This commit is contained in:
Jesper Wilhelmsson 2016-02-18 18:10:35 +01:00
commit 6f4adc7c72
65 changed files with 1141 additions and 355 deletions

View file

@ -54,6 +54,7 @@ define_pd_global(intx, INTPRESSURE, 25);
define_pd_global(intx, InteriorEntryAlignment, 16);
define_pd_global(intx, NewSizeThreadIncrease, ScaleForWordSize(4*K));
define_pd_global(intx, LoopUnrollLimit, 60);
define_pd_global(intx, LoopPercentProfileLimit, 10);
// InitialCodeCacheSize derived from specjbb2000 run.
define_pd_global(intx, InitialCodeCacheSize, 2496*K); // Integral multiple of CodeCacheExpansionSize
define_pd_global(intx, CodeCacheExpansionSize, 64*K);

View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2015, 2016, 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
@ -30,39 +30,151 @@
#include "vmreg_aarch64.inline.hpp"
jint CodeInstaller::pd_next_offset(NativeInstruction* inst, jint pc_offset, Handle method, TRAPS) {
Unimplemented();
return 0;
if (inst->is_call() || inst->is_jump() || inst->is_blr()) {
return pc_offset + NativeCall::instruction_size;
} else if (inst->is_general_jump()) {
return pc_offset + NativeGeneralJump::instruction_size;
} else {
JVMCI_ERROR_0("unsupported type of instruction for call site");
}
}
void CodeInstaller::pd_patch_OopConstant(int pc_offset, Handle constant, TRAPS) {
Unimplemented();
address pc = _instructions->start() + pc_offset;
Handle obj = HotSpotObjectConstantImpl::object(constant);
jobject value = JNIHandles::make_local(obj());
if (HotSpotObjectConstantImpl::compressed(constant)) {
int oop_index = _oop_recorder->find_index(value);
RelocationHolder rspec = oop_Relocation::spec(oop_index);
_instructions->relocate(pc, rspec, 1);
Unimplemented();
} else {
NativeMovConstReg* move = nativeMovConstReg_at(pc);
move->set_data((intptr_t) value);
int oop_index = _oop_recorder->find_index(value);
RelocationHolder rspec = oop_Relocation::spec(oop_index);
_instructions->relocate(pc, rspec);
}
}
void CodeInstaller::pd_patch_MetaspaceConstant(int pc_offset, Handle constant, TRAPS) {
Unimplemented();
address pc = _instructions->start() + pc_offset;
if (HotSpotMetaspaceConstantImpl::compressed(constant)) {
narrowKlass narrowOop = record_narrow_metadata_reference(constant, CHECK);
TRACE_jvmci_3("relocating (narrow metaspace constant) at " PTR_FORMAT "/0x%x", p2i(pc), narrowOop);
Unimplemented();
} else {
NativeMovConstReg* move = nativeMovConstReg_at(pc);
Metadata* reference = record_metadata_reference(constant, CHECK);
move->set_data((intptr_t) reference);
TRACE_jvmci_3("relocating (metaspace constant) at " PTR_FORMAT "/" PTR_FORMAT, p2i(pc), p2i(reference));
}
}
void CodeInstaller::pd_patch_DataSectionReference(int pc_offset, int data_offset) {
Unimplemented();
void CodeInstaller::pd_patch_DataSectionReference(int pc_offset, int data_offset, TRAPS) {
address pc = _instructions->start() + pc_offset;
NativeInstruction* inst = nativeInstruction_at(pc);
if (inst->is_adr_aligned()) {
address dest = _constants->start() + data_offset;
_instructions->relocate(pc, section_word_Relocation::spec((address) dest, CodeBuffer::SECT_CONSTS));
TRACE_jvmci_3("relocating at " PTR_FORMAT " (+%d) with destination at %d", p2i(pc), pc_offset, data_offset);
} else {
JVMCI_ERROR("unknown load or move instruction at " PTR_FORMAT, p2i(pc));
}
}
void CodeInstaller::pd_relocate_ForeignCall(NativeInstruction* inst, jlong foreign_call_destination, TRAPS) {
Unimplemented();
address pc = (address) inst;
if (inst->is_call()) {
NativeCall* call = nativeCall_at(pc);
call->set_destination((address) foreign_call_destination);
_instructions->relocate(call->instruction_address(), runtime_call_Relocation::spec());
} else if (inst->is_jump()) {
NativeJump* jump = nativeJump_at(pc);
jump->set_jump_destination((address) foreign_call_destination);
_instructions->relocate(jump->instruction_address(), runtime_call_Relocation::spec());
} else if (inst->is_general_jump()) {
NativeGeneralJump* jump = nativeGeneralJump_at(pc);
jump->set_jump_destination((address) foreign_call_destination);
_instructions->relocate(jump->instruction_address(), runtime_call_Relocation::spec());
} else {
JVMCI_ERROR("unknown call or jump instruction at " PTR_FORMAT, p2i(pc));
}
TRACE_jvmci_3("relocating (foreign call) at " PTR_FORMAT, p2i(inst));
}
void CodeInstaller::pd_relocate_JavaMethod(Handle hotspot_method, jint pc_offset, TRAPS) {
Unimplemented();
#ifdef ASSERT
Method* method = NULL;
// we need to check, this might also be an unresolved method
if (hotspot_method->is_a(HotSpotResolvedJavaMethodImpl::klass())) {
method = getMethodFromHotSpotMethod(hotspot_method());
}
#endif
switch (_next_call_type) {
case INLINE_INVOKE:
break;
case INVOKEVIRTUAL:
case INVOKEINTERFACE: {
assert(method == NULL || !method->is_static(), "cannot call static method with invokeinterface");
NativeCall* call = nativeCall_at(_instructions->start() + pc_offset);
call->set_destination(SharedRuntime::get_resolve_virtual_call_stub());
_instructions->relocate(call->instruction_address(), virtual_call_Relocation::spec(_invoke_mark_pc));
break;
}
case INVOKESTATIC: {
assert(method == NULL || method->is_static(), "cannot call non-static method with invokestatic");
NativeCall* call = nativeCall_at(_instructions->start() + pc_offset);
call->set_destination(SharedRuntime::get_resolve_static_call_stub());
_instructions->relocate(call->instruction_address(), relocInfo::static_call_type);
break;
}
case INVOKESPECIAL: {
assert(method == NULL || !method->is_static(), "cannot call static method with invokespecial");
NativeCall* call = nativeCall_at(_instructions->start() + pc_offset);
call->set_destination(SharedRuntime::get_resolve_opt_virtual_call_stub());
_instructions->relocate(call->instruction_address(), relocInfo::opt_virtual_call_type);
break;
}
default:
JVMCI_ERROR("invalid _next_call_type value");
break;
}
}
void CodeInstaller::pd_relocate_poll(address pc, jint mark, TRAPS) {
Unimplemented();
switch (mark) {
case POLL_NEAR:
JVMCI_ERROR("unimplemented");
break;
case POLL_FAR:
_instructions->relocate(pc, relocInfo::poll_type);
break;
case POLL_RETURN_NEAR:
JVMCI_ERROR("unimplemented");
break;
case POLL_RETURN_FAR:
_instructions->relocate(pc, relocInfo::poll_return_type);
break;
default:
JVMCI_ERROR("invalid mark value");
break;
}
}
// convert JVMCI register indices (as used in oop maps) to HotSpot registers
VMReg CodeInstaller::get_hotspot_reg(jint jvmci_reg, TRAPS) {
return NULL;
if (jvmci_reg < RegisterImpl::number_of_registers) {
return as_Register(jvmci_reg)->as_VMReg();
} else {
jint floatRegisterNumber = jvmci_reg - RegisterImpl::number_of_registers;
if (floatRegisterNumber < FloatRegisterImpl::number_of_registers) {
return as_FloatRegister(floatRegisterNumber)->as_VMReg();
}
JVMCI_ERROR_NULL("invalid register number: %d", jvmci_reg);
}
}
bool CodeInstaller::is_general_purpose_reg(VMReg hotspotRegister) {
return false;
return !hotspotRegister->is_FloatRegister();
}

