1 /*
  2  * Copyright (c) 2000, 2019, Oracle and/or its affiliates. All rights reserved.
  3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  4  *
  5  * This code is free software; you can redistribute it and/or modify it
  6  * under the terms of the GNU General Public License version 2 only, as
  7  * published by the Free Software Foundation.
  8  *
  9  * This code is distributed in the hope that it will be useful, but WITHOUT
 10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 12  * version 2 for more details (a copy is included in the LICENSE file that
 13  * accompanied this code).
 14  *
 15  * You should have received a copy of the GNU General Public License version
 16  * 2 along with this work; if not, write to the Free Software Foundation,
 17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 18  *
 19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 20  * or visit www.oracle.com if you need additional information or have any
 21  * questions.
 22  *
 23  */
 24 
 25 #ifndef CPU_X86_GLOBALS_X86_HPP
 26 #define CPU_X86_GLOBALS_X86_HPP
 27 
 28 #include "utilities/globalDefinitions.hpp"
 29 #include "utilities/macros.hpp"
 30 
 31 // Sets the default values for platform dependent flags used by the runtime system.
 32 // (see globals.hpp)
 33 
 34 define_pd_global(bool, ImplicitNullChecks,       true);  // Generate code for implicit null checks
 35 define_pd_global(bool, TrapBasedNullChecks,      false); // Not needed on x86.
 36 define_pd_global(bool, UncommonNullCast,         true);  // Uncommon-trap NULLs passed to check cast
 37 
 38 define_pd_global(uintx, CodeCacheSegmentSize,    64 TIERED_ONLY(+64)); // Tiered compilation has large code-entry alignment.
 39 // See 4827828 for this change. There is no globals_core_i486.hpp. I can't
 40 // assign a different value for C2 without touching a number of files. Use
 41 // #ifdef to minimize the change as it's late in Mantis. -- FIXME.
 42 // c1 doesn't have this problem because the fix to 4858033 assures us
 43 // the the vep is aligned at CodeEntryAlignment whereas c2 only aligns
 44 // the uep and the vep doesn't get real alignment but just slops on by
 45 // only assured that the entry instruction meets the 5 byte size requirement.
 46 #if COMPILER2_OR_JVMCI
 47 define_pd_global(intx, CodeEntryAlignment,       32);
 48 #else
 49 define_pd_global(intx, CodeEntryAlignment,       16);
 50 #endif // COMPILER2_OR_JVMCI
 51 define_pd_global(intx, OptoLoopAlignment,        16);
 52 define_pd_global(intx, InlineFrequencyCount,     100);
 53 define_pd_global(intx, InlineSmallCode,          1000);
 54 
 55 #define DEFAULT_STACK_YELLOW_PAGES (NOT_WINDOWS(2) WINDOWS_ONLY(3))
 56 #define DEFAULT_STACK_RED_PAGES (1)
 57 #define DEFAULT_STACK_RESERVED_PAGES (NOT_WINDOWS(1) WINDOWS_ONLY(0))
 58 
 59 #define MIN_STACK_YELLOW_PAGES DEFAULT_STACK_YELLOW_PAGES
 60 #define MIN_STACK_RED_PAGES DEFAULT_STACK_RED_PAGES
 61 #define MIN_STACK_RESERVED_PAGES (0)
 62 
 63 #ifdef _LP64
 64 // Java_java_net_SocketOutputStream_socketWrite0() uses a 64k buffer on the
 65 // stack if compiled for unix and LP64. To pass stack overflow tests we need
 66 // 20 shadow pages.
 67 #define DEFAULT_STACK_SHADOW_PAGES (NOT_WIN64(20) WIN64_ONLY(7) DEBUG_ONLY(+2))
 68 // For those clients that do not use write socket, we allow
 69 // the min range value to be below that of the default
 70 #define MIN_STACK_SHADOW_PAGES (NOT_WIN64(10) WIN64_ONLY(7) DEBUG_ONLY(+2))
 71 #else
 72 #define DEFAULT_STACK_SHADOW_PAGES (4 DEBUG_ONLY(+5))
 73 #define MIN_STACK_SHADOW_PAGES DEFAULT_STACK_SHADOW_PAGES
 74 #endif // _LP64
 75 
 76 define_pd_global(intx, StackYellowPages, DEFAULT_STACK_YELLOW_PAGES);
 77 define_pd_global(intx, StackRedPages, DEFAULT_STACK_RED_PAGES);
 78 define_pd_global(intx, StackShadowPages, DEFAULT_STACK_SHADOW_PAGES);
 79 define_pd_global(intx, StackReservedPages, DEFAULT_STACK_RESERVED_PAGES);
 80 
 81 define_pd_global(bool, RewriteBytecodes,     true);
 82 define_pd_global(bool, RewriteFrequentPairs, true);
 83 
 84 define_pd_global(uintx, TypeProfileLevel, 111);
 85 
 86 define_pd_global(bool, CompactStrings, true);
 87 
 88 define_pd_global(bool, PreserveFramePointer, false);
 89 
 90 define_pd_global(intx, InitArrayShortSize, 8*BytesPerLong);
 91 
 92 #if defined(_LP64) || defined(_WINDOWS)
 93 define_pd_global(bool, ThreadLocalHandshakes, true);
 94 #else
 95 // get_thread() is slow on linux 32 bit, therefore off by default
 96 define_pd_global(bool, ThreadLocalHandshakes, false);
 97 #endif
 98 
 99 #define ARCH_FLAGS(develop, \
100                    product, \
101                    diagnostic, \
102                    experimental, \
103                    notproduct, \
104                    range, \
105                    constraint, \
106                    writeable) \
107                                                                             \
108   develop(bool, IEEEPrecision, true,                                        \
109           "Enables IEEE precision (for INTEL only)")                        \
110                                                                             \
111   product(bool, UseStoreImmI16, true,                                       \
112           "Use store immediate 16-bits value instruction on x86")           \
113                                                                             \
114   product(intx, UseAVX, 3,                                                  \
115           "Highest supported AVX instructions set on x86/x64")              \
116           range(0, 99)                                                      \
117                                                                             \
118   product(bool, UseCLMUL, false,                                            \
119           "Control whether CLMUL instructions can be used on x86/x64")      \
120                                                                             \
121   diagnostic(bool, UseIncDec, true,                                         \
122           "Use INC, DEC instructions on x86")                               \
123                                                                             \
124   product(bool, UseNewLongLShift, false,                                    \
125           "Use optimized bitwise shift left")                               \
126                                                                             \
127   product(bool, UseAddressNop, false,                                       \
128           "Use '0F 1F [addr]' NOP instructions on x86 cpus")                \
129                                                                             \
130   product(bool, UseXmmLoadAndClearUpper, true,                              \
131           "Load low part of XMM register and clear upper part")             \
132                                                                             \
133   product(bool, UseXmmRegToRegMoveAll, false,                               \
134           "Copy all XMM register bits when moving value between registers") \
