src/cpu/ppc/vm/vm_version_ppc.cpp

Tue, 04 Feb 2020 18:13:14 +0800

author
aoqi
date
Tue, 04 Feb 2020 18:13:14 +0800
changeset 9806
758c07667682
parent 9756
2be326848943
parent 9788
44ef77ad417c
permissions
-rw-r--r--

Merge

     1 /*
     2  * Copyright (c) 1997, 2018, Oracle and/or its affiliates. All rights reserved.
     3  * Copyright 2012, 2018 SAP AG. All rights reserved.
     4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
     5  *
     6  * This code is free software; you can redistribute it and/or modify it
     7  * under the terms of the GNU General Public License version 2 only, as
     8  * published by the Free Software Foundation.
     9  *
    10  * This code is distributed in the hope that it will be useful, but WITHOUT
    11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
    12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
    13  * version 2 for more details (a copy is included in the LICENSE file that
    14  * accompanied this code).
    15  *
    16  * You should have received a copy of the GNU General Public License version
    17  * 2 along with this work; if not, write to the Free Software Foundation,
    18  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
    19  *
    20  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
    21  * or visit www.oracle.com if you need additional information or have any
    22  * questions.
    23  *
    24  */
    26 #include "precompiled.hpp"
    27 #include "asm/assembler.inline.hpp"
    28 #include "asm/macroAssembler.inline.hpp"
    29 #include "compiler/disassembler.hpp"
    30 #include "memory/resourceArea.hpp"
    31 #include "runtime/java.hpp"
    32 #include "runtime/stubCodeGenerator.hpp"
    33 #include "utilities/defaultStream.hpp"
    34 #include "vm_version_ppc.hpp"
    35 #ifdef TARGET_OS_FAMILY_aix
    36 # include "os_aix.inline.hpp"
    37 #endif
    38 #ifdef TARGET_OS_FAMILY_linux
    39 # include "os_linux.inline.hpp"
    40 #endif
    42 # include <sys/sysinfo.h>
    44 int VM_Version::_features = VM_Version::unknown_m;
    45 int VM_Version::_measured_cache_line_size = 128; // default value
    46 const char* VM_Version::_features_str = "";
    47 bool VM_Version::_is_determine_features_test_running = false;
    48 uint64_t VM_Version::_dscr_val = 0;
    50 #define MSG(flag)   \
    51   if (flag && !FLAG_IS_DEFAULT(flag))                                  \
    52       jio_fprintf(defaultStream::error_stream(),                       \
    53                   "warning: -XX:+" #flag " requires -XX:+UseSIGTRAP\n" \
    54                   "         -XX:+" #flag " will be disabled!\n");
    56 void VM_Version::initialize() {
    58   // Test which instructions are supported and measure cache line size.
    59   determine_features();
    61   // If PowerArchitecturePPC64 hasn't been specified explicitly determine from features.
    62   if (FLAG_IS_DEFAULT(PowerArchitecturePPC64)) {
    63     if (VM_Version::has_lqarx()) {
    64       FLAG_SET_ERGO(uintx, PowerArchitecturePPC64, 8);
    65     } else if (VM_Version::has_popcntw()) {
    66       FLAG_SET_ERGO(uintx, PowerArchitecturePPC64, 7);
    67     } else if (VM_Version::has_cmpb()) {
    68       FLAG_SET_ERGO(uintx, PowerArchitecturePPC64, 6);
    69     } else if (VM_Version::has_popcntb()) {
    70       FLAG_SET_ERGO(uintx, PowerArchitecturePPC64, 5);
    71     } else {
    72       FLAG_SET_ERGO(uintx, PowerArchitecturePPC64, 0);
    73     }
    74   }
    75   guarantee(PowerArchitecturePPC64 == 0 || PowerArchitecturePPC64 == 5 ||
    76             PowerArchitecturePPC64 == 6 || PowerArchitecturePPC64 == 7 ||
    77             PowerArchitecturePPC64 == 8,
    78             "PowerArchitecturePPC64 should be 0, 5, 6, 7, or 8");
    80   // Power 8: Configure Data Stream Control Register.