View file

@ -136,7 +136,7 @@ void NativeMovConstReg::set_data(intptr_t x) {
MacroAssembler::pd_patch_instruction(instruction_address(), (address)x);
ICache::invalidate_range(instruction_address(), instruction_size);
}
};
}
void NativeMovConstReg::print() {
tty->print_cr(PTR_FORMAT ": mov reg, " INTPTR_FORMAT,
@ -208,6 +208,32 @@ void NativeJump::set_jump_destination(address dest) {
//-------------------------------------------------------------------
address NativeGeneralJump::jump_destination() const {
NativeMovConstReg* move = nativeMovConstReg_at(instruction_address());
address dest = (address) move->data();
// We use jump to self as the unresolved address which the inline
// cache code (and relocs) know about
// return -1 if jump to self
dest = (dest == (address) this) ? (address) -1 : dest;
return dest;
}
void NativeGeneralJump::set_jump_destination(address dest) {
NativeMovConstReg* move = nativeMovConstReg_at(instruction_address());
// We use jump to self as the unresolved address which the inline
// cache code (and relocs) know about
if (dest == (address) -1) {
dest = instruction_address();
}
move->set_data((uintptr_t) dest);
};
//-------------------------------------------------------------------
bool NativeInstruction::is_safepoint_poll() {
// a safepoint_poll is implemented in two steps as either
//
@ -249,6 +275,22 @@ bool NativeInstruction::is_ldrw_to_zr(address instr) {
Instruction_aarch64::extract(insn, 4, 0) == 0b11111);
}
bool NativeInstruction::is_general_jump() {
if (is_movz()) {
NativeInstruction* inst1 = nativeInstruction_at(addr_at(instruction_size * 1));
if (inst1->is_movk()) {
NativeInstruction* inst2 = nativeInstruction_at(addr_at(instruction_size * 2));
if (inst2->is_movk()) {
NativeInstruction* inst3 = nativeInstruction_at(addr_at(instruction_size * 3));
if (inst3->is_blr()) {
return true;
}
}
}
}
return false;
}
bool NativeInstruction::is_movz() {
return Instruction_aarch64::extract(int_at(0), 30, 23) == 0b10100101;
}