135                                                                             \
136   product(bool, UseXmmI2D, false,                                           \
137           "Use SSE2 CVTDQ2PD instruction to convert Integer to Double")     \
138                                                                             \
139   product(bool, UseXmmI2F, false,                                           \
140           "Use SSE2 CVTDQ2PS instruction to convert Integer to Float")      \
141                                                                             \
142   product(bool, UseUnalignedLoadStores, false,                              \
143           "Use SSE2 MOVDQU instruction for Arraycopy")                      \
144                                                                             \
145   product(bool, UseXMMForObjInit, false,                                    \
146           "Use XMM/YMM MOVDQU instruction for Object Initialization")       \
147                                                                             \
148   product(bool, UseFastStosb, false,                                        \
149           "Use fast-string operation for zeroing: rep stosb")               \
150                                                                             \
151   /* Use Restricted Transactional Memory for lock eliding */                \
152   product(bool, UseRTMLocking, false,                                       \
153           "Enable RTM lock eliding for inflated locks in compiled code")    \
154                                                                             \
155   experimental(bool, UseRTMForStackLocks, false,                            \
156           "Enable RTM lock eliding for stack locks in compiled code")       \
157                                                                             \
158   product(bool, UseRTMDeopt, false,                                         \
159           "Perform deopt and recompilation based on RTM abort ratio")       \
160                                                                             \
161   product(int, RTMRetryCount, 5,                                            \
162           "Number of RTM retries on lock abort or busy")                    \
163           range(0, max_jint)                                                \
164                                                                             \
165   experimental(int, RTMSpinLoopCount, 100,                                  \
166           "Spin count for lock to become free before RTM retry")            \
167           range(0, max_jint)                                                \
168                                                                             \
169   experimental(int, RTMAbortThreshold, 1000,                                \
170           "Calculate abort ratio after this number of aborts")              \
171           range(0, max_jint)                                                \
172                                                                             \
173   experimental(int, RTMLockingThreshold, 10000,                             \
174           "Lock count at which to do RTM lock eliding without "             \
175           "abort ratio calculation")                                        \
176           range(0, max_jint)                                                \
177                                                                             \
178   experimental(int, RTMAbortRatio, 50,                                      \
179           "Lock abort ratio at which to stop use RTM lock eliding")         \
180           range(0, 100) /* natural range */                                 \
181                                                                             \
182   experimental(int, RTMTotalCountIncrRate, 64,                              \
183           "Increment total RTM attempted lock count once every n times")    \
184           range(1, max_jint)                                                \
185           constraint(RTMTotalCountIncrRateConstraintFunc,AfterErgo)         \
186                                                                             \
187   experimental(intx, RTMLockingCalculationDelay, 0,                         \
188           "Number of milliseconds to wait before start calculating aborts " \
189           "for RTM locking")                                                \
190                                                                             \
191   experimental(bool, UseRTMXendForLockBusy, true,                           \
192           "Use RTM Xend instead of Xabort when lock busy")                  \
193                                                                             \
194   /* assembler */                                                           \
195   product(bool, UseCountLeadingZerosInstruction, false,                     \
196           "Use count leading zeros instruction")                            \
197                                                                             \
198   product(bool, UseCountTrailingZerosInstruction, false,                    \
199           "Use count trailing zeros instruction")                           \
200                                                                             \
201   product(bool, UseSSE42Intrinsics, false,                                  \
202           "SSE4.2 versions of intrinsics")                                  \
203                                                                             \
204   product(bool, UseBMI1Instructions, false,                                 \
205           "Use BMI1 instructions")                                          \
206                                                                             \
207   product(bool, UseBMI2Instructions, false,                                 \
208           "Use BMI2 instructions")                                          \
209                                                                             \
210   diagnostic(bool, UseLibmIntrinsic, true,                                  \
211           "Use Libm Intrinsics")                                            \
212                                                                             \
213   /* Minimum array size in bytes to use AVX512 intrinsics */                \
214   /* for copy, inflate and fill which don't bail out early based on any */  \
215   /* condition. When this value is set to zero compare operations like */   \
216   /* compare, vectorizedMismatch, compress can also use AVX512 intrinsics.*/\
217   diagnostic(int, AVX3Threshold, 4096,                                      \
218              "Minimum array size in bytes to use AVX512 intrinsics"         \
219              "for copy, inflate and fill. When this value is set as zero"   \
220              "compare operations can also use AVX512 intrinsics.")          \
221           range(0, max_jint)
222 #endif // CPU_X86_GLOBALS_X86_HPP