void C1_MacroAssembler::initialize_object(
Register obj, // result: pointer to object after successful allocation
Register klass, // object klass
Register var_size_in_bytes, // object size in bytes if unknown at compile time; invalid otherwise
int con_size_in_bytes, // object size in bytes if known at compile time
Register t1, // temp register
Register t2 // temp register
) {
const int hdr_size_in_bytes = instanceOopDesc::header_size() * HeapWordSize;
initialize_header(obj, klass, noreg, t1, t2);
#ifdef ASSERT
{
lwz(t1, in_bytes(Klass::layout_helper_offset()), klass);
if (var_size_in_bytes != noreg) {
cmpw(CCR0, t1, var_size_in_bytes);
} else {
cmpwi(CCR0, t1, con_size_in_bytes);
}
asm_assert_eq("bad size in initialize_object", 0x753);
}
#endif
// Initialize body.
if (var_size_in_bytes != noreg) {
// Use a loop.
addi(t1, obj, hdr_size_in_bytes); // Compute address of first element.
addi(t2, var_size_in_bytes, -hdr_size_in_bytes); // Compute size of body.
initialize_body(t1, t2);
} else if (con_size_in_bytes > hdr_size_in_bytes) {
// Use a loop.
initialize_body(obj, t1, t2, con_size_in_bytes, hdr_size_in_bytes);
}
if (CURRENT_ENV->dtrace_alloc_probes()) {
Unimplemented();
// assert(obj == O0, "must be");
// call(CAST_FROM_FN_PTR(address, Runtime1::entry_for(Runtime1::dtrace_object_alloc_id)),
// relocInfo::runtime_call_type);
}
verify_oop(obj);
}
// Interpreter intrinsic for WeakReference.get().
// 1. Don't push a full blown frame and go on dispatching, but fetch the value
// into R8 and return quickly
// 2. If G1 is active we *must* execute this intrinsic for corrrectness:
// It contains a GC barrier which puts the reference into the satb buffer
// to indicate that someone holds a strong reference to the object the
// weak ref points to!
address InterpreterGenerator::generate_Reference_get_entry(void) {
// Code: _aload_0, _getfield, _areturn
// parameter size = 1
//
// The code that gets generated by this routine is split into 2 parts:
// 1. the "intrinsified" code for G1 (or any SATB based GC),
// 2. the slow path - which is an expansion of the regular method entry.
//
// Notes:
// * In the G1 code we do not check whether we need to block for
// a safepoint. If G1 is enabled then we must execute the specialized
// code for Reference.get (except when the Reference object is null)
// so that we can log the value in the referent field with an SATB
// update buffer.
// If the code for the getfield template is modified so that the
// G1 pre-barrier code is executed when the current method is
// Reference.get() then going through the normal method entry
// will be fine.
// * The G1 code can, however, check the receiver object (the instance
// of java.lang.Reference) and jump to the slow path if null. If the
// Reference object is null then we obviously cannot fetch the referent
// and so we don't need to call the G1 pre-barrier. Thus we can use the
// regular method entry code to generate the NPE.
//
// This code is based on generate_accessor_enty.
address entry = __ pc();
const int referent_offset = java_lang_ref_Reference::referent_offset;
guarantee(referent_offset > 0, "referent offset not initialized");
if (UseG1GC) {
Label slow_path;
// Debugging not possible, so can't use __ skip_if_jvmti_mode(slow_path, GR31_SCRATCH);
// In the G1 code we don't check if we need to reach a safepoint. We
// continue and the thread will safepoint at the next bytecode dispatch.
// If the receiver is null then it is OK to jump to the slow path.
__ ld(R3_RET, Interpreter::stackElementSize, CC_INTERP_ONLY(R17_tos) NOT_CC_INTERP(R15_esp)); // get receiver
// Check if receiver == NULL and go the slow path.
__ cmpdi(CCR0, R3_RET, 0);
__ beq(CCR0, slow_path);
// Load the value of the referent field.
__ load_heap_oop(R3_RET, referent_offset, R3_RET);
// Generate the G1 pre-barrier code to log the value of
// the referent field in an SATB buffer. Note with
// these parameters the pre-barrier does not generate
// the load of the previous value.
// Restore caller sp for c2i case.
#ifdef ASSERT
__ ld(R9_ARG7, 0, R1_SP);
__ ld(R10_ARG8, 0, R21_sender_SP);
__ cmpd(CCR0, R9_ARG7, R10_ARG8);
__ asm_assert_eq("backlink", 0x544);
#endif // ASSERT
__ mr(R1_SP, R21_sender_SP); // Cut the stack back to where the caller started.
__ g1_write_barrier_pre(noreg, // obj
noreg, // offset
R3_RET, // pre_val
R11_scratch1, // tmp
R12_scratch2, // tmp
true); // needs_frame
__ blr();
// Generate regular method entry.