    81   if (PowerArchitecturePPC64 >= 8) {
    82     config_dscr();
    83   }
    85   if (!UseSIGTRAP) {
    86     MSG(TrapBasedICMissChecks);
    87     MSG(TrapBasedNotEntrantChecks);
    88     MSG(TrapBasedNullChecks);
    89     FLAG_SET_ERGO(bool, TrapBasedNotEntrantChecks, false);
    90     FLAG_SET_ERGO(bool, TrapBasedNullChecks,       false);
    91     FLAG_SET_ERGO(bool, TrapBasedICMissChecks,     false);
    92   }
    94 #ifdef COMPILER2
    95   if (!UseSIGTRAP) {
    96     MSG(TrapBasedRangeChecks);
    97     FLAG_SET_ERGO(bool, TrapBasedRangeChecks, false);
    98   }
   100   // On Power6 test for section size.
   101   if (PowerArchitecturePPC64 == 6) {
   102     determine_section_size();
   103   // TODO: PPC port } else {
   104   // TODO: PPC port PdScheduling::power6SectorSize = 0x20;
   105   }
   107   MaxVectorSize = 8;
   108 #endif
   110   // Create and print feature-string.
   111   char buf[(num_features+1) * 16]; // Max 16 chars per feature.
   112   jio_snprintf(buf, sizeof(buf),
   113                "ppc64%s%s%s%s%s%s%s%s%s%s%s%s%s%s",
   114                (has_fsqrt()   ? " fsqrt"   : ""),
   115                (has_isel()    ? " isel"    : ""),
   116                (has_lxarxeh() ? " lxarxeh" : ""),
   117                (has_cmpb()    ? " cmpb"    : ""),
   118                //(has_mftgpr()? " mftgpr"  : ""),
   119                (has_popcntb() ? " popcntb" : ""),
   120                (has_popcntw() ? " popcntw" : ""),
   121                (has_fcfids()  ? " fcfids"  : ""),
   122                (has_vand()    ? " vand"    : ""),
   123                (has_lqarx()   ? " lqarx"   : ""),
   124                (has_vcipher() ? " aes"     : ""),
   125                (has_vpmsumb() ? " vpmsumb" : ""),
   126                (has_mfdscr()  ? " mfdscr"  : ""),
   127                (has_vsx()     ? " vsx"     : ""),
   128                (has_vshasig() ? " sha"     : "")
   129                // Make sure number of %s matches num_features!
   130               );
   131   _features_str = strdup(buf);
   132   if (Verbose) {
   133     print_features();
   134   }
   136   // PPC64 supports 8-byte compare-exchange operations (see
   137   // Atomic::cmpxchg and StubGenerator::generate_atomic_cmpxchg_ptr)
   138   // and 'atomic long memory ops' (see Unsafe_GetLongVolatile).
   139   _supports_cx8 = true;
   141   UseSSE = 0; // Only on x86 and x64
   143   intx cache_line_size = _measured_cache_line_size;
   145   if (FLAG_IS_DEFAULT(AllocatePrefetchStyle)) AllocatePrefetchStyle = 1;
   147   if (AllocatePrefetchStyle == 4) {
   148     AllocatePrefetchStepSize = cache_line_size; // Need exact value.
   149     if (FLAG_IS_DEFAULT(AllocatePrefetchLines)) AllocatePrefetchLines = 12; // Use larger blocks by default.
   150     if (AllocatePrefetchDistance < 0) AllocatePrefetchDistance = 2*cache_line_size; // Default is not defined?
   151   } else {
   152     if (cache_line_size > AllocatePrefetchStepSize) AllocatePrefetchStepSize = cache_line_size;
   153     if (FLAG_IS_DEFAULT(AllocatePrefetchLines)) AllocatePrefetchLines = 3; // Optimistic value.
   154     if (AllocatePrefetchDistance < 0) AllocatePrefetchDistance = 3*cache_line_size; // Default is not defined?
   155   }
   157   assert(AllocatePrefetchLines > 0, "invalid value");
   158   if (AllocatePrefetchLines < 1) { // Set valid value in product VM.
