg_class ymm21_reg(XMM21, XMM21b, XMM21c, XMM21d, XMM21e, XMM21f, XMM21g, XMM21h); -reg_class zmm21_reg(XMM21, XMM21b, XMM21c, XMM21d, XMM21e, XMM21f, XMM21g, XMM21h, XMM21i, XMM21j, XMM21k, XMM21l, XMM21m, XMM21n, XMM21o, XMM21p); - -reg_class xmm22_reg(XMM22, XMM22b, XMM22c, XMM22d); -reg_class ymm22_reg(XMM22, XMM22b, XMM22c, XMM22d, XMM22e, XMM22f, XMM22g, XMM22h); -reg_class zmm22_reg(XMM22, XMM22b, XMM22c, XMM22d, XMM22e, XMM22f, XMM22g, XMM22h, XMM22i, XMM22j, XMM22k, XMM22l, XMM22m, XMM22n, XMM22o, XMM22p); - -reg_class xmm23_reg(XMM23, XMM23b, XMM23c, XMM23d); -reg_class ymm23_reg(XMM23, XMM23b, XMM23c, XMM23d, XMM23e, XMM23f, XMM23g, XMM23h); -reg_class zmm23_reg(XMM23, XMM23b, XMM23c, XMM23d, XMM23e, XMM23f, XMM23g, XMM23h, XMM23i, XMM23j, XMM23k, XMM23l, XMM23m, XMM23n, XMM23o, XMM23p); - -reg_class xmm24_reg(XMM24, XMM24b, XMM24c, XMM24d); -reg_class ymm24_reg(XMM24, XMM24b, XMM24c, XMM24d, XMM24e, XMM24f, XMM24g, XMM24h); -reg_class zmm24_reg(XMM24, XMM24b, XMM24c, XMM24d, XMM24e, XMM24f, XMM24g, XMM24h, XMM24i, XMM24j, XMM24k, XMM24l, XMM24m, XMM24n, XMM24o, XMM24p); - -reg_class xmm25_reg(XMM25, XMM25b, XMM25c, XMM25d); -reg_class ymm25_reg(XMM25, XMM25b, XMM25c, XMM25d, XMM25e, XMM25f, XMM25g, XMM25h); -reg_class zmm25_reg(XMM25, XMM25b, XMM25c, XMM25d, XMM25e, XMM25f, XMM25g, XMM25h, XMM25i, XMM25j, XMM25k, XMM25l, XMM25m, XMM25n, XMM25o, XMM25p); - -reg_class xmm26_reg(XMM26, XMM26b, XMM26c, XMM26d); -reg_class ymm26_reg(XMM26, XMM26b, XMM26c, XMM26d, XMM26e, XMM26f, XMM26g, XMM26h); -reg_class zmm26_reg(XMM26, XMM26b, XMM26c, XMM26d, XMM26e, XMM26f, XMM26g, XMM26h, XMM26i, XMM26j, XMM26k, XMM26l, XMM26m, XMM26n, XMM26o, XMM26p); - -reg_class xmm27_reg(XMM27, XMM27b, XMM27c, XMM27d); -reg_class ymm27_reg(XMM27, XMM27b, XMM27c, XMM27d, XMM27e, XMM27f, XMM27g, XMM27h); -reg_class zmm27_reg(XMM27, XMM27b, XMM27c, XMM27d, XMM27e, XMM27f, XMM27g, XMM27h, XMM27i, XMM27j, XMM27k, XMM27l, XMM27m, XMM27n, XMM27o, XMM27p); - -reg_class xmm28_reg(XMM28, XMM28b, XMM28c, XMM28d); -reg_class ymm28_reg(XMM28, XMM28b, XMM28c, XMM28d, XMM28e, XMM28f, XMM28g, XMM28h); -reg_class zmm28_reg(XMM28, XMM28b, XMM28c, XMM28d, XMM28e, XMM28f, XMM28g, XMM28h, XMM28i, XMM28j, XMM28k, XMM28l, XMM28m, XMM28n, XMM28o, XMM28p); - -reg_class xmm29_reg(XMM29, XMM29b, XMM29c, XMM29d); -reg_class ymm29_reg(XMM29, XMM29b, XMM29c, XMM29d, XMM29e, XMM29f, XMM29g, XMM29h); -reg_class zmm29_reg(XMM29, XMM29b, XMM29c, XMM29d, XMM29e, XMM29f, XMM29g, XMM29h, XMM29i, XMM29j, XMM29k, XMM29l, XMM29m, XMM29n, XMM29o, XMM29p); - -reg_class xmm30_reg(XMM30, XMM30b, XMM30c, XMM30d); -reg_class ymm30_reg(XMM30, XMM30b, XMM30c, XMM30d, XMM30e, XMM30f, XMM30g, XMM30h); -reg_class zmm30_reg(XMM30, XMM30b, XMM30c, XMM30d, XMM30e, XMM30f, XMM30g, XMM30h, XMM30i, XMM30j, XMM30k, XMM30l, XMM30m, XMM30n, XMM30o, XMM30p); - -reg_class xmm31_reg(XMM31, XMM31b, XMM31c, XMM31d); -reg_class ymm31_reg(XMM31, XMM31b, XMM31c, XMM31d, XMM31e, XMM31f, XMM31g, XMM31h); -reg_class zmm31_reg(XMM31, XMM31b, XMM31c, XMM31d, XMM31e, XMM31f, XMM31g, XMM31h, XMM31i, XMM31j, XMM31k, XMM31l, XMM31m, XMM31n, XMM31o, XMM31p); - -#endif - %} @@ -1800,8 +1668,8 @@ return (UseAVX > 2) ? 6 : 4; } -static int vec_spill_helper(CodeBuffer *cbuf, bool do_size, bool is_load, - int stack_offset, int reg, uint ireg, outputStream* st) { +int vec_spill_helper(CodeBuffer *cbuf, bool do_size, bool is_load, + int stack_offset, int reg, uint ireg, outputStream* st) { // In 64-bit VM size calculation is very complex. Emitting instructions // into scratch buffer is used to get size in 64-bit VM. LP64_ONLY( assert(!do_size, "this method calculates size only for 32-bit VM"); )
--- a/src/hotspot/cpu/x86/x86_64.ad Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/cpu/x86/x86_64.ad Mon Oct 14 21:01:25 2019 +0000 @@ -1058,8 +1058,8 @@ static int vec_mov_helper(CodeBuffer *cbuf, bool do_size, int src_lo, int dst_lo, int src_hi, int dst_hi, uint ireg, outputStream* st); -static int vec_spill_helper(CodeBuffer *cbuf, bool do_size, bool is_load, - int stack_offset, int reg, uint ireg, outputStream* st); +int vec_spill_helper(CodeBuffer *cbuf, bool do_size, bool is_load, + int stack_offset, int reg, uint ireg, outputStream* st); static void vec_stack_to_stack_helper(CodeBuffer *cbuf, int src_offset, int dst_offset, uint ireg, outputStream* st) { @@ -4260,200 +4260,6 @@ %} %} -// Operands for bound floating pointer register arguments -operand rxmm0() %{ - constraint(ALLOC_IN_RC(xmm0_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm1() %{ - constraint(ALLOC_IN_RC(xmm1_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm2() %{ - constraint(ALLOC_IN_RC(xmm2_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm3() %{ - constraint(ALLOC_IN_RC(xmm3_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm4() %{ - constraint(ALLOC_IN_RC(xmm4_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm5() %{ - constraint(ALLOC_IN_RC(xmm5_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm6() %{ - constraint(ALLOC_IN_RC(xmm6_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm7() %{ - constraint(ALLOC_IN_RC(xmm7_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm8() %{ - constraint(ALLOC_IN_RC(xmm8_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm9() %{ - constraint(ALLOC_IN_RC(xmm9_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm10() %{ - constraint(ALLOC_IN_RC(xmm10_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm11() %{ - constraint(ALLOC_IN_RC(xmm11_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm12() %{ - constraint(ALLOC_IN_RC(xmm12_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm13() %{ - constraint(ALLOC_IN_RC(xmm13_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm14() %{ - constraint(ALLOC_IN_RC(xmm14_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm15() %{ - constraint(ALLOC_IN_RC(xmm15_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm16() %{ - constraint(ALLOC_IN_RC(xmm16_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm17() %{ - constraint(ALLOC_IN_RC(xmm17_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm18() %{ - constraint(ALLOC_IN_RC(xmm18_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm19() %{ - constraint(ALLOC_IN_RC(xmm19_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm20() %{ - constraint(ALLOC_IN_RC(xmm20_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm21() %{ - constraint(ALLOC_IN_RC(xmm21_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm22() %{ - constraint(ALLOC_IN_RC(xmm22_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm23() %{ - constraint(ALLOC_IN_RC(xmm23_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm24() %{ - constraint(ALLOC_IN_RC(xmm24_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm25() %{ - constraint(ALLOC_IN_RC(xmm25_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm26() %{ - constraint(ALLOC_IN_RC(xmm26_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm27() %{ - constraint(ALLOC_IN_RC(xmm27_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm28() %{ - constraint(ALLOC_IN_RC(xmm28_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm29() %{ - constraint(ALLOC_IN_RC(xmm29_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm30() %{ - constraint(ALLOC_IN_RC(xmm30_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} -operand rxmm31() %{ - constraint(ALLOC_IN_RC(xmm31_reg)); - match(VecX); - format%{%} - interface(REG_INTER); -%} - //----------OPERAND CLASSES---------------------------------------------------- // Operand Classes are groups of operands that are used as to simplify // instruction definitions by not requiring the AD writer to specify separate @@ -5346,6 +5152,7 @@ instruct loadP(rRegP dst, memory mem) %{ match(Set dst (LoadP mem)); + predicate(n->as_Load()->barrier_data() == 0); ins_cost(125); // XXX format %{ "movq $dst, $mem\t# ptr" %} @@ -7794,6 +7601,7 @@ rax_RegP oldval, rRegP newval, rFlagsReg cr) %{ + predicate(n->as_LoadStore()->barrier_data() == 0); match(Set cr (StorePConditional heap_top_ptr (Binary oldval newval))); format %{ "cmpxchgq $heap_top_ptr, $newval\t# (ptr) " @@ -7845,7 +7653,7 @@ rax_RegP oldval, rRegP newval, rFlagsReg cr) %{ - predicate(VM_Version::supports_cx8()); + predicate(VM_Version::supports_cx8() && n->as_LoadStore()->barrier_data() == 0); match(Set res (CompareAndSwapP mem_ptr (Binary oldval newval))); match(Set res (WeakCompareAndSwapP mem_ptr (Binary oldval newval))); effect(KILL cr, KILL oldval); @@ -8087,7 +7895,7 @@ rax_RegP oldval, rRegP newval, rFlagsReg cr) %{ - predicate(VM_Version::supports_cx8()); + predicate(VM_Version::supports_cx8() && n->as_LoadStore()->barrier_data() == 0); match(Set oldval (CompareAndExchangeP mem_ptr (Binary oldval newval))); effect(KILL cr); @@ -8232,6 +8040,7 @@ instruct xchgP( memory mem, rRegP newval) %{ match(Set newval (GetAndSetP mem newval)); + predicate(n->as_LoadStore()->barrier_data() == 0); format %{ "XCHGQ $newval,[$mem]" %} ins_encode %{ __ xchgq($newval$$Register, $mem$$Address); @@ -11974,6 +11783,7 @@ instruct compP_rReg_mem(rFlagsRegU cr, rRegP op1, memory op2) %{ match(Set cr (CmpP op1 (LoadP op2))); + predicate(n->in(2)->as_Load()->barrier_data() == 0); ins_cost(500); // XXX format %{ "cmpq $op1, $op2\t# ptr" %} @@ -11999,7 +11809,8 @@ // and raw pointers have no anti-dependencies. instruct compP_mem_rReg(rFlagsRegU cr, rRegP op1, memory op2) %{ - predicate(n->in(2)->in(2)->bottom_type()->reloc() == relocInfo::none); + predicate(n->in(2)->in(2)->bottom_type()->reloc() == relocInfo::none && + n->in(2)->as_Load()->barrier_data() == 0); match(Set cr (CmpP op1 (LoadP op2))); format %{ "cmpq $op1, $op2\t# raw ptr" %} @@ -12024,7 +11835,8 @@ // any compare to a zero should be eq/neq. instruct testP_mem(rFlagsReg cr, memory op, immP0 zero) %{ - predicate(!UseCompressedOops || (CompressedOops::base() != NULL)); + predicate((!UseCompressedOops || (CompressedOops::base() != NULL)) && + n->in(1)->as_Load()->barrier_data() == 0); match(Set cr (CmpP (LoadP op) zero)); ins_cost(500); // XXX @@ -12037,7 +11849,9 @@ instruct testP_mem_reg0(rFlagsReg cr, memory mem, immP0 zero) %{ - predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL)); + predicate(UseCompressedOops && (CompressedOops::base() == NULL) && + (CompressedKlassPointers::base() == NULL) && + n->in(1)->as_Load()->barrier_data() == 0); match(Set cr (CmpP (LoadP mem) zero)); format %{ "cmpq R12, $mem\t# ptr (R12_heapbase==0)" %}
--- a/src/hotspot/os/aix/os_aix.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/os/aix/os_aix.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -132,18 +132,6 @@ #define ERROR_MP_VMGETINFO_CLAIMS_NO_SUPPORT_FOR_64K 103 // excerpts from systemcfg.h that might be missing on older os levels -#ifndef PV_5_Compat - #define PV_5_Compat 0x0F8000 /* Power PC 5 */ -#endif -#ifndef PV_6 - #define PV_6 0x100000 /* Power PC 6 */ -#endif -#ifndef PV_6_1 - #define PV_6_1 0x100001 /* Power PC 6 DD1.x */ -#endif -#ifndef PV_6_Compat - #define PV_6_Compat 0x108000 /* Power PC 6 */ -#endif #ifndef PV_7 #define PV_7 0x200000 /* Power PC 7 */ #endif @@ -156,6 +144,13 @@ #ifndef PV_8_Compat #define PV_8_Compat 0x308000 /* Power PC 8 */ #endif +#ifndef PV_9 + #define PV_9 0x400000 /* Power PC 9 */ +#endif +#ifndef PV_9_Compat + #define PV_9_Compat 0x408000 /* Power PC 9 */ +#endif + static address resolve_function_descriptor_to_code_pointer(address p); @@ -1386,15 +1381,7 @@ void os::print_os_info(outputStream* st) { st->print("OS:"); - st->print("uname:"); - struct utsname name; - uname(&name); - st->print(name.sysname); st->print(" "); - st->print(name.nodename); st->print(" "); - st->print(name.release); st->print(" "); - st->print(name.version); st->print(" "); - st->print(name.machine); - st->cr(); + os::Posix::print_uname_info(st); uint32_t ver = os::Aix::os_version(); st->print_cr("AIX kernel version %u.%u.%u.%u", @@ -1402,16 +1389,12 @@ os::Posix::print_rlimit_info(st); + os::Posix::print_load_average(st); + // _SC_THREAD_THREADS_MAX is the maximum number of threads within a process. long tmax = sysconf(_SC_THREAD_THREADS_MAX); st->print_cr("maximum #threads within a process:%ld", tmax); - // load average - st->print("load average:"); - double loadavg[3] = {-1.L, -1.L, -1.L}; - os::loadavg(loadavg, 3); - st->print_cr("%0.02f %0.02f %0.02f", loadavg[0], loadavg[1], loadavg[2]); - // print wpar info libperfstat::wparinfo_t wi; if (libperfstat::get_wparinfo(&wi)) { @@ -1504,6 +1487,9 @@ void os::get_summary_cpu_info(char* buf, size_t buflen) { // read _system_configuration.version switch (_system_configuration.version) { + case PV_9: + strncpy(buf, "Power PC 9", buflen); + break; case PV_8: strncpy(buf, "Power PC 8", buflen); break; @@ -1537,6 +1523,9 @@ case PV_8_Compat: strncpy(buf, "PV_8_Compat", buflen); break; + case PV_9_Compat: + strncpy(buf, "PV_9_Compat", buflen); + break; default: strncpy(buf, "unknown", buflen); } @@ -2767,7 +2756,7 @@ os::SuspendResume::State state = osthread->sr.suspended(); if (state == os::SuspendResume::SR_SUSPENDED) { sigset_t suspend_set; // signals for sigsuspend() - + sigemptyset(&suspend_set); // get current set of blocked signals and unblock resume signal pthread_sigmask(SIG_BLOCK, NULL, &suspend_set); sigdelset(&suspend_set, SR_signum); @@ -3053,6 +3042,7 @@ // try to honor the signal mask sigset_t oset; + sigemptyset(&oset); pthread_sigmask(SIG_SETMASK, &(actp->sa_mask), &oset); // call into the chained handler @@ -3063,7 +3053,7 @@ } // restore the signal mask - pthread_sigmask(SIG_SETMASK, &oset, 0); + pthread_sigmask(SIG_SETMASK, &oset, NULL); } // Tell jvm's signal handler the signal is taken care of. return true; @@ -4020,7 +4010,7 @@ void os::pause() { char filename[MAX_PATH]; if (PauseAtStartupFile && PauseAtStartupFile[0]) { - jio_snprintf(filename, MAX_PATH, PauseAtStartupFile); + jio_snprintf(filename, MAX_PATH, "%s", PauseAtStartupFile); } else { jio_snprintf(filename, MAX_PATH, "./vm.paused.%d", current_process_id()); }
--- a/src/hotspot/os/bsd/os_bsd.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/os/bsd/os_bsd.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -3671,7 +3671,7 @@ void os::pause() { char filename[MAX_PATH]; if (PauseAtStartupFile && PauseAtStartupFile[0]) { - jio_snprintf(filename, MAX_PATH, PauseAtStartupFile); + jio_snprintf(filename, MAX_PATH, "%s", PauseAtStartupFile); } else { jio_snprintf(filename, MAX_PATH, "./vm.paused.%d", current_process_id()); }
--- a/src/hotspot/os/posix/os_posix.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/os/posix/os_posix.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -373,8 +373,12 @@ void os::Posix::print_load_average(outputStream* st) { st->print("load average:"); double loadavg[3]; - os::loadavg(loadavg, 3); - st->print("%0.02f %0.02f %0.02f", loadavg[0], loadavg[1], loadavg[2]); + int res = os::loadavg(loadavg, 3); + if (res != -1) { + st->print("%0.02f %0.02f %0.02f", loadavg[0], loadavg[1], loadavg[2]); + } else { + st->print(" Unavailable"); + } st->cr(); }
--- a/src/hotspot/os/solaris/os_solaris.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/os/solaris/os_solaris.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -4428,7 +4428,7 @@ void os::pause() { char filename[MAX_PATH]; if (PauseAtStartupFile && PauseAtStartupFile[0]) { - jio_snprintf(filename, MAX_PATH, PauseAtStartupFile); + jio_snprintf(filename, MAX_PATH, "%s", PauseAtStartupFile); } else { jio_snprintf(filename, MAX_PATH, "./vm.paused.%d", current_process_id()); }
--- a/src/hotspot/os/windows/os_windows.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/os/windows/os_windows.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -4975,7 +4975,7 @@ void os::pause() { char filename[MAX_PATH]; if (PauseAtStartupFile && PauseAtStartupFile[0]) { - jio_snprintf(filename, MAX_PATH, PauseAtStartupFile); + jio_snprintf(filename, MAX_PATH, "%s", PauseAtStartupFile); } else { jio_snprintf(filename, MAX_PATH, "./vm.paused.%d", current_process_id()); }
--- a/src/hotspot/share/adlc/formssel.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/adlc/formssel.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -773,11 +773,6 @@ !strcmp(_matrule->_rChild->_opType,"CheckCastPP") || !strcmp(_matrule->_rChild->_opType,"GetAndSetP") || !strcmp(_matrule->_rChild->_opType,"GetAndSetN") || -#if INCLUDE_ZGC - !strcmp(_matrule->_rChild->_opType,"ZGetAndSetP") || - !strcmp(_matrule->_rChild->_opType,"ZCompareAndExchangeP") || - !strcmp(_matrule->_rChild->_opType,"LoadBarrierSlowReg") || -#endif #if INCLUDE_SHENANDOAHGC !strcmp(_matrule->_rChild->_opType,"ShenandoahCompareAndExchangeP") || !strcmp(_matrule->_rChild->_opType,"ShenandoahCompareAndExchangeN") || @@ -3510,9 +3505,6 @@ "StoreCM", "GetAndSetB", "GetAndSetS", "GetAndAddI", "GetAndSetI", "GetAndSetP", "GetAndAddB", "GetAndAddS", "GetAndAddL", "GetAndSetL", "GetAndSetN", -#if INCLUDE_ZGC - "ZGetAndSetP", "ZCompareAndSwapP", "ZCompareAndExchangeP", "ZWeakCompareAndSwapP", -#endif "ClearArray" }; int cnt = sizeof(needs_ideal_memory_list)/sizeof(char*);
--- a/src/hotspot/share/aot/aotCodeHeap.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/aot/aotCodeHeap.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -37,6 +37,7 @@ #include "memory/allocation.inline.hpp" #include "memory/universe.hpp" #include "oops/compressedOops.hpp" +#include "oops/klass.inline.hpp" #include "oops/method.inline.hpp" #include "runtime/deoptimization.hpp" #include "runtime/handles.inline.hpp"
--- a/src/hotspot/share/aot/aotCompiledMethod.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/aot/aotCompiledMethod.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -32,6 +32,7 @@ #include "compiler/compilerOracle.hpp" #include "gc/shared/cardTableBarrierSet.hpp" #include "gc/shared/collectedHeap.hpp" +#include "oops/klass.inline.hpp" #include "oops/method.inline.hpp" #include "runtime/frame.inline.hpp" #include "runtime/handles.inline.hpp"
--- a/src/hotspot/share/c1/c1_GraphBuilder.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/c1/c1_GraphBuilder.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -33,13 +33,13 @@ #include "ci/ciKlass.hpp" #include "ci/ciMemberName.hpp" #include "ci/ciUtilities.inline.hpp" +#include "compiler/compilationPolicy.hpp" #include "compiler/compileBroker.hpp" #include "interpreter/bytecode.hpp" #include "jfr/jfrEvents.hpp" #include "memory/resourceArea.hpp" #include "oops/oop.inline.hpp" #include "runtime/sharedRuntime.hpp" -#include "runtime/compilationPolicy.hpp" #include "runtime/vm_version.hpp" #include "utilities/bitMap.inline.hpp"
--- a/src/hotspot/share/c1/c1_LIRGenerator.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/c1/c1_LIRGenerator.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -36,6 +36,7 @@ #include "ci/ciUtilities.hpp" #include "gc/shared/barrierSet.hpp" #include "gc/shared/c1/barrierSetC1.hpp" +#include "oops/klass.inline.hpp" #include "runtime/arguments.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/stubRoutines.hpp"
--- a/src/hotspot/share/c1/c1_Runtime1.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/c1/c1_Runtime1.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -37,6 +37,7 @@ #include "code/pcDesc.hpp" #include "code/scopeDesc.hpp" #include "code/vtableStubs.hpp" +#include "compiler/compilationPolicy.hpp" #include "compiler/disassembler.hpp" #include "gc/shared/barrierSet.hpp" #include "gc/shared/c1/barrierSetC1.hpp" @@ -55,7 +56,6 @@ #include "oops/oop.inline.hpp" #include "runtime/atomic.hpp" #include "runtime/biasedLocking.hpp" -#include "runtime/compilationPolicy.hpp" #include "runtime/fieldDescriptor.inline.hpp" #include "runtime/frame.inline.hpp" #include "runtime/handles.inline.hpp"
--- a/src/hotspot/share/ci/ciEnv.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/ci/ciEnv.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -238,6 +238,7 @@ _jvmti_can_access_local_variables = JvmtiExport::can_access_local_variables(); _jvmti_can_post_on_exceptions = JvmtiExport::can_post_on_exceptions(); _jvmti_can_pop_frame = JvmtiExport::can_pop_frame(); + _jvmti_can_get_owned_monitor_info = JvmtiExport::can_get_owned_monitor_info(); } bool ciEnv::jvmti_state_changed() const { @@ -262,6 +263,10 @@ JvmtiExport::can_pop_frame()) { return true; } + if (!_jvmti_can_get_owned_monitor_info && + JvmtiExport::can_get_owned_monitor_info()) { + return true; + } return false; }
--- a/src/hotspot/share/ci/ciEnv.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/ci/ciEnv.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -73,6 +73,7 @@ bool _jvmti_can_access_local_variables; bool _jvmti_can_post_on_exceptions; bool _jvmti_can_pop_frame; + bool _jvmti_can_get_owned_monitor_info; // includes can_get_owned_monitor_stack_depth_info // Cache DTrace flags bool _dtrace_extended_probes; @@ -347,6 +348,7 @@ } bool jvmti_can_hotswap_or_post_breakpoint() const { return _jvmti_can_hotswap_or_post_breakpoint; } bool jvmti_can_post_on_exceptions() const { return _jvmti_can_post_on_exceptions; } + bool jvmti_can_get_owned_monitor_info() const { return _jvmti_can_get_owned_monitor_info; } // Cache DTrace flags void cache_dtrace_flags();
--- a/src/hotspot/share/classfile/classLoader.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/classfile/classLoader.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -57,7 +57,6 @@ #include "oops/symbol.hpp" #include "prims/jvm_misc.hpp" #include "runtime/arguments.hpp" -#include "runtime/compilationPolicy.hpp" #include "runtime/handles.inline.hpp" #include "runtime/init.hpp" #include "runtime/interfaceSupport.inline.hpp"