View file

@ -54,11 +54,22 @@ class NativeInstruction VALUE_OBJ_CLASS_SPEC {
friend class Relocation;
friend bool is_NativeCallTrampolineStub_at(address);
public:
enum { instruction_size = 4 };
enum {
instruction_size = 4
};
juint encoding() const {
return uint_at(0);
}
bool is_blr() const { return (encoding() & 0xfffffc1f) == 0xd63f0000; }
bool is_adr_aligned() const { return (encoding() & 0xff000000) == 0x10000000; } // adr Xn, <label>, where label is aligned to 4 bytes (address of instruction).
inline bool is_nop();
inline bool is_illegal();
inline bool is_return();
bool is_jump();
bool is_general_jump();
inline bool is_jump_or_nop();
inline bool is_cond_jump();
bool is_safepoint_poll();
@ -341,11 +352,15 @@ class NativeMovRegMemPatching: public NativeMovRegMem {
// An interface for accessing/manipulating native leal instruction of form:
// leal reg, [reg + offset]
class NativeLoadAddress: public NativeMovRegMem {
static const bool has_rex = true;
static const int rex_size = 1;
public:
class NativeLoadAddress: public NativeInstruction {
enum AArch64_specific_constants {
instruction_size = 4,
instruction_offset = 0,
data_offset = 0,
next_instruction_offset = 4
};
public:
void verify();
void print ();
@ -398,6 +413,10 @@ public:
data_offset = 0,
next_instruction_offset = 4 * 4
};
address jump_destination() const;
void set_jump_destination(address dest);
static void insert_unconditional(address code_pos, address entry);
static void replace_mt_safe(address instr_addr, address code_buffer);
static void verify();

View file

@ -54,6 +54,7 @@ define_pd_global(intx, RegisterCostAreaRatio, 16000);
define_pd_global(bool, UseTLAB, true);
define_pd_global(bool, ResizeTLAB, true);
define_pd_global(intx, LoopUnrollLimit, 60);
define_pd_global(intx, LoopPercentProfileLimit, 10);
// Peephole and CISC spilling both break the graph, and so make the
// scheduler sick.

View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2015, 2016, 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
@ -42,7 +42,7 @@ void CodeInstaller::pd_patch_MetaspaceConstant(int pc_offset, Handle constant, T
Unimplemented();
}
void CodeInstaller::pd_patch_DataSectionReference(int pc_offset, int data_offset) {
void CodeInstaller::pd_patch_DataSectionReference(int pc_offset, int data_offset, TRAPS) {
Unimplemented();
}

View file

@ -52,6 +52,7 @@ define_pd_global(intx, RegisterCostAreaRatio, 12000);
define_pd_global(bool, UseTLAB, true);
define_pd_global(bool, ResizeTLAB, true);
define_pd_global(intx, LoopUnrollLimit, 60); // Design center runs on 1.3.1
define_pd_global(intx, LoopPercentProfileLimit, 10);
define_pd_global(intx, MinJumpTableSize, 5);
// Peephole and CISC spilling both break the graph, and so makes the

View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2013, 2015, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013, 2016, 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
@ -85,7 +85,7 @@ void CodeInstaller::pd_patch_MetaspaceConstant(int pc_offset, Handle constant, T
}
}
void CodeInstaller::pd_patch_DataSectionReference(int pc_offset, int data_offset) {
void CodeInstaller::pd_patch_DataSectionReference(int pc_offset, int data_offset, TRAPS) {
address pc = _instructions->start() + pc_offset;
NativeInstruction* inst = nativeInstruction_at(pc);
NativeInstruction* inst1 = nativeInstruction_at(pc + 4);