   159     AllocatePrefetchLines = 1; // Conservative value.
   160   }
   162   if (AllocatePrefetchStyle == 3 && AllocatePrefetchDistance < cache_line_size) {
   163     AllocatePrefetchStyle = 1; // Fall back if inappropriate.
   164   }
   166   assert(AllocatePrefetchStyle >= 0, "AllocatePrefetchStyle should be positive");
   168   // Implementation does not use any of the vector instructions
   169   // available with Power8. Their exploitation is still pending.
   170   if (!UseCRC32Intrinsics) {
   171     if (FLAG_IS_DEFAULT(UseCRC32Intrinsics)) {
   172       FLAG_SET_DEFAULT(UseCRC32Intrinsics, true);
   173     }
   174   }
   176   // The AES intrinsic stubs require AES instruction support.
   177   if (has_vcipher()) {
   178     if (FLAG_IS_DEFAULT(UseAES)) {
   179       UseAES = true;
   180     }
   181   } else if (UseAES) {
   182     if (!FLAG_IS_DEFAULT(UseAES))
   183       warning("AES instructions are not available on this CPU");
   184     FLAG_SET_DEFAULT(UseAES, false);
   185   }
   187   if (UseAES && has_vcipher()) {
   188     if (FLAG_IS_DEFAULT(UseAESIntrinsics)) {
   189       UseAESIntrinsics = true;
   190     }
   191   } else if (UseAESIntrinsics) {
   192     if (!FLAG_IS_DEFAULT(UseAESIntrinsics))
   193       warning("AES intrinsics are not available on this CPU");
   194     FLAG_SET_DEFAULT(UseAESIntrinsics, false);
   195   }
   197   if (UseGHASHIntrinsics) {
   198     warning("GHASH intrinsics are not available on this CPU");
   199     FLAG_SET_DEFAULT(UseGHASHIntrinsics, false);
   200   }
   202   if (has_vshasig()) {
   203     if (FLAG_IS_DEFAULT(UseSHA)) {
   204       UseSHA = true;
   205     }
   206   } else if (UseSHA) {
   207     if (!FLAG_IS_DEFAULT(UseSHA))
   208       warning("SHA instructions are not available on this CPU");
   209     FLAG_SET_DEFAULT(UseSHA, false);
   210   }
   212   if (UseSHA1Intrinsics) {
   213     warning("Intrinsics for SHA-1 crypto hash functions not available on this CPU.");
   214     FLAG_SET_DEFAULT(UseSHA1Intrinsics, false);
   215   }
   217   if (UseSHA && has_vshasig()) {
   218     if (FLAG_IS_DEFAULT(UseSHA256Intrinsics)) {
   219       FLAG_SET_DEFAULT(UseSHA256Intrinsics, true);
   220     }
   221   } else if (UseSHA256Intrinsics) {
   222     warning("Intrinsics for SHA-224 and SHA-256 crypto hash functions not available on this CPU.");
   223     FLAG_SET_DEFAULT(UseSHA256Intrinsics, false);
   224   }
   226   if (UseSHA && has_vshasig()) {
   227     if (FLAG_IS_DEFAULT(UseSHA512Intrinsics)) {
   228       FLAG_SET_DEFAULT(UseSHA512Intrinsics, true);
   229     }
   230   } else if (UseSHA512Intrinsics) {
   231     warning("Intrinsics for SHA-384 and SHA-512 crypto hash functions not available on this CPU.");
   232     FLAG_SET_DEFAULT(UseSHA512Intrinsics, false);
   233   }
   235   if (!(UseSHA1Intrinsics || UseSHA256Intrinsics || UseSHA512Intrinsics)) {
   236     FLAG_SET_DEFAULT(UseSHA, false);
   237   }
   239   if (FLAG_IS_DEFAULT(UseMontgomeryMultiplyIntrinsic)) {
   240     UseMontgomeryMultiplyIntrinsic = true;
   241   }
   242   if (FLAG_IS_DEFAULT(UseMontgomerySquareIntrinsic)) {
   243     UseMontgomerySquareIntrinsic = true;
   244   }
   245 }
   247 void VM_Version::print_features() {
   248   tty->print_cr("Version: %s cache_line_size = %d", cpu_features(), (int) get_cache_line_size());
   249 }
   251 #ifdef COMPILER2
   252 // Determine section size on power6: If section size is 8 instructions,
   253 // there should be a difference between the two testloops of ~15 %. If
   254 // no difference is detected the section is assumed to be 32 instructions.