--- a/src/hotspot/share/classfile/classLoader.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/classfile/classLoader.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -237,6 +237,8 @@ CDS_ONLY(static ClassPathEntry* app_classpath_entries() {return _app_classpath_entries;}) CDS_ONLY(static ClassPathEntry* module_path_entries() {return _module_path_entries;}) + static bool has_bootclasspath_append() { return _first_append_entry != NULL; } + protected: // Initialization: // - setup the boot loader's system class path
--- a/src/hotspot/share/classfile/javaClasses.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/classfile/javaClasses.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -377,20 +377,24 @@ if (_to_java_string_fn == NULL) { void *lib_handle = os::native_java_library(); - _to_java_string_fn = CAST_TO_FN_PTR(to_java_string_fn_t, os::dll_lookup(lib_handle, "NewStringPlatform")); + _to_java_string_fn = CAST_TO_FN_PTR(to_java_string_fn_t, os::dll_lookup(lib_handle, "JNU_NewStringPlatform")); if (_to_java_string_fn == NULL) { fatal("NewStringPlatform missing"); } } jstring js = NULL; - { JavaThread* thread = (JavaThread*)THREAD; - assert(thread->is_Java_thread(), "must be java thread"); + { + assert(THREAD->is_Java_thread(), "must be java thread"); + JavaThread* thread = (JavaThread*)THREAD; HandleMark hm(thread); ThreadToNativeFromVM ttn(thread); js = (_to_java_string_fn)(thread->jni_environment(), str); } - return Handle(THREAD, JNIHandles::resolve(js)); + + Handle native_platform_string(THREAD, JNIHandles::resolve(js)); + JNIHandles::destroy_local(js); // destroy local JNIHandle. + return native_platform_string; } // Converts a Java String to a native C string that can be used for
--- a/src/hotspot/share/classfile/systemDictionary.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/classfile/systemDictionary.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -1205,10 +1205,8 @@ TempNewSymbol pkg_name = NULL; PackageEntry* pkg_entry = NULL; ModuleEntry* mod_entry = NULL; - const char* pkg_string = NULL; pkg_name = InstanceKlass::package_from_name(class_name, CHECK_false); if (pkg_name != NULL) { - pkg_string = pkg_name->as_C_string(); if (loader_data != NULL) { pkg_entry = loader_data->packages()->lookup_only(pkg_name); } @@ -1245,7 +1243,7 @@ // 3. or, the class is from an unamed module if (!ent->is_modules_image() && ik->is_shared_boot_class()) { // the class is from the -Xbootclasspath/a - if (pkg_string == NULL || + if (pkg_name == NULL || pkg_entry == NULL || pkg_entry->in_unnamed_module()) { assert(mod_entry == NULL || @@ -1257,8 +1255,7 @@ return false; } else { bool res = SystemDictionaryShared::is_shared_class_visible_for_classloader( - ik, class_loader, pkg_string, pkg_name, - pkg_entry, mod_entry, CHECK_(false)); + ik, class_loader, pkg_name, pkg_entry, mod_entry, CHECK_(false)); return res; } } @@ -1432,6 +1429,11 @@ // a named package within the unnamed module. In all cases, // limit visibility to search for the class only in the boot // loader's append path. + if (!ClassLoader::has_bootclasspath_append()) { + // If there is no bootclasspath append entry, no need to continue + // searching. + return NULL; + } search_only_bootloader_append = true; } }
--- a/src/hotspot/share/classfile/systemDictionaryShared.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/classfile/systemDictionaryShared.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -657,7 +657,6 @@ bool SystemDictionaryShared::is_shared_class_visible_for_classloader( InstanceKlass* ik, Handle class_loader, - const char* pkg_string, Symbol* pkg_name, PackageEntry* pkg_entry, ModuleEntry* mod_entry, @@ -684,7 +683,7 @@ } } else if (SystemDictionary::is_system_class_loader(class_loader())) { assert(ent != NULL, "shared class for system loader should have valid SharedClassPathEntry"); - if (pkg_string == NULL) { + if (pkg_name == NULL) { // The archived class is in the unnamed package. Currently, the boot image // does not contain any class in the unnamed package. assert(!ent->is_modules_image(), "Class in the unnamed package must be from the classpath"); @@ -906,14 +905,9 @@ return NULL; } - const RunTimeSharedClassInfo* record = find_record(&_unregistered_dictionary, class_name); + const RunTimeSharedClassInfo* record = find_record(&_unregistered_dictionary, &_dynamic_unregistered_dictionary, class_name); if (record == NULL) { - if (DynamicArchive::is_mapped()) { - record = find_record(&_dynamic_unregistered_dictionary, class_name); - } - if (record == NULL) { - return NULL; - } + return NULL; } int clsfile_size = cfs->length(); @@ -1413,29 +1407,34 @@ } const RunTimeSharedClassInfo* -SystemDictionaryShared::find_record(RunTimeSharedDictionary* dict, Symbol* name) { - if (UseSharedSpaces) { - unsigned int hash = primitive_hash<Symbol*>(name); - return dict->lookup(name, hash, 0); - } else { +SystemDictionaryShared::find_record(RunTimeSharedDictionary* static_dict, RunTimeSharedDictionary* dynamic_dict, Symbol* name) { + if (!UseSharedSpaces || !name->is_shared()) { + // The names of all shared classes must also be a shared Symbol. return NULL; } + + unsigned int hash = primitive_hash<Symbol*>(name); + const RunTimeSharedClassInfo* record = NULL; + if (!MetaspaceShared::is_shared_dynamic(name)) { + // The names of all shared classes in the static dict must also be in the + // static archive + record = static_dict->lookup(name, hash, 0); + } + + if (record == NULL && DynamicArchive::is_mapped()) { + record = dynamic_dict->lookup(name, hash, 0); + } + + return record; } InstanceKlass* SystemDictionaryShared::find_builtin_class(Symbol* name) { - const RunTimeSharedClassInfo* record = find_record(&_builtin_dictionary, name); - if (record) { + const RunTimeSharedClassInfo* record = find_record(&_builtin_dictionary, &_dynamic_builtin_dictionary, name); + if (record != NULL) { return record->_klass; + } else { + return NULL; } - - if (DynamicArchive::is_mapped()) { - record = find_record(&_dynamic_builtin_dictionary, name); - if (record) { - return record->_klass; - } - } - - return NULL; } void SystemDictionaryShared::update_shared_entry(InstanceKlass* k, int id) {
--- a/src/hotspot/share/classfile/systemDictionaryShared.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/classfile/systemDictionaryShared.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -223,7 +223,9 @@ public: static InstanceKlass* find_builtin_class(Symbol* class_name); - static const RunTimeSharedClassInfo* find_record(RunTimeSharedDictionary* dict, Symbol* name); + static const RunTimeSharedClassInfo* find_record(RunTimeSharedDictionary* static_dict, + RunTimeSharedDictionary* dynamic_dict, + Symbol* name); static bool has_platform_or_app_classes(); @@ -240,7 +242,6 @@ static bool is_sharing_possible(ClassLoaderData* loader_data); static bool is_shared_class_visible_for_classloader(InstanceKlass* ik, Handle class_loader, - const char* pkg_string, Symbol* pkg_name, PackageEntry* pkg_entry, ModuleEntry* mod_entry,
--- a/src/hotspot/share/classfile/verificationType.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/classfile/verificationType.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -28,6 +28,7 @@ #include "classfile/verificationType.hpp" #include "classfile/verifier.hpp" #include "logging/log.hpp" +#include "oops/klass.inline.hpp" #include "runtime/handles.inline.hpp" VerificationType VerificationType::from_tag(u1 tag) {
--- a/src/hotspot/share/code/codeCache.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/code/codeCache.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -33,6 +33,7 @@ #include "code/icBuffer.hpp" #include "code/nmethod.hpp" #include "code/pcDesc.hpp" +#include "compiler/compilationPolicy.hpp" #include "compiler/compileBroker.hpp" #include "jfr/jfrEvents.hpp" #include "logging/log.hpp" @@ -46,7 +47,6 @@ #include "oops/oop.inline.hpp" #include "oops/verifyOopClosure.hpp" #include "runtime/arguments.hpp" -#include "runtime/compilationPolicy.hpp" #include "runtime/deoptimization.hpp" #include "runtime/handles.inline.hpp" #include "runtime/icache.hpp"
--- a/src/hotspot/share/code/compiledIC.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/code/compiledIC.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -741,4 +741,22 @@ tty->cr(); } +void CompiledDirectStaticCall::verify_mt_safe(const methodHandle& callee, address entry, + NativeMovConstReg* method_holder, + NativeJump* jump) { + // A generated lambda form might be deleted from the Lambdaform + // cache in MethodTypeForm. If a jit compiled lambdaform method + // becomes not entrant and the cache access returns null, the new + // resolve will lead to a new generated LambdaForm. + Method* old_method = reinterpret_cast<Method*>(method_holder->data()); + assert(old_method == NULL || old_method == callee() || + callee->is_compiled_lambda_form() || + !old_method->method_holder()->is_loader_alive() || + old_method->is_old(), // may be race patching deoptimized nmethod due to redefinition. + "a) MT-unsafe modification of inline cache"); + + address destination = jump->jump_destination(); + assert(destination == (address)-1 || destination == entry, + "b) MT-unsafe modification of inline cache"); +} #endif // !PRODUCT
--- a/src/hotspot/share/code/compiledIC.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/code/compiledIC.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -402,6 +402,9 @@ // Also used by CompiledIC void set_to_interpreted(const methodHandle& callee, address entry); + void verify_mt_safe(const methodHandle& callee, address entry, + NativeMovConstReg* method_holder, + NativeJump* jump) PRODUCT_RETURN; #if INCLUDE_AOT void set_to_far(const methodHandle& callee, address entry); #endif
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/src/hotspot/share/compiler/compilationPolicy.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -0,0 +1,506 @@ +/* + * Copyright (c) 2000, 2019, 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. + * + */ + +#include "precompiled.hpp" +#include "classfile/classLoaderDataGraph.inline.hpp" +#include "code/compiledIC.hpp" +#include "code/nmethod.hpp" +#include "code/scopeDesc.hpp" +#include "compiler/compilationPolicy.hpp" +#include "compiler/tieredThresholdPolicy.hpp" +#include "interpreter/interpreter.hpp" +#include "memory/resourceArea.hpp" +#include "oops/methodData.hpp" +#include "oops/method.inline.hpp" +#include "oops/oop.inline.hpp" +#include "prims/nativeLookup.hpp" +#include "runtime/frame.hpp" +#include "runtime/handles.inline.hpp" +#include "runtime/stubRoutines.hpp" +#include "runtime/thread.hpp" +#include "runtime/vframe.hpp" +#include "runtime/vmOperations.hpp" +#include "utilities/events.hpp" +#include "utilities/globalDefinitions.hpp" + +#ifdef COMPILER1 +#include "c1/c1_Compiler.hpp" +#endif +#ifdef COMPILER2 +#include "opto/c2compiler.hpp" +#endif + +CompilationPolicy* CompilationPolicy::_policy; + +// Determine compilation policy based on command line argument +void compilationPolicy_init() { + #ifdef TIERED + if (TieredCompilation) { + CompilationPolicy::set_policy(new TieredThresholdPolicy()); + } else { + CompilationPolicy::set_policy(new SimpleCompPolicy()); + } + #else + CompilationPolicy::set_policy(new SimpleCompPolicy()); + #endif + + CompilationPolicy::policy()->initialize(); +} + +// Returns true if m must be compiled before executing it +// This is intended to force compiles for methods (usually for +// debugging) that would otherwise be interpreted for some reason. +bool CompilationPolicy::must_be_compiled(const methodHandle& m, int comp_level) { + // Don't allow Xcomp to cause compiles in replay mode + if (ReplayCompiles) return false; + + if (m->has_compiled_code()) return false; // already compiled + if (!can_be_compiled(m, comp_level)) return false; + + return !UseInterpreter || // must compile all methods + (UseCompiler && AlwaysCompileLoopMethods && m->has_loops() && CompileBroker::should_compile_new_jobs()); // eagerly compile loop methods +} + +void CompilationPolicy::compile_if_required(const methodHandle& selected_method, TRAPS) { + if (must_be_compiled(selected_method)) { + // This path is unusual, mostly used by the '-Xcomp' stress test mode. + + // Note: with several active threads, the must_be_compiled may be true + // while can_be_compiled is false; remove assert + // assert(CompilationPolicy::can_be_compiled(selected_method), "cannot compile"); + if (!THREAD->can_call_java() || THREAD->is_Compiler_thread()) { + // don't force compilation, resolve was on behalf of compiler + return; + } + if (selected_method->method_holder()->is_not_initialized()) { + // 'is_not_initialized' means not only '!is_initialized', but also that + // initialization has not been started yet ('!being_initialized') + // Do not force compilation of methods in uninitialized classes. + // Note that doing this would throw an assert later, + // in CompileBroker::compile_method. + // We sometimes use the link resolver to do reflective lookups + // even before classes are initialized. + return; + } + CompileBroker::compile_method(selected_method, InvocationEntryBci, + CompilationPolicy::policy()->initial_compile_level(), + methodHandle(), 0, CompileTask::Reason_MustBeCompiled, CHECK); + } +} + +// Returns true if m is allowed to be compiled +bool CompilationPolicy::can_be_compiled(const methodHandle& m, int comp_level) { + // allow any levels for WhiteBox + assert(WhiteBoxAPI || comp_level == CompLevel_all || is_compile(comp_level), "illegal compilation level"); + + if (m->is_abstract()) return false; + if (DontCompileHugeMethods && m->code_size() > HugeMethodLimit) return false; + + // Math intrinsics should never be compiled as this can lead to + // monotonicity problems because the interpreter will prefer the + // compiled code to the intrinsic version. This can't happen in + // production because the invocation counter can't be incremented + // but we shouldn't expose the system to this problem in testing + // modes. + if (!AbstractInterpreter::can_be_compiled(m)) { + return false; + } + if (comp_level == CompLevel_all) { + if (TieredCompilation) { + // enough to be compilable at any level for tiered + return !m->is_not_compilable(CompLevel_simple) || !m->is_not_compilable(CompLevel_full_optimization); + } else { + // must be compilable at available level for non-tiered + return !m->is_not_compilable(CompLevel_highest_tier); + } + } else if (is_compile(comp_level)) { + return !m->is_not_compilable(comp_level); + } + return false; +} + +// Returns true if m is allowed to be osr compiled +bool CompilationPolicy::can_be_osr_compiled(const methodHandle& m, int comp_level) { + bool result = false; + if (comp_level == CompLevel_all) { + if (TieredCompilation) { + // enough to be osr compilable at any level for tiered + result = !m->is_not_osr_compilable(CompLevel_simple) || !m->is_not_osr_compilable(CompLevel_full_optimization); + } else { + // must be osr compilable at available level for non-tiered + result = !m->is_not_osr_compilable(CompLevel_highest_tier); + } + } else if (is_compile(comp_level)) { + result = !m->is_not_osr_compilable(comp_level); + } + return (result && can_be_compiled(m, comp_level)); +} + +bool CompilationPolicy::is_compilation_enabled() { + // NOTE: CompileBroker::should_compile_new_jobs() checks for UseCompiler + return CompileBroker::should_compile_new_jobs(); +} + +CompileTask* CompilationPolicy::select_task_helper(CompileQueue* compile_queue) { + // Remove unloaded methods from the queue + for (CompileTask* task = compile_queue->first(); task != NULL; ) { + CompileTask* next = task->next(); + if (task->is_unloaded()) { + compile_queue->remove_and_mark_stale(task); + } + task = next; + } +#if INCLUDE_JVMCI + if (UseJVMCICompiler && !BackgroundCompilation) { + /* + * In blocking compilation mode, the CompileBroker will make + * compilations submitted by a JVMCI compiler thread non-blocking. These + * compilations should be scheduled after all blocking compilations + * to service non-compiler related compilations sooner and reduce the + * chance of such compilations timing out. + */ + for (CompileTask* task = compile_queue->first(); task != NULL; task = task->next()) { + if (task->is_blocking()) { + return task; + } + } + } +#endif + return compile_queue->first(); +} + +#ifndef PRODUCT +void SimpleCompPolicy::trace_osr_completion(nmethod* osr_nm) { + if (TraceOnStackReplacement) { + if (osr_nm == NULL) tty->print_cr("compilation failed"); + else tty->print_cr("nmethod " INTPTR_FORMAT, p2i(osr_nm)); + } +} +#endif // !PRODUCT + +void SimpleCompPolicy::initialize() { + // Setup the compiler thread numbers + if (CICompilerCountPerCPU) { + // Example: if CICompilerCountPerCPU is true, then we get + // max(log2(8)-1,1) = 2 compiler threads on an 8-way machine. + // May help big-app startup time. + _compiler_count = MAX2(log2_int(os::active_processor_count())-1,1); + // Make sure there is enough space in the code cache to hold all the compiler buffers + size_t buffer_size = 1; +#ifdef COMPILER1 + buffer_size = is_client_compilation_mode_vm() ? Compiler::code_buffer_size() : buffer_size; +#endif +#ifdef COMPILER2 + buffer_size = is_server_compilation_mode_vm() ? C2Compiler::initial_code_buffer_size() : buffer_size; +#endif + int max_count = (ReservedCodeCacheSize - (CodeCacheMinimumUseSpace DEBUG_ONLY(* 3))) / (int)buffer_size; + if (_compiler_count > max_count) { + // Lower the compiler count such that all buffers fit into the code cache + _compiler_count = MAX2(max_count, 1); + } + FLAG_SET_ERGO(CICompilerCount, _compiler_count); + } else { + _compiler_count = CICompilerCount; + } +} + +// Note: this policy is used ONLY if TieredCompilation is off. +// compiler_count() behaves the following way: +// - with TIERED build (with both COMPILER1 and COMPILER2 defined) it should return +// zero for the c1 compilation levels in server compilation mode runs +// and c2 compilation levels in client compilation mode runs. +// - with COMPILER2 not defined it should return zero for c2 compilation levels. +// - with COMPILER1 not defined it should return zero for c1 compilation levels. +// - if neither is defined - always return zero. +int SimpleCompPolicy::compiler_count(CompLevel comp_level) { + assert(!TieredCompilation, "This policy should not be used with TieredCompilation"); + if (COMPILER2_PRESENT(is_server_compilation_mode_vm() && is_c2_compile(comp_level) ||) + is_client_compilation_mode_vm() && is_c1_compile(comp_level)) { + return _compiler_count; + } + return 0; +} + +void SimpleCompPolicy::reset_counter_for_invocation_event(const methodHandle& m) { + // Make sure invocation and backedge counter doesn't overflow again right away + // as would be the case for native methods. + + // BUT also make sure the method doesn't look like it was never executed. + // Set carry bit and reduce counter's value to min(count, CompileThreshold/2). + MethodCounters* mcs = m->method_counters(); + assert(mcs != NULL, "MethodCounters cannot be NULL for profiling"); + mcs->invocation_counter()->set_carry(); + mcs->backedge_counter()->set_carry(); + + assert(!m->was_never_executed(), "don't reset to 0 -- could be mistaken for never-executed"); +} + +void SimpleCompPolicy::reset_counter_for_back_branch_event(const methodHandle& m) { + // Delay next back-branch event but pump up invocation counter to trigger + // whole method compilation. + MethodCounters* mcs = m->method_counters(); + assert(mcs != NULL, "MethodCounters cannot be NULL for profiling"); + InvocationCounter* i = mcs->invocation_counter(); + InvocationCounter* b = mcs->backedge_counter(); + + // Don't set invocation_counter's value too low otherwise the method will + // look like immature (ic < ~5300) which prevents the inlining based on + // the type profiling. + i->set(i->state(), CompileThreshold); + // Don't reset counter too low - it is used to check if OSR method is ready. + b->set(b->state(), CompileThreshold / 2); +} + +// +// CounterDecay +// +// Iterates through invocation counters and decrements them. This +// is done at each safepoint. +// +class CounterDecay : public AllStatic { + static jlong _last_timestamp; + static void do_method(Method* m) { + MethodCounters* mcs = m->method_counters(); + if (mcs != NULL) { + mcs->invocation_counter()->decay(); + } + } +public: + static void decay(); + static bool is_decay_needed() { + return (os::javaTimeMillis() - _last_timestamp) > CounterDecayMinIntervalLength; + } +}; + +jlong CounterDecay::_last_timestamp = 0; + +void CounterDecay::decay() { + _last_timestamp = os::javaTimeMillis(); + + // This operation is going to be performed only at the end of a safepoint + // and hence GC's will not be going on, all Java mutators are suspended + // at this point and hence SystemDictionary_lock is also not needed. + assert(SafepointSynchronize::is_at_safepoint(), "can only be executed at a safepoint"); + size_t nclasses = ClassLoaderDataGraph::num_instance_classes(); + size_t classes_per_tick = nclasses * (CounterDecayMinIntervalLength * 1e-3 / + CounterHalfLifeTime); + for (size_t i = 0; i < classes_per_tick; i++) { + InstanceKlass* k = ClassLoaderDataGraph::try_get_next_class(); + if (k != NULL) { + k->methods_do(do_method); + } + } +} + +// Called at the end of the safepoint +void SimpleCompPolicy::do_safepoint_work() { + if(UseCounterDecay && CounterDecay::is_decay_needed()) { + CounterDecay::decay(); + } +} + +void SimpleCompPolicy::reprofile(ScopeDesc* trap_scope, bool is_osr) { + ScopeDesc* sd = trap_scope; + MethodCounters* mcs; + InvocationCounter* c; + for (; !sd->is_top(); sd = sd->sender()) { + mcs = sd->method()->method_counters(); + if (mcs != NULL) { + // Reset ICs of inlined methods, since they can trigger compilations also. + mcs->invocation_counter()->reset(); + } + } + mcs = sd->method()->method_counters(); + if (mcs != NULL) { + c = mcs->invocation_counter(); + if (is_osr) { + // It was an OSR method, so bump the count higher. + c->set(c->state(), CompileThreshold); + } else { + c->reset(); + } + mcs->backedge_counter()->reset(); + } +} + +// This method can be called by any component of the runtime to notify the policy +// that it's recommended to delay the compilation of this method. +void SimpleCompPolicy::delay_compilation(Method* method) { + MethodCounters* mcs = method->method_counters(); + if (mcs != NULL) { + mcs->invocation_counter()->decay(); + mcs->backedge_counter()->decay(); + } +} + +void SimpleCompPolicy::disable_compilation(Method* method) { + MethodCounters* mcs = method->method_counters(); + if (mcs != NULL) { + mcs->invocation_counter()->set_state(InvocationCounter::wait_for_nothing); + mcs->backedge_counter()->set_state(InvocationCounter::wait_for_nothing); + } +} + +CompileTask* SimpleCompPolicy::select_task(CompileQueue* compile_queue) { + return select_task_helper(compile_queue); +} + +bool SimpleCompPolicy::is_mature(Method* method) { + MethodData* mdo = method->method_data(); + assert(mdo != NULL, "Should be"); + uint current = mdo->mileage_of(method); + uint initial = mdo->creation_mileage(); + if (current < initial) + return true; // some sort of overflow + uint target; + if (ProfileMaturityPercentage <= 0) + target = (uint) -ProfileMaturityPercentage; // absolute value + else + target = (uint)( (ProfileMaturityPercentage * CompileThreshold) / 100 ); + return (current >= initial + target); +} + +nmethod* SimpleCompPolicy::event(const methodHandle& method, const methodHandle& inlinee, int branch_bci, + int bci, CompLevel comp_level, CompiledMethod* nm, JavaThread* thread) { + assert(comp_level == CompLevel_none, "This should be only called from the interpreter"); + NOT_PRODUCT(trace_frequency_counter_overflow(method, branch_bci, bci)); + if (JvmtiExport::can_post_interpreter_events() && thread->is_interp_only_mode()) { + // If certain JVMTI events (e.g. frame pop event) are requested then the + // thread is forced to remain in interpreted code. This is + // implemented partly by a check in the run_compiled_code + // section of the interpreter whether we should skip running + // compiled code, and partly by skipping OSR compiles for + // interpreted-only threads. + if (bci != InvocationEntryBci) { + reset_counter_for_back_branch_event(method); + return NULL; + } + } + if (ReplayCompiles) { + // Don't trigger other compiles in testing mode + if (bci == InvocationEntryBci) { + reset_counter_for_invocation_event(method); + } else { + reset_counter_for_back_branch_event(method); + } + return NULL; + } + + if (bci == InvocationEntryBci) { + // when code cache is full, compilation gets switched off, UseCompiler + // is set to false + if (!method->has_compiled_code() && UseCompiler) { + method_invocation_event(method, thread); + } else { + // Force counter overflow on method entry, even if no compilation + // happened. (The method_invocation_event call does this also.) + reset_counter_for_invocation_event(method); + } + // compilation at an invocation overflow no longer goes and retries test for + // compiled method. We always run the loser of the race as interpreted. + // so return NULL + return NULL; + } else { + // counter overflow in a loop => try to do on-stack-replacement + nmethod* osr_nm = method->lookup_osr_nmethod_for(bci, CompLevel_highest_tier, true); + NOT_PRODUCT(trace_osr_request(method, osr_nm, bci)); + // when code cache is full, we should not compile any more... + if (osr_nm == NULL && UseCompiler) { + method_back_branch_event(method, bci, thread); + osr_nm = method->lookup_osr_nmethod_for(bci, CompLevel_highest_tier, true); + } + if (osr_nm == NULL) { + reset_counter_for_back_branch_event(method); + return NULL; + } + return osr_nm; + } + return NULL; +} + +#ifndef PRODUCT +void SimpleCompPolicy::trace_frequency_counter_overflow(const methodHandle& m, int branch_bci, int bci) { + if (TraceInvocationCounterOverflow) { + MethodCounters* mcs = m->method_counters(); + assert(mcs != NULL, "MethodCounters cannot be NULL for profiling"); + InvocationCounter* ic = mcs->invocation_counter(); + InvocationCounter* bc = mcs->backedge_counter(); + ResourceMark rm; + if (bci == InvocationEntryBci) { + tty->print("comp-policy cntr ovfl @ %d in entry of ", bci); + } else { + tty->print("comp-policy cntr ovfl @ %d in loop of ", bci); + } + m->print_value(); + tty->cr(); + ic->print(); + bc->print(); + if (ProfileInterpreter) { + if (bci != InvocationEntryBci) { + MethodData* mdo = m->method_data(); + if (mdo != NULL) { + ProfileData *pd = mdo->bci_to_data(branch_bci); + if (pd == NULL) { + tty->print_cr("back branch count = N/A (missing ProfileData)"); + } else { + tty->print_cr("back branch count = %d", pd->as_JumpData()->taken()); + } + } + } + } + } +} + +void SimpleCompPolicy::trace_osr_request(const methodHandle& method, nmethod* osr, int bci) { + if (TraceOnStackReplacement) { + ResourceMark rm; + tty->print(osr != NULL ? "Reused OSR entry for " : "Requesting OSR entry for "); + method->print_short_name(tty); + tty->print_cr(" at bci %d", bci); + } +} +#endif // !PRODUCT + +void SimpleCompPolicy::method_invocation_event(const methodHandle& m, JavaThread* thread) { + const int comp_level = CompLevel_highest_tier; + const int hot_count = m->invocation_count(); + reset_counter_for_invocation_event(m); + + if (is_compilation_enabled() && can_be_compiled(m, comp_level)) { + CompiledMethod* nm = m->code(); + if (nm == NULL ) { + CompileBroker::compile_method(m, InvocationEntryBci, comp_level, m, hot_count, CompileTask::Reason_InvocationCount, thread); + } + } +} + +void SimpleCompPolicy::method_back_branch_event(const methodHandle& m, int bci, JavaThread* thread) { + const int comp_level = CompLevel_highest_tier; + const int hot_count = m->backedge_count(); + + if (is_compilation_enabled() && can_be_osr_compiled(m, comp_level)) { + CompileBroker::compile_method(m, bci, comp_level, m, hot_count, CompileTask::Reason_BackedgeCount, thread); + NOT_PRODUCT(trace_osr_completion(m->lookup_osr_nmethod_for(bci, comp_level, true));) + } +}
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/src/hotspot/share/compiler/compilationPolicy.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -0,0 +1,113 @@ +/* + * Copyright (c) 2000, 2019, 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. + * + */ + +#ifndef SHARE_COMPILER_COMPILATIONPOLICY_HPP +#define SHARE_COMPILER_COMPILATIONPOLICY_HPP + +#include "code/nmethod.hpp" +#include "compiler/compileBroker.hpp" +#include "memory/allocation.hpp" +#include "runtime/vmOperations.hpp" +#include "utilities/growableArray.hpp" + +// The CompilationPolicy selects which method (if any) should be compiled. +// It also decides which methods must always be compiled (i.e., are never +// interpreted). +class CompileTask; +class CompileQueue; + +class CompilationPolicy : public CHeapObj<mtCompiler> { + static CompilationPolicy* _policy; + + // m must be compiled before executing it + static bool must_be_compiled(const methodHandle& m, int comp_level = CompLevel_all); + +public: + // If m must_be_compiled then request a compilation from the CompileBroker. + // This supports the -Xcomp option. + static void compile_if_required(const methodHandle& m, TRAPS); + + // m is allowed to be compiled + static bool can_be_compiled(const methodHandle& m, int comp_level = CompLevel_all); + // m is allowed to be osr compiled + static bool can_be_osr_compiled(const methodHandle& m, int comp_level = CompLevel_all); + static bool is_compilation_enabled(); + static void set_policy(CompilationPolicy* policy) { _policy = policy; } + static CompilationPolicy* policy() { return _policy; } + + static CompileTask* select_task_helper(CompileQueue* compile_queue); + + // Return initial compile level that is used with Xcomp + virtual CompLevel initial_compile_level() = 0; + virtual int compiler_count(CompLevel comp_level) = 0; + // main notification entry, return a pointer to an nmethod if the OSR is required, + // returns NULL otherwise. + virtual nmethod* event(const methodHandle& method, const methodHandle& inlinee, int branch_bci, int bci, CompLevel comp_level, CompiledMethod* nm, JavaThread* thread) = 0; + // safepoint() is called at the end of the safepoint + virtual void do_safepoint_work() = 0; + // reprofile request + virtual void reprofile(ScopeDesc* trap_scope, bool is_osr) = 0; + // delay_compilation(method) can be called by any component of the runtime to notify the policy + // that it's recommended to delay the compilation of this method. + virtual void delay_compilation(Method* method) = 0; + // disable_compilation() is called whenever the runtime decides to disable compilation of the + // specified method. + virtual void disable_compilation(Method* method) = 0; + // Select task is called by CompileBroker. The queue is guaranteed to have at least one + // element and is locked. The function should select one and return it. + virtual CompileTask* select_task(CompileQueue* compile_queue) = 0; + // Tell the runtime if we think a given method is adequately profiled. + virtual bool is_mature(Method* method) = 0; + // Do policy initialization + virtual void initialize() = 0; + virtual bool should_not_inline(ciEnv* env, ciMethod* method) { return false; } +}; + +// A simple compilation policy. +class SimpleCompPolicy : public CompilationPolicy { + int _compiler_count; + private: + static void trace_frequency_counter_overflow(const methodHandle& m, int branch_bci, int bci); + static void trace_osr_request(const methodHandle& method, nmethod* osr, int bci); + static void trace_osr_completion(nmethod* osr_nm); + void reset_counter_for_invocation_event(const methodHandle& method); + void reset_counter_for_back_branch_event(const methodHandle& method); + void method_invocation_event(const methodHandle& m, JavaThread* thread); + void method_back_branch_event(const methodHandle& m, int bci, JavaThread* thread); + public: + SimpleCompPolicy() : _compiler_count(0) { } + virtual CompLevel initial_compile_level() { return CompLevel_highest_tier; } + virtual int compiler_count(CompLevel comp_level); + virtual void do_safepoint_work(); + virtual void reprofile(ScopeDesc* trap_scope, bool is_osr); + virtual void delay_compilation(Method* method); + virtual void disable_compilation(Method* method); + virtual bool is_mature(Method* method); + virtual void initialize(); + virtual CompileTask* select_task(CompileQueue* compile_queue); + virtual nmethod* event(const methodHandle& method, const methodHandle& inlinee, int branch_bci, int bci, CompLevel comp_level, CompiledMethod* nm, JavaThread* thread); +}; + + +#endif // SHARE_COMPILER_COMPILATIONPOLICY_HPP
--- a/src/hotspot/share/compiler/compileBroker.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/compiler/compileBroker.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -30,6 +30,7 @@ #include "code/codeCache.hpp" #include "code/codeHeapState.hpp" #include "code/dependencyContext.hpp" +#include "compiler/compilationPolicy.hpp" #include "compiler/compileBroker.hpp" #include "compiler/compileLog.hpp" #include "compiler/compilerOracle.hpp" @@ -48,7 +49,6 @@ #include "prims/whitebox.hpp" #include "runtime/arguments.hpp" #include "runtime/atomic.hpp" -#include "runtime/compilationPolicy.hpp" #include "runtime/handles.inline.hpp" #include "runtime/init.hpp" #include "runtime/interfaceSupport.inline.hpp"
--- a/src/hotspot/share/compiler/compilerDirectives.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/compiler/compilerDirectives.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -66,8 +66,7 @@ cflags(VectorizeDebug, uintx, 0, VectorizeDebug) \ cflags(CloneMapDebug, bool, false, CloneMapDebug) \ cflags(IGVPrintLevel, intx, PrintIdealGraphLevel, IGVPrintLevel) \ - cflags(MaxNodeLimit, intx, MaxNodeLimit, MaxNodeLimit) \ -ZGC_ONLY(cflags(ZTraceLoadBarriers, bool, false, ZTraceLoadBarriers)) + cflags(MaxNodeLimit, intx, MaxNodeLimit, MaxNodeLimit) #else #define compilerdirectives_c2_flags(cflags) #endif
--- a/src/hotspot/share/compiler/oopMap.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/compiler/oopMap.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -48,29 +48,25 @@ // OopMapStream -OopMapStream::OopMapStream(OopMap* oop_map, int oop_types_mask) { +OopMapStream::OopMapStream(OopMap* oop_map) { _stream = new CompressedReadStream(oop_map->write_stream()->buffer()); - _mask = oop_types_mask; _size = oop_map->omv_count(); _position = 0; _valid_omv = false; } -OopMapStream::OopMapStream(const ImmutableOopMap* oop_map, int oop_types_mask) { +OopMapStream::OopMapStream(const ImmutableOopMap* oop_map) { _stream = new CompressedReadStream(oop_map->data_addr()); - _mask = oop_types_mask; _size = oop_map->count(); _position = 0; _valid_omv = false; } void OopMapStream::find_next() { - while(_position++ < _size) { + if (_position++ < _size) { _omv.read_from(_stream); - if(((int)_omv.type() & _mask) > 0) { - _valid_omv = true; - return; - } + _valid_omv = true; + return; } _valid_omv = false; } @@ -140,16 +136,7 @@ assert( _locs_used[reg->value()] == OopMapValue::unused_value, "cannot insert twice" ); debug_only( _locs_used[reg->value()] = x; ) - OopMapValue o(reg, x); - - if(x == OopMapValue::callee_saved_value) { - // This can never be a stack location, so we don't need to transform it. - assert(optional->is_reg(), "Trying to callee save a stack location"); - o.set_content_reg(optional); - } else if(x == OopMapValue::derived_oop_value) { - o.set_content_reg(optional); - } - + OopMapValue o(reg, x, optional); o.write_on(write_stream()); increment_count(); } @@ -160,11 +147,6 @@ } -void OopMap::set_value(VMReg reg) { - // At this time, we don't need value entries in our OopMap. -} - - void OopMap::set_narrowoop(VMReg reg) { set_xxx(reg, OopMapValue::narrowoop_value, VMRegImpl::Bad()); } @@ -328,7 +310,7 @@ // changed before derived pointer offset has been collected) OopMapValue omv; { - OopMapStream oms(map,OopMapValue::derived_oop_value); + OopMapStream oms(map); if (!oms.is_done()) { #ifndef TIERED COMPILER1_PRESENT(ShouldNotReachHere();) @@ -340,27 +322,28 @@ #endif // !TIERED do { omv = oms.current(); - oop* loc = fr->oopmapreg_to_location(omv.reg(),reg_map); - guarantee(loc != NULL, "missing saved register"); - oop *derived_loc = loc; - oop *base_loc = fr->oopmapreg_to_location(omv.content_reg(), reg_map); - // Ignore NULL oops and decoded NULL narrow oops which - // equal to CompressedOops::base() when a narrow oop - // implicit null check is used in compiled code. - // The narrow_oop_base could be NULL or be the address - // of the page below heap depending on compressed oops mode. - if (base_loc != NULL && *base_loc != NULL && !CompressedOops::is_base(*base_loc)) { - derived_oop_fn(base_loc, derived_loc); + if (omv.type() == OopMapValue::derived_oop_value) { + oop* loc = fr->oopmapreg_to_location(omv.reg(),reg_map); + guarantee(loc != NULL, "missing saved register"); + oop *derived_loc = loc; + oop *base_loc = fr->oopmapreg_to_location(omv.content_reg(), reg_map); + // Ignore NULL oops and decoded NULL narrow oops which + // equal to CompressedOops::base() when a narrow oop + // implicit null check is used in compiled code. + // The narrow_oop_base could be NULL or be the address + // of the page below heap depending on compressed oops mode. + if (base_loc != NULL && *base_loc != NULL && !CompressedOops::is_base(*base_loc)) { + derived_oop_fn(base_loc, derived_loc); + } } oms.next(); } while (!oms.is_done()); } } - // We want coop and oop oop_types - int mask = OopMapValue::oop_value | OopMapValue::narrowoop_value; { - for (OopMapStream oms(map,mask); !oms.is_done(); oms.next()) { + // We want coop and oop oop_types + for (OopMapStream oms(map); !oms.is_done(); oms.next()) { omv = oms.current(); oop* loc = fr->oopmapreg_to_location(omv.reg(),reg_map); // It should be an error if no location can be found for a @@ -436,12 +419,14 @@ assert(map != NULL, "no ptr map found"); DEBUG_ONLY(int nof_callee = 0;) - for (OopMapStream oms(map, OopMapValue::callee_saved_value); !oms.is_done(); oms.next()) { + for (OopMapStream oms(map); !oms.is_done(); oms.next()) { OopMapValue omv = oms.current(); - VMReg reg = omv.content_reg(); - oop* loc = fr->oopmapreg_to_location(omv.reg(), reg_map); - reg_map->set_location(reg, (address) loc); - DEBUG_ONLY(nof_callee++;) + if (omv.type() == OopMapValue::callee_saved_value) { + VMReg reg = omv.content_reg(); + oop* loc = fr->oopmapreg_to_location(omv.reg(), reg_map); + reg_map->set_location(reg, (address) loc); + DEBUG_ONLY(nof_callee++;) + } } // Check that runtime stubs save all callee-saved registers @@ -452,25 +437,6 @@ #endif // COMPILER2 } -//============================================================================= -// Non-Product code - -#ifndef PRODUCT - -bool ImmutableOopMap::has_derived_pointer() const { -#if !defined(TIERED) && !INCLUDE_JVMCI - COMPILER1_PRESENT(return false); -#endif // !TIERED -#if COMPILER2_OR_JVMCI - OopMapStream oms(this,OopMapValue::derived_oop_value); - return oms.is_done(); -#else - return false; -#endif // COMPILER2_OR_JVMCI -} - -#endif //PRODUCT - // Printing code is present in product build for -XX:+PrintAssembly. static
--- a/src/hotspot/share/compiler/oopMap.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/compiler/oopMap.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -53,7 +53,7 @@ public: // Constants - enum { type_bits = 4, + enum { type_bits = 2, register_bits = BitsPerShort - type_bits }; enum { type_shift = 0, @@ -64,19 +64,41 @@ register_mask = right_n_bits(register_bits), register_mask_in_place = register_mask << register_shift }; - enum oop_types { // must fit in type_bits - unused_value =0, // powers of 2, for masking OopMapStream - oop_value = 1, - narrowoop_value = 2, - callee_saved_value = 4, - derived_oop_value= 8 }; + enum oop_types { + oop_value, + narrowoop_value, + callee_saved_value, + derived_oop_value, + unused_value = -1 // Only used as a sentinel value + }; // Constructors OopMapValue () { set_value(0); set_content_reg(VMRegImpl::Bad()); } - OopMapValue (VMReg reg, oop_types t) { set_reg_type(reg, t); set_content_reg(VMRegImpl::Bad()); } - OopMapValue (VMReg reg, oop_types t, VMReg reg2) { set_reg_type(reg, t); set_content_reg(reg2); } - OopMapValue (CompressedReadStream* stream) { read_from(stream); } + OopMapValue (VMReg reg, oop_types t, VMReg reg2) { + set_reg_type(reg, t); + set_content_reg(reg2); + } + + private: + void set_reg_type(VMReg p, oop_types t) { + set_value((p->value() << register_shift) | t); + assert(reg() == p, "sanity check" ); + assert(type() == t, "sanity check" ); + } + void set_content_reg(VMReg r) { + if (is_callee_saved()) { + // This can never be a stack location, so we don't need to transform it. + assert(r->is_reg(), "Trying to callee save a stack location"); + } else if (is_derived_oop()) { + assert (r->is_valid(), "must have a valid VMReg"); + } else { + assert (!r->is_valid(), "valid VMReg not allowed"); + } + _content_reg = r->value(); + } + + public: // Archiving void write_on(CompressedWriteStream* stream) { stream->write_int(value()); @@ -94,15 +116,10 @@ // Querying bool is_oop() { return mask_bits(value(), type_mask_in_place) == oop_value; } - bool is_narrowoop() { return mask_bits(value(), type_mask_in_place) == narrowoop_value; } + bool is_narrowoop() { return mask_bits(value(), type_mask_in_place) == narrowoop_value; } bool is_callee_saved() { return mask_bits(value(), type_mask_in_place) == callee_saved_value; } bool is_derived_oop() { return mask_bits(value(), type_mask_in_place) == derived_oop_value; } - void set_oop() { set_value((value() & register_mask_in_place) | oop_value); } - void set_narrowoop() { set_value((value() & register_mask_in_place) | narrowoop_value); } - void set_callee_saved() { set_value((value() & register_mask_in_place) | callee_saved_value); } - void set_derived_oop() { set_value((value() & register_mask_in_place) | derived_oop_value); } - VMReg reg() const { return VMRegImpl::as_VMReg(mask_bits(value(), register_mask_in_place) >> register_shift); } oop_types type() const { return (oop_types)mask_bits(value(), type_mask_in_place); } @@ -110,15 +127,7 @@ return (p->value() == (p->value() & register_mask)); } - void set_reg_type(VMReg p, oop_types t) { - set_value((p->value() << register_shift) | t); - assert(reg() == p, "sanity check" ); - assert(type() == t, "sanity check" ); - } - - VMReg content_reg() const { return VMRegImpl::as_VMReg(_content_reg, true); } - void set_content_reg(VMReg r) { _content_reg = r->value(); } // Physical location queries bool is_register_loc() { return reg()->is_reg(); } @@ -156,6 +165,8 @@ enum DeepCopyToken { _deep_copy_token }; OopMap(DeepCopyToken, OopMap* source); // used only by deep_copy + void set_xxx(VMReg reg, OopMapValue::oop_types x, VMReg optional); + public: OopMap(int frame_size, int arg_count); @@ -173,19 +184,14 @@ // frame_size units are stack-slots (4 bytes) NOT intptr_t; we can name odd // slots to hold 4-byte values like ints and floats in the LP64 build. void set_oop ( VMReg local); - void set_value( VMReg local); void set_narrowoop(VMReg local); - void set_dead ( VMReg local); void set_callee_saved( VMReg local, VMReg caller_machine_register ); void set_derived_oop ( VMReg local, VMReg derived_from_local_register ); - void set_xxx(VMReg reg, OopMapValue::oop_types x, VMReg optional); int heap_size() const; void copy_data_to(address addr) const; OopMap* deep_copy(); - bool has_derived_pointer() const PRODUCT_RETURN0; - bool legal_vm_reg_name(VMReg local) { return OopMapValue::legal_vm_reg_name(local); } @@ -269,7 +275,6 @@ public: ImmutableOopMap(const OopMap* oopmap); - bool has_derived_pointer() const PRODUCT_RETURN0; int count() const { return _count; } #ifdef ASSERT int nr_of_bytes() const; // this is an expensive operation, only used in debug builds @@ -334,7 +339,6 @@ class OopMapStream : public StackObj { private: CompressedReadStream* _stream; - int _mask; int _size; int _position; bool _valid_omv; @@ -342,8 +346,8 @@ void find_next(); public: - OopMapStream(OopMap* oop_map, int oop_types_mask = OopMapValue::type_mask_in_place); - OopMapStream(const ImmutableOopMap* oop_map, int oop_types_mask = OopMapValue::type_mask_in_place); + OopMapStream(OopMap* oop_map); + OopMapStream(const ImmutableOopMap* oop_map); bool is_done() { if(!_valid_omv) { find_next(); } return !_valid_omv; } void next() { find_next(); } OopMapValue current() { return _omv; }
