view src/cpu/aarch64/vm/vm_version_aarch64.cpp @ 10905:f57189b7648d

8257192: Integrate AArch64 JIT port into 8u 7009641: Don't fail VM when CodeCache is full 8073108: [AArch64] Use x86 and SPARC CPU instructions for GHASH acceleration 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space (AArch64 changes) 8131779: AARCH64: add Montgomery multiply intrinsic 8132875: AArch64: Fix error introduced into AArch64 CodeCache by commit for 8130309 8135018: AARCH64: Missing memory barriers for CMS collector 8145320: Create unsafe_arraycopy and generic_arraycopy for AArch64 8148328: aarch64: redundant lsr instructions in stub code. 8148783: aarch64: SEGV running SpecJBB2013 8148948: aarch64: generate_copy_longs calls align() incorrectly 8149080: AArch64: Recognise disjoint array copy in stub code 8149365: aarch64: memory copy does not prefetch on backwards copy 8149907: aarch64: use load/store pair instructions in call_stub 8150038: aarch64: make use of CBZ and CBNZ when comparing narrow pointer with zero 8150045: arraycopy causes segfaults in SATB during garbage collection 8150082: aarch64: optimise small array copy 8150229: aarch64: pipeline class for several instructions is not set correctly 8150313: aarch64: optimise array copy using SIMD instructions 8150394: aarch64: add support for 8.1 LSE CAS instructions 8150652: Remove unused code in AArch64 back end 8151340: aarch64: prefetch the destination word for write prior to ldxr/stxr loops. 8151502: optimize pd_disjoint_words and pd_conjoint_words 8151775: aarch64: add support for 8.1 LSE atomic operations 8152537: aarch64: Make use of CBZ and CBNZ when comparing unsigned values with zero. 8152840: aarch64: improve _unsafe_arraycopy stub routine 8153172: aarch64: hotspot crashes after the 8.1 LSE patch is merged 8153713: aarch64: improve short array clearing using store pair 8153797: aarch64: Add Arrays.fill stub code 8154413: AArch64: Better byte behaviour 8154537: AArch64: some integer rotate instructions are never emitted 8154739: AArch64: TemplateTable::fast_xaccess loads in wrong mode 8155015: Aarch64: bad assert in spill generation code 8155100: AArch64: Relax alignment requirement for byte_map_base 8155612: Aarch64: vector nodes need to support misaligned offset 8155617: aarch64: ClearArray does not use DC ZVA 8155627: Enable SA on AArch64 8155653: TestVectorUnalignedOffset.java not pushed with 8155612 8156731: aarch64: java/util/Arrays/Correct.java fails due to _generic_arraycopy stub routine 8157841: aarch64: prefetch ignores cache line size 8157906: aarch64: some more integer rotate instructions are never emitted 8158913: aarch64: SEGV running Spark terasort 8159052: aarch64: optimise unaligned copies in pd_disjoint_words and pd_conjoint_words 8159063: aarch64: optimise unaligned array copy long 8160748: [AArch64] Inconsistent types for ideal_reg 8161072: AArch64: jtreg compiler/uncommontrap/TestDeoptOOM failure 8161190: AArch64: Fix overflow in immediate cmp instruction 8164113: AArch64: follow-up the fix for 8161598 8165673: AArch64: Fix JNI floating point argument handling 8167200: AArch64: Broken stack pointer adjustment in interpreter 8167421: AArch64: in one core system, fatal error: Illegal threadstate encountered 8167595: AArch64: SEGV in stub code cipherBlockChaining_decryptAESCrypt 8168699: Validate special case invocations [AArch64 support] 8168888: Port 8160591: Improve internal array handling to AArch64. 8170100: AArch64: Crash in C1-compiled code accessing References 8170188: jtreg test compiler/types/TestMeetIncompatibleInterfaceArrays.java causes JVM crash 8170873: PPC64/aarch64: Poor StrictMath performance due to non-optimized compilation 8171537: aarch64: compiler/c1/Test6849574.java generates guarantee failure in C1 8172881: AArch64: assertion failure: the int pressure is incorrect 8173472: AArch64: C1 comparisons with null only use 32-bit instructions 8176100: [AArch64] [REDO][REDO] G1 Needs pre barrier on dereference of weak JNI handles 8177661: Correct ad rule output register types from iRegX to iRegXNoSp 8179954: AArch64: C1 