   255 void VM_Version::determine_section_size() {
   257   int unroll = 80;
   259   const int code_size = (2* unroll * 32 + 100)*BytesPerInstWord;
   261   // Allocate space for the code.
   262   ResourceMark rm;
   263   CodeBuffer cb("detect_section_size", code_size, 0);
   264   MacroAssembler* a = new MacroAssembler(&cb);
   266   uint32_t *code = (uint32_t *)a->pc();
   267   // Emit code.
   268   void (*test1)() = (void(*)())(void *)a->function_entry();
   270   Label l1;
   272   a->li(R4, 1);
   273   a->sldi(R4, R4, 28);
   274   a->b(l1);
   275   a->align(CodeEntryAlignment);
   277   a->bind(l1);
   279   for (int i = 0; i < unroll; i++) {
   280     // Schleife 1
   281     // ------- sector 0 ------------
   282     // ;; 0
   283     a->nop();                   // 1
   284     a->fpnop0();                // 2
   285     a->fpnop1();                // 3
   286     a->addi(R4,R4, -1); // 4
   288     // ;;  1
   289     a->nop();                   // 5
   290     a->fmr(F6, F6);             // 6
   291     a->fmr(F7, F7);             // 7
   292     a->endgroup();              // 8
   293     // ------- sector 8 ------------
   295     // ;;  2
   296     a->nop();                   // 9
   297     a->nop();                   // 10
   298     a->fmr(F8, F8);             // 11
   299     a->fmr(F9, F9);             // 12
   301     // ;;  3
   302     a->nop();                   // 13
   303     a->fmr(F10, F10);           // 14
   304     a->fmr(F11, F11);           // 15
   305     a->endgroup();              // 16
   306     // -------- sector 16 -------------
   308     // ;;  4
   309     a->nop();                   // 17
   310     a->nop();                   // 18
   311     a->fmr(F15, F15);           // 19
   312     a->fmr(F16, F16);           // 20
   314     // ;;  5
   315     a->nop();                   // 21
   316     a->fmr(F17, F17);           // 22
   317     a->fmr(F18, F18);           // 23
   318     a->endgroup();              // 24
   319     // ------- sector 24  ------------
   321     // ;;  6
   322     a->nop();                   // 25
   323     a->nop();                   // 26
   324     a->fmr(F19, F19);           // 27
   325     a->fmr(F20, F20);           // 28
   327     // ;;  7
   328     a->nop();                   // 29
   329     a->fmr(F21, F21);           // 30
   330     a->fmr(F22, F22);           // 31
   331     a->brnop0();                // 32
   333     // ------- sector 32 ------------
   334   }
   336   // ;; 8
   337   a->cmpdi(CCR0, R4, unroll);   // 33
   338   a->bge(CCR0, l1);             // 34
   339   a->blr();
   341   // Emit code.
   342   void (*test2)() = (void(*)())(void *)a->function_entry();
   343   // uint32_t *code = (uint32_t *)a->pc();
   345   Label l2;
   347   a->li(R4, 1);
   348   a->sldi(R4, R4, 28);
   349   a->b(l2);
   350   a->align(CodeEntryAlignment);
   352   a->bind(l2);
   354   for (int i = 0; i < unroll; i++) {
   355     // Schleife 2
   356     // ------- sector 0 ------------
   357     // ;; 0
   358     a->brnop0();                  // 1
   359     a->nop();                     // 2
   360     //a->cmpdi(CCR0, R4, unroll);
   361     a->fpnop0();                  // 3
   362     a->fpnop1();                  // 4
   363     a->addi(R4,R4, -1);           // 5
   365     // ;; 1
   367     a->nop();                     // 6
   368     a->fmr(F6, F6);               // 7
   369     a->fmr(F7, F7);               // 8
   370     // ------- sector 8 ---------------
   372     // ;; 2
   373     a->endgroup();                // 9
   375     // ;; 3
   376     a->nop();                     // 10
   377     a->nop();                     // 11
   378     a->fmr(F8, F8);               // 12
   380     // ;; 4