View file

@ -2015,23 +2015,33 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
int vep_offset = ((intptr_t)__ pc()) - start;
#ifdef COMPILER1
if (InlineObjectHash && method->intrinsic_id() == vmIntrinsics::_hashCode) {
// Object.hashCode can pull the hashCode from the header word
// instead of doing a full VM transition once it's been computed.
// Since hashCode is usually polymorphic at call sites we can't do
// this optimization at the call site without a lot of work.
if ((InlineObjectHash && method->intrinsic_id() == vmIntrinsics::_hashCode) || (method->intrinsic_id() == vmIntrinsics::_identityHashCode)) {
// Object.hashCode, System.identityHashCode can pull the hashCode from the
// header word instead of doing a full VM transition once it's been computed.
// Since hashCode is usually polymorphic at call sites we can't do this
// optimization at the call site without a lot of work.
Label slowCase;
Register receiver = O0;
Label done;
Register obj_reg = O0;
Register result = O0;
Register header = G3_scratch;
Register hash = G3_scratch; // overwrite header value with hash value
Register mask = G1; // to get hash field from header
// Unlike for Object.hashCode, System.identityHashCode is static method and
// gets object as argument instead of the receiver.
if (method->intrinsic_id() == vmIntrinsics::_identityHashCode) {
assert(method->is_static(), "method should be static");
// return 0 for null reference input
__ br_null(obj_reg, false, Assembler::pn, done);
__ delayed()->mov(obj_reg, hash);
}
// Read the header and build a mask to get its hash field. Give up if the object is not unlocked.
// We depend on hash_mask being at most 32 bits and avoid the use of
// hash_mask_in_place because it could be larger than 32 bits in a 64-bit
// vm: see markOop.hpp.
__ ld_ptr(receiver, oopDesc::mark_offset_in_bytes(), header);
__ ld_ptr(obj_reg, oopDesc::mark_offset_in_bytes(), header);
__ sethi(markOopDesc::hash_mask, mask);
__ btst(markOopDesc::unlocked_value, header);
__ br(Assembler::zero, false, Assembler::pn, slowCase);
@ -2054,6 +2064,7 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
__ delayed()->nop();
// leaf return.
__ bind(done);
__ retl();
__ delayed()->mov(hash, result);
__ bind(slowCase);