and C2 volatile accesses are not sequentially consistent 8182581: aarch64: fix for crash caused by earlyret of compiled method 8183925: [AArch64] Decouple crash protection from watcher thread 8186325: AArch64: jtreg test hotspot/test/gc/g1/TestJNIWeakG1/TestJNIWeakG1.java SEGV 8187224: aarch64: some inconsistency between aarch64_ad.m4 and aarch64.ad 8189170: [AArch64] Add option to disable stack overflow checking in primordial thread for use with JNI_CreateJavaJVM 8193133: Assertion failure because 0xDEADDEAD can be in-heap 8195685: AArch64 port of 8174962: Better interface invocations 8195859: AArch64: vtableStubs gtest fails after 8174962 8196136: AArch64: Correct register use in patch for JDK-8194686 8196221: AArch64: Mistake in committed patch for JDK-8195859 8199712: [AArch64] Flight Recorder 8203481: Incorrect constraint for unextended_sp in frame:safe_for_sender 8203699: java/lang/invoke/SpecialInterfaceCall fails with SIGILL on aarch64 8205421: AARCH64: StubCodeMark should be placed after alignment 8206163: AArch64: incorrect code generation for StoreCM 8207345: Trampoline generation code reads from uninitialized memory 8207838: AArch64: Float registers incorrectly restored in JNI call 8209413: AArch64: NPE in clhsdb jstack command 8209414: [AArch64] method handle invocation does not respect JVMTI interp_only mode 8209415: Fix JVMTI test failure HS202 8209420: Track membars for volatile accesses so they can be properly optimized 8209835: Aarch64: elide barriers on all volatile operations 8210425: [AArch64] sharedRuntimeTrig/sharedRuntimeTrans compiled without optimization 8211064: [AArch64] Interpreter and c1 don't correctly handle jboolean results in native calls 8211233: MemBarNode::trailing_membar() and MemBarNode::leading_membar() need to handle dying subgraphs better 8213134: AArch64: vector shift failed with MaxVectorSize=8 8213419: [AArch64] C2 may hang in MulLNode::Ideal()/MulINode::Ideal() with gcc 8.2.1 8214857: "bad trailing membar" assert failure at memnode.cpp:3220 8215951: AArch64: jtreg test vmTestbase/nsk/jvmti/PopFrame/popframe005 segfaults 8215961: jdk/jfr/event/os/TestCPUInformation.java fails on AArch64 8216350: AArch64: monitor unlock fast path not called 8216989: CardTableBarrierSetAssembler::gen_write_ref_array_post_barrier() does not check for zero length on AARCH64 8217368: AArch64: C2 recursive stack locking optimisation not triggered 8218185: aarch64: missing LoadStore barrier in TemplateTable::putfield_or_static 8219011: Implement MacroAssembler::warn method on AArch64 8219635: aarch64: missing LoadStore barrier in TemplateTable::fast_storefield 8221220: AArch64: Add StoreStore membar explicitly for Volatile Writes in TemplateTable 8221658: aarch64: add necessary predicate for ubfx patterns 8224671: AArch64: mauve System.arraycopy test failure 8224828: aarch64: rflags is not correct after safepoint poll 8224851: AArch64: fix warnings and errors with Clang and GCC 8.3 8224880: AArch64: java/javac error with AllocatePrefetchDistance 8228400: Remove built-in AArch64 simulator 8228406: Superfluous change in chaitin.hpp 8228593: Revert explicit JDK 7 support additions 8228716: Revert InstanceKlass::print_on debug additions 8228718: Revert incorrect backport of JDK-8129757 to 8-aarch64 8228725: AArch64: Purge method call format support 8228747: Revert "unused" attribute from test_arraycopy_func 8228767: Revert ResourceMark additions 8228770: Revert development hsdis changes 8229123: Revert build fixes for aarch64/zero 8229124: Revert disassembler.cpp changes 8229145: Revert TemplateTable::bytecode() visibility change 8233839: aarch64: missing memory barrier in NewObjectArrayStub and NewTypeArrayStub 8237512: AArch64: aarch64TestHook leaks a BufferBlob 8246482: Build failures with +JFR -PCH 8247979: aarch64: missing side effect of killing flags for clearArray_reg_reg 8248219: aarch64: missing memory barrier in fast_storefield and fast_accessfield Reviewed-by: shade, aph
author andrew
date Mon, 01 Feb 2021 03:48:36 +0000
parents
children 306a4643e4d2
line wrap: on
line source