   381     a->fmr(F9, F9);               // 13
   382     a->nop();                     // 14
   383     a->fmr(F10, F10);             // 15
   385     // ;; 5
   386     a->fmr(F11, F11);             // 16
   387     // -------- sector 16 -------------
   389     // ;; 6
   390     a->endgroup();                // 17
   392     // ;; 7
   393     a->nop();                     // 18
   394     a->nop();                     // 19
   395     a->fmr(F15, F15);             // 20
   397     // ;; 8
   398     a->fmr(F16, F16);             // 21
   399     a->nop();                     // 22
   400     a->fmr(F17, F17);             // 23
   402     // ;; 9
   403     a->fmr(F18, F18);             // 24
   404     // -------- sector 24 -------------
   406     // ;; 10
   407     a->endgroup();                // 25
   409     // ;; 11
   410     a->nop();                     // 26
   411     a->nop();                     // 27
   412     a->fmr(F19, F19);             // 28
   414     // ;; 12
   415     a->fmr(F20, F20);             // 29
   416     a->nop();                     // 30
   417     a->fmr(F21, F21);             // 31
   419     // ;; 13
   420     a->fmr(F22, F22);             // 32
   421   }
   423   // -------- sector 32 -------------
   424   // ;; 14
   425   a->cmpdi(CCR0, R4, unroll); // 33
   426   a->bge(CCR0, l2);           // 34
   428   a->blr();
   429   uint32_t *code_end = (uint32_t *)a->pc();
   430   a->flush();
   432   double loop1_seconds,loop2_seconds, rel_diff;
   433   uint64_t start1, stop1;
   435   start1 = os::current_thread_cpu_time(false);
   436   (*test1)();
   437   stop1 = os::current_thread_cpu_time(false);
   438   loop1_seconds = (stop1- start1) / (1000 *1000 *1000.0);
   441   start1 = os::current_thread_cpu_time(false);
   442   (*test2)();
   443   stop1 = os::current_thread_cpu_time(false);
   445   loop2_seconds = (stop1 - start1) / (1000 *1000 *1000.0);
   447   rel_diff = (loop2_seconds - loop1_seconds) / loop1_seconds *100;
   449   if (PrintAssembly) {
   450     ttyLocker ttyl;
   451     tty->print_cr("Decoding section size detection stub at " INTPTR_FORMAT " before execution:", p2i(code));
   452     Disassembler::decode((u_char*)code, (u_char*)code_end, tty);
   453     tty->print_cr("Time loop1 :%f", loop1_seconds);
   454     tty->print_cr("Time loop2 :%f", loop2_seconds);
   455     tty->print_cr("(time2 - time1) / time1 = %f %%", rel_diff);
   457     if (rel_diff > 12.0) {
   458       tty->print_cr("Section Size 8 Instructions");
   459     } else{
   460       tty->print_cr("Section Size 32 Instructions or Power5");
   461     }
   462   }
   464 #if 0 // TODO: PPC port
   465   // Set sector size (if not set explicitly).
   466   if (FLAG_IS_DEFAULT(Power6SectorSize128PPC64)) {
   467     if (rel_diff > 12.0) {
   468       PdScheduling::power6SectorSize = 0x20;
   469     } else {
   470       PdScheduling::power6SectorSize = 0x80;
   471     }
   472   } else if (Power6SectorSize128PPC64) {
   473     PdScheduling::power6SectorSize = 0x80;
   474   } else {
   475     PdScheduling::power6SectorSize = 0x20;
   476   }
   477 #endif
   478   if (UsePower6SchedulerPPC64) Unimplemented();
   479 }
   480 #endif // COMPILER2
   482 void VM_Version::determine_features() {
   483 #if defined(ABI_ELFv2)
   484   const int code_size = (num_features+1+2*7)*BytesPerInstWord; // TODO(asmundak): calculation is incorrect.
   485 #else
   486   // 7 InstWords for each call (function descriptor + blr instruction).
   487   const int code_size = (num_features+1+2*7)*BytesPerInstWord;
   488 #endif
   489   int features = 0;
   491   // create test area
   492   enum { BUFFER_SIZE = 2*4*K }; // Needs to be >=2* max cache line size (cache line size can't exceed min page size).