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/src/hotspot/share/compiler/tieredThresholdPolicy.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -0,0 +1,1005 @@ +/* + * Copyright (c) 2010, 2019, 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. + * + */ + +#include "precompiled.hpp" +#include "compiler/compileBroker.hpp" +#include "compiler/compilerOracle.hpp" +#include "compiler/tieredThresholdPolicy.hpp" +#include "memory/resourceArea.hpp" +#include "runtime/arguments.hpp" +#include "runtime/handles.inline.hpp" +#include "runtime/safepoint.hpp" +#include "runtime/safepointVerifiers.hpp" +#include "code/scopeDesc.hpp" +#include "oops/method.inline.hpp" +#if INCLUDE_JVMCI +#include "jvmci/jvmci.hpp" +#endif + +#ifdef TIERED + +#include "c1/c1_Compiler.hpp" +#include "opto/c2compiler.hpp" + +template<CompLevel level> +bool TieredThresholdPolicy::call_predicate_helper(int i, int b, double scale, Method* method) { + double threshold_scaling; + if (CompilerOracle::has_option_value(method, "CompileThresholdScaling", threshold_scaling)) { + scale *= threshold_scaling; + } + switch(level) { + case CompLevel_aot: + return (i >= Tier3AOTInvocationThreshold * scale) || + (i >= Tier3AOTMinInvocationThreshold * scale && i + b >= Tier3AOTCompileThreshold * scale); + case CompLevel_none: + case CompLevel_limited_profile: + return (i >= Tier3InvocationThreshold * scale) || + (i >= Tier3MinInvocationThreshold * scale && i + b >= Tier3CompileThreshold * scale); + case CompLevel_full_profile: + return (i >= Tier4InvocationThreshold * scale) || + (i >= Tier4MinInvocationThreshold * scale && i + b >= Tier4CompileThreshold * scale); + } + return true; +} + +template<CompLevel level> +bool TieredThresholdPolicy::loop_predicate_helper(int i, int b, double scale, Method* method) { + double threshold_scaling; + if (CompilerOracle::has_option_value(method, "CompileThresholdScaling", threshold_scaling)) { + scale *= threshold_scaling; + } + switch(level) { + case CompLevel_aot: + return b >= Tier3AOTBackEdgeThreshold * scale; + case CompLevel_none: + case CompLevel_limited_profile: + return b >= Tier3BackEdgeThreshold * scale; + case CompLevel_full_profile: + return b >= Tier4BackEdgeThreshold * scale; + } + return true; +} + +// Simple methods are as good being compiled with C1 as C2. +// Determine if a given method is such a case. +bool TieredThresholdPolicy::is_trivial(Method* method) { + if (method->is_accessor() || + method->is_constant_getter()) { + return true; + } + return false; +} + +bool TieredThresholdPolicy::should_compile_at_level_simple(Method* method) { + if (TieredThresholdPolicy::is_trivial(method)) { + return true; + } +#if INCLUDE_JVMCI + if (UseJVMCICompiler) { + AbstractCompiler* comp = CompileBroker::compiler(CompLevel_full_optimization); + if (comp != NULL && comp->is_jvmci() && ((JVMCICompiler*) comp)->force_comp_at_level_simple(method)) { + return true; + } + } +#endif + return false; +} + +CompLevel TieredThresholdPolicy::comp_level(Method* method) { + CompiledMethod *nm = method->code(); + if (nm != NULL && nm->is_in_use()) { + return (CompLevel)nm->comp_level(); + } + return CompLevel_none; +} + +void TieredThresholdPolicy::print_counters(const char* prefix, const methodHandle& mh) { + int invocation_count = mh->invocation_count(); + int backedge_count = mh->backedge_count(); + MethodData* mdh = mh->method_data(); + int mdo_invocations = 0, mdo_backedges = 0; + int mdo_invocations_start = 0, mdo_backedges_start = 0; + if (mdh != NULL) { + mdo_invocations = mdh->invocation_count(); + mdo_backedges = mdh->backedge_count(); + mdo_invocations_start = mdh->invocation_count_start(); + mdo_backedges_start = mdh->backedge_count_start(); + } + tty->print(" %stotal=%d,%d %smdo=%d(%d),%d(%d)", prefix, + invocation_count, backedge_count, prefix, + mdo_invocations, mdo_invocations_start, + mdo_backedges, mdo_backedges_start); + tty->print(" %smax levels=%d,%d", prefix, + mh->highest_comp_level(), mh->highest_osr_comp_level()); +} + +// Print an event. +void TieredThresholdPolicy::print_event(EventType type, const methodHandle& mh, const methodHandle& imh, + int bci, CompLevel level) { + bool inlinee_event = mh() != imh(); + + ttyLocker tty_lock; + tty->print("%lf: [", os::elapsedTime()); + + switch(type) { + case CALL: + tty->print("call"); + break; + case LOOP: + tty->print("loop"); + break; + case COMPILE: + tty->print("compile"); + break; + case REMOVE_FROM_QUEUE: + tty->print("remove-from-queue"); + break; + case UPDATE_IN_QUEUE: + tty->print("update-in-queue"); + break; + case REPROFILE: + tty->print("reprofile"); + break; + case MAKE_NOT_ENTRANT: + tty->print("make-not-entrant"); + break; + default: + tty->print("unknown"); + } + + tty->print(" level=%d ", level); + + ResourceMark rm; + char *method_name = mh->name_and_sig_as_C_string(); + tty->print("[%s", method_name); + if (inlinee_event) { + char *inlinee_name = imh->name_and_sig_as_C_string(); + tty->print(" [%s]] ", inlinee_name); + } + else tty->print("] "); + tty->print("@%d queues=%d,%d", bci, CompileBroker::queue_size(CompLevel_full_profile), + CompileBroker::queue_size(CompLevel_full_optimization)); + + print_specific(type, mh, imh, bci, level); + + if (type != COMPILE) { + print_counters("", mh); + if (inlinee_event) { + print_counters("inlinee ", imh); + } + tty->print(" compilable="); + bool need_comma = false; + if (!mh->is_not_compilable(CompLevel_full_profile)) { + tty->print("c1"); + need_comma = true; + } + if (!mh->is_not_osr_compilable(CompLevel_full_profile)) { + if (need_comma) tty->print(","); + tty->print("c1-osr"); + need_comma = true; + } + if (!mh->is_not_compilable(CompLevel_full_optimization)) { + if (need_comma) tty->print(","); + tty->print("c2"); + need_comma = true; + } + if (!mh->is_not_osr_compilable(CompLevel_full_optimization)) { + if (need_comma) tty->print(","); + tty->print("c2-osr"); + } + tty->print(" status="); + if (mh->queued_for_compilation()) { + tty->print("in-queue"); + } else tty->print("idle"); + } + tty->print_cr("]"); +} + +void TieredThresholdPolicy::initialize() { + int count = CICompilerCount; + bool c1_only = TieredStopAtLevel < CompLevel_full_optimization; +#ifdef _LP64 + // Turn on ergonomic compiler count selection + if (FLAG_IS_DEFAULT(CICompilerCountPerCPU) && FLAG_IS_DEFAULT(CICompilerCount)) { + FLAG_SET_DEFAULT(CICompilerCountPerCPU, true); + } + if (CICompilerCountPerCPU) { + // Simple log n seems to grow too slowly for tiered, try something faster: log n * log log n + int log_cpu = log2_int(os::active_processor_count()); + int loglog_cpu = log2_int(MAX2(log_cpu, 1)); + count = MAX2(log_cpu * loglog_cpu * 3 / 2, 2); + // Make sure there is enough space in the code cache to hold all the compiler buffers + size_t c1_size = Compiler::code_buffer_size(); + size_t c2_size = C2Compiler::initial_code_buffer_size(); + size_t buffer_size = c1_only ? c1_size : (c1_size/3 + 2*c2_size/3); + int max_count = (ReservedCodeCacheSize - (CodeCacheMinimumUseSpace DEBUG_ONLY(* 3))) / (int)buffer_size; + if (count > max_count) { + // Lower the compiler count such that all buffers fit into the code cache + count = MAX2(max_count, c1_only ? 1 : 2); + } + FLAG_SET_ERGO(CICompilerCount, count); + } +#else + // On 32-bit systems, the number of compiler threads is limited to 3. + // On these systems, the virtual address space available to the JVM + // is usually limited to 2-4 GB (the exact value depends on the platform). + // As the compilers (especially C2) can consume a large amount of + // memory, scaling the number of compiler threads with the number of + // available cores can result in the exhaustion of the address space + /// available to the VM and thus cause the VM to crash. + if (FLAG_IS_DEFAULT(CICompilerCount)) { + count = 3; + FLAG_SET_ERGO(CICompilerCount, count); + } +#endif + + if (c1_only) { + // No C2 compiler thread required + set_c1_count(count); + } else { + set_c1_count(MAX2(count / 3, 1)); + set_c2_count(MAX2(count - c1_count(), 1)); + } + assert(count == c1_count() + c2_count(), "inconsistent compiler thread count"); + + // Some inlining tuning +#ifdef X86 + if (FLAG_IS_DEFAULT(InlineSmallCode)) { + FLAG_SET_DEFAULT(InlineSmallCode, 2000); + } +#endif + +#if defined SPARC || defined AARCH64 + if (FLAG_IS_DEFAULT(InlineSmallCode)) { + FLAG_SET_DEFAULT(InlineSmallCode, 2500); + } +#endif + + set_increase_threshold_at_ratio(); + set_start_time(os::javaTimeMillis()); +} + +void TieredThresholdPolicy::set_carry_if_necessary(InvocationCounter *counter) { + if (!counter->carry() && counter->count() > InvocationCounter::count_limit / 2) { + counter->set_carry_flag(); + } +} + +// Set carry flags on the counters if necessary +void TieredThresholdPolicy::handle_counter_overflow(Method* method) { + MethodCounters *mcs = method->method_counters(); + if (mcs != NULL) { + set_carry_if_necessary(mcs->invocation_counter()); + set_carry_if_necessary(mcs->backedge_counter()); + } + MethodData* mdo = method->method_data(); + if (mdo != NULL) { + set_carry_if_necessary(mdo->invocation_counter()); + set_carry_if_necessary(mdo->backedge_counter()); + } +} + +// Called with the queue locked and with at least one element +CompileTask* TieredThresholdPolicy::select_task(CompileQueue* compile_queue) { + CompileTask *max_blocking_task = NULL; + CompileTask *max_task = NULL; + Method* max_method = NULL; + jlong t = os::javaTimeMillis(); + // Iterate through the queue and find a method with a maximum rate. + for (CompileTask* task = compile_queue->first(); task != NULL;) { + CompileTask* next_task = task->next(); + Method* method = task->method(); + // If a method was unloaded or has been stale for some time, remove it from the queue. + // Blocking tasks and tasks submitted from whitebox API don't become stale + if (task->is_unloaded() || (task->can_become_stale() && is_stale(t, TieredCompileTaskTimeout, method) && !is_old(method))) { + if (!task->is_unloaded()) { + if (PrintTieredEvents) { + print_event(REMOVE_FROM_QUEUE, method, method, task->osr_bci(), (CompLevel) task->comp_level()); + } + method->clear_queued_for_compilation(); + } + compile_queue->remove_and_mark_stale(task); + task = next_task; + continue; + } + update_rate(t, method); + if (max_task == NULL || compare_methods(method, max_method)) { + // Select a method with the highest rate + max_task = task; + max_method = method; + } + + if (task->is_blocking()) { + if (max_blocking_task == NULL || compare_methods(method, max_blocking_task->method())) { + max_blocking_task = task; + } + } + + task = next_task; + } + + if (max_blocking_task != NULL) { + // In blocking compilation mode, the CompileBroker will make + // compilations submitted by a JVMCI compiler thread non-blocking. These + // compilations should be scheduled after all blocking compilations + // to service non-compiler related compilations sooner and reduce the + // chance of such compilations timing out. + max_task = max_blocking_task; + max_method = max_task->method(); + } + + if (max_task != NULL && max_task->comp_level() == CompLevel_full_profile && + TieredStopAtLevel > CompLevel_full_profile && + max_method != NULL && is_method_profiled(max_method)) { + max_task->set_comp_level(CompLevel_limited_profile); + + if (CompileBroker::compilation_is_complete(max_method, max_task->osr_bci(), CompLevel_limited_profile)) { + if (PrintTieredEvents) { + print_event(REMOVE_FROM_QUEUE, max_method, max_method, max_task->osr_bci(), (CompLevel)max_task->comp_level()); + } + compile_queue->remove_and_mark_stale(max_task); + max_method->clear_queued_for_compilation(); + return NULL; + } + + if (PrintTieredEvents) { + print_event(UPDATE_IN_QUEUE, max_method, max_method, max_task->osr_bci(), (CompLevel)max_task->comp_level()); + } + } + + return max_task; +} + +void TieredThresholdPolicy::reprofile(ScopeDesc* trap_scope, bool is_osr) { + for (ScopeDesc* sd = trap_scope;; sd = sd->sender()) { + if (PrintTieredEvents) { + methodHandle mh(sd->method()); + print_event(REPROFILE, mh, mh, InvocationEntryBci, CompLevel_none); + } + MethodData* mdo = sd->method()->method_data(); + if (mdo != NULL) { + mdo->reset_start_counters(); + } + if (sd->is_top()) break; + } +} + +nmethod* TieredThresholdPolicy::event(const methodHandle& method, const methodHandle& inlinee, + int branch_bci, int bci, CompLevel comp_level, CompiledMethod* nm, JavaThread* thread) { + if (comp_level == CompLevel_none && + JvmtiExport::can_post_interpreter_events() && + thread->is_interp_only_mode()) { + return NULL; + } + if (ReplayCompiles) { + // Don't trigger other compiles in testing mode + return NULL; + } + + handle_counter_overflow(method()); + if (method() != inlinee()) { + handle_counter_overflow(inlinee()); + } + + if (PrintTieredEvents) { + print_event(bci == InvocationEntryBci ? CALL : LOOP, method, inlinee, bci, comp_level); + } + + if (bci == InvocationEntryBci) { + method_invocation_event(method, inlinee, comp_level, nm, thread); + } else { + // method == inlinee if the event originated in the main method + method_back_branch_event(method, inlinee, bci, comp_level, nm, thread); + // Check if event led to a higher level OSR compilation + CompLevel expected_comp_level = comp_level; + if (inlinee->is_not_osr_compilable(expected_comp_level)) { + // It's not possble to reach the expected level so fall back to simple. + expected_comp_level = CompLevel_simple; + } + nmethod* osr_nm = inlinee->lookup_osr_nmethod_for(bci, expected_comp_level, false); + assert(osr_nm == NULL || osr_nm->comp_level() >= expected_comp_level, "lookup_osr_nmethod_for is broken"); + if (osr_nm != NULL) { + // Perform OSR with new nmethod + return osr_nm; + } + } + return NULL; +} + +// Check if the method can be compiled, change level if necessary +void TieredThresholdPolicy::compile(const methodHandle& mh, int bci, CompLevel level, JavaThread* thread) { + assert(level <= TieredStopAtLevel, "Invalid compilation level"); + if (level == CompLevel_none) { + return; + } + if (level == CompLevel_aot) { + if (mh->has_aot_code()) { + if (PrintTieredEvents) { + print_event(COMPILE, mh, mh, bci, level); + } + MutexLocker ml(Compile_lock); + NoSafepointVerifier nsv; + if (mh->has_aot_code() && mh->code() != mh->aot_code()) { + mh->aot_code()->make_entrant(); + if (mh->has_compiled_code()) { + mh->code()->make_not_entrant(); + } + MutexLocker pl(CompiledMethod_lock, Mutex::_no_safepoint_check_flag); + Method::set_code(mh, mh->aot_code()); + } + } + return; + } + + // Check if the method can be compiled. If it cannot be compiled with C1, continue profiling + // in the interpreter and then compile with C2 (the transition function will request that, + // see common() ). If the method cannot be compiled with C2 but still can with C1, compile it with + // pure C1. + if ((bci == InvocationEntryBci && !can_be_compiled(mh, level))) { + if (level == CompLevel_full_optimization && can_be_compiled(mh, CompLevel_simple)) { + compile(mh, bci, CompLevel_simple, thread); + } + return; + } + if ((bci != InvocationEntryBci && !can_be_osr_compiled(mh, level))) { + if (level == CompLevel_full_optimization && can_be_osr_compiled(mh, CompLevel_simple)) { + nmethod* osr_nm = mh->lookup_osr_nmethod_for(bci, CompLevel_simple, false); + if (osr_nm != NULL && osr_nm->comp_level() > CompLevel_simple) { + // Invalidate the existing OSR nmethod so that a compile at CompLevel_simple is permitted. + osr_nm->make_not_entrant(); + } + compile(mh, bci, CompLevel_simple, thread); + } + return; + } + if (bci != InvocationEntryBci && mh->is_not_osr_compilable(level)) { + return; + } + if (!CompileBroker::compilation_is_in_queue(mh)) { + if (PrintTieredEvents) { + print_event(COMPILE, mh, mh, bci, level); + } + submit_compile(mh, bci, level, thread); + } +} + +// Update the rate and submit compile +void TieredThresholdPolicy::submit_compile(const methodHandle& mh, int bci, CompLevel level, JavaThread* thread) { + int hot_count = (bci == InvocationEntryBci) ? mh->invocation_count() : mh->backedge_count(); + update_rate(os::javaTimeMillis(), mh()); + CompileBroker::compile_method(mh, bci, level, mh, hot_count, CompileTask::Reason_Tiered, thread); +} + +// Print an event. +void TieredThresholdPolicy::print_specific(EventType type, const methodHandle& mh, const methodHandle& imh, + int bci, CompLevel level) { + tty->print(" rate="); + if (mh->prev_time() == 0) tty->print("n/a"); + else tty->print("%f", mh->rate()); + + tty->print(" k=%.2lf,%.2lf", threshold_scale(CompLevel_full_profile, Tier3LoadFeedback), + threshold_scale(CompLevel_full_optimization, Tier4LoadFeedback)); + +} + +// update_rate() is called from select_task() while holding a compile queue lock. +void TieredThresholdPolicy::update_rate(jlong t, Method* m) { + // Skip update if counters are absent. + // Can't allocate them since we are holding compile queue lock. + if (m->method_counters() == NULL) return; + + if (is_old(m)) { + // We don't remove old methods from the queue, + // so we can just zero the rate. + m->set_rate(0); + return; + } + + // We don't update the rate if we've just came out of a safepoint. + // delta_s is the time since last safepoint in milliseconds. + jlong delta_s = t - SafepointTracing::end_of_last_safepoint_epoch_ms(); + jlong delta_t = t - (m->prev_time() != 0 ? m->prev_time() : start_time()); // milliseconds since the last measurement + // How many events were there since the last time? + int event_count = m->invocation_count() + m->backedge_count(); + int delta_e = event_count - m->prev_event_count(); + + // We should be running for at least 1ms. + if (delta_s >= TieredRateUpdateMinTime) { + // And we must've taken the previous point at least 1ms before. + if (delta_t >= TieredRateUpdateMinTime && delta_e > 0) { + m->set_prev_time(t); + m->set_prev_event_count(event_count); + m->set_rate((float)delta_e / (float)delta_t); // Rate is events per millisecond + } else { + if (delta_t > TieredRateUpdateMaxTime && delta_e == 0) { + // If nothing happened for 25ms, zero the rate. Don't modify prev values. + m->set_rate(0); + } + } + } +} + +// Check if this method has been stale for a given number of milliseconds. +// See select_task(). +bool TieredThresholdPolicy::is_stale(jlong t, jlong timeout, Method* m) { + jlong delta_s = t - SafepointTracing::end_of_last_safepoint_epoch_ms(); + jlong delta_t = t - m->prev_time(); + if (delta_t > timeout && delta_s > timeout) { + int event_count = m->invocation_count() + m->backedge_count(); + int delta_e = event_count - m->prev_event_count(); + // Return true if there were no events. + return delta_e == 0; + } + return false; +} + +// We don't remove old methods from the compile queue even if they have +// very low activity. See select_task(). +bool TieredThresholdPolicy::is_old(Method* method) { + return method->invocation_count() > 50000 || method->backedge_count() > 500000; +} + +double TieredThresholdPolicy::weight(Method* method) { + return (double)(method->rate() + 1) * + (method->invocation_count() + 1) * (method->backedge_count() + 1); +} + +// Apply heuristics and return true if x should be compiled before y +bool TieredThresholdPolicy::compare_methods(Method* x, Method* y) { + if (x->highest_comp_level() > y->highest_comp_level()) { + // recompilation after deopt + return true; + } else + if (x->highest_comp_level() == y->highest_comp_level()) { + if (weight(x) > weight(y)) { + return true; + } + } + return false; +} + +// Is method profiled enough? +bool TieredThresholdPolicy::is_method_profiled(Method* method) { + MethodData* mdo = method->method_data(); + if (mdo != NULL) { + int i = mdo->invocation_count_delta(); + int b = mdo->backedge_count_delta(); + return call_predicate_helper<CompLevel_full_profile>(i, b, 1, method); + } + return false; +} + +double TieredThresholdPolicy::threshold_scale(CompLevel level, int feedback_k) { + double queue_size = CompileBroker::queue_size(level); + int comp_count = compiler_count(level); + double k = queue_size / (feedback_k * comp_count) + 1; + + // Increase C1 compile threshold when the code cache is filled more + // than specified by IncreaseFirstTierCompileThresholdAt percentage. + // The main intention is to keep enough free space for C2 compiled code + // to achieve peak performance if the code cache is under stress. + if ((TieredStopAtLevel == CompLevel_full_optimization) && (level != CompLevel_full_optimization)) { + double current_reverse_free_ratio = CodeCache::reverse_free_ratio(CodeCache::get_code_blob_type(level)); + if (current_reverse_free_ratio > _increase_threshold_at_ratio) { + k *= exp(current_reverse_free_ratio - _increase_threshold_at_ratio); + } + } + return k; +} + +// Call and loop predicates determine whether a transition to a higher +// compilation level should be performed (pointers to predicate functions +// are passed to common()). +// Tier?LoadFeedback is basically a coefficient that determines of +// how many methods per compiler thread can be in the queue before +// the threshold values double. +bool TieredThresholdPolicy::loop_predicate(int i, int b, CompLevel cur_level, Method* method) { + switch(cur_level) { + case CompLevel_aot: { + double k = threshold_scale(CompLevel_full_profile, Tier3LoadFeedback); + return loop_predicate_helper<CompLevel_aot>(i, b, k, method); + } + case CompLevel_none: + case CompLevel_limited_profile: { + double k = threshold_scale(CompLevel_full_profile, Tier3LoadFeedback); + return loop_predicate_helper<CompLevel_none>(i, b, k, method); + } + case CompLevel_full_profile: { + double k = threshold_scale(CompLevel_full_optimization, Tier4LoadFeedback); + return loop_predicate_helper<CompLevel_full_profile>(i, b, k, method); + } + default: + return true; + } +} + +bool TieredThresholdPolicy::call_predicate(int i, int b, CompLevel cur_level, Method* method) { + switch(cur_level) { + case CompLevel_aot: { + double k = threshold_scale(CompLevel_full_profile, Tier3LoadFeedback); + return call_predicate_helper<CompLevel_aot>(i, b, k, method); + } + case CompLevel_none: + case CompLevel_limited_profile: { + double k = threshold_scale(CompLevel_full_profile, Tier3LoadFeedback); + return call_predicate_helper<CompLevel_none>(i, b, k, method); + } + case CompLevel_full_profile: { + double k = threshold_scale(CompLevel_full_optimization, Tier4LoadFeedback); + return call_predicate_helper<CompLevel_full_profile>(i, b, k, method); + } + default: + return true; + } +} + +// Determine is a method is mature. +bool TieredThresholdPolicy::is_mature(Method* method) { + if (should_compile_at_level_simple(method)) return true; + MethodData* mdo = method->method_data(); + if (mdo != NULL) { + int i = mdo->invocation_count(); + int b = mdo->backedge_count(); + double k = ProfileMaturityPercentage / 100.0; + return call_predicate_helper<CompLevel_full_profile>(i, b, k, method) || + loop_predicate_helper<CompLevel_full_profile>(i, b, k, method); + } + return false; +} + +// If a method is old enough and is still in the interpreter we would want to +// start profiling without waiting for the compiled method to arrive. +// We also take the load on compilers into the account. +bool TieredThresholdPolicy::should_create_mdo(Method* method, CompLevel cur_level) { + if (cur_level == CompLevel_none && + CompileBroker::queue_size(CompLevel_full_optimization) <= + Tier3DelayOn * compiler_count(CompLevel_full_optimization)) { + int i = method->invocation_count(); + int b = method->backedge_count(); + double k = Tier0ProfilingStartPercentage / 100.0; + return call_predicate_helper<CompLevel_none>(i, b, k, method) || loop_predicate_helper<CompLevel_none>(i, b, k, method); + } + return false; +} + +// Inlining control: if we're compiling a profiled method with C1 and the callee +// is known to have OSRed in a C2 version, don't inline it. +bool TieredThresholdPolicy::should_not_inline(ciEnv* env, ciMethod* callee) { + CompLevel comp_level = (CompLevel)env->comp_level(); + if (comp_level == CompLevel_full_profile || + comp_level == CompLevel_limited_profile) { + return callee->highest_osr_comp_level() == CompLevel_full_optimization; + } + return false; +} + +// Create MDO if necessary. +void TieredThresholdPolicy::create_mdo(const methodHandle& mh, JavaThread* THREAD) { + if (mh->is_native() || + mh->is_abstract() || + mh->is_accessor() || + mh->is_constant_getter()) { + return; + } + if (mh->method_data() == NULL) { + Method::build_interpreter_method_data(mh, CHECK_AND_CLEAR); + } +} + + +/* + * Method states: + * 0 - interpreter (CompLevel_none) + * 1 - pure C1 (CompLevel_simple) + * 2 - C1 with invocation and backedge counting (CompLevel_limited_profile) + * 3 - C1 with full profiling (CompLevel_full_profile) + * 4 - C2 (CompLevel_full_optimization) + * + * Common state transition patterns: + * a. 0 -> 3 -> 4. + * The most common path. But note that even in this straightforward case + * profiling can start at level 0 and finish at level 3. + * + * b. 0 -> 2 -> 3 -> 4. + * This case occurs when the load on C2 is deemed too high. So, instead of transitioning + * into state 3 directly and over-profiling while a method is in the C2 queue we transition to + * level 2 and wait until the load on C2 decreases. This path is disabled for OSRs. + * + * c. 0 -> (3->2) -> 4. + * In this case we enqueue a method for compilation at level 3, but the C1 queue is long enough + * to enable the profiling to fully occur at level 0. In this case we change the compilation level + * of the method to 2 while the request is still in-queue, because it'll allow it to run much faster + * without full profiling while c2 is compiling. + * + * d. 0 -> 3 -> 1 or 0 -> 2 -> 1. + * After a method was once compiled with C1 it can be identified as trivial and be compiled to + * level 1. These transition can also occur if a method can't be compiled with C2 but can with C1. + * + * e. 0 -> 4. + * This can happen if a method fails C1 compilation (it will still be profiled in the interpreter) + * or because of a deopt that didn't require reprofiling (compilation won't happen in this case because + * the compiled version already exists). + * + * Note that since state 0 can be reached from any other state via deoptimization different loops + * are possible. + * + */ + +// Common transition function. Given a predicate determines if a method should transition to another level. +CompLevel TieredThresholdPolicy::common(Predicate p, Method* method, CompLevel cur_level, bool disable_feedback) { + CompLevel next_level = cur_level; + int i = method->invocation_count(); + int b = method->backedge_count(); + + if (should_compile_at_level_simple(method)) { + next_level = CompLevel_simple; + } else { + switch(cur_level) { + default: break; + case CompLevel_aot: { + // If we were at full profile level, would we switch to full opt? + if (common(p, method, CompLevel_full_profile, disable_feedback) == CompLevel_full_optimization) { + next_level = CompLevel_full_optimization; + } else if (disable_feedback || (CompileBroker::queue_size(CompLevel_full_optimization) <= + Tier3DelayOff * compiler_count(CompLevel_full_optimization) && + (this->*p)(i, b, cur_level, method))) { + next_level = CompLevel_full_profile; + } + } + break; + case CompLevel_none: + // If we were at full profile level, would we switch to full opt? + if (common(p, method, CompLevel_full_profile, disable_feedback) == CompLevel_full_optimization) { + next_level = CompLevel_full_optimization; + } else if ((this->*p)(i, b, cur_level, method)) { +#if INCLUDE_JVMCI + if (EnableJVMCI && UseJVMCICompiler) { + // Since JVMCI takes a while to warm up, its queue inevitably backs up during + // early VM execution. As of 2014-06-13, JVMCI's inliner assumes that the root + // compilation method and all potential inlinees have mature profiles (which + // includes type profiling). If it sees immature profiles, JVMCI's inliner + // can perform pathologically bad (e.g., causing OutOfMemoryErrors due to + // exploring/inlining too many graphs). Since a rewrite of the inliner is + // in progress, we simply disable the dialing back heuristic for now and will + // revisit this decision once the new inliner is completed. + next_level = CompLevel_full_profile; + } else +#endif + { + // C1-generated fully profiled code is about 30% slower than the limited profile + // code that has only invocation and backedge counters. The observation is that + // if C2 queue is large enough we can spend too much time in the fully profiled code + // while waiting for C2 to pick the method from the queue. To alleviate this problem + // we introduce a feedback on the C2 queue size. If the C2 queue is sufficiently long + // we choose to compile a limited profiled version and then recompile with full profiling + // when the load on C2 goes down. + if (!disable_feedback && CompileBroker::queue_size(CompLevel_full_optimization) > + Tier3DelayOn * compiler_count(CompLevel_full_optimization)) { + next_level = CompLevel_limited_profile; + } else { + next_level = CompLevel_full_profile; + } + } + } + break; + case CompLevel_limited_profile: + if (is_method_profiled(method)) { + // Special case: we got here because this method was fully profiled in the interpreter. + next_level = CompLevel_full_optimization; + } else { + MethodData* mdo = method->method_data(); + if (mdo != NULL) { + if (mdo->would_profile()) { + if (disable_feedback || (CompileBroker::queue_size(CompLevel_full_optimization) <= + Tier3DelayOff * compiler_count(CompLevel_full_optimization) && + (this->*p)(i, b, cur_level, method))) { + next_level = CompLevel_full_profile; + } + } else { + next_level = CompLevel_full_optimization; + } + } else { + // If there is no MDO we need to profile + if (disable_feedback || (CompileBroker::queue_size(CompLevel_full_optimization) <= + Tier3DelayOff * compiler_count(CompLevel_full_optimization) && + (this->*p)(i, b, cur_level, method))) { + next_level = CompLevel_full_profile; + } + } + } + break; + case CompLevel_full_profile: + { + MethodData* mdo = method->method_data(); + if (mdo != NULL) { + if (mdo->would_profile()) { + int mdo_i = mdo->invocation_count_delta(); + int mdo_b = mdo->backedge_count_delta(); + if ((this->*p)(mdo_i, mdo_b, cur_level, method)) { + next_level = CompLevel_full_optimization; + } + } else { + next_level = CompLevel_full_optimization; + } + } + } + break; + } + } + return MIN2(next_level, (CompLevel)TieredStopAtLevel); +} + +// Determine if a method should be compiled with a normal entry point at a different level. +CompLevel TieredThresholdPolicy::call_event(Method* method, CompLevel cur_level, JavaThread * thread) { + CompLevel osr_level = MIN2((CompLevel) method->highest_osr_comp_level(), + common(&TieredThresholdPolicy::loop_predicate, method, cur_level, true)); + CompLevel next_level = common(&TieredThresholdPolicy::call_predicate, method, cur_level); + + // If OSR method level is greater than the regular method level, the levels should be + // equalized by raising the regular method level in order to avoid OSRs during each + // invocation of the method. + if (osr_level == CompLevel_full_optimization && cur_level == CompLevel_full_profile) { + MethodData* mdo = method->method_data(); + guarantee(mdo != NULL, "MDO should not be NULL"); + if (mdo->invocation_count() >= 1) { + next_level = CompLevel_full_optimization; + } + } else { + next_level = MAX2(osr_level, next_level); + } + return next_level; +} + +// Determine if we should do an OSR compilation of a given method. +CompLevel TieredThresholdPolicy::loop_event(Method* method, CompLevel cur_level, JavaThread* thread) { + CompLevel next_level = common(&TieredThresholdPolicy::loop_predicate, method, cur_level, true); + if (cur_level == CompLevel_none) { + // If there is a live OSR method that means that we deopted to the interpreter + // for the transition. + CompLevel osr_level = MIN2((CompLevel)method->highest_osr_comp_level(), next_level); + if (osr_level > CompLevel_none) { + return osr_level; + } + } + return next_level; +} + +bool TieredThresholdPolicy::maybe_switch_to_aot(const methodHandle& mh, CompLevel cur_level, CompLevel next_level, JavaThread* thread) { + if (UseAOT) { + if (cur_level == CompLevel_full_profile || cur_level == CompLevel_none) { + // If the current level is full profile or interpreter and we're switching to any other level, + // activate the AOT code back first so that we won't waste time overprofiling. + compile(mh, InvocationEntryBci, CompLevel_aot, thread); + // Fall through for JIT compilation. + } + if (next_level == CompLevel_limited_profile && cur_level != CompLevel_aot && mh->has_aot_code()) { + // If the next level is limited profile, use the aot code (if there is any), + // since it's essentially the same thing. + compile(mh, InvocationEntryBci, CompLevel_aot, thread); + // Not need to JIT, we're done. + return true; + } + } + return false; +} + + +// Handle the invocation event. +void TieredThresholdPolicy::method_invocation_event(const methodHandle& mh, const methodHandle& imh, + CompLevel level, CompiledMethod* nm, JavaThread* thread) { + if (should_create_mdo(mh(), level)) { + create_mdo(mh, thread); + } + CompLevel next_level = call_event(mh(), level, thread); + if (next_level != level) { + if (maybe_switch_to_aot(mh, level, next_level, thread)) { + // No JITting necessary + return; + } + if (is_compilation_enabled() && !CompileBroker::compilation_is_in_queue(mh)) { + compile(mh, InvocationEntryBci, next_level, thread); + } + } +} + +// Handle the back branch event. Notice that we can compile the method +// with a regular entry from here. +void TieredThresholdPolicy::method_back_branch_event(const methodHandle& mh, const methodHandle& imh, + int bci, CompLevel level, CompiledMethod* nm, JavaThread* thread) { + if (should_create_mdo(mh(), level)) { + create_mdo(mh, thread); + } + // Check if MDO should be created for the inlined method + if (should_create_mdo(imh(), level)) { + create_mdo(imh, thread); + } + + if (is_compilation_enabled()) { + CompLevel next_osr_level = loop_event(imh(), level, thread); + CompLevel max_osr_level = (CompLevel)imh->highest_osr_comp_level(); + // At the very least compile the OSR version + if (!CompileBroker::compilation_is_in_queue(imh) && (next_osr_level != level)) { + compile(imh, bci, next_osr_level, thread); + } + + // Use loop event as an opportunity to also check if there's been + // enough calls. + CompLevel cur_level, next_level; + if (mh() != imh()) { // If there is an enclosing method + if (level == CompLevel_aot) { + // Recompile the enclosing method to prevent infinite OSRs. Stay at AOT level while it's compiling. + if (max_osr_level != CompLevel_none && !CompileBroker::compilation_is_in_queue(mh)) { + compile(mh, InvocationEntryBci, MIN2((CompLevel)TieredStopAtLevel, CompLevel_full_profile), thread); + } + } else { + // Current loop event level is not AOT + guarantee(nm != NULL, "Should have nmethod here"); + cur_level = comp_level(mh()); + next_level = call_event(mh(), cur_level, thread); + + if (max_osr_level == CompLevel_full_optimization) { + // The inlinee OSRed to full opt, we need to modify the enclosing method to avoid deopts + bool make_not_entrant = false; + if (nm->is_osr_method()) { + // This is an osr method, just make it not entrant and recompile later if needed + make_not_entrant = true; + } else { + if (next_level != CompLevel_full_optimization) { + // next_level is not full opt, so we need to recompile the + // enclosing method without the inlinee + cur_level = CompLevel_none; + make_not_entrant = true; + } + } + if (make_not_entrant) { + if (PrintTieredEvents) { + int osr_bci = nm->is_osr_method() ? nm->osr_entry_bci() : InvocationEntryBci; + print_event(MAKE_NOT_ENTRANT, mh(), mh(), osr_bci, level); + } + nm->make_not_entrant(); + } + } + // Fix up next_level if necessary to avoid deopts + if (next_level == CompLevel_limited_profile && max_osr_level == CompLevel_full_profile) { + next_level = CompLevel_full_profile; + } + if (cur_level != next_level) { + if (!maybe_switch_to_aot(mh, cur_level, next_level, thread) && !CompileBroker::compilation_is_in_queue(mh)) { + compile(mh, InvocationEntryBci, next_level, thread); + } + } + } + } else { + cur_level = comp_level(mh()); + next_level = call_event(mh(), cur_level, thread); + if (next_level != cur_level) { + if (!maybe_switch_to_aot(mh, cur_level, next_level, thread) && !CompileBroker::compilation_is_in_queue(mh)) { + compile(mh, InvocationEntryBci, next_level, thread); + } + } + } + } +} + +#endif
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/src/hotspot/share/compiler/tieredThresholdPolicy.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -0,0 +1,278 @@ +/* + * Copyright (c) 2010, 2019, 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. + * + */ + +#ifndef SHARE_COMPILER_TIEREDTHRESHOLDPOLICY_HPP +#define SHARE_COMPILER_TIEREDTHRESHOLDPOLICY_HPP + +#include "code/nmethod.hpp" +#include "compiler/compilationPolicy.hpp" +#include "oops/methodData.hpp" +#include "utilities/globalDefinitions.hpp" + +#ifdef TIERED + +class CompileTask; +class CompileQueue; +/* + * The system supports 5 execution levels: + * * level 0 - interpreter + * * level 1 - C1 with full optimization (no profiling) + * * level 2 - C1 with invocation and backedge counters + * * level 3 - C1 with full profiling (level 2 + MDO) + * * level 4 - C2 + * + * Levels 0, 2 and 3 periodically notify the runtime about the current value of the counters + * (invocation counters and backedge counters). The frequency of these notifications is + * different at each level. These notifications are used by the policy to decide what transition + * to make. + * + * Execution starts at level 0 (interpreter), then the policy can decide either to compile the + * method at level 3 or level 2. The decision is based on the following factors: + * 1. The length of the C2 queue determines the next level. The observation is that level 2 + * is generally faster than level 3 by about 30%, therefore we would want to minimize the time + * a method spends at level 3. We should only spend the time at level 3 that is necessary to get + * adequate profiling. So, if the C2 queue is long enough it is more beneficial to go first to + * level 2, because if we transitioned to level 3 we would be stuck there until our C2 compile + * request makes its way through the long queue. When the load on C2 recedes we are going to + * recompile at level 3 and start gathering profiling information. + * 2. The length of C1 queue is used to dynamically adjust the thresholds, so as to introduce + * additional filtering if the compiler is overloaded. The rationale is that by the time a + * method gets compiled it can become unused, so it doesn't make sense to put too much onto the + * queue. + * + * After profiling is completed at level 3 the transition is made to level 4. Again, the length + * of the C2 queue is used as a feedback to adjust the thresholds. + * + * After the first C1 compile some basic information is determined about the code like the number + * of the blocks and the number of the loops. Based on that it can be decided that a method + * is trivial and compiling it with C1 will yield the same code. In this case the method is + * compiled at level 1 instead of 4. + * + * We also support profiling at level 0. If C1 is slow enough to produce the level 3 version of + * the code and the C2 queue is sufficiently small we can decide to start profiling in the + * interpreter (and continue profiling in the compiled code once the level 3 version arrives). + * If the profiling at level 0 is fully completed before level 3 version is produced, a level 2 + * version is compiled instead in order to run faster waiting for a level 4 version. + * + * Compile queues are implemented as priority queues - for each method in the queue we compute + * the event rate (the number of invocation and backedge counter increments per unit of time). + * When getting an element off the queue we pick the one with the largest rate. Maintaining the + * rate also allows us to remove stale methods (the ones that got on the queue but stopped + * being used shortly after that). +*/ + +/* Command line options: + * - Tier?InvokeNotifyFreqLog and Tier?BackedgeNotifyFreqLog control the frequency of method + * invocation and backedge notifications. Basically every n-th invocation or backedge a mutator thread + * makes a call into the runtime. + * + * - Tier?InvocationThreshold, Tier?CompileThreshold, Tier?BackEdgeThreshold, Tier?MinInvocationThreshold control + * compilation thresholds. + * Level 2 thresholds are not used and are provided for option-compatibility and potential future use. + * Other thresholds work as follows: + * + * Transition from interpreter (level 0) to C1 with full profiling (level 3) happens when + * the following predicate is true (X is the level): + * + * i > TierXInvocationThreshold * s || (i > TierXMinInvocationThreshold * s && i + b > TierXCompileThreshold * s), + * + * where $i$ is the number of method invocations, $b$ number of backedges and $s$ is the scaling + * coefficient that will be discussed further. + * The intuition is to equalize the time that is spend profiling each method. + * The same predicate is used to control the transition from level 3 to level 4 (C2). It should be + * noted though that the thresholds are relative. Moreover i and b for the 0->3 transition come + * from Method* and for 3->4 transition they come from MDO (since profiled invocations are + * counted separately). Finally, if a method does not contain anything worth profiling, a transition + * from level 3 to level 4 occurs without considering thresholds (e.g., with fewer invocations than + * what is specified by Tier4InvocationThreshold). + * + * OSR transitions are controlled simply with b > TierXBackEdgeThreshold * s predicates. + * + * - Tier?LoadFeedback options are used to automatically scale the predicates described above depending + * on the compiler load. The scaling coefficients are computed as follows: + * + * s = queue_size_X / (TierXLoadFeedback * compiler_count_X) + 1, + * + * where queue_size_X is the current size of the compiler queue of level X, and compiler_count_X + * is the number of level X compiler threads. + * + * Basically these parameters describe how many methods should be in the compile queue + * per compiler thread before the scaling coefficient increases by one. + * + * This feedback provides the mechanism to automatically control the flow of compilation requests + * depending on the machine speed, mutator load and other external factors. + * + * - Tier3DelayOn and Tier3DelayOff parameters control another important feedback loop. + * Consider the following observation: a method compiled with full profiling (level 3) + * is about 30% slower than a method at level 2 (just invocation and backedge counters, no MDO). + * Normally, the following transitions will occur: 0->3->4. The problem arises when the C2 queue + * gets congested and the 3->4 