__ bind(slow_path);
__ branch_to_entry(Interpreter::entry_for_kind(Interpreter::zerolocals), R11_scratch1);
__ flush();
return entry;
} else {
return generate_accessor_entry();
}
}
// Call an accessor method (assuming it is resolved, otherwise drop into
// vanilla (slow path) entry.
address InterpreterGenerator::generate_accessor_entry(void) {
if (!UseFastAccessorMethods && (!FLAG_IS_ERGO(UseFastAccessorMethods))) {
return NULL;
}
Label Lslow_path, Lacquire;
const Register
Rclass_or_obj = R3_ARG1,
Rconst_method = R4_ARG2,
Rcodes = Rconst_method,
Rcpool_cache = R5_ARG3,
Rscratch = R11_scratch1,
Rjvmti_mode = Rscratch,
Roffset = R12_scratch2,
Rflags = R6_ARG4,
Rbtable = R7_ARG5;
static address branch_table[number_of_states];
address entry = __ pc();
// Check for safepoint:
// Ditch this, real man don't need safepoint checks.
// Also check for JVMTI mode
// Check for null obj, take slow path if so.
__ ld(Rclass_or_obj, Interpreter::stackElementSize, CC_INTERP_ONLY(R17_tos) NOT_CC_INTERP(R15_esp));
__ lwz(Rjvmti_mode, thread_(interp_only_mode));
__ cmpdi(CCR1, Rclass_or_obj, 0);
__ cmpwi(CCR0, Rjvmti_mode, 0);
__ crorc(/*CCR0 eq*/2, /*CCR1 eq*/4+2, /*CCR0 eq*/2);
__ beq(CCR0, Lslow_path); // this==null or jvmti_mode!=0
// Do 2 things in parallel:
// 1. Load the index out of the first instruction word, which looks like this:
// <0x2a><0xb4><index (2 byte, native endianess)>.
// 2. Load constant pool cache base.
__ ld(Rconst_method, in_bytes(Method::const_offset()), R19_method);
__ ld(Rcpool_cache, in_bytes(ConstMethod::constants_offset()), Rconst_method);
__ lhz(Rcodes, in_bytes(ConstMethod::codes_offset()) + 2, Rconst_method); // Lower half of 32 bit field.
__ ld(Rcpool_cache, ConstantPool::cache_offset_in_bytes(), Rcpool_cache);
// Get the const pool entry by means of <index>.
const int codes_shift = exact_log2(in_words(ConstantPoolCacheEntry::size()) * BytesPerWord);
__ slwi(Rscratch, Rcodes, codes_shift); // (codes&0xFFFF)<<codes_shift
__ add(Rcpool_cache, Rscratch, Rcpool_cache);
// Check if cpool cache entry is resolved.
// We are resolved if the indices offset contains the current bytecode.
ByteSize cp_base_offset = ConstantPoolCache::base_offset();
// Big Endian:
__ lbz(Rscratch, in_bytes(cp_base_offset) + in_bytes(ConstantPoolCacheEntry::indices_offset()) + 7 - 2, Rcpool_cache);
__ cmpwi(CCR0, Rscratch, Bytecodes::_getfield);
__ bne(CCR0, Lslow_path);
__ isync(); // Order succeeding loads wrt. load of _indices field from cpool_cache.
// Finally, start loading the value: Get cp cache entry into regs.
__ ld(Rflags, in_bytes(cp_base_offset) + in_bytes(ConstantPoolCacheEntry::flags_offset()), Rcpool_cache);
__ ld(Roffset, in_bytes(cp_base_offset) + in_bytes(ConstantPoolCacheEntry::f2_offset()), Rcpool_cache);
// Following code is from templateTable::getfield_or_static
// Load pointer to branch table
__ load_const_optimized(Rbtable, (address)branch_table, Rscratch);
// Get volatile flag
__ rldicl(Rscratch, Rflags, 64-ConstantPoolCacheEntry::is_volatile_shift, 63); // extract volatile bit
// note: sync is needed before volatile load on PPC64
// Check field type
__ rldicl(Rflags, Rflags, 64-ConstantPoolCacheEntry::tos_state_shift, 64-ConstantPoolCacheEntry::tos_state_bits);
#ifdef ASSERT
Label LFlagInvalid;