   493   char test_area[BUFFER_SIZE];
   494   char *mid_of_test_area = &test_area[BUFFER_SIZE>>1];
   496   // Allocate space for the code.
   497   ResourceMark rm;
   498   CodeBuffer cb("detect_cpu_features", code_size, 0);
   499   MacroAssembler* a = new MacroAssembler(&cb);
   501   // Must be set to true so we can generate the test code.
   502   _features = VM_Version::all_features_m;
   504   // Emit code.
   505   void (*test)(address addr, uint64_t offset)=(void(*)(address addr, uint64_t offset))(void *)a->function_entry();
   506   uint32_t *code = (uint32_t *)a->pc();
   507   // Don't use R0 in ldarx.
   508   // Keep R3_ARG1 unmodified, it contains &field (see below).
   509   // Keep R4_ARG2 unmodified, it contains offset = 0 (see below).
   510   a->fsqrt(F3, F4);                            // code[0] -> fsqrt_m
   511   a->fsqrts(F3, F4);                           // code[1] -> fsqrts_m
   512   a->isel(R7, R5, R6, 0);                      // code[2] -> isel_m
   513   a->ldarx_unchecked(R7, R3_ARG1, R4_ARG2, 1); // code[3] -> lxarx_m
   514   a->cmpb(R7, R5, R6);                         // code[4] -> bcmp
   515   //a->mftgpr(R7, F3);                         // code[5] -> mftgpr
   516   a->popcntb(R7, R5);                          // code[6] -> popcntb
   517   a->popcntw(R7, R5);                          // code[7] -> popcntw
   518   a->fcfids(F3, F4);                           // code[8] -> fcfids
   519   a->vand(VR0, VR0, VR0);                      // code[9] -> vand
   520   a->lqarx_unchecked(R7, R3_ARG1, R4_ARG2, 1); // code[10] -> lqarx_m
   521   a->vcipher(VR0, VR1, VR2);                   // code[11] -> vcipher
   522   a->vpmsumb(VR0, VR1, VR2);                   // code[12] -> vpmsumb
   523   a->mfdscr(R0);                               // code[13] -> mfdscr
   524   a->lxvd2x(VSR0, R3_ARG1);                    // code[14] -> vsx
   525   a->vshasigmaw(VR0, VR1, 1, 0xF);             // code[15] -> vshasig
   526   a->blr();
   528   // Emit function to set one cache line to zero. Emit function descriptor and get pointer to it.
   529   void (*zero_cacheline_func_ptr)(char*) = (void(*)(char*))(void *)a->function_entry();
   530   a->dcbz(R3_ARG1); // R3_ARG1 = addr
   531   a->blr();
   533   uint32_t *code_end = (uint32_t *)a->pc();
   534   a->flush();
   535   _features = VM_Version::unknown_m;
   537   // Print the detection code.
   538   if (PrintAssembly) {
   539     ttyLocker ttyl;
   540     tty->print_cr("Decoding cpu-feature detection stub at " INTPTR_FORMAT " before execution:", p2i(code));
   541     Disassembler::decode((u_char*)code, (u_char*)code_end, tty);
   542   }
   544   // Measure cache line size.
   545   memset(test_area, 0xFF, BUFFER_SIZE); // Fill test area with 0xFF.
   546   (*zero_cacheline_func_ptr)(mid_of_test_area); // Call function which executes dcbz to the middle.
   547   int count = 0; // count zeroed bytes
   548   for (int i = 0; i < BUFFER_SIZE; i++) if (test_area[i] == 0) count++;
   549   guarantee(is_power_of_2(count), "cache line size needs to be a power of 2");
   550   _measured_cache_line_size = count;
   552   // Execute code. Illegal instructions will be replaced by 0 in the signal handler.
   553   VM_Version::_is_determine_features_test_running = true;
   554   (*test)((address)mid_of_test_area, (uint64_t)0);
   555   VM_Version::_is_determine_features_test_running = false;
   557   // determine which instructions are legal.