transition is delayed. While the method is the C2 queue it continues + * executing at level 3 for much longer time than is required by the predicate and at suboptimal speed. + * The idea is to dynamically change the behavior of the system in such a way that if a substantial + * load on C2 is detected we would first do the 0->2 transition allowing a method to run faster. + * And then when the load decreases to allow 2->3 transitions. + * + * Tier3Delay* parameters control this switching mechanism. + * Tier3DelayOn is the number of methods in the C2 queue per compiler thread after which the policy + * no longer does 0->3 transitions but does 0->2 transitions instead. + * Tier3DelayOff switches the original behavior back when the number of methods in the C2 queue + * per compiler thread falls below the specified amount. + * The hysteresis is necessary to avoid jitter. + * + * - TieredCompileTaskTimeout is the amount of time an idle method can spend in the compile queue. + * Basically, since we use the event rate d(i + b)/dt as a value of priority when selecting a method to + * compile from the compile queue, we also can detect stale methods for which the rate has been + * 0 for some time in the same iteration. Stale methods can appear in the queue when an application + * abruptly changes its behavior. + * + * - TieredStopAtLevel, is used mostly for testing. It allows to bypass the policy logic and stick + * to a given level. For example it's useful to set TieredStopAtLevel = 1 in order to compile everything + * with pure c1. + * + * - Tier0ProfilingStartPercentage allows the interpreter to start profiling when the inequalities in the + * 0->3 predicate are already exceeded by the given percentage but the level 3 version of the + * method is still not ready. We can even go directly from level 0 to 4 if c1 doesn't produce a compiled + * version in time. This reduces the overall transition to level 4 and decreases the startup time. + * Note that this behavior is also guarded by the Tier3Delay mechanism: when the c2 queue is too long + * these is not reason to start profiling prematurely. + * + * - TieredRateUpdateMinTime and TieredRateUpdateMaxTime are parameters of the rate computation. + * Basically, the rate is not computed more frequently than TieredRateUpdateMinTime and is considered + * to be zero if no events occurred in TieredRateUpdateMaxTime. + */ + +class TieredThresholdPolicy : public CompilationPolicy { + jlong _start_time; + int _c1_count, _c2_count; + + // Check if the counter is big enough and set carry (effectively infinity). + inline void set_carry_if_necessary(InvocationCounter *counter); + // Set carry flags in the counters (in Method* and MDO). + inline void handle_counter_overflow(Method* method); + // Call and loop predicates determine whether a transition to a higher compilation + // level should be performed (pointers to predicate functions are passed to common_TF(). + // Predicates also take compiler load into account. + typedef bool (TieredThresholdPolicy::*Predicate)(int i, int b, CompLevel cur_level, Method* method); + bool call_predicate(int i, int b, CompLevel cur_level, Method* method); + bool loop_predicate(int i, int b, CompLevel cur_level, Method* method); + // Common transition function. Given a predicate determines if a method should transition to another level. + CompLevel common(Predicate p, Method* method, CompLevel cur_level, bool disable_feedback = false); + // Transition functions. + // call_event determines if a method should be compiled at a different + // level with a regular invocation entry. + CompLevel call_event(Method* method, CompLevel cur_level, JavaThread* thread); + // loop_event checks if a method should be OSR compiled at a different + // level. + CompLevel loop_event(Method* method, CompLevel cur_level, JavaThread* thread); + void print_counters(const char* prefix, const methodHandle& mh); + // Has a method been long around? + // We don't remove old methods from the compile queue even if they have + // very low activity (see select_task()). + inline bool is_old(Method* method); + // Was a given method inactive for a given number of milliseconds. + // If it is, we would remove it from the queue (see select_task()). + inline bool is_stale(jlong t, jlong timeout, Method* m); + // Compute the weight of the method for the compilation scheduling + inline double weight(Method* method); + // Apply heuristics and return true if x should be compiled before y + inline bool compare_methods(Method* x, Method* y); + // Compute event rate for a given method. The rate is the number of event (invocations + backedges) + // per millisecond. + inline void update_rate(jlong t, Method* m); + // Compute threshold scaling coefficient + inline double threshold_scale(CompLevel level, int feedback_k); + // If a method is old enough and is still in the interpreter we would want to + // start profiling without waiting for the compiled method to arrive. This function + // determines whether we should do that. + inline bool should_create_mdo(Method* method, CompLevel cur_level); + // Create MDO if necessary. + void create_mdo(const methodHandle& mh, JavaThread* thread); + // Is method profiled enough? + bool is_method_profiled(Method* method); + + double _increase_threshold_at_ratio; + + bool maybe_switch_to_aot(const methodHandle& mh, CompLevel cur_level, CompLevel next_level, JavaThread* thread); + + int c1_count() const { return _c1_count; } + int c2_count() const { return _c2_count; } + void set_c1_count(int x) { _c1_count = x; } + void set_c2_count(int x) { _c2_count = x; } + + enum EventType { CALL, LOOP, COMPILE, REMOVE_FROM_QUEUE, UPDATE_IN_QUEUE, REPROFILE, MAKE_NOT_ENTRANT }; + void print_event(EventType type, const methodHandle& mh, const methodHandle& imh, int bci, CompLevel level); + // Print policy-specific information if necessary + void print_specific(EventType type, const methodHandle& mh, const methodHandle& imh, int bci, CompLevel level); + // Check if the method can be compiled, change level if necessary + void compile(const methodHandle& mh, int bci, CompLevel level, JavaThread* thread); + // Submit a given method for compilation + void submit_compile(const methodHandle& mh, int bci, CompLevel level, JavaThread* thread); + // Simple methods are as good being compiled with C1 as C2. + // This function tells if it's such a function. + inline static bool is_trivial(Method* method); + // Force method to be compiled at CompLevel_simple? + inline static bool should_compile_at_level_simple(Method* method); + + // Predicate helpers are used by .*_predicate() methods as well as others. + // They check the given counter values, multiplied by the scale against the thresholds. + template<CompLevel level> static inline bool call_predicate_helper(int i, int b, double scale, Method* method); + template<CompLevel level> static inline bool loop_predicate_helper(int i, int b, double scale, Method* method); + + // Get a compilation level for a given method. + static CompLevel comp_level(Method* method); + void method_invocation_event(const methodHandle& method, const methodHandle& inlinee, + CompLevel level, CompiledMethod* nm, JavaThread* thread); + void method_back_branch_event(const methodHandle& method, const methodHandle& inlinee, + int bci, CompLevel level, CompiledMethod* nm, JavaThread* thread); + + void set_increase_threshold_at_ratio() { _increase_threshold_at_ratio = 100 / (100 - (double)IncreaseFirstTierCompileThresholdAt); } + void set_start_time(jlong t) { _start_time = t; } + jlong start_time() const { return _start_time; } + +public: + TieredThresholdPolicy() : _start_time(0), _c1_count(0), _c2_count(0) { } + virtual int compiler_count(CompLevel comp_level) { + if (is_c1_compile(comp_level)) return c1_count(); + if (is_c2_compile(comp_level)) return c2_count(); + return 0; + } + virtual CompLevel initial_compile_level() { return MIN2((CompLevel)TieredStopAtLevel, CompLevel_initial_compile); } + virtual void do_safepoint_work() { } + virtual void delay_compilation(Method* method) { } + virtual void disable_compilation(Method* method) { } + virtual void reprofile(ScopeDesc* trap_scope, bool is_osr); + virtual nmethod* event(const methodHandle& method, const methodHandle& inlinee, + int branch_bci, int bci, CompLevel comp_level, CompiledMethod* nm, JavaThread* thread); + // Select task is called by CompileBroker. We should return a task or NULL. + virtual CompileTask* select_task(CompileQueue* compile_queue); + // Tell the runtime if we think a given method is adequately profiled. + virtual bool is_mature(Method* method); + // Initialize: set compiler thread count + virtual void initialize(); + virtual bool should_not_inline(ciEnv* env, ciMethod* callee); +}; + +#endif // TIERED + +#endif // SHARE_COMPILER_TIEREDTHRESHOLDPOLICY_HPP
--- a/src/hotspot/share/gc/g1/g1Analytics.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1Analytics.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -78,6 +78,8 @@ _alloc_rate_ms_seq(new TruncatedSeq(TruncatedSeqLength)), _prev_collection_pause_end_ms(0.0), _rs_length_diff_seq(new TruncatedSeq(TruncatedSeqLength)), + _concurrent_refine_rate_ms_seq(new TruncatedSeq(TruncatedSeqLength)), + _logged_cards_rate_ms_seq(new TruncatedSeq(TruncatedSeqLength)), _cost_per_logged_card_ms_seq(new TruncatedSeq(TruncatedSeqLength)), _cost_scan_hcc_seq(new TruncatedSeq(TruncatedSeqLength)), _young_cards_per_entry_ratio_seq(new TruncatedSeq(TruncatedSeqLength)), @@ -102,6 +104,10 @@ int index = MIN2(ParallelGCThreads - 1, 7u); _rs_length_diff_seq->add(rs_length_diff_defaults[index]); + // Start with inverse of maximum STW cost. + _concurrent_refine_rate_ms_seq->add(1/cost_per_logged_card_ms_defaults[0]); + // Some applications have very low rates for logging cards. + _logged_cards_rate_ms_seq->add(0.0); _cost_per_logged_card_ms_seq->add(cost_per_logged_card_ms_defaults[index]); _cost_scan_hcc_seq->add(0.0); _young_cards_per_entry_ratio_seq->add(young_cards_per_entry_ratio_defaults[index]); @@ -159,6 +165,14 @@ (pause_time_ms * _recent_prev_end_times_for_all_gcs_sec->num()) / interval_ms; } +void G1Analytics::report_concurrent_refine_rate_ms(double cards_per_ms) { + _concurrent_refine_rate_ms_seq->add(cards_per_ms); +} + +void G1Analytics::report_logged_cards_rate_ms(double cards_per_ms) { + _logged_cards_rate_ms_seq->add(cards_per_ms); +} + void G1Analytics::report_cost_per_logged_card_ms(double cost_per_logged_card_ms) { _cost_per_logged_card_ms_seq->add(cost_per_logged_card_ms); } @@ -223,6 +237,14 @@ return get_new_prediction(_alloc_rate_ms_seq); } +double G1Analytics::predict_concurrent_refine_rate_ms() const { + return get_new_prediction(_concurrent_refine_rate_ms_seq); +} + +double G1Analytics::predict_logged_cards_rate_ms() const { + return get_new_prediction(_logged_cards_rate_ms_seq); +} + double G1Analytics::predict_cost_per_logged_card_ms() const { return get_new_prediction(_cost_per_logged_card_ms_seq); }
--- a/src/hotspot/share/gc/g1/g1Analytics.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1Analytics.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -46,6 +46,8 @@ double _prev_collection_pause_end_ms; TruncatedSeq* _rs_length_diff_seq; + TruncatedSeq* _concurrent_refine_rate_ms_seq; + TruncatedSeq* _logged_cards_rate_ms_seq; TruncatedSeq* _cost_per_logged_card_ms_seq; TruncatedSeq* _cost_scan_hcc_seq; TruncatedSeq* _young_cards_per_entry_ratio_seq; @@ -99,6 +101,8 @@ void report_concurrent_mark_remark_times_ms(double ms); void report_concurrent_mark_cleanup_times_ms(double ms); void report_alloc_rate_ms(double alloc_rate); + void report_concurrent_refine_rate_ms(double cards_per_ms); + void report_logged_cards_rate_ms(double cards_per_ms); void report_cost_per_logged_card_ms(double cost_per_logged_card_ms); void report_cost_scan_hcc(double cost_scan_hcc); void report_cost_per_remset_card_ms(double cost_per_remset_card_ms, bool for_young_gc); @@ -116,6 +120,8 @@ double predict_alloc_rate_ms() const; int num_alloc_rate_ms() const; + double predict_concurrent_refine_rate_ms() const; + double predict_logged_cards_rate_ms() const; double predict_cost_per_logged_card_ms() const; double predict_scan_hcc_ms() const;
--- a/src/hotspot/share/gc/g1/g1CollectionSet.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1CollectionSet.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -409,7 +409,7 @@ guarantee(target_pause_time_ms > 0.0, "target_pause_time_ms = %1.6lf should be positive", target_pause_time_ms); - size_t pending_cards = _policy->pending_cards(); + size_t pending_cards = _policy->pending_cards_at_gc_start(); double base_time_ms = _policy->predict_base_elapsed_time_ms(pending_cards); double time_remaining_ms = MAX2(target_pause_time_ms - base_time_ms, 0.0);
--- a/src/hotspot/share/gc/g1/g1ConcurrentRefine.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1ConcurrentRefine.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -412,6 +412,22 @@ dcqs.notify_if_necessary(); } +G1ConcurrentRefine::RefinementStats G1ConcurrentRefine::total_refinement_stats() const { + struct CollectData : public ThreadClosure { + Tickspan _total_time; + size_t _total_cards; + CollectData() : _total_time(), _total_cards(0) {} + virtual void do_thread(Thread* t) { + G1ConcurrentRefineThread* crt = static_cast<G1ConcurrentRefineThread*>(t); + _total_time += crt->total_refinement_time(); + _total_cards += crt->total_refined_cards(); + } + } collector; + // Cast away const so we can call non-modifying closure on threads. + const_cast<G1ConcurrentRefine*>(this)->threads_do(&collector); + return RefinementStats(collector._total_time, collector._total_cards); +} + size_t G1ConcurrentRefine::activation_threshold(uint worker_id) const { Thresholds thresholds = calc_thresholds(_green_zone, _yellow_zone, worker_id); return activation_level(thresholds); @@ -432,7 +448,8 @@ } } -bool G1ConcurrentRefine::do_refinement_step(uint worker_id) { +bool G1ConcurrentRefine::do_refinement_step(uint worker_id, + size_t* total_refined_cards) { G1DirtyCardQueueSet& dcqs = G1BarrierSet::dirty_card_queue_set(); size_t curr_cards = dcqs.num_cards(); @@ -448,5 +465,6 @@ // Process the next buffer, if there are enough left. return dcqs.refine_completed_buffer_concurrently(worker_id + worker_id_offset(), - deactivation_threshold(worker_id)); + deactivation_threshold(worker_id), + total_refined_cards); }
--- a/src/hotspot/share/gc/g1/g1ConcurrentRefine.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1ConcurrentRefine.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -27,6 +27,7 @@ #include "memory/allocation.hpp" #include "utilities/globalDefinitions.hpp" +#include "utilities/ticks.hpp" // Forward decl class G1ConcurrentRefine; @@ -118,11 +119,22 @@ // Adjust refinement thresholds based on work done during the pause and the goal time. void adjust(double logged_cards_scan_time, size_t processed_logged_cards, double goal_ms); + struct RefinementStats { + Tickspan _time; + size_t _cards; + RefinementStats(Tickspan time, size_t cards) : _time(time), _cards(cards) {} + }; + + RefinementStats total_refinement_stats() const; + // Cards in the dirty card queue set. size_t activation_threshold(uint worker_id) const; size_t deactivation_threshold(uint worker_id) const; - // Perform a single refinement step. Called by the refinement threads when woken up. - bool do_refinement_step(uint worker_id); + + // Perform a single refinement step; called by the refinement + // threads. Returns true if there was refinement work available. + // Increments *total_refined_cards. + bool do_refinement_step(uint worker_id, size_t* total_refined_cards); // Iterate over all concurrent refinement threads applying the given closure. void threads_do(ThreadClosure *tc);
--- a/src/hotspot/share/gc/g1/g1ConcurrentRefineThread.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1ConcurrentRefineThread.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -37,6 +37,8 @@ ConcurrentGCThread(), _vtime_start(0.0), _vtime_accum(0.0), + _total_refinement_time(), + _total_refined_cards(0), _worker_id(worker_id), _active(false), _monitor(NULL), @@ -101,11 +103,12 @@ break; } - size_t buffers_processed = 0; log_debug(gc, refine)("Activated worker %d, on threshold: " SIZE_FORMAT ", current: " SIZE_FORMAT, _worker_id, _cr->activation_threshold(_worker_id), G1BarrierSet::dirty_card_queue_set().num_cards()); + size_t start_total_refined_cards = _total_refined_cards; // For logging. + { SuspendibleThreadSetJoiner sts_join; @@ -115,20 +118,22 @@ continue; // Re-check for termination after yield delay. } - if (!_cr->do_refinement_step(_worker_id)) { - break; + Ticks start_time = Ticks::now(); + if (!_cr->do_refinement_step(_worker_id, &_total_refined_cards)) { + break; // No cards to process. } - ++buffers_processed; + _total_refinement_time += (Ticks::now() - start_time); } } deactivate(); log_debug(gc, refine)("Deactivated worker %d, off threshold: " SIZE_FORMAT - ", current: " SIZE_FORMAT ", buffers processed: " - SIZE_FORMAT, + ", current: " SIZE_FORMAT ", refined cards: " + SIZE_FORMAT ", total refined cards: " SIZE_FORMAT, _worker_id, _cr->deactivation_threshold(_worker_id), G1BarrierSet::dirty_card_queue_set().num_cards(), - buffers_processed); + _total_refined_cards - start_total_refined_cards, + _total_refined_cards); if (os::supports_vtime()) { _vtime_accum = (os::elapsedVTime() - _vtime_start);
--- a/src/hotspot/share/gc/g1/g1ConcurrentRefineThread.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1ConcurrentRefineThread.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -26,6 +26,7 @@ #define SHARE_GC_G1_G1CONCURRENTREFINETHREAD_HPP #include "gc/shared/concurrentGCThread.hpp" +#include "utilities/ticks.hpp" // Forward Decl. class G1ConcurrentRefine; @@ -38,6 +39,10 @@ double _vtime_start; // Initial virtual time. double _vtime_accum; // Accumulated virtual time. + + Tickspan _total_refinement_time; + size_t _total_refined_cards; + uint _worker_id; bool _active; @@ -61,6 +66,9 @@ // Activate this thread. void activate(); + Tickspan total_refinement_time() const { return _total_refinement_time; } + size_t total_refined_cards() const { return _total_refined_cards; } + // Total virtual time so far. double vtime_accum() { return _vtime_accum; } };
--- a/src/hotspot/share/gc/g1/g1DirtyCardQueue.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1DirtyCardQueue.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -37,6 +37,7 @@ #include "runtime/atomic.hpp" #include "runtime/flags/flagSetting.hpp" #include "runtime/mutexLocker.hpp" +#include "runtime/os.hpp" #include "runtime/safepoint.hpp" #include "runtime/thread.inline.hpp" #include "runtime/threadSMR.hpp" @@ -62,6 +63,9 @@ } } +// Assumed to be zero by concurrent threads. +static uint par_ids_start() { return 0; } + G1DirtyCardQueueSet::G1DirtyCardQueueSet(Monitor* cbl_mon, BufferNode::Allocator* allocator) : PtrQueueSet(allocator), @@ -73,15 +77,16 @@ _process_completed_buffers(false), _max_cards(MaxCardsUnlimited), _max_cards_padding(0), - _free_ids(0, num_par_ids()), - _processed_buffers_mut(0), - _processed_buffers_rs_thread(0) + _free_ids(par_ids_start(), num_par_ids()), + _mutator_refined_cards_counters(NEW_C_HEAP_ARRAY(size_t, num_par_ids(), mtGC)) { + ::memset(_mutator_refined_cards_counters, 0, num_par_ids() * sizeof(size_t)); _all_active = true; } G1DirtyCardQueueSet::~G1DirtyCardQueueSet() { abandon_completed_buffers(); + FREE_C_HEAP_ARRAY(size_t, _mutator_refined_cards_counters); } // Determines how many mutator threads can process the buffers in parallel. @@ -89,6 +94,14 @@ return (uint)os::initial_active_processor_count(); } +size_t G1DirtyCardQueueSet::total_mutator_refined_cards() const { + size_t sum = 0; + for (uint i = 0; i < num_par_ids(); ++i) { + sum += _mutator_refined_cards_counters[i]; + } + return sum; +} + void G1DirtyCardQueueSet::handle_zero_index_for_thread(Thread* t) { G1ThreadLocalData::dirty_card_queue(t).handle_zero_index(); } @@ -213,7 +226,9 @@ return result; } -bool G1DirtyCardQueueSet::refine_buffer(BufferNode* node, uint worker_id) { +bool G1DirtyCardQueueSet::refine_buffer(BufferNode* node, + uint worker_id, + size_t* total_refined_cards) { G1RemSet* rem_set = G1CollectedHeap::heap()->rem_set(); size_t size = buffer_size(); void** buffer = BufferNode::make_buffer_from_node(node); @@ -223,6 +238,7 @@ CardTable::CardValue* cp = static_cast<CardTable::CardValue*>(buffer[i]); rem_set->refine_card_concurrently(cp, worker_id); } + *total_refined_cards += (i - node->index()); node->set_index(i); return i == size; } @@ -260,25 +276,27 @@ bool G1DirtyCardQueueSet::mut_process_buffer(BufferNode* node) { uint worker_id = _free_ids.claim_par_id(); // temporarily claim an id - bool result = refine_buffer(node, worker_id); + uint counter_index = worker_id - par_ids_start(); + size_t* counter = &_mutator_refined_cards_counters[counter_index]; + bool result = refine_buffer(node, worker_id, counter); _free_ids.release_par_id(worker_id); // release the id if (result) { assert_fully_consumed(node, buffer_size()); - Atomic::inc(&_processed_buffers_mut); } return result; } -bool G1DirtyCardQueueSet::refine_completed_buffer_concurrently(uint worker_id, size_t stop_at) { +bool G1DirtyCardQueueSet::refine_completed_buffer_concurrently(uint worker_id, + size_t stop_at, + size_t* total_refined_cards) { BufferNode* node = get_completed_buffer(stop_at); if (node == NULL) { return false; - } else if (refine_buffer(node, worker_id)) { + } else if (refine_buffer(node, worker_id, total_refined_cards)) { assert_fully_consumed(node, buffer_size()); // Done with fully processed buffer. deallocate_buffer(node); - Atomic::inc(&_processed_buffers_rs_thread); return true; } else { // Return partially processed buffer to the queue.