View file

@ -2361,7 +2361,7 @@ void Assembler::movdqa(XMMRegister dst, Address src) {
void Assembler::movdqu(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ true);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
attributes.set_address_attributes(/* tuple_type */ EVEX_FVM, /* input_size_in_bits */ EVEX_NObit);
simd_prefix(dst, xnoreg, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x6F);
@ -2398,7 +2398,7 @@ void Assembler::vmovdqu(XMMRegister dst, XMMRegister src) {
void Assembler::vmovdqu(XMMRegister dst, Address src) {
assert(UseAVX > 0, "");
InstructionMark im(this);
InstructionAttr attributes(AVX_256bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ true);
InstructionAttr attributes(AVX_256bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
attributes.set_address_attributes(/* tuple_type */ EVEX_FVM, /* input_size_in_bits */ EVEX_NObit);
vex_prefix(src, 0, dst->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x6F);
@ -2486,7 +2486,7 @@ void Assembler::evmovdqul(XMMRegister dst, XMMRegister src, int vector_len) {
void Assembler::evmovdqul(XMMRegister dst, Address src, int vector_len) {
assert(VM_Version::supports_evex(), "");
InstructionMark im(this);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ true);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false , /* uses_vl */ true);
attributes.set_address_attributes(/* tuple_type */ EVEX_FVM, /* input_size_in_bits */ EVEX_NObit);
vex_prefix(src, 0, dst->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x6F);
@ -2515,7 +2515,7 @@ void Assembler::evmovdquq(XMMRegister dst, XMMRegister src, int vector_len) {
void Assembler::evmovdquq(XMMRegister dst, Address src, int vector_len) {
assert(VM_Version::supports_evex(), "");
InstructionMark im(this);
InstructionAttr attributes(vector_len, /* vex_w */ true, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ true);
InstructionAttr attributes(vector_len, /* vex_w */ true, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
attributes.set_address_attributes(/* tuple_type */ EVEX_FVM, /* input_size_in_bits */ EVEX_NObit);
vex_prefix(src, 0, dst->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x6F);
@ -2640,7 +2640,7 @@ void Assembler::movsbl(Register dst, Register src) { // movsxb
void Assembler::movsd(XMMRegister dst, XMMRegister src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x10);
emit_int8((unsigned char)(0xC0 | encode));
@ -2649,7 +2649,7 @@ void Assembler::movsd(XMMRegister dst, XMMRegister src) {
void Assembler::movsd(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_64bit);
simd_prefix(dst, xnoreg, src, VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x10);
@ -2668,7 +2668,7 @@ void Assembler::movsd(Address dst, XMMRegister src) {
void Assembler::movss(XMMRegister dst, XMMRegister src) {
NOT_LP64(assert(VM_Version::supports_sse(), ""));
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x10);
emit_int8((unsigned char)(0xC0 | encode));
@ -2677,7 +2677,7 @@ void Assembler::movss(XMMRegister dst, XMMRegister src) {
void Assembler::movss(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_32bit);
simd_prefix(dst, xnoreg, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x10);
@ -2782,7 +2782,7 @@ void Assembler::mull(Register src) {
void Assembler::mulsd(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_64bit);
simd_prefix(dst, dst, src, VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x59);
@ -2791,7 +2791,7 @@ void Assembler::mulsd(XMMRegister dst, Address src) {
void Assembler::mulsd(XMMRegister dst, XMMRegister src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x59);
emit_int8((unsigned char)(0xC0 | encode));
@ -2800,7 +2800,7 @@ void Assembler::mulsd(XMMRegister dst, XMMRegister src) {
void Assembler::mulss(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_32bit);
simd_prefix(dst, dst, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x59);
@ -2809,7 +2809,7 @@ void Assembler::mulss(XMMRegister dst, Address src) {
void Assembler::mulss(XMMRegister dst, XMMRegister src) {
NOT_LP64(assert(VM_Version::supports_sse(), ""));
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x59);
emit_int8((unsigned char)(0xC0 | encode));
@ -3993,7 +3993,7 @@ void Assembler::smovl() {
void Assembler::sqrtsd(XMMRegister dst, XMMRegister src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x51);
emit_int8((unsigned char)(0xC0 | encode));
@ -4002,7 +4002,7 @@ void Assembler::sqrtsd(XMMRegister dst, XMMRegister src) {
void Assembler::sqrtsd(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_64bit);
simd_prefix(dst, dst, src, VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x51);
@ -4011,7 +4011,7 @@ void Assembler::sqrtsd(XMMRegister dst, Address src) {
void Assembler::sqrtss(XMMRegister dst, XMMRegister src) {
NOT_LP64(assert(VM_Version::supports_sse(), ""));
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x51);
emit_int8((unsigned char)(0xC0 | encode));
@ -4024,7 +4024,7 @@ void Assembler::std() {
void Assembler::sqrtss(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_32bit);
simd_prefix(dst, dst, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x51);
@ -4078,7 +4078,7 @@ void Assembler::subl(Register dst, Register src) {
void Assembler::subsd(XMMRegister dst, XMMRegister src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x5C);
emit_int8((unsigned char)(0xC0 | encode));
@ -4087,7 +4087,7 @@ void Assembler::subsd(XMMRegister dst, XMMRegister src) {
void Assembler::subsd(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_64bit);
simd_prefix(dst, dst, src, VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x5C);
@ -4096,7 +4096,7 @@ void Assembler::subsd(XMMRegister dst, Address src) {
void Assembler::subss(XMMRegister dst, XMMRegister src) {
NOT_LP64(assert(VM_Version::supports_sse(), ""));
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false , /* uses_vl */ false);
int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x5C);
emit_int8((unsigned char)(0xC0 | encode));
@ -4105,7 +4105,7 @@ void Assembler::subss(XMMRegister dst, XMMRegister src) {
void Assembler::subss(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse(), ""));
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* rex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_32bit);
simd_prefix(dst, dst, src, VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x5C);
@ -4293,7 +4293,7 @@ void Assembler::xorb(Register dst, Address src) {
void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_64bit);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
vex_prefix(src, nds_enc, dst->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
@ -4303,7 +4303,7 @@ void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, Address src) {
void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
int encode = vex_prefix_and_encode(dst->encoding(), nds_enc, src->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x58);
@ -4313,7 +4313,7 @@ void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
void Assembler::vaddss(XMMRegister dst, XMMRegister nds, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_32bit);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
vex_prefix(src, nds_enc, dst->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
@ -4323,7 +4323,7 @@ void Assembler::vaddss(XMMRegister dst, XMMRegister nds, Address src) {
void Assembler::vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
