Mercurial > hg > truffle
view make/solaris/makefiles/sparcWorks.make @ 453:c96030fff130
6684579: SoftReference processing can be made more efficient
Summary: For current soft-ref clearing policies, we can decide at marking time if a soft-reference will definitely not be cleared, postponing the decision of whether it will definitely be cleared to the final reference processing phase. This can be especially beneficial in the case of concurrent collectors where the marking is usually concurrent but reference processing is usually not.
Reviewed-by: jmasa
author | ysr |
---|---|
date | Thu, 20 Nov 2008 16:56:09 -0800 |
parents | 4d20a3aaf1ab |
children | 2bf529ef0adb |
line wrap: on
line source
# # Copyright 1998-2008 Sun Microsystems, Inc. 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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, # CA 95054 USA or visit www.sun.com if you need additional information or # have any questions. # # # Compiler-specific flags for sparcworks. # tell make which C and C++ compilers to use CC = cc CPP = CC # Note that this 'as' is an older version of the Sun Studio 'fbe', and will # use the older style options. The 'fbe' options will match 'cc' and 'CC'. AS = /usr/ccs/bin/as NM = /usr/ccs/bin/nm NAWK = /bin/nawk REORDER_FLAG = -xF # Check for the versions of C++ and C compilers ($CPP and $CC) used. # Get the last thing on the line that looks like x.x+ (x is a digit). COMPILER_REV := \ $(shell $(CPP) -V 2>&1 | sed -n 's/^.*[ ,\t]C++[ ,\t]\([1-9]\.[0-9][0-9]*\).*/\1/p') C_COMPILER_REV := \ $(shell $(CC) -V 2>&1 | sed -n 's/^.*[ ,\t]C[ ,\t]\([1-9]\.[0-9][0-9]*\).*/\1/p') # Pick which compiler is validated ifeq ($(JDK_MINOR_VERSION),6) # Validated compiler for JDK6 is SS11 (5.8) VALIDATED_COMPILER_REV := 5.8 VALIDATED_C_COMPILER_REV := 5.8 else # Validated compiler for JDK7 is SS12 (5.9) VALIDATED_COMPILER_REV := 5.9 VALIDATED_C_COMPILER_REV := 5.9 endif # Warning messages about not using the above validated version ENFORCE_COMPILER_REV${ENFORCE_COMPILER_REV} := ${VALIDATED_COMPILER_REV} ifneq (${COMPILER_REV},${ENFORCE_COMPILER_REV}) dummy_target_to_enforce_compiler_rev:=\ $(shell echo >&2 WARNING: You are using CC version ${COMPILER_REV} \ and should be using version ${ENFORCE_COMPILER_REV}. Set ENFORCE_COMPILER_REV=${COMPILER_REV} to avoid this warning.) endif ENFORCE_C_COMPILER_REV${ENFORCE_C_COMPILER_REV} := ${VALIDATED_C_COMPILER_REV} ifneq (${C_COMPILER_REV},${ENFORCE_C_COMPILER_REV}) dummy_target_to_enforce_c_compiler_rev:=\ $(shell echo >&2 WARNING: You are using cc version ${C_COMPILER_REV} \ and should be using version ${ENFORCE_C_COMPILER_REV}. Set ENFORCE_C_COMPILER_REV=${C_COMPILER_REV} to avoid this warning.) endif COMPILER_REV_NUMERIC := $(shell echo $(COMPILER_REV) | awk -F. '{ print $$1 * 100 + $$2 }') # Fail the build if __fabsf is used. __fabsf exists only in Solaris 8 2/04 # and newer; objects with a dependency on this symbol will not run on older # Solaris 8. JVM_FAIL_IF_UNDEFINED = __fabsf JVM_CHECK_SYMBOLS = $(NM) -u -p $(LIBJVM.o) | \ $(NAWK) -v f="${JVM_FAIL_IF_UNDEFINED}" \ 'BEGIN { c=split(f,s); rc=0; } \ /:$$/ { file = $$1; } \ /[^:]$$/ { for(n=1;n<=c;++n) { \ if($$1==s[n]) { \ printf("JVM_CHECK_SYMBOLS: %s contains illegal symbol %s\n", \ file,$$1); \ rc=1; \ } \ } \ } \ END { exit rc; }' LINK_LIB.CC/PRE_HOOK += $(JVM_CHECK_SYMBOLS) || exit 1; # Some interfaces (_lwp_create) changed with LP64 and Solaris 7 SOLARIS_7_OR_LATER := \ $(shell uname -r | awk -F. '{ if ($$2 >= 7) print "-DSOLARIS_7_OR_LATER"; }') CFLAGS += ${SOLARIS_7_OR_LATER} # New architecture options started in SS12 (5.9), we need both styles to build. # The older arch options for SS11 (5.8) or older and also for /usr/ccs/bin/as. # Note: SS12 default for 32bit sparc is now the same as v8plus, so the # settings below have changed all SS12 32bit sparc builds to be v8plus. # The older SS11 (5.8) settings have remained as they always have been. ifeq ($(TYPE),COMPILER2) ARCHFLAG_OLD/sparc = -xarch=v8plus else ifeq ($(TYPE),TIERED) ARCHFLAG_OLD/sparc = -xarch=v8plus else ARCHFLAG_OLD/sparc = -xarch=v8 endif endif ARCHFLAG_NEW/sparc = -m32 -xarch=sparc ARCHFLAG_OLD/sparcv9 = -xarch=v9 ARCHFLAG_NEW/sparcv9 = -m64 -xarch=sparc ARCHFLAG_OLD/i486 = ARCHFLAG_NEW/i486 = -m32 ARCHFLAG_OLD/amd64 = -xarch=amd64 ARCHFLAG_NEW/amd64 = -m64 # Select the ARCHFLAGs and other SS12 (5.9) options ifeq ($(shell expr $(COMPILER_REV_NUMERIC) \>= 509), 1) ARCHFLAG/sparc = $(ARCHFLAG_NEW/sparc) ARCHFLAG/sparcv9 = $(ARCHFLAG_NEW/sparcv9) ARCHFLAG/i486 = $(ARCHFLAG_NEW/i486) ARCHFLAG/amd64 = $(ARCHFLAG_NEW/amd64) else ARCHFLAG/sparc = $(ARCHFLAG_OLD/sparc) ARCHFLAG/sparcv9 = $(ARCHFLAG_OLD/sparcv9) ARCHFLAG/i486 = $(ARCHFLAG_OLD/i486) ARCHFLAG/amd64 = $(ARCHFLAG_OLD/amd64) endif # ARCHFLAGS for the current build arch ARCHFLAG = $(ARCHFLAG/$(BUILDARCH)) AS_ARCHFLAG = $(ARCHFLAG_OLD/$(BUILDARCH)) # Optional sub-directory in /usr/lib where BUILDARCH libraries are kept. ISA_DIR=$(ISA_DIR/$(BUILDARCH)) ISA_DIR/sparcv9=/sparcv9 ISA_DIR/amd64=/amd64 # Use these to work around compiler bugs: OPT_CFLAGS/SLOWER=-xO3 OPT_CFLAGS/O2=-xO2 OPT_CFLAGS/NOOPT=-xO1 ################################################# # Begin current (>=5.6) Forte compiler options # ################################################# ifeq ($(shell expr $(COMPILER_REV_NUMERIC) \>= 506), 1) ifeq ("${Platform_arch}", "sparc") # We MUST allow data alignment of 4 for sparc (sparcv9 is ok at 8s) ifndef LP64 CFLAGS += -xmemalign=4s endif endif endif ################################################# # Begin current (>=5.5) Forte compiler options # ################################################# ifeq ($(shell expr $(COMPILER_REV_NUMERIC) \>= 505), 1) CFLAGS += $(ARCHFLAG) AOUT_FLAGS += $(ARCHFLAG) LIB_FLAGS += $(ARCHFLAG) LFLAGS += $(ARCHFLAG) ifeq ("${Platform_arch}", "sparc") # Flags for Optimization # [phh] Commented out pending verification that we do indeed want # to potentially bias against u1 and u3 targets. #CFLAGS += -xchip=ultra2 OPT_CFLAGS=-xO4 $(EXTRA_OPT_CFLAGS) endif # sparc ifeq ("${Platform_arch_model}", "x86_32") OPT_CFLAGS=-xtarget=pentium $(EXTRA_OPT_CFLAGS) # UBE (CC 5.5) has bug 4923569 with -xO4 OPT_CFLAGS+=-xO3 endif # 32bit x86 ifeq ("${Platform_arch_model}", "x86_64") ASFLAGS += $(AS_ARCHFLAG) CFLAGS += $(ARCHFLAG/amd64) # this one seemed useless LFLAGS_VM += $(ARCHFLAG/amd64) # this one worked LFLAGS += $(ARCHFLAG/amd64) AOUT_FLAGS += $(ARCHFLAG/amd64) # -xO3 is faster than -xO4 on specjbb with SS10 compiler OPT_CFLAGS=-xO4 $(EXTRA_OPT_CFLAGS) endif # 64bit x86 # Inline functions CFLAGS += $(GAMMADIR)/src/os_cpu/solaris_${Platform_arch}/vm/solaris_${Platform_arch_model}.il # no more exceptions CFLAGS/NOEX=-features=no%except # avoid compilation problems arising from fact that C++ compiler tries # to search for external template definition by just compiling additional # source files in th same context CFLAGS += -template=no%extdef # Reduce code bloat by reverting back to 5.0 behavior for static initializers CFLAGS += -features=no%split_init # Use -D_Crun_inline_placement so we don't get references to # __1c2n6FIpv_0_ or void*operator new(unsigned,void*) # This avoids the hard requirement of the newer Solaris C++ runtime patches. # NOTE: This is an undocumented feature of the SS10 compiler. See 6306698. CFLAGS += -D_Crun_inline_placement # PIC is safer for SPARC, and is considerably slower # a file foo.o which wants to compile -pic can set "PICFLAG/foo.o = -PIC" PICFLAG = -KPIC PICFLAG/DEFAULT = $(PICFLAG) # [RGV] Need to figure which files to remove to get link to work #PICFLAG/BETTER = -pic PICFLAG/BETTER = $(PICFLAG/DEFAULT) PICFLAG/BYFILE = $(PICFLAG/$@)$(PICFLAG/DEFAULT$(PICFLAG/$@)) # Use $(MAPFLAG:FILENAME=real_file_name) to specify a map file. MAPFLAG = -M FILENAME # Use $(SONAMEFLAG:SONAME=soname) to specify the intrinsic name of a shared obj SONAMEFLAG = -h SONAME # Build shared library SHARED_FLAG = -G # We don't need libCstd.so and librwtools7.so, only libCrun.so CFLAGS += -library=%none LFLAGS += -library=%none LFLAGS += -mt endif # COMPILER_REV_NUMERIC >= 505 ###################################### # End 5.5 Forte compiler options # ###################################### ###################################### # Begin 5.2 Forte compiler options # ###################################### ifeq ($(COMPILER_REV_NUMERIC), 502) CFLAGS += $(ARCHFLAG) AOUT_FLAGS += $(ARCHFLAG) LIB_FLAGS += $(ARCHFLAG) LFLAGS += $(ARCHFLAG) ifeq ("${Platform_arch}", "sparc") # Flags for Optimization # [phh] Commented out pending verification that we do indeed want # to potentially bias against u1 and u3 targets. #CFLAGS += -xchip=ultra2 ifdef LP64 # SC5.0 tools on v9 are flakey at -xO4 # [phh] Is this still true for 6.1? OPT_CFLAGS=-xO3 $(EXTRA_OPT_CFLAGS) else OPT_CFLAGS=-xO4 $(EXTRA_OPT_CFLAGS) endif CFLAGS += $(GAMMADIR)/src/os_cpu/solaris_sparc/vm/solaris_sparc.il endif # sparc ifeq ("${Platform_arch_model}", "x86_32") OPT_CFLAGS=-xtarget=pentium $(EXTRA_OPT_CFLAGS) # SC5.0 tools on x86 are flakey at -xO4 # [phh] Is this still true for 6.1? OPT_CFLAGS+=-xO3 CFLAGS += $(GAMMADIR)/src/os_cpu/solaris_x86/vm/solaris_x86_32.il endif # 32bit x86 # no more exceptions CFLAGS/NOEX=-noex # Reduce code bloat by reverting back to 5.0 behavior for static initializers CFLAGS += -Qoption ccfe -one_static_init # PIC is safer for SPARC, and is considerably slower # a file foo.o which wants to compile -pic can set "PICFLAG/foo.o = -PIC" PICFLAG = -KPIC PICFLAG/DEFAULT = $(PICFLAG) # [RGV] Need to figure which files to remove to get link to work #PICFLAG/BETTER = -pic PICFLAG/BETTER = $(PICFLAG/DEFAULT) PICFLAG/BYFILE = $(PICFLAG/$@)$(PICFLAG/DEFAULT$(PICFLAG/$@)) # Would be better if these weren't needed, since we link with CC, but # at present removing them causes run-time errors LFLAGS += -library=Crun LIBS += -library=Crun -lCrun endif # COMPILER_REV_NUMERIC == 502 ################################## # End 5.2 Forte compiler options # ################################## ################################## # Begin old 5.1 compiler options # ################################## ifeq ($(COMPILER_REV_NUMERIC), 501) _JUNK_ := $(shell echo >&2 \ "*** ERROR: sparkWorks.make incomplete for 5.1 compiler") @exit 1 endif ################################## # End old 5.1 compiler options # ################################## ################################## # Begin old 5.0 compiler options # ################################## ifeq (${COMPILER_REV_NUMERIC}, 500) # Had to hoist this higher apparently because of other changes. Must # come before -xarch specification. # NOTE: native says optimize for the machine doing the compile, bad news. CFLAGS += -xtarget=native CFLAGS += $(ARCHFLAG) AOUT_FLAGS += $(ARCHFLAG) LIB_FLAGS += $(ARCHFLAG) LFLAGS += $(ARCHFLAG) CFLAGS += -library=iostream LFLAGS += -library=iostream -library=Crun LIBS += -library=iostream -library=Crun -lCrun # Flags for Optimization ifdef LP64 # SC5.0 tools on v9 are flakey at -xO4 OPT_CFLAGS=-xO3 $(EXTRA_OPT_CFLAGS) else OPT_CFLAGS=-xO4 $(EXTRA_OPT_CFLAGS) endif ifeq ("${Platform_arch}", "sparc") CFLAGS += $(GAMMADIR)/src/os_cpu/solaris_sparc/vm/atomic_solaris_sparc.il endif # sparc ifeq ("${Platform_arch_model}", "x86_32") OPT_CFLAGS=-xtarget=pentium $(EXTRA_OPT_CFLAGS) ifeq ("${COMPILER_REV_NUMERIC}", "500") # SC5.0 tools on x86 are flakey at -xO4 OPT_CFLAGS+=-xO3 else OPT_CFLAGS+=-xO4 endif CFLAGS += $(GAMMADIR)/src/os_cpu/solaris_x86/vm/solaris_x86_32.il endif # 32bit x86 # The following options run into misaligned ldd problem (raj) #OPT_CFLAGS = -fast -O4 $(ARCHFLAG/sparc) -xchip=ultra # no more exceptions CFLAGS/NOEX=-noex # PIC is safer for SPARC, and is considerably slower # a file foo.o which wants to compile -pic can set "PICFLAG/foo.o = -PIC" PICFLAG = -PIC PICFLAG/DEFAULT = $(PICFLAG) # [RGV] Need to figure which files to remove to get link to work #PICFLAG/BETTER = -pic PICFLAG/BETTER = $(PICFLAG/DEFAULT) PICFLAG/BYFILE = $(PICFLAG/$@)$(PICFLAG/DEFAULT$(PICFLAG/$@)) endif # COMPILER_REV_NUMERIC = 500 ################################ # End old 5.0 compiler options # ################################ ifeq ("${COMPILER_REV_NUMERIC}", "402") # 4.2 COMPILERS SHOULD NO LONGER BE USED _JUNK_ := $(shell echo >&2 \ "*** ERROR: SC4.2 compilers are not supported by this code base!") @exit 1 endif # do not include shared lib path in a.outs AOUT_FLAGS += -norunpath LFLAGS_VM = -norunpath -z noversion # need position-indep-code for shared libraries # (ild appears to get errors on PIC code, so we'll try non-PIC for debug) ifeq ($(PICFLAGS),DEFAULT) VM_PICFLAG/LIBJVM = $(PICFLAG/DEFAULT) else VM_PICFLAG/LIBJVM = $(PICFLAG/BYFILE) endif VM_PICFLAG/AOUT = VM_PICFLAG = $(VM_PICFLAG/$(LINK_INTO)) CFLAGS += $(VM_PICFLAG) # less dynamic linking (no PLTs, please) #LIB_FLAGS += $(LINK_MODE) # %%%%% despite -znodefs, -Bsymbolic gets link errors -- Rose LINK_MODE = $(LINK_MODE/$(VERSION)) LINK_MODE/debug = LINK_MODE/optimized = -Bsymbolic -znodefs # Have thread local errnos ifeq ($(shell expr $(COMPILER_REV_NUMERIC) \>= 505), 1) CFLAGS += -mt else CFLAGS += -D_REENTRANT endif ifdef CC_INTERP # C++ Interpreter CFLAGS += -DCC_INTERP endif # Flags for Debugging DEBUG_CFLAGS = -g FASTDEBUG_CFLAGS = -g0 # The -g0 setting allows the C++ frontend to inline, which is a