/*
 * Copyright (c) 2013, Red Hat Inc.
 * Copyright (c) 1997, 2015, 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 "asm/macroAssembler.hpp"
#include "asm/macroAssembler.inline.hpp"
#include "memory/resourceArea.hpp"
#include "runtime/java.hpp"
#include "runtime/stubCodeGenerator.hpp"
#include "vm_version_aarch64.hpp"
#ifdef TARGET_OS_FAMILY_linux
# include "os_linux.inline.hpp"
#endif

#include <sys/auxv.h>
#include <asm/hwcap.h>

#ifndef HWCAP_AES
#define HWCAP_AES   (1<<3)
#endif

#ifndef HWCAP_SHA1
#define HWCAP_SHA1  (1<<5)
#endif

#ifndef HWCAP_SHA2
#define HWCAP_SHA2  (1<<6)
#endif

#ifndef HWCAP_CRC32
#define HWCAP_CRC32 (1<<7)
#endif

#ifndef HWCAP_ATOMICS
#define HWCAP_ATOMICS (1<<8)
#endif

int VM_Version::_cpu;
int VM_Version::_model;
int VM_Version::_model2;
int VM_Version::_variant;
int VM_Version::_revision;
int VM_Version::_stepping;
int VM_Version::_cpuFeatures;
const char*           VM_Version::_features_str = "";
VM_Version::PsrInfo VM_Version::_psr_info   = { 0, };

static BufferBlob* stub_blob;
static const int stub_size = 550;

extern "C" {
  typedef void (*getPsrInfo_stub_t)(void*);
}
static getPsrInfo_stub_t getPsrInfo_stub = NULL;


class VM_Version_StubGenerator: public StubCodeGenerator {
 public:

  VM_Version_StubGenerator(CodeBuffer *c) : StubCodeGenerator(c) {}

  address generate_getPsrInfo() {
    StubCodeMark mark(this, "VM_Version", "getPsrInfo_stub");
#   define __ _masm->
    address start = __ pc();

    // void getPsrInfo(VM_Version::PsrInfo* psr_info);

    address entry = __ pc();

    __ enter();

    __ get_dczid_el0(rscratch1);
    __ strw(rscratch1, Address(c_rarg0, in_bytes(VM_Version::dczid_el0_offset())));

    __ get_ctr_el0(rscratch1);
    __ strw(rscratch1, Address(c_rarg0, in_bytes(VM_Version::ctr_el0_offset())));

    __ leave();
    __ ret(lr);

#   undef __

    return start;
  }
};


void VM_Version::get_processor_features() {
  _supports_cx8 = true;
  _supports_atomic_getset4 = true;
  _supports_atomic_getadd4 = true;
  _supports_atomic_getset8 = true;
  _supports_atomic_getadd8 = true;

  getPsrInfo_stub(&_psr_info);

  int dcache_line = VM_Version::dcache_line_size();

  // Limit AllocatePrefetchDistance so that it does not exceed the
  // constraint in AllocatePrefetchDistanceConstraintFunc.
  if (FLAG_IS_DEFAULT(AllocatePrefetchDistance))
    FLAG_SET_DEFAULT(AllocatePrefetchDistance, MIN2(512, 3*dcache_line));

  if (FLAG_IS_DEFAULT(AllocatePrefetchStepSize))
    FLAG_SET_DEFAULT(AllocatePrefetchStepSize, dcache_line);
  if (FLAG_IS_DEFAULT(PrefetchScanIntervalInBytes))
    FLAG_SET_DEFAULT(PrefetchScanIntervalInBytes, 3*dcache_line);
  if (FLAG_IS_DEFAULT(PrefetchCopyIntervalInBytes))
    FLAG_SET_DEFAULT(PrefetchCopyIntervalInBytes, 3*dcache_line);

  if (PrefetchCopyIntervalInBytes != -1 &&
       ((PrefetchCopyIntervalInBytes & 7) || (PrefetchCopyIntervalInBytes >= 32768))) {
    warning("PrefetchCopyIntervalInBytes must be -1, or a multiple of 8 and < 32768");
    PrefetchCopyIntervalInBytes &= ~7;
    if (PrefetchCopyIntervalInBytes >= 32768)
      PrefetchCopyIntervalInBytes = 32760;
  }
  FLAG_SET_DEFAULT(UseSSE42Intrinsics, true);