int encode = vex_prefix_and_encode(dst->encoding(), nds_enc, src->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x58);
@ -4333,7 +4333,7 @@ void Assembler::vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_64bit);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
vex_prefix(src, nds_enc, dst->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
@ -4343,7 +4343,7 @@ void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, Address src) {
void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
int encode = vex_prefix_and_encode(dst->encoding(), nds_enc, src->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x5E);
@ -4353,7 +4353,7 @@ void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
void Assembler::vdivss(XMMRegister dst, XMMRegister nds, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_32bit);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
vex_prefix(src, nds_enc, dst->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
@ -4363,7 +4363,7 @@ void Assembler::vdivss(XMMRegister dst, XMMRegister nds, Address src) {
void Assembler::vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
int encode = vex_prefix_and_encode(dst->encoding(), nds_enc, src->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x5E);
@ -4373,7 +4373,7 @@ void Assembler::vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_64bit);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
vex_prefix(src, nds_enc, dst->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
@ -4383,7 +4383,7 @@ void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, Address src) {
void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
int encode = vex_prefix_and_encode(dst->encoding(), nds_enc, src->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x59);
@ -4393,7 +4393,7 @@ void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
void Assembler::vmulss(XMMRegister dst, XMMRegister nds, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_32bit);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
vex_prefix(src, nds_enc, dst->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
@ -4403,7 +4403,7 @@ void Assembler::vmulss(XMMRegister dst, XMMRegister nds, Address src) {
void Assembler::vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
int encode = vex_prefix_and_encode(dst->encoding(), nds_enc, src->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x59);
@ -4413,7 +4413,7 @@ void Assembler::vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_64bit);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
vex_prefix(src, nds_enc, dst->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
@ -4423,7 +4423,7 @@ void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, Address src) {
void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ VM_Version::supports_evex(), /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
int encode = vex_prefix_and_encode(dst->encoding(), nds_enc, src->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F, &attributes);
emit_int8(0x5C);
@ -4433,7 +4433,7 @@ void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
void Assembler::vsubss(XMMRegister dst, XMMRegister nds, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_32bit);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
vex_prefix(src, nds_enc, dst->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
@ -4443,7 +4443,7 @@ void Assembler::vsubss(XMMRegister dst, XMMRegister nds, Address src) {
void Assembler::vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ true, /* uses_vl */ false);
InstructionAttr attributes(AVX_128bit, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ false);
int nds_enc = nds->is_valid() ? nds->encoding() : 0;
int encode = vex_prefix_and_encode(dst->encoding(), nds_enc, src->encoding(), VEX_SIMD_F3, VEX_OPCODE_0F, &attributes);
emit_int8(0x5C);
@ -5901,7 +5901,7 @@ void Assembler::vpbroadcastw(XMMRegister dst, XMMRegister src) {
// duplicate 1-byte integer data from src into 16||32|64 locations in dest : requires AVX512BW and AVX512VL
void Assembler::evpbroadcastb(XMMRegister dst, XMMRegister src, int vector_len) {
assert(VM_Version::supports_evex(), "");
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ _legacy_mode_bw, /* no_mask_reg */ true, /* uses_vl */ true);
int encode = vex_prefix_and_encode(dst->encoding(), 0, src->encoding(), VEX_SIMD_66, VEX_OPCODE_0F_38, &attributes);
emit_int8(0x78);
emit_int8((unsigned char)(0xC0 | encode));
@ -5911,7 +5911,7 @@ void Assembler::evpbroadcastb(XMMRegister dst, Address src, int vector_len) {
assert(VM_Version::supports_evex(), "");
assert(dst != xnoreg, "sanity");
InstructionMark im(this);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ _legacy_mode_bw, /* no_mask_reg */ true, /* uses_vl */ true);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_8bit);
// swap src<->dst for encoding
vex_prefix(src, dst->encoding(), dst->encoding(), VEX_SIMD_66, VEX_OPCODE_0F_38, &attributes);
@ -5922,7 +5922,7 @@ void Assembler::evpbroadcastb(XMMRegister dst, Address src, int vector_len) {
// duplicate 2-byte integer data from src into 8|16||32 locations in dest : requires AVX512BW and AVX512VL
void Assembler::evpbroadcastw(XMMRegister dst, XMMRegister src, int vector_len) {
assert(VM_Version::supports_evex(), "");
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ _legacy_mode_bw, /* no_mask_reg */ true, /* uses_vl */ true);
int encode = vex_prefix_and_encode(dst->encoding(), 0, src->encoding(), VEX_SIMD_66, VEX_OPCODE_0F_38, &attributes);
emit_int8(0x79);
emit_int8((unsigned char)(0xC0 | encode));
@ -5932,7 +5932,7 @@ void Assembler::evpbroadcastw(XMMRegister dst, Address src, int vector_len) {
assert(VM_Version::supports_evex(), "");
assert(dst != xnoreg, "sanity");
InstructionMark im(this);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ _legacy_mode_bw, /* no_mask_reg */ true, /* uses_vl */ true);
attributes.set_address_attributes(/* tuple_type */ EVEX_T1S, /* input_size_in_bits */ EVEX_16bit);
// swap src<->dst for encoding
vex_prefix(src, dst->encoding(), dst->encoding(), VEX_SIMD_66, VEX_OPCODE_0F_38, &attributes);
@ -6027,7 +6027,7 @@ void Assembler::evpbroadcastsd(XMMRegister dst, Address src, int vector_len) {
// duplicate 1-byte integer data from src into 16||32|64 locations in dest : requires AVX512BW and AVX512VL
void Assembler::evpbroadcastb(XMMRegister dst, Register src, int vector_len) {
assert(VM_Version::supports_evex(), "");
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ _legacy_mode_bw, /* no_mask_reg */ true, /* uses_vl */ true);
int encode = vex_prefix_and_encode(dst->encoding(), 0, src->encoding(), VEX_SIMD_66, VEX_OPCODE_0F_38, &attributes);
emit_int8(0x7A);
emit_int8((unsigned char)(0xC0 | encode));
@ -6036,7 +6036,7 @@ void Assembler::evpbroadcastb(XMMRegister dst, Register src, int vector_len) {
// duplicate 2-byte integer data from src into 8|16||32 locations in dest : requires AVX512BW and AVX512VL
void Assembler::evpbroadcastw(XMMRegister dst, Register src, int vector_len) {
assert(VM_Version::supports_evex(), "");
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ false, /* no_mask_reg */ false, /* uses_vl */ true);
InstructionAttr attributes(vector_len, /* vex_w */ false, /* legacy_mode */ _legacy_mode_bw, /* no_mask_reg */ true, /* uses_vl */ true);
int encode = vex_prefix_and_encode(dst->encoding(), 0, src->encoding(), VEX_SIMD_66, VEX_OPCODE_0F_38, &attributes);
emit_int8(0x7B);
emit_int8((unsigned char)(0xC0 | encode));