--- a/src/hotspot/share/gc/g1/g1DirtyCardQueue.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1DirtyCardQueue.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -78,14 +78,15 @@ void abandon_completed_buffers(); - // Refine the cards in "node" from it's index to buffer_size. + // Refine the cards in "node" from its index to buffer_size. // Stops processing if SuspendibleThreadSet::should_yield() is true. // Returns true if the entire buffer was processed, false if there // is a pending yield request. The node's index is updated to exclude // the processed elements, e.g. up to the element before processing // stopped, or one past the last element if the entire buffer was - // processed. - bool refine_buffer(BufferNode* node, uint worker_id); + // processed. Increments *total_refined_cards by the number of cards + // processed and removed from the buffer. + bool refine_buffer(BufferNode* node, uint worker_id, size_t* total_refined_cards); bool mut_process_buffer(BufferNode* node); @@ -97,10 +98,9 @@ G1FreeIdSet _free_ids; - // The number of completed buffers processed by mutator and rs thread, - // respectively. - jint _processed_buffers_mut; - jint _processed_buffers_rs_thread; + // Array of cumulative dirty cards refined by mutator threads. + // Array has an entry per id in _free_ids. + size_t* _mutator_refined_cards_counters; public: G1DirtyCardQueueSet(Monitor* cbl_mon, BufferNode::Allocator* allocator); @@ -158,7 +158,12 @@ // Stops processing a buffer if SuspendibleThreadSet::should_yield(), // returning the incompletely processed buffer to the completed buffer // list, for later processing of the remainder. - bool refine_completed_buffer_concurrently(uint worker_id, size_t stop_at); + // + // Increments *total_refined_cards by the number of cards processed and + // removed from the buffer. + bool refine_completed_buffer_concurrently(uint worker_id, + size_t stop_at, + size_t* total_refined_cards); // If a full collection is happening, reset partial logs, and release // completed ones: the full collection will make them all irrelevant. @@ -181,13 +186,8 @@ return _max_cards_padding; } - jint processed_buffers_mut() { - return _processed_buffers_mut; - } - jint processed_buffers_rs_thread() { - return _processed_buffers_rs_thread; - } - + // Total dirty cards refined by mutator threads. + size_t total_mutator_refined_cards() const; }; inline G1DirtyCardQueueSet* G1DirtyCardQueue::dirty_card_qset() const {
--- a/src/hotspot/share/gc/g1/g1Policy.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1Policy.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -70,7 +70,11 @@ _free_regions_at_end_of_collection(0), _max_rs_length(0), _rs_length_prediction(0), - _pending_cards(0), + _pending_cards_at_gc_start(0), + _pending_cards_at_prev_gc_end(0), + _total_mutator_refined_cards(0), + _total_concurrent_refined_cards(0), + _total_concurrent_refinement_time(), _bytes_allocated_in_old_since_last_gc(0), _initial_mark_to_mixed(), _collection_set(NULL), @@ -442,6 +446,7 @@ collector_state()->set_in_young_only_phase(false); collector_state()->set_in_full_gc(true); _collection_set->clear_candidates(); + record_concurrent_refinement_data(true /* is_full_collection */); } void G1Policy::record_full_collection_end() { @@ -472,12 +477,67 @@ _survivor_surv_rate_group->reset(); update_young_list_max_and_target_length(); update_rs_length_prediction(); + _pending_cards_at_prev_gc_end = _g1h->pending_card_num(); _bytes_allocated_in_old_since_last_gc = 0; record_pause(FullGC, _full_collection_start_sec, end_sec); } +void G1Policy::record_concurrent_refinement_data(bool is_full_collection) { + _pending_cards_at_gc_start = _g1h->pending_card_num(); + + // Record info about concurrent refinement thread processing. + G1ConcurrentRefine* cr = _g1h->concurrent_refine(); + G1ConcurrentRefine::RefinementStats cr_stats = cr->total_refinement_stats(); + + Tickspan cr_time = cr_stats._time - _total_concurrent_refinement_time; + _total_concurrent_refinement_time = cr_stats._time; + + size_t cr_cards = cr_stats._cards - _total_concurrent_refined_cards; + _total_concurrent_refined_cards = cr_stats._cards; + + // Don't update rate if full collection. We could be in an implicit full + // collection after a non-full collection failure, in which case there + // wasn't any mutator/cr-thread activity since last recording. And if + // we're in an explicit full collection, the time since the last GC can + // be arbitrarily short, so not a very good sample. Similarly, don't + // update the rate if the current sample is empty or time is zero. + if (!is_full_collection && (cr_cards > 0) && (cr_time > Tickspan())) { + double rate = cr_cards / (cr_time.seconds() * MILLIUNITS); + _analytics->report_concurrent_refine_rate_ms(rate); + } + + // Record info about mutator thread processing. + G1DirtyCardQueueSet& dcqs = G1BarrierSet::dirty_card_queue_set(); + size_t mut_total_cards = dcqs.total_mutator_refined_cards(); + size_t mut_cards = mut_total_cards - _total_mutator_refined_cards; + _total_mutator_refined_cards = mut_total_cards; + + // Record mutator's card logging rate. + // Don't update if full collection; see above. + if (!is_full_collection) { + size_t total_cards = _pending_cards_at_gc_start + cr_cards + mut_cards; + assert(_pending_cards_at_prev_gc_end <= total_cards, + "untracked cards: last pending: " SIZE_FORMAT + ", pending: " SIZE_FORMAT ", conc refine: " SIZE_FORMAT + ", mut refine:" SIZE_FORMAT, + _pending_cards_at_prev_gc_end, _pending_cards_at_gc_start, + cr_cards, mut_cards); + size_t logged_cards = total_cards - _pending_cards_at_prev_gc_end; + double logging_start_time = _analytics->prev_collection_pause_end_ms(); + double logging_end_time = Ticks::now().seconds() * MILLIUNITS; + double logging_time = logging_end_time - logging_start_time; + // Unlike above for conc-refine rate, here we should not require a + // non-empty sample, since an application could go some time with only + // young-gen or filtered out writes. But we'll ignore unusually short + // sample periods, as they may just pollute the predictions. + if (logging_time > 1.0) { // Require > 1ms sample time. + _analytics->report_logged_cards_rate_ms(logged_cards / logging_time); + } + } +} + void G1Policy::record_collection_pause_start(double start_time_sec) { // We only need to do this here as the policy will only be applied // to the GC we're about to start. so, no point is calculating this @@ -490,7 +550,8 @@ assert_used_and_recalculate_used_equal(_g1h); phase_times()->record_cur_collection_start_sec(start_time_sec); - _pending_cards = _g1h->pending_card_num(); + + record_concurrent_refinement_data(false /* is_full_collection */); _collection_set->reset_bytes_used_before(); _bytes_copied_during_gc = 0; @@ -744,7 +805,7 @@ // after the mixed gc phase. // During mixed gc we do not use them for young gen sizing. if (this_pause_was_young_only) { - _analytics->report_pending_cards((double) _pending_cards); + _analytics->report_pending_cards((double) _pending_cards_at_gc_start); _analytics->report_rs_length((double) _max_rs_length); } } @@ -798,6 +859,7 @@ scan_logged_cards_time_goal_ms -= scan_hcc_time_ms; } + _pending_cards_at_prev_gc_end = _g1h->pending_card_num(); double const logged_cards_time = logged_cards_processing_time(); log_debug(gc, ergo, refine)("Concurrent refinement times: Logged Cards Scan time goal: %1.2fms Logged Cards Scan time: %1.2fms HCC time: %1.2fms",
--- a/src/hotspot/share/gc/g1/g1Policy.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1Policy.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -100,7 +100,11 @@ size_t _rs_length_prediction; - size_t _pending_cards; + size_t _pending_cards_at_gc_start; + size_t _pending_cards_at_prev_gc_end; + size_t _total_mutator_refined_cards; + size_t _total_concurrent_refined_cards; + Tickspan _total_concurrent_refinement_time; // The amount of allocated bytes in old gen during the last mutator and the following // young GC phase. @@ -244,7 +248,15 @@ uint base_free_regions, double target_pause_time_ms) const; public: - size_t pending_cards() const { return _pending_cards; } + size_t pending_cards_at_gc_start() const { return _pending_cards_at_gc_start; } + + size_t total_concurrent_refined_cards() const { + return _total_concurrent_refined_cards; + } + + size_t total_mutator_refined_cards() const { + return _total_mutator_refined_cards; + } // Calculate the minimum number of old regions we'll add to the CSet // during a mixed GC. @@ -283,6 +295,9 @@ void record_pause(PauseKind kind, double start, double end); // Indicate that we aborted marking before doing any mixed GCs. void abort_time_to_mixed_tracking(); + + void record_concurrent_refinement_data(bool is_full_collection); + public: G1Policy(STWGCTimer* gc_timer);
--- a/src/hotspot/share/gc/g1/g1RemSet.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1RemSet.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -489,7 +489,6 @@ _scan_state(new G1RemSetScanState()), _prev_period_summary(), _g1h(g1h), - _num_conc_refined_cards(0), _ct(ct), _g1p(_g1h->policy()), _hot_card_cache(hot_card_cache) { @@ -1377,7 +1376,6 @@ G1ConcurrentRefineOopClosure conc_refine_cl(_g1h, worker_id); if (r->oops_on_memregion_seq_iterate_careful<false>(dirty_region, &conc_refine_cl) != NULL) { - _num_conc_refined_cards++; // Unsynchronized update, only used for logging. return; }
--- a/src/hotspot/share/gc/g1/g1RemSet.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1RemSet.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -61,7 +61,6 @@ G1RemSetSummary _prev_period_summary; G1CollectedHeap* _g1h; - size_t _num_conc_refined_cards; // Number of cards refined concurrently to the mutator. G1CardTable* _ct; G1Policy* _g1p; @@ -125,8 +124,6 @@ // Print accumulated summary info from the last time called. void print_periodic_summary_info(const char* header, uint period_count); - size_t num_conc_refined_cards() const { return _num_conc_refined_cards; } - // Rebuilds the remembered set by scanning from bottom to TARS for all regions // using the given work gang. void rebuild_rem_set(G1ConcurrentMark* cm, WorkGang* workers, uint worker_id_offset);
--- a/src/hotspot/share/gc/g1/g1RemSetSummary.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1RemSetSummary.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -27,6 +27,7 @@ #include "gc/g1/g1ConcurrentRefine.hpp" #include "gc/g1/g1ConcurrentRefineThread.hpp" #include "gc/g1/g1DirtyCardQueue.hpp" +#include "gc/g1/g1Policy.hpp" #include "gc/g1/g1RemSet.hpp" #include "gc/g1/g1RemSetSummary.hpp" #include "gc/g1/g1YoungRemSetSamplingThread.hpp" @@ -53,18 +54,17 @@ }; void G1RemSetSummary::update() { - _num_conc_refined_cards = _rem_set->num_conc_refined_cards(); - G1DirtyCardQueueSet& dcqs = G1BarrierSet::dirty_card_queue_set(); - _num_processed_buf_mutator = dcqs.processed_buffers_mut(); - _num_processed_buf_rs_threads = dcqs.processed_buffers_rs_thread(); + G1CollectedHeap* g1h = G1CollectedHeap::heap(); + + const G1Policy* policy = g1h->policy(); + _total_mutator_refined_cards = policy->total_mutator_refined_cards(); + _total_concurrent_refined_cards = policy->total_concurrent_refined_cards(); _num_coarsenings = HeapRegionRemSet::n_coarsenings(); - G1CollectedHeap* g1h = G1CollectedHeap::heap(); - G1ConcurrentRefine* cg1r = g1h->concurrent_refine(); if (_rs_threads_vtimes != NULL) { GetRSThreadVTimeClosure p(this); - cg1r->threads_do(&p); + g1h->concurrent_refine()->threads_do(&p); } set_sampling_thread_vtime(g1h->sampling_thread()->vtime_accum()); } @@ -83,9 +83,8 @@ G1RemSetSummary::G1RemSetSummary() : _rem_set(NULL), - _num_conc_refined_cards(0), - _num_processed_buf_mutator(0), - _num_processed_buf_rs_threads(0), + _total_mutator_refined_cards(0), + _total_concurrent_refined_cards(0), _num_coarsenings(0), _num_vtimes(G1ConcurrentRefine::max_num_threads()), _rs_threads_vtimes(NEW_C_HEAP_ARRAY(double, _num_vtimes, mtGC)), @@ -96,9 +95,8 @@ G1RemSetSummary::G1RemSetSummary(G1RemSet* rem_set) : _rem_set(rem_set), - _num_conc_refined_cards(0), - _num_processed_buf_mutator(0), - _num_processed_buf_rs_threads(0), + _total_mutator_refined_cards(0), + _total_concurrent_refined_cards(0), _num_coarsenings(0), _num_vtimes(G1ConcurrentRefine::max_num_threads()), _rs_threads_vtimes(NEW_C_HEAP_ARRAY(double, _num_vtimes, mtGC)), @@ -114,12 +112,10 @@ assert(other != NULL, "just checking"); assert(_num_vtimes == other->_num_vtimes, "just checking"); - _num_conc_refined_cards = other->num_conc_refined_cards(); + _total_mutator_refined_cards = other->total_mutator_refined_cards(); + _total_concurrent_refined_cards = other->total_concurrent_refined_cards(); - _num_processed_buf_mutator = other->num_processed_buf_mutator(); - _num_processed_buf_rs_threads = other->num_processed_buf_rs_threads(); - - _num_coarsenings = other->_num_coarsenings; + _num_coarsenings = other->num_coarsenings(); memcpy(_rs_threads_vtimes, other->_rs_threads_vtimes, sizeof(double) * _num_vtimes); @@ -130,10 +126,8 @@ assert(other != NULL, "just checking"); assert(_num_vtimes == other->_num_vtimes, "just checking"); - _num_conc_refined_cards = other->num_conc_refined_cards() - _num_conc_refined_cards; - - _num_processed_buf_mutator = other->num_processed_buf_mutator() - _num_processed_buf_mutator; - _num_processed_buf_rs_threads = other->num_processed_buf_rs_threads() - _num_processed_buf_rs_threads; + _total_mutator_refined_cards = other->total_mutator_refined_cards() - _total_mutator_refined_cards; + _total_concurrent_refined_cards = other->total_concurrent_refined_cards() - _total_concurrent_refined_cards; _num_coarsenings = other->num_coarsenings() - _num_coarsenings; @@ -356,16 +350,15 @@ void G1RemSetSummary::print_on(outputStream* out) { out->print_cr(" Recent concurrent refinement statistics"); - out->print_cr(" Processed " SIZE_FORMAT " cards concurrently", num_conc_refined_cards()); - out->print_cr(" Of " SIZE_FORMAT " completed buffers:", num_processed_buf_total()); - out->print_cr(" " SIZE_FORMAT_W(8) " (%5.1f%%) by concurrent RS threads.", - num_processed_buf_total(), - percent_of(num_processed_buf_rs_threads(), num_processed_buf_total())); + out->print_cr(" Of " SIZE_FORMAT " refined cards:", total_refined_cards()); + out->print_cr(" " SIZE_FORMAT_W(8) " (%5.1f%%) by concurrent refinement threads.", + total_concurrent_refined_cards(), + percent_of(total_concurrent_refined_cards(), total_refined_cards())); out->print_cr(" " SIZE_FORMAT_W(8) " (%5.1f%%) by mutator threads.", - num_processed_buf_mutator(), - percent_of(num_processed_buf_mutator(), num_processed_buf_total())); + total_mutator_refined_cards(), + percent_of(total_mutator_refined_cards(), total_refined_cards())); out->print_cr(" Did " SIZE_FORMAT " coarsenings.", num_coarsenings()); - out->print_cr(" Concurrent RS threads times (s)"); + out->print_cr(" Concurrent refinement threads times (s)"); out->print(" "); for (uint i = 0; i < _num_vtimes; i++) { out->print(" %5.2f", rs_thread_vtime(i));
--- a/src/hotspot/share/gc/g1/g1RemSetSummary.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/g1RemSetSummary.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -38,9 +38,8 @@ G1RemSet* _rem_set; - size_t _num_conc_refined_cards; - size_t _num_processed_buf_mutator; - size_t _num_processed_buf_rs_threads; + size_t _total_mutator_refined_cards; + size_t _total_concurrent_refined_cards; size_t _num_coarsenings; @@ -76,20 +75,16 @@ return _sampling_thread_vtime; } - size_t num_conc_refined_cards() const { - return _num_conc_refined_cards; + size_t total_mutator_refined_cards() const { + return _total_mutator_refined_cards; } - size_t num_processed_buf_mutator() const { - return _num_processed_buf_mutator; + size_t total_concurrent_refined_cards() const { + return _total_concurrent_refined_cards; } - size_t num_processed_buf_rs_threads() const { - return _num_processed_buf_rs_threads; - } - - size_t num_processed_buf_total() const { - return num_processed_buf_mutator() + num_processed_buf_rs_threads(); + size_t total_refined_cards() const { + return total_mutator_refined_cards() + total_concurrent_refined_cards(); } size_t num_coarsenings() const {
--- a/src/hotspot/share/gc/g1/heapRegionRemSet.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/heapRegionRemSet.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -186,10 +186,6 @@ _collision_list_next(NULL) {} - inline void add_card_work(CardIdx_t from_card, bool par); - - inline void add_reference_work(OopOrNarrowOopStar from, bool par); - public: // We need access in order to union things into the base table. BitMap* bm() { return &_bm; } @@ -206,12 +202,8 @@ inline void add_reference(OopOrNarrowOopStar from); - inline void seq_add_reference(OopOrNarrowOopStar from); - inline void add_card(CardIdx_t from_card_index); - void seq_add_card(CardIdx_t from_card_index); - // (Destructively) union the bitmap of the current table into the given // bitmap (which is assumed to be of the same size.) void union_bitmap_into(BitMap* bm) { @@ -381,12 +373,6 @@ _state = Complete; } - // Used in the sequential case. - void add_reference(OopOrNarrowOopStar from) { - add_reference(from, 0); - } - - // Used in the parallel case. void add_reference(OopOrNarrowOopStar from, uint tid) { RemSetState state = _state; if (state == Untracked) {
--- a/src/hotspot/share/gc/g1/heapRegionRemSet.inline.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/g1/heapRegionRemSet.inline.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -35,20 +35,13 @@ _other_regions.iterate(cl); } -inline void PerRegionTable::add_card_work(CardIdx_t from_card, bool par) { - if (!_bm.at(from_card)) { - if (par) { - if (_bm.par_set_bit(from_card)) { - Atomic::inc(&_occupied); - } - } else { - _bm.set_bit(from_card); - _occupied++; - } +inline void PerRegionTable::add_card(CardIdx_t from_card_index) { + if (_bm.par_set_bit(from_card_index)) { + Atomic::inc(&_occupied); } } -inline void PerRegionTable::add_reference_work(OopOrNarrowOopStar from, bool par) { +inline void PerRegionTable::add_reference(OopOrNarrowOopStar from) { // Must make this robust in case "from" is not in "_hr", because of // concurrency. @@ -58,26 +51,10 @@ // and adding a bit to the new table is never incorrect. if (loc_hr->is_in_reserved(from)) { CardIdx_t from_card = OtherRegionsTable::card_within_region(from, loc_hr); - add_card_work(from_card, par); + add_card(from_card); } } -inline void PerRegionTable::add_card(CardIdx_t from_card_index) { - add_card_work(from_card_index, /*parallel*/ true); -} - -inline void PerRegionTable::seq_add_card(CardIdx_t from_card_index) { - add_card_work(from_card_index, /*parallel*/ false); -} - -inline void PerRegionTable::add_reference(OopOrNarrowOopStar from) { - add_reference_work(from, /*parallel*/ true); -} - -inline void PerRegionTable::seq_add_reference(OopOrNarrowOopStar from) { - add_reference_work(from, /*parallel*/ false); -} - inline void PerRegionTable::init(HeapRegion* hr, bool clear_links_to_all_list) { if (clear_links_to_all_list) { set_next(NULL);
--- a/src/hotspot/share/gc/shared/c2/barrierSetC2.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shared/c2/barrierSetC2.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -264,7 +264,7 @@ virtual void clone_at_expansion(PhaseMacroExpand* phase, ArrayCopyNode* ac) const; // Support for GC barriers emitted during parsing - virtual bool has_load_barriers() const { return false; } + virtual bool has_load_barrier_nodes() const { return false; } virtual bool is_gc_barrier_node(Node* node) const { return false; } virtual Node* step_over_gc_barrier(Node* c) const { return c; } virtual Node* step_over_gc_barrier_ctrl(Node* c) const { return c; } @@ -287,13 +287,9 @@ virtual bool is_gc_specific_loop_opts_pass(LoopOptsMode mode) const { return false; } virtual bool has_special_unique_user(const Node* node) const { return false; } - virtual bool needs_anti_dependence_check(const Node* node) const { return true; } - - virtual void barrier_insertion_phase(Compile* C, PhaseIterGVN &igvn) const { } enum CompilePhase { BeforeOptimize, - BeforeLateInsertion, BeforeMacroExpand, BeforeCodeGen }; @@ -320,6 +316,10 @@ virtual Node* split_if_pre(PhaseIdealLoop* phase, Node* n) const { return NULL; } virtual bool build_loop_late_post(PhaseIdealLoop* phase, Node* n) const { return false; } virtual bool sink_node(PhaseIdealLoop* phase, Node* n, Node* x, Node* x_ctrl, Node* n_ctrl) const { return false; } + + virtual void late_barrier_analysis() const { } + virtual int estimate_stub_size() const { return 0; } + virtual void emit_stubs(CodeBuffer& cb) const { } }; #endif // SHARE_GC_SHARED_C2_BARRIERSETC2_HPP