big win. # Special global options for SS12 ifeq ($(COMPILER_REV_NUMERIC),509) # There appears to be multiple issues with the new Dwarf2 debug format, so # we tell the compiler to use the older 'stabs' debug format all the time. # Note that this needs to be used in optimized compiles too to be 100%. # This is a workaround for SS12 (5.9) bug 6694600 CFLAGS += -xdebugformat=stabs endif # Enable the following CFLAGS additions if you need to compare the # built ELF objects. # # The -g option makes static data global and the "-Qoption ccfe # -xglobalstatic" option tells the compiler to not globalize static # data using a unique globalization prefix. Instead force the use of # a static globalization prefix based on the source filepath so the # objects from two identical compilations are the same. #DEBUG_CFLAGS += -Qoption ccfe -xglobalstatic #FASTDEBUG_CFLAGS += -Qoption ccfe -xglobalstatic ifeq (${COMPILER_REV_NUMERIC}, 502) COMPILER_DATE := $(shell $(CPP) -V 2>&1 | sed -n '/^.*[ ]C++[ ]\([1-9]\.[0-9][0-9]*\)/p' | awk '{ print $$NF; }') ifeq (${COMPILER_DATE}, 2001/01/31) # disable -g0 in fastdebug since SC6.1 dated 2001/01/31 seems to be buggy # use an innocuous value because it will get -g if it's empty FASTDEBUG_CFLAGS = -c endif endif # Uncomment or 'gmake CFLAGS_BROWSE=-sbfast' to get source browser information. # CFLAGS_BROWSE = -sbfast CFLAGS += $(CFLAGS_BROWSE) # ILD is gone as of SS11 (5.8), not supportted in SS10 (5.7) ifeq ($(shell expr $(COMPILER_REV_NUMERIC) \< 507), 1) # use ild when debugging (but when optimizing we want reproducible results) ILDFLAG = $(ILDFLAG/$(VERSION)) ILDFLAG/debug = -xildon ILDFLAG/optimized = AOUT_FLAGS += $(ILDFLAG) endif # Where to put the *.o files (a.out, or shared library)? LINK_INTO = $(LINK_INTO/$(VERSION)) LINK_INTO/debug = LIBJVM LINK_INTO/optimized = LIBJVM # We link the debug version into the a.out because: # 1. ild works on a.out but not shared libraries, and using ild # can cut rebuild times by 25% for small changes. (ILD is gone in SS11) # 2. dbx cannot gracefully set breakpoints in shared libraries # # apply this setting to link into the shared library even in the debug version: ifdef LP64 LINK_INTO = LIBJVM else #LINK_INTO = LIBJVM endif MCS = /usr/ccs/bin/mcs STRIP = /usr/ccs/bin/strip # Solaris platforms collect lots of redundant file-ident lines, # to the point of wasting a significant percentage of file space. # (The text is stored in ELF .comment sections, contributed by # all "#pragma ident" directives in header and source files.) # This command "compresses" the .comment sections simply by # removing repeated lines. The data can be extracted from # binaries in the field by using "mcs -p libjvm.so" or the older # command "what libjvm.so". LINK_LIB.CC/POST_HOOK += $(MCS) -c $@ || exit 1; # (The exit 1 is necessary to cause a build failure if the command fails and # multiple commands are strung together, and the final semicolon is necessary # since the hook must terminate itself as a valid command.) # Also, strip debug and line number information (worth about 1.7Mb). STRIP_LIB.CC/POST_HOOK = $(STRIP) -x $@ || exit 1; # STRIP_LIB.CC/POST_HOOK is incorporated into LINK_LIB.CC/POST_HOOK # in certain configurations, such as product.make. Other configurations, # such as debug.make, do not include the strip operation.