View file

@ -2044,11 +2044,11 @@ private:
class InstructionAttr {
public:
InstructionAttr(
int vector_len,
bool rex_vex_w,
bool legacy_mode,
bool no_reg_mask,
bool uses_vl)
int vector_len, // The length of vector to be applied in encoding - for both AVX and EVEX
bool rex_vex_w, // Width of data: if 32-bits or less, false, else if 64-bit or specially defined, true
bool legacy_mode, // Details if either this instruction is conditionally encoded to AVX or earlier if true else possibly EVEX
bool no_reg_mask, // when true, k0 is used when EVEX encoding is chosen, else k1 is used under the same condition
bool uses_vl) // This instruction may have legacy constraints based on vector length for EVEX
:
_avx_vector_len(vector_len),
_rex_vex_w(rex_vex_w),

View file

@ -46,6 +46,7 @@ define_pd_global(intx, OnStackReplacePercentage, 140);
define_pd_global(intx, ConditionalMoveLimit, 3);
define_pd_global(intx, FreqInlineSize, 325);
define_pd_global(intx, MinJumpTableSize, 10);
define_pd_global(intx, LoopPercentProfileLimit, 30);
#ifdef AMD64
define_pd_global(intx, INTPRESSURE, 13);
define_pd_global(intx, FLOATPRESSURE, 14);

View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2013, 2015, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013, 2016, 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
@ -101,7 +101,7 @@ void CodeInstaller::pd_patch_MetaspaceConstant(int pc_offset, Handle constant, T
}
}
void CodeInstaller::pd_patch_DataSectionReference(int pc_offset, int data_offset) {
void CodeInstaller::pd_patch_DataSectionReference(int pc_offset, int data_offset, TRAPS) {
address pc = _instructions->start() + pc_offset;
address operand = Assembler::locate_operand(pc, Assembler::disp32_operand);

View file

@ -7840,7 +7840,9 @@ void MacroAssembler::string_compare(Register str1, Register str2,
Label COMPARE_WIDE_VECTORS_LOOP_FAILED; // used only _LP64 && AVX3
int stride, stride2, adr_stride, adr_stride1, adr_stride2;
int stride2x2 = 0x40;
Address::ScaleFactor scale, scale1, scale2;
Address::ScaleFactor scale = Address::no_scale;
Address::ScaleFactor scale1 = Address::no_scale;
Address::ScaleFactor scale2 = Address::no_scale;
if (ae != StrIntrinsicNode::LL) {
stride2x2 = 0x20;
@ -7894,9 +7896,9 @@ void MacroAssembler::string_compare(Register str1, Register str2,
stride = 8;
}
} else {
scale = Address::no_scale; // not used
scale1 = Address::times_1;
scale2 = Address::times_2;
// scale not used
stride = 8;
}