--- a/src/hotspot/share/gc/shenandoah/c2/shenandoahBarrierSetC2.hpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/c2/shenandoahBarrierSetC2.hpp Mon Oct 14 21:01:25 2019 +0000 @@ -103,7 +103,7 @@ static const TypeFunc* write_ref_field_pre_entry_Type(); static const TypeFunc* shenandoah_clone_barrier_Type(); static const TypeFunc* shenandoah_load_reference_barrier_Type(); - virtual bool has_load_barriers() const { return true; } + virtual bool has_load_barrier_nodes() const { return true; } // This is the entry-point for the backend to perform accesses through the Access API. virtual void clone_at_expansion(PhaseMacroExpand* phase, ArrayCopyNode* ac) const;
--- a/src/hotspot/share/gc/shenandoah/heuristics/shenandoahAdaptiveHeuristics.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/heuristics/shenandoahAdaptiveHeuristics.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -66,9 +66,12 @@ size_t min_garbage = free_target > actual_free ? (free_target - actual_free) : 0; size_t max_cset = (size_t)((1.0 * capacity / 100 * ShenandoahEvacReserve) / ShenandoahEvacWaste); - log_info(gc, ergo)("Adaptive CSet Selection. Target Free: " SIZE_FORMAT "M, Actual Free: " - SIZE_FORMAT "M, Max CSet: " SIZE_FORMAT "M, Min Garbage: " SIZE_FORMAT "M", - free_target / M, actual_free / M, max_cset / M, min_garbage / M); + log_info(gc, ergo)("Adaptive CSet Selection. Target Free: " SIZE_FORMAT "%s, Actual Free: " + SIZE_FORMAT "%s, Max CSet: " SIZE_FORMAT "%s, Min Garbage: " SIZE_FORMAT "%s", + byte_size_in_proper_unit(free_target), proper_unit_for_byte_size(free_target), + byte_size_in_proper_unit(actual_free), proper_unit_for_byte_size(actual_free), + byte_size_in_proper_unit(max_cset), proper_unit_for_byte_size(max_cset), + byte_size_in_proper_unit(min_garbage), proper_unit_for_byte_size(min_garbage)); // Better select garbage-first regions QuickSort::sort<RegionData>(data, (int)size, compare_by_garbage, false); @@ -119,8 +122,9 @@ // anything else. size_t min_threshold = capacity / 100 * ShenandoahMinFreeThreshold; if (available < min_threshold) { - log_info(gc)("Trigger: Free (" SIZE_FORMAT "M) is below minimum threshold (" SIZE_FORMAT "M)", - available / M, min_threshold / M); + log_info(gc)("Trigger: Free (" SIZE_FORMAT "%s) is below minimum threshold (" SIZE_FORMAT "%s)", + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(min_threshold), proper_unit_for_byte_size(min_threshold)); return true; } @@ -129,8 +133,10 @@ if (_gc_times_learned < max_learn) { size_t init_threshold = capacity / 100 * ShenandoahInitFreeThreshold; if (available < init_threshold) { - log_info(gc)("Trigger: Learning " SIZE_FORMAT " of " SIZE_FORMAT ". Free (" SIZE_FORMAT "M) is below initial threshold (" SIZE_FORMAT "M)", - _gc_times_learned + 1, max_learn, available / M, init_threshold / M); + log_info(gc)("Trigger: Learning " SIZE_FORMAT " of " SIZE_FORMAT ". Free (" SIZE_FORMAT "%s) is below initial threshold (" SIZE_FORMAT "%s)", + _gc_times_learned + 1, max_learn, + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(init_threshold), proper_unit_for_byte_size(init_threshold)); return true; } } @@ -154,10 +160,15 @@ double allocation_rate = heap->bytes_allocated_since_gc_start() / time_since_last; if (average_gc > allocation_headroom / allocation_rate) { - log_info(gc)("Trigger: Average GC time (%.2f ms) is above the time for allocation rate (%.2f MB/s) to deplete free headroom (" SIZE_FORMAT "M)", - average_gc * 1000, allocation_rate / M, allocation_headroom / M); - log_info(gc, ergo)("Free headroom: " SIZE_FORMAT "M (free) - " SIZE_FORMAT "M (spike) - " SIZE_FORMAT "M (penalties) = " SIZE_FORMAT "M", - available / M, spike_headroom / M, penalties / M, allocation_headroom / M); + log_info(gc)("Trigger: Average GC time (%.2f ms) is above the time for allocation rate (%.0f %sB/s) to deplete free headroom (" SIZE_FORMAT "%s)", + average_gc * 1000, + byte_size_in_proper_unit(allocation_rate), proper_unit_for_byte_size(allocation_rate), + byte_size_in_proper_unit(allocation_headroom), proper_unit_for_byte_size(allocation_headroom)); + log_info(gc, ergo)("Free headroom: " SIZE_FORMAT "%s (free) - " SIZE_FORMAT "%s (spike) - " SIZE_FORMAT "%s (penalties) = " SIZE_FORMAT "%s", + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(spike_headroom), proper_unit_for_byte_size(spike_headroom), + byte_size_in_proper_unit(penalties), proper_unit_for_byte_size(penalties), + byte_size_in_proper_unit(allocation_headroom), proper_unit_for_byte_size(allocation_headroom)); return true; }
--- a/src/hotspot/share/gc/shenandoah/heuristics/shenandoahCompactHeuristics.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/heuristics/shenandoahCompactHeuristics.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -59,21 +59,24 @@ size_t min_threshold = capacity / 100 * ShenandoahMinFreeThreshold; if (available < min_threshold) { - log_info(gc)("Trigger: Free (" SIZE_FORMAT "M) is below minimum threshold (" SIZE_FORMAT "M)", - available / M, min_threshold / M); + log_info(gc)("Trigger: Free (" SIZE_FORMAT "%s) is below minimum threshold (" SIZE_FORMAT "%s)", + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(min_threshold), proper_unit_for_byte_size(min_threshold)); return true; } if (available < threshold_bytes_allocated) { - log_info(gc)("Trigger: Free (" SIZE_FORMAT "M) is lower than allocated recently (" SIZE_FORMAT "M)", - available / M, threshold_bytes_allocated / M); + log_info(gc)("Trigger: Free (" SIZE_FORMAT "%s) is lower than allocated recently (" SIZE_FORMAT "%s)", + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(threshold_bytes_allocated), proper_unit_for_byte_size(threshold_bytes_allocated)); return true; } size_t bytes_allocated = heap->bytes_allocated_since_gc_start(); if (bytes_allocated > threshold_bytes_allocated) { - log_info(gc)("Trigger: Allocated since last cycle (" SIZE_FORMAT "M) is larger than allocation threshold (" SIZE_FORMAT "M)", - bytes_allocated / M, threshold_bytes_allocated / M); + log_info(gc)("Trigger: Allocated since last cycle (" SIZE_FORMAT "%s) is larger than allocation threshold (" SIZE_FORMAT "%s)", + byte_size_in_proper_unit(bytes_allocated), proper_unit_for_byte_size(bytes_allocated), + byte_size_in_proper_unit(threshold_bytes_allocated), proper_unit_for_byte_size(threshold_bytes_allocated)); return true; } @@ -86,8 +89,9 @@ // Do not select too large CSet that would overflow the available free space size_t max_cset = actual_free * 3 / 4; - log_info(gc, ergo)("CSet Selection. Actual Free: " SIZE_FORMAT "M, Max CSet: " SIZE_FORMAT "M", - actual_free / M, max_cset / M); + log_info(gc, ergo)("CSet Selection. Actual Free: " SIZE_FORMAT "%s, Max CSet: " SIZE_FORMAT "%s", + byte_size_in_proper_unit(actual_free), proper_unit_for_byte_size(actual_free), + byte_size_in_proper_unit(max_cset), proper_unit_for_byte_size(max_cset)); size_t threshold = ShenandoahHeapRegion::region_size_bytes() * ShenandoahGarbageThreshold / 100;
--- a/src/hotspot/share/gc/shenandoah/heuristics/shenandoahPassiveHeuristics.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/heuristics/shenandoahPassiveHeuristics.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -60,8 +60,9 @@ size_t available = MAX2(capacity / 100 * ShenandoahEvacReserve, actual_free); size_t max_cset = (size_t)(available / ShenandoahEvacWaste); - log_info(gc, ergo)("CSet Selection. Actual Free: " SIZE_FORMAT "M, Max CSet: " SIZE_FORMAT "M", - actual_free / M, max_cset / M); + log_info(gc, ergo)("CSet Selection. Actual Free: " SIZE_FORMAT "%s, Max CSet: " SIZE_FORMAT "%s", + byte_size_in_proper_unit(actual_free), proper_unit_for_byte_size(actual_free), + byte_size_in_proper_unit(max_cset), proper_unit_for_byte_size(max_cset)); size_t threshold = ShenandoahHeapRegion::region_size_bytes() * ShenandoahGarbageThreshold / 100;
--- a/src/hotspot/share/gc/shenandoah/heuristics/shenandoahStaticHeuristics.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/heuristics/shenandoahStaticHeuristics.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -57,8 +57,9 @@ size_t threshold_available = capacity / 100 * ShenandoahFreeThreshold; if (available < threshold_available) { - log_info(gc)("Trigger: Free (" SIZE_FORMAT "M) is below free threshold (" SIZE_FORMAT "M)", - available / M, threshold_available / M); + log_info(gc)("Trigger: Free (" SIZE_FORMAT "%s) is below free threshold (" SIZE_FORMAT "%s)", + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(threshold_available), proper_unit_for_byte_size(threshold_available)); return true; } return ShenandoahHeuristics::should_start_gc();
--- a/src/hotspot/share/gc/shenandoah/heuristics/shenandoahTraversalHeuristics.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/heuristics/shenandoahTraversalHeuristics.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -100,9 +100,12 @@ size_t min_garbage = free_target > actual_free ? (free_target - actual_free) : 0; size_t max_cset = (size_t)((1.0 * capacity / 100 * ShenandoahEvacReserve) / ShenandoahEvacWaste); - log_info(gc, ergo)("Adaptive CSet Selection. Target Free: " SIZE_FORMAT "M, Actual Free: " - SIZE_FORMAT "M, Max CSet: " SIZE_FORMAT "M, Min Garbage: " SIZE_FORMAT "M", - free_target / M, actual_free / M, max_cset / M, min_garbage / M); + log_info(gc, ergo)("Adaptive CSet Selection. Target Free: " SIZE_FORMAT "%s, Actual Free: " + SIZE_FORMAT "%s, Max CSet: " SIZE_FORMAT "%s, Min Garbage: " SIZE_FORMAT "%s", + byte_size_in_proper_unit(free_target), proper_unit_for_byte_size(free_target), + byte_size_in_proper_unit(actual_free), proper_unit_for_byte_size(actual_free), + byte_size_in_proper_unit(max_cset), proper_unit_for_byte_size(max_cset), + byte_size_in_proper_unit(min_garbage), proper_unit_for_byte_size(min_garbage)); // Better select garbage-first regions, and then older ones QuickSort::sort<RegionData>(data, (int) cnt, compare_by_garbage_then_alloc_seq_ascending, false); @@ -190,8 +193,9 @@ // anything else. size_t min_threshold = capacity / 100 * ShenandoahMinFreeThreshold; if (available < min_threshold) { - log_info(gc)("Trigger: Free (" SIZE_FORMAT "M) is below minimum threshold (" SIZE_FORMAT "M)", - available / M, min_threshold / M); + log_info(gc)("Trigger: Free (" SIZE_FORMAT "%s) is below minimum threshold (" SIZE_FORMAT "%s)", + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(min_threshold), proper_unit_for_byte_size(min_threshold)); return true; } @@ -200,8 +204,10 @@ if (_gc_times_learned < max_learn) { size_t init_threshold = capacity / 100 * ShenandoahInitFreeThreshold; if (available < init_threshold) { - log_info(gc)("Trigger: Learning " SIZE_FORMAT " of " SIZE_FORMAT ". Free (" SIZE_FORMAT "M) is below initial threshold (" SIZE_FORMAT "M)", - _gc_times_learned + 1, max_learn, available / M, init_threshold / M); + log_info(gc)("Trigger: Learning " SIZE_FORMAT " of " SIZE_FORMAT ". Free (" SIZE_FORMAT "%s) is below initial threshold (" SIZE_FORMAT "%s)", + _gc_times_learned + 1, max_learn, + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(init_threshold), proper_unit_for_byte_size(init_threshold)); return true; } } @@ -223,10 +229,15 @@ double allocation_rate = heap->bytes_allocated_since_gc_start() / time_since_last; if (average_gc > allocation_headroom / allocation_rate) { - log_info(gc)("Trigger: Average GC time (%.2f ms) is above the time for allocation rate (%.2f MB/s) to deplete free headroom (" SIZE_FORMAT "M)", - average_gc * 1000, allocation_rate / M, allocation_headroom / M); - log_info(gc, ergo)("Free headroom: " SIZE_FORMAT "M (free) - " SIZE_FORMAT "M (spike) - " SIZE_FORMAT "M (penalties) = " SIZE_FORMAT "M", - available / M, spike_headroom / M, penalties / M, allocation_headroom / M); + log_info(gc)("Trigger: Average GC time (%.2f ms) is above the time for allocation rate (%.0f %sB/s) to deplete free headroom (" SIZE_FORMAT "%s)", + average_gc * 1000, + byte_size_in_proper_unit(allocation_rate), proper_unit_for_byte_size(allocation_rate), + byte_size_in_proper_unit(allocation_headroom), proper_unit_for_byte_size(allocation_headroom)); + log_info(gc, ergo)("Free headroom: " SIZE_FORMAT "%s (free) - " SIZE_FORMAT "%s (spike) - " SIZE_FORMAT "%s (penalties) = " SIZE_FORMAT "%s", + byte_size_in_proper_unit(available), proper_unit_for_byte_size(available), + byte_size_in_proper_unit(spike_headroom), proper_unit_for_byte_size(spike_headroom), + byte_size_in_proper_unit(penalties), proper_unit_for_byte_size(penalties), + byte_size_in_proper_unit(allocation_headroom), proper_unit_for_byte_size(allocation_headroom)); return true; } else if (ShenandoahHeuristics::should_start_gc()) { return true;
--- a/src/hotspot/share/gc/shenandoah/shenandoahFreeSet.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/shenandoahFreeSet.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -491,8 +491,12 @@ size_t max_humongous = max_contig * ShenandoahHeapRegion::region_size_bytes(); size_t free = capacity() - used(); - ls.print("Free: " SIZE_FORMAT "M (" SIZE_FORMAT " regions), Max regular: " SIZE_FORMAT "K, Max humongous: " SIZE_FORMAT "K, ", - total_free / M, mutator_count(), max / K, max_humongous / K); + ls.print("Free: " SIZE_FORMAT "%s (" SIZE_FORMAT " regions), Max regular: " SIZE_FORMAT "%s, Max humongous: " SIZE_FORMAT "%s, ", + byte_size_in_proper_unit(total_free), proper_unit_for_byte_size(total_free), + mutator_count(), + byte_size_in_proper_unit(max), proper_unit_for_byte_size(max), + byte_size_in_proper_unit(max_humongous), proper_unit_for_byte_size(max_humongous) + ); size_t frag_ext; if (free > 0) { @@ -525,8 +529,10 @@ } } - ls.print_cr("Evacuation Reserve: " SIZE_FORMAT "M (" SIZE_FORMAT " regions), Max regular: " SIZE_FORMAT "K", - total_free / M, collector_count(), max / K); + ls.print_cr("Evacuation Reserve: " SIZE_FORMAT "%s (" SIZE_FORMAT " regions), Max regular: " SIZE_FORMAT "%s", + byte_size_in_proper_unit(total_free), proper_unit_for_byte_size(total_free), + collector_count(), + byte_size_in_proper_unit(max), proper_unit_for_byte_size(max)); } } }
--- a/src/hotspot/share/gc/shenandoah/shenandoahHeap.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/shenandoahHeap.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -522,10 +522,14 @@ void ShenandoahHeap::print_on(outputStream* st) const { st->print_cr("Shenandoah Heap"); - st->print_cr(" " SIZE_FORMAT "K total, " SIZE_FORMAT "K committed, " SIZE_FORMAT "K used", - max_capacity() / K, committed() / K, used() / K); - st->print_cr(" " SIZE_FORMAT " x " SIZE_FORMAT"K regions", - num_regions(), ShenandoahHeapRegion::region_size_bytes() / K); + st->print_cr(" " SIZE_FORMAT "%s total, " SIZE_FORMAT "%s committed, " SIZE_FORMAT "%s used", + byte_size_in_proper_unit(max_capacity()), proper_unit_for_byte_size(max_capacity()), + byte_size_in_proper_unit(committed()), proper_unit_for_byte_size(committed()), + byte_size_in_proper_unit(used()), proper_unit_for_byte_size(used())); + st->print_cr(" " SIZE_FORMAT " x " SIZE_FORMAT"%s regions", + num_regions(), + byte_size_in_proper_unit(ShenandoahHeapRegion::region_size_bytes()), + proper_unit_for_byte_size(ShenandoahHeapRegion::region_size_bytes())); st->print("Status: "); if (has_forwarded_objects()) st->print("has forwarded objects, "); @@ -959,7 +963,8 @@ ShenandoahConcurrentEvacuateRegionObjectClosure cl(_sh); ShenandoahHeapRegion* r; while ((r =_cs->claim_next()) != NULL) { - assert(r->has_live(), "all-garbage regions are reclaimed early"); + assert(r->has_live(), "Region " SIZE_FORMAT " should have been reclaimed early", r->region_number()); + assert(r->is_conc_move_allowed(), "Region " SIZE_FORMAT " should be movable", r->region_number()); _sh->marked_object_iterate(r, &cl); if (ShenandoahPacing) {
--- a/src/hotspot/share/gc/shenandoah/shenandoahHeapRegion.cpp Fri Oct 04 13:44:11 2019 -0700 +++ b/src/hotspot/share/gc/shenandoah/shenandoahHeapRegion.cpp Mon Oct 14 21:01:25 2019 +0000 @@ -527,29 +527,35 @@ size_t region_size; if (FLAG_IS_DEFAULT(ShenandoahHeapRegionSize)) { if (ShenandoahMinRegionSize > max_heap_size / MIN_NUM_REGIONS) { - err_msg message("Max heap size (" SIZE_FORMAT "K) is too low to afford the minimum number " - "of regions (" SIZE_FORMAT ") of minimum region size (" SIZE_FORMAT "K).", - max_heap_size/K, MIN_NUM_REGIONS, ShenandoahMinRegionSize/K); + err_msg message("Max heap size (" SIZE_FORMAT "%s) is too low to afford the minimum number " + "of regions (" SIZE_FORMAT ") of minimum region size (" SIZE_FORMAT "%s).", + byte_size_in_proper_unit(max_heap_size), proper_unit_for_byte_size(max_heap_size), + MIN_NUM_REGIONS, + byte_size_in_proper_unit(ShenandoahMinRegionSize), proper_unit_for_byte_size(ShenandoahMinRegionSize)); vm_exit_during_initialization("Invalid -XX:ShenandoahMinRegionSize option", message); } if (ShenandoahMinRegionSize < MIN_REGION_SIZE) { - err_msg message("" SIZE_FORMAT "K should not be lower than minimum region size (" SIZE_FORMAT "K).", - ShenandoahMinRegionSize/K, MIN_REGION_SIZE/K); + err_msg message("" SIZE_FORMAT "%s should not be lower than minimum region size (" SIZE_FORMAT "%s).", + byte_size_in_proper_unit(ShenandoahMinRegionSize), proper_unit_for_byte_size(ShenandoahMinRegionSize), + byte_size_in_proper_unit(MIN_REGION_SIZE), proper_unit_for_byte_size(MIN_REGION_SIZE)); vm_exit_during_initialization("Invalid -XX:ShenandoahMinRegionSize option", message); } if (ShenandoahMinRegionSize < MinTLABSize) { - err_msg message("" SIZE_FORMAT "K should not be lower than TLAB size size (" SIZE_FORMAT "K).", - ShenandoahMinRegionSize/K, MinTLABSize/K); + err_msg message("" SIZE_FORMAT "%s should not be lower than TLAB size size (" SIZE_FORMAT "%s).", + byte_size_in_proper_unit(ShenandoahMinRegionSize), proper_unit_for_byte_size(ShenandoahMinRegionSize), + byte_size_in_proper_unit(MinTLABSize), proper_unit_for_byte_size(MinTLABSize)); vm_exit_during_initialization("Invalid -XX:ShenandoahMinRegionSize option", message); } if (ShenandoahMaxRegionSize < MIN_REGION_SIZE) { - err_msg message("" SIZE_FORMAT "K should not be lower than min region size (" SIZE_FORMAT "K).", - ShenandoahMaxRegionSize/K, MIN_REGION_SIZE/K); + err_msg message("" SIZE_FORMAT "%s should not be lower than min region size (" SIZE_FORMAT "%s).", + byte_size_in_proper_unit(ShenandoahMaxRegionSize), proper_unit_for_byte_size(ShenandoahMaxRegionSize), + byte_size_in_proper_unit(MIN_REGION_SIZE), proper_unit_for_byte_size(MIN_REGION_SIZE)); vm_exit_during_initialization("Invalid -XX:ShenandoahMaxRegionSize option", message); } if (ShenandoahMinRegionSize > ShenandoahMaxRegionSize) { - err_msg message("Minimum (" SIZE_FORMAT "K) should be larger than maximum (" SIZE_FORMAT "K).", - ShenandoahMinRegionSize/K, ShenandoahMaxRegionSize/K); + err_msg message("Minimum (" SIZE_FORMAT "%s) should be larger than maximum (" SIZE_FORMAT "%s).", + byte_size_in_proper_unit(ShenandoahMinRegionSize), proper_unit_for_byte_size(ShenandoahMinRegionSize), + byte_size_in_proper_unit(ShenandoahMaxRegionSize), proper_unit_for_byte_size(ShenandoahMaxRegionSize)); vm_exit_during_initialization("Invalid -XX:ShenandoahMinRegionSize or -XX:ShenandoahMaxRegionSize", message); } @@ -563,19 +569,23 @@ } else { if (ShenandoahHeapRegionSize > max_heap_size / MIN_NUM_REGIONS) { - err_msg message("Max heap size (" SIZE_FORMAT "K) is too low to afford the minimum number " - "of regions (" SIZE_FORMAT ") of requested size (" SIZE_FORMAT "K).", - max_heap_size/K, MIN_NUM_REGIONS, ShenandoahHeapRegionSize/K); + err_msg message("Max heap size (" SIZE_FORMAT "%s) is too low to afford the minimum number " + "of regions (" SIZE_FORMAT ") of requested size (" SIZE_FORMAT "%s).", + byte_size_in_proper_unit(max_heap_size), proper_unit_for_byte_size(max_heap_size), + MIN_NUM_REGIONS, + byte_size_in_proper_unit(ShenandoahHeapRegionSize), proper_unit_for_byte_size(ShenandoahHeapRegionSize)); vm_exit_during_initialization("Invalid -XX:ShenandoahHeapRegionSize option", message); } if (ShenandoahHeapRegionSize < ShenandoahMinRegionSize) { - err_msg message("Heap region size (" SIZE_FORMAT "K) should be larger than min region size (" SIZE_FORMAT "K).", - ShenandoahHeapRegionSize/K, ShenandoahMinRegionSize/K); + err_msg message("Heap region size (" SIZE_FORMAT "%s) should be larger than min region size (" SIZE_FORMAT "%s).", + byte_size_in_proper_unit(ShenandoahHeapRegionSize), proper_unit_for_byte_size(ShenandoahHeapRegionSize), + byte_size_in_proper_unit(ShenandoahMinRegionSize), proper_unit_for_byte_size(ShenandoahMinRegionSize)); vm_exit_during_initialization("Invalid -XX:ShenandoahHeapRegionSize option", message); } if (ShenandoahHeapRegionSize > ShenandoahMaxRegionSize) { - err_msg message("Heap region size (" SIZE_FORMAT "K) should be lower than max region size (" SIZE_FORMAT "K).", - ShenandoahHeapRegionSize/K, ShenandoahMaxRegionSize/K); + err_msg message("Heap region size (" SIZE_FORMAT "%s) should be lower than max region size (" SIZE_FORMAT "%s).",