  unsigned long auxv = getauxval(AT_HWCAP);

  char buf[512];

  strcpy(buf, "simd");
  if (auxv & HWCAP_CRC32) strcat(buf, ", crc");
  if (auxv & HWCAP_AES)   strcat(buf, ", aes");
  if (auxv & HWCAP_SHA1)  strcat(buf, ", sha1");
  if (auxv & HWCAP_SHA2)  strcat(buf, ", sha256");
  if (auxv & HWCAP_ATOMICS) strcat(buf, ", lse");

  _features_str = strdup(buf);
  _cpuFeatures = auxv;

  int cpu_lines = 0;
  if (FILE *f = fopen("/proc/cpuinfo", "r")) {
    char buf[128], *p;
    while (fgets(buf, sizeof (buf), f) != NULL) {
      if ((p = strchr(buf, ':')) != NULL) {
        long v = strtol(p+1, NULL, 0);
        if (strncmp(buf, "CPU implementer", sizeof "CPU implementer" - 1) == 0) {
          _cpu = v;
          cpu_lines++;
        } else if (strncmp(buf, "CPU variant", sizeof "CPU variant" - 1) == 0) {
          _variant = v;
        } else if (strncmp(buf, "CPU part", sizeof "CPU part" - 1) == 0) {
          if (_model != v)  _model2 = _model;
          _model = v;
        } else if (strncmp(buf, "CPU revision", sizeof "CPU revision" - 1) == 0) {
          _revision = v;
        }
      }
    }
    fclose(f);
  }

  // Enable vendor specific features
  if (_cpu == CPU_CAVIUM) {
    if (_variant == 0) _cpuFeatures |= CPU_DMB_ATOMICS;
    if (FLAG_IS_DEFAULT(AvoidUnalignedAccesses)) {
      FLAG_SET_DEFAULT(AvoidUnalignedAccesses, true);
    }
    if (FLAG_IS_DEFAULT(UseSIMDForMemoryOps)) {
      FLAG_SET_DEFAULT(UseSIMDForMemoryOps, (_variant > 0));
    }
  }
  if (_cpu == CPU_ARM && (_model == 0xd03 || _model2 == 0xd03)) _cpuFeatures |= CPU_A53MAC;
  if (_cpu == CPU_ARM && (_model == 0xd07 || _model2 == 0xd07)) _cpuFeatures |= CPU_STXR_PREFETCH;
  // If an olde style /proc/cpuinfo (cpu_lines == 1) then if _model is an A57 (0xd07)
  // we assume the worst and assume we could be on a big little system and have
  // undisclosed A53 cores which we could be swapped to at any stage
  if (_cpu == CPU_ARM && cpu_lines == 1 && _model == 0xd07) _cpuFeatures |= CPU_A53MAC;

  if (FLAG_IS_DEFAULT(UseCRC32)) {
    UseCRC32 = (auxv & HWCAP_CRC32) != 0;
  }
  if (UseCRC32 && (auxv & HWCAP_CRC32) == 0) {
    warning("UseCRC32 specified, but not supported on this CPU");
  }

  if (auxv & HWCAP_ATOMICS) {
    if (FLAG_IS_DEFAULT(UseLSE))
      FLAG_SET_DEFAULT(UseLSE, true);
  } else {
    if (UseLSE) {
      warning("UseLSE specified, but not supported on this CPU");
    }
  }

  if (auxv & HWCAP_AES) {
    UseAES = UseAES || FLAG_IS_DEFAULT(UseAES);
    UseAESIntrinsics =
        UseAESIntrinsics || (UseAES && FLAG_IS_DEFAULT(UseAESIntrinsics));
    if (UseAESIntrinsics && !UseAES) {
      warning("UseAESIntrinsics enabled, but UseAES not, enabling");
      UseAES = true;
    }
  } else {
    if (UseAES) {
      warning("UseAES specified, but not supported on this CPU");
    }
    if (UseAESIntrinsics) {
      warning("UseAESIntrinsics specified, but not supported on this CPU");
    }
  }