View file

@ -0,0 +1,91 @@
/*
* Copyright (c) 2016, 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 "runtime/sharedRuntime.hpp"
#include "vmreg_x86.inline.hpp"
#ifdef COMPILER1
#include "c1/c1_Runtime1.hpp"
#endif //COMPILER1
#define __ masm->
#ifdef COMPILER1
// ---------------------------------------------------------------------------
// Object.hashCode, System.identityHashCode can pull the hashCode from the
// header word instead of doing a full VM transition once it's been computed.
// Since hashCode is usually polymorphic at call sites we can't do this
// optimization at the call site without a lot of work.
void SharedRuntime::inline_check_hashcode_from_object_header(MacroAssembler* masm,
methodHandle method,
Register obj_reg,
Register result) {
Label slowCase;
// Unlike for Object.hashCode, System.identityHashCode is static method and
// gets object as argument instead of the receiver.
if (method->intrinsic_id() == vmIntrinsics::_identityHashCode) {
Label Continue;
// return 0 for null reference input
__ cmpptr(obj_reg, (int32_t)NULL_WORD);
__ jcc(Assembler::notEqual, Continue);
__ xorptr(result, result);
__ ret(0);
__ bind(Continue);
}
__ movptr(result, Address(obj_reg, oopDesc::mark_offset_in_bytes()));
// check if locked
__ testptr(result, markOopDesc::unlocked_value);
__ jcc(Assembler::zero, slowCase);
if (UseBiasedLocking) {
// Check if biased and fall through to runtime if so
__ testptr(result, markOopDesc::biased_lock_bit_in_place);
__ jcc(Assembler::notZero, slowCase);
}
// get hash
#ifdef _LP64
// Read the header and build a mask to get its hash field.
// Depend on hash_mask being at most 32 bits and avoid the use of hash_mask_in_place
// because it could be larger than 32 bits in a 64-bit vm. See markOop.hpp.
__ shrptr(result, markOopDesc::hash_shift);
__ andptr(result, markOopDesc::hash_mask);
#else
__ andptr(result, markOopDesc::hash_mask_in_place);
#endif //_LP64
// test if hashCode exists
__ jcc(Assembler::zero, slowCase);
#ifndef _LP64
__ shrptr(result, markOopDesc::hash_shift);
#endif
__ ret(0);
__ bind(slowCase);
}
#endif //COMPILER1

View file

@ -1754,34 +1754,10 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
int vep_offset = ((intptr_t)__ pc()) - start;
#ifdef COMPILER1
if (InlineObjectHash && method->intrinsic_id() == vmIntrinsics::_hashCode) {
// Object.hashCode can pull the hashCode from the header word
// instead of doing a full VM transition once it's been computed.
// Since hashCode is usually polymorphic at call sites we can't do
// this optimization at the call site without a lot of work.
Label slowCase;
Register receiver = rcx;
Register result = rax;
__ movptr(result, Address(receiver, oopDesc::mark_offset_in_bytes()));
// check if locked
__ testptr(result, markOopDesc::unlocked_value);
__ jcc (Assembler::zero, slowCase);
if (UseBiasedLocking) {
// Check if biased and fall through to runtime if so
__ testptr(result, markOopDesc::biased_lock_bit_in_place);
__ jcc (Assembler::notZero, slowCase);
}
// get hash
__ andptr(result, markOopDesc::hash_mask_in_place);
// test if hashCode exists
__ jcc (Assembler::zero, slowCase);
__ shrptr(result, markOopDesc::hash_shift);
__ ret(0);
__ bind (slowCase);
}
// For Object.hashCode, System.identityHashCode try to pull hashCode from object header if available.
if ((InlineObjectHash && method->intrinsic_id() == vmIntrinsics::_hashCode) || (method->intrinsic_id() == vmIntrinsics::_identityHashCode)) {
inline_check_hashcode_from_object_header(masm, method, rcx /*obj_reg*/, rax /*result*/);
}
#endif // COMPILER1
// The instruction at the verified entry point must be 5 bytes or longer

View file

@ -2058,6 +2058,13 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
int vep_offset = ((intptr_t)__ pc()) - start;
#ifdef COMPILER1
// For Object.hashCode, System.identityHashCode try to pull hashCode from object header if available.
if ((InlineObjectHash && method->intrinsic_id() == vmIntrinsics::_hashCode) || (method->intrinsic_id() == vmIntrinsics::_identityHashCode)) {
inline_check_hashcode_from_object_header(masm, method, j_rarg0 /*obj_reg*/, rax /*result*/);
}
#endif // COMPILER1
// The instruction at the verified entry point must be 5 bytes or longer
// because it can be patched on the fly by make_non_entrant. The stack bang
// instruction fits that requirement.