__ cmpldi(CCR0, Rflags, number_of_states);
__ bge(CCR0, LFlagInvalid);
__ ld(R9_ARG7, 0, R1_SP);
__ ld(R10_ARG8, 0, R21_sender_SP);
__ cmpd(CCR0, R9_ARG7, R10_ARG8);
__ asm_assert_eq("backlink", 0x543);
#endif // ASSERT
__ mr(R1_SP, R21_sender_SP); // Cut the stack back to where the caller started.
// Load from branch table and dispatch (volatile case: one instruction ahead)
__ sldi(Rflags, Rflags, LogBytesPerWord);
__ cmpwi(CCR6, Rscratch, 1); // volatile?
if (support_IRIW_for_not_multiple_copy_atomic_cpu) {
__ sldi(Rscratch, Rscratch, exact_log2(BytesPerInstWord)); // volatile ? size of 1 instruction : 0
}
__ ldx(Rbtable, Rbtable, Rflags);
if (support_IRIW_for_not_multiple_copy_atomic_cpu) {
__ subf(Rbtable, Rscratch, Rbtable); // point to volatile/non-volatile entry point
}
__ mtctr(Rbtable);
__ bctr();
#ifdef ASSERT
__ bind(LFlagInvalid);
__ stop("got invalid flag", 0x6541);
bool all_uninitialized = true,
all_initialized = true;
for (int i = 0; i<number_of_states; ++i) {
all_uninitialized = all_uninitialized && (branch_table[i] == NULL);
all_initialized = all_initialized && (branch_table[i] != NULL);
}
assert(all_uninitialized != all_initialized, "consistency"); // either or
__ fence(); // volatile entry point (one instruction before non-volatile_entry point)
if (branch_table[vtos] == 0) branch_table[vtos] = __ pc(); // non-volatile_entry point
if (branch_table[dtos] == 0) branch_table[dtos] = __ pc(); // non-volatile_entry point
if (branch_table[ftos] == 0) branch_table[ftos] = __ pc(); // non-volatile_entry point
__ stop("unexpected type", 0x6551);
#endif
if (branch_table[itos] == 0) { // generate only once
__ align(32, 28, 28); // align load
__ fence(); // volatile entry point (one instruction before non-volatile_entry point)
branch_table[itos] = __ pc(); // non-volatile_entry point
__ lwax(R3_RET, Rclass_or_obj, Roffset);
__ beq(CCR6, Lacquire);
__ blr();
}
if (branch_table[ltos] == 0) { // generate only once
__ align(32, 28, 28); // align load
__ fence(); // volatile entry point (one instruction before non-volatile_entry point)
branch_table[ltos] = __ pc(); // non-volatile_entry point
__ ldx(R3_RET, Rclass_or_obj, Roffset);
__ beq(CCR6, Lacquire);
__ blr();
}
if (branch_table[btos] == 0) { // generate only once
__ align(32, 28, 28); // align load
__ fence(); // volatile entry point (one instruction before non-volatile_entry point)
branch_table[btos] = __ pc(); // non-volatile_entry point
__ lbzx(R3_RET, Rclass_or_obj, Roffset);
__ extsb(R3_RET, R3_RET);
__ beq(CCR6, Lacquire);
__ blr();
}
if (branch_table[ctos] == 0) { // generate only once
__ align(32, 28, 28); // align load
__ fence(); // volatile entry point (one instruction before non-volatile_entry point)
branch_table[ctos] = __ pc(); // non-volatile_entry point
__ lhzx(R3_RET, Rclass_or_obj, Roffset);
__ beq(CCR6, Lacquire);
__ blr();
}
if (branch_table[stos] == 0) { // generate only once
__ align(32, 28, 28); // align load
__ fence(); // volatile entry point (one instruction before non-volatile_entry point)
branch_table[stos] = __ pc(); // non-volatile_entry point
__ lhax(R3_RET, Rclass_or_obj, Roffset);
__ beq(CCR6, Lacquire);
__ blr();
}
if (branch_table[atos] == 0) { // generate only once
__ align(32, 28, 28); // align load
__ fence(); // volatile entry point (one instruction before non-volatile_entry point)
branch_table[atos] = __ pc(); // non-volatile_entry point
__ load_heap_oop(R3_RET, (RegisterOrConstant)Roffset, Rclass_or_obj);
__ verify_oop(R3_RET);
//__ dcbt(R3_RET); // prefetch
__ beq(CCR6, Lacquire);
__ blr();
}
__ align(32, 12);
__ bind(Lacquire);
__ twi_0(R3_RET);
__ isync(); // acquire
__ blr();
#ifdef ASSERT
for (int i = 0; i<number_of_states; ++i) {
assert(branch_table[i], "accessor_entry initialization");
//tty->print_cr("accessor_entry: branch_table[%d] = 0x%llx (opcode 0x%llx)", i, branch_table[i], *((unsigned int*)branch_table[i]));
}
#endif
__ bind(Lslow_path);
__ branch_to_entry(Interpreter::entry_for_kind(Interpreter::zerolocals), Rscratch);
__ flush();
return entry;
}