  if (UseGHASHIntrinsics) {
    warning("GHASH intrinsics are not available on this CPU");
    FLAG_SET_DEFAULT(UseGHASHIntrinsics, false);
  }

  if (FLAG_IS_DEFAULT(UseCRC32Intrinsics)) {
    UseCRC32Intrinsics = true;
  }

  if (auxv & (HWCAP_SHA1 | HWCAP_SHA2)) {
    if (FLAG_IS_DEFAULT(UseSHA)) {
      FLAG_SET_DEFAULT(UseSHA, true);
    }
  } else if (UseSHA) {
    warning("SHA instructions are not available on this CPU");
    FLAG_SET_DEFAULT(UseSHA, false);
  }

  if (!UseSHA) {
    FLAG_SET_DEFAULT(UseSHA1Intrinsics, false);
    FLAG_SET_DEFAULT(UseSHA256Intrinsics, false);
    FLAG_SET_DEFAULT(UseSHA512Intrinsics, false);
  } else {
    if (auxv & HWCAP_SHA1) {
      if (FLAG_IS_DEFAULT(UseSHA1Intrinsics)) {
        FLAG_SET_DEFAULT(UseSHA1Intrinsics, true);
      }
    } else if (UseSHA1Intrinsics) {
      warning("SHA1 instruction is not available on this CPU.");
      FLAG_SET_DEFAULT(UseSHA1Intrinsics, false);
    }
    if (auxv & HWCAP_SHA2) {
      if (FLAG_IS_DEFAULT(UseSHA256Intrinsics)) {
        FLAG_SET_DEFAULT(UseSHA256Intrinsics, true);
      }
    } else if (UseSHA256Intrinsics) {
      warning("SHA256 instruction (for SHA-224 and SHA-256) is not available on this CPU.");
      FLAG_SET_DEFAULT(UseSHA256Intrinsics, false);
    }
    if (UseSHA512Intrinsics) {
      warning("SHA512 instruction (for SHA-384 and SHA-512) is not available on this CPU.");
      FLAG_SET_DEFAULT(UseSHA512Intrinsics, false);
    }
  }

  if (is_zva_enabled()) {
    if (FLAG_IS_DEFAULT(UseBlockZeroing)) {
      FLAG_SET_DEFAULT(UseBlockZeroing, true);
    }
    if (FLAG_IS_DEFAULT(BlockZeroingLowLimit)) {
      FLAG_SET_DEFAULT(BlockZeroingLowLimit, 4 * VM_Version::zva_length());
    }
  } else if (UseBlockZeroing) {
    warning("DC ZVA is not available on this CPU");
    FLAG_SET_DEFAULT(UseBlockZeroing, false);
  }

  if (FLAG_IS_DEFAULT(UseMultiplyToLenIntrinsic)) {
    UseMultiplyToLenIntrinsic = true;
  }

  if (FLAG_IS_DEFAULT(UseBarriersForVolatile)) {
    UseBarriersForVolatile = (_cpuFeatures & CPU_DMB_ATOMICS) != 0;
  }

  if (FLAG_IS_DEFAULT(UsePopCountInstruction)) {
    UsePopCountInstruction = true;
  }

  if (FLAG_IS_DEFAULT(UseMontgomeryMultiplyIntrinsic)) {
    UseMontgomeryMultiplyIntrinsic = true;
  }
  if (FLAG_IS_DEFAULT(UseMontgomerySquareIntrinsic)) {
    UseMontgomerySquareIntrinsic = true;
  }

#ifdef COMPILER2
  if (FLAG_IS_DEFAULT(OptoScheduling)) {
    OptoScheduling = true;
  }
#else
  if (ReservedCodeCacheSize > 128*M) {
    vm_exit_during_initialization("client compiler does not support ReservedCodeCacheSize > 128M");
  }
#endif
}

void VM_Version::initialize() {
  ResourceMark rm;

  stub_blob = BufferBlob::create("getPsrInfo_stub", stub_size);
  if (stub_blob == NULL) {
    vm_exit_during_initialization("Unable to allocate getPsrInfo_stub");
  }

  CodeBuffer c(stub_blob);
  VM_Version_StubGenerator g(&c);
  getPsrInfo_stub = CAST_TO_FN_PTR(getPsrInfo_stub_t,
                                   g.generate_getPsrInfo());

  get_processor_features();
}