   558   int feature_cntr = 0;
   559   if (code[feature_cntr++]) features |= fsqrt_m;
   560   if (code[feature_cntr++]) features |= fsqrts_m;
   561   if (code[feature_cntr++]) features |= isel_m;
   562   if (code[feature_cntr++]) features |= lxarxeh_m;
   563   if (code[feature_cntr++]) features |= cmpb_m;
   564   //if(code[feature_cntr++])features |= mftgpr_m;
   565   if (code[feature_cntr++]) features |= popcntb_m;
   566   if (code[feature_cntr++]) features |= popcntw_m;
   567   if (code[feature_cntr++]) features |= fcfids_m;
   568   if (code[feature_cntr++]) features |= vand_m;
   569   if (code[feature_cntr++]) features |= lqarx_m;
   570   if (code[feature_cntr++]) features |= vcipher_m;
   571   if (code[feature_cntr++]) features |= vpmsumb_m;
   572   if (code[feature_cntr++]) features |= mfdscr_m;
   573   if (code[feature_cntr++]) features |= vsx_m;
   574   if (code[feature_cntr++]) features |= vshasig_m;
   576   // Print the detection code.
   577   if (PrintAssembly) {
   578     ttyLocker ttyl;
   579     tty->print_cr("Decoding cpu-feature detection stub at " INTPTR_FORMAT " after execution:", p2i(code));
   580     Disassembler::decode((u_char*)code, (u_char*)code_end, tty);
   581   }
   583   _features = features;
   584 }
   586 // Power 8: Configure Data Stream Control Register.
   587 void VM_Version::config_dscr() {
   588   assert(has_lqarx(), "Only execute on Power 8 or later!");
   590   // 7 InstWords for each call (function descriptor + blr instruction).
   591   const int code_size = (2+2*7)*BytesPerInstWord;
   593   // Allocate space for the code.
   594   ResourceMark rm;
   595   CodeBuffer cb("config_dscr", code_size, 0);
   596   MacroAssembler* a = new MacroAssembler(&cb);
   598   // Emit code.
   599   uint64_t (*get_dscr)() = (uint64_t(*)())(void *)a->function_entry();
   600   uint32_t *code = (uint32_t *)a->pc();
   601   a->mfdscr(R3);
   602   a->blr();
   604   void (*set_dscr)(long) = (void(*)(long))(void *)a->function_entry();
   605   a->mtdscr(R3);
   606   a->blr();
   608   uint32_t *code_end = (uint32_t *)a->pc();
   609   a->flush();
   611   // Print the detection code.
   612   if (PrintAssembly) {
   613     ttyLocker ttyl;
   614     tty->print_cr("Decoding dscr configuration stub at " INTPTR_FORMAT " before execution:", p2i(code));
   615     Disassembler::decode((u_char*)code, (u_char*)code_end, tty);
   616   }
   618   // Apply the configuration if needed.
   619   _dscr_val = (*get_dscr)();
   620   if (Verbose) {
   621     tty->print_cr("dscr value was 0x%lx" , _dscr_val);
   622   }
   623   bool change_requested = false;
   624   if (DSCR_PPC64 != (uintx)-1) {
   625     _dscr_val = DSCR_PPC64;
   626     change_requested = true;
   627   }
   628   if (DSCR_DPFD_PPC64 <= 7) {
   629     uint64_t mask = 0x7;
   630     if ((_dscr_val & mask) != DSCR_DPFD_PPC64) {
   631       _dscr_val = (_dscr_val & ~mask) | (DSCR_DPFD_PPC64);
   632       change_requested = true;
   633     }
   634   }
   635   if (DSCR_URG_PPC64 <= 7) {
   636     uint64_t mask = 0x7 << 6;
   637     if ((_dscr_val & mask) != DSCR_DPFD_PPC64 << 6) {
   638       _dscr_val = (_dscr_val & ~mask) | (DSCR_URG_PPC64 << 6);
   639       change_requested = true;
   640     }
   641   }
   642   if (change_requested) {
   643     (*set_dscr)(_dscr_val);
   644     if (Verbose) {
   645       tty->print_cr("dscr was set to 0x%lx" , (*get_dscr)());
   646     }
   647   }
   648 }
   650 static int saved_features = 0;
   652 void VM_Version::allow_all() {
   653   saved_features = _features;
   654   _features      = all_features_m;
   655 }
   657 void VM_Version::revert() {
   658   _features = saved_features;
   659 }

mercurial