static int lua_loadfile(lua_State *L, const std::string& fname, const std::string& relativename)
{
lua_filestream lfs(fname);
//lua uses '@' to know that this is a file (as opposed to something loaded via loadstring )
std::string chunkname = '@' + relativename;
LOG_LUA << "starting to read from " << fname << "\n";
return lua_load(L, &lua_filestream::lua_read_data, &lfs, chunkname.c_str(), "t");
}
address CppInterpreterGenerator::generate_stack_to_native_abi_converter(
BasicType type)
{
const Register stack = r5;
address start = __ pc();
switch (type) {
case T_VOID:
break;
case T_BOOLEAN:
case T_CHAR:
case T_BYTE:
case T_SHORT:
case T_INT:
__ load (stack, STATE(_stack));
__ lwa (r3, Address(stack, wordSize));
break;
case T_LONG:
__ load (stack, STATE(_stack));
__ load (r3, Address(stack, wordSize));
#ifdef PPC32
__ load (r4, Address(stack, wordSize * 2));
#endif
break;
case T_FLOAT:
__ load (stack, STATE(_stack));
__ lfs (f1, Address(stack, wordSize));
break;
case T_DOUBLE:
__ load (stack, STATE(_stack));
__ lfd (f1, Address(stack, wordSize));
break;
case T_OBJECT:
__ load (stack, STATE(_stack));
__ load (r3, Address(stack, wordSize));
__ verify_oop (r3);
break;
default:
ShouldNotReachHere();
}
__ blr ();
return start;
}
seg::Model* load_model(const char * model_path, const char * lexicon_path = NULL) {
if ((NULL == model_path)&&(NULL == lexicon_path)) {
return NULL;
}
seg::Model *mdl = new seg::Model;
if (NULL != model_path) {
std::ifstream mfs(model_path, std::ifstream::binary);
if (mfs) {
if (!mdl->load(mfs)) {
delete mdl;
mdl = 0;
return NULL;
}
} else {
delete mdl;
mdl = 0;
return NULL;
}
}
if (NULL != lexicon_path) {
std::ifstream lfs(lexicon_path);
if (lfs) {
std::string buffer;
while (std::getline(lfs, buffer)) {
buffer = ltp::strutils::chomp(buffer);
if (buffer.size() == 0) {
continue;
}
mdl->external_lexicon.set(buffer.c_str(), true);
}
}
}
beg_tag0 = mdl->labels.index( seg::__b__ );
beg_tag1 = mdl->labels.index( seg::__s__ );
if (!rule) {
rule = new seg::rulebase::RuleBase(mdl->labels);
}
return mdl;
}
static void load_model_constrain(Model * model , const char * lexicon_file = NULL) {
if (NULL != lexicon_file) {
std::ifstream lfs(lexicon_file);
if (lfs) {
std::string buffer;
std::vector<std::string> key_values;
int key_values_size;
std::string key;
int value;
Bitset * original_bitset;
while (std::getline(lfs, buffer)) {
buffer = ltp::strutils::chomp(buffer);
if (buffer.size() == 0) {
continue;
}
Bitset values;
key_values = ltp::strutils::split(buffer);
key_values_size = key_values.size();
if(key_values_size == 0 || key_values_size == 1) {
continue;
}
key = ltp::strutils::chartypes::sbc2dbc_x(key_values[0]);
for(int i=1;i<key_values_size;i++){
value = model->labels.index(key_values[i]);
if (value != -1){
if(!(values.set(value))) {
WARNING_LOG("Tag named %s for word %s add external lexicon error.",key_values[i].c_str(),key_values[0].c_str());
}
}
else {
WARNING_LOG("Tag named %s for word %s is not existed in LTP labels set.",key_values[i].c_str(),key_values[0].c_str());
}
}
if(!values.empty()) {
original_bitset = model->external_lexicon.get(key.c_str());
if(original_bitset){
original_bitset->merge(values);
}
else{
model->external_lexicon.set(key.c_str(),values);
}
}
}
}
}
}//end func load_model_constrain
int main( void )
{
printf("snoc list:\n");
char name[] = "Kate";
snoc_list res0 =
snoc(
snoc(
snoc(
snoc(NULL, name),
name + 1),
name + 2),
name + 3);
sprint( print_char, res0 );
printf("\n");
printf("\nntree :\n");
ntree res1 = br(snoc( snoc( NULL, lf(1) ), lfs( 3, 2, 3, 4 )));
nprint( res1 );
printf("\n");
return 0;
}
bool load(const char * model_file, const char * lexicon_file = NULL) {
std::ifstream mfs(model_file, std::ifstream::binary);
if (!mfs) {
return false;
}
model = new ltp::segmentor::Model;
if (!model->load(mfs)) {
delete model;
return false;
}
if (NULL != lexicon_file) {
std::ifstream lfs(lexicon_file);
if (lfs) {
std::string buffer;
while (std::getline(lfs, buffer)) {
buffer = ltp::strutils::chomp(buffer);
if (buffer.size() == 0) {
continue;
}
model->external_lexicon.set(buffer.c_str(), true);
}
}
}
// don't need to allocate a decoder
// one sentence, one decoder
baseAll = new ltp::segmentor::rulebase::RuleBase(model->labels);
beg_tag0 = model->labels.index( ltp::segmentor::__b__ );
beg_tag1 = model->labels.index( ltp::segmentor::__s__ );
return true;
}
address AbstractInterpreterGenerator::generate_slow_signature_handler() {
// Slow_signature handler that respects the PPC C calling conventions.
//
// We get called by the native entry code with our output register
// area == 8. First we call InterpreterRuntime::get_result_handler
// to copy the pointer to the signature string temporarily to the
// first C-argument and to return the result_handler in
// R3_RET. Since native_entry will copy the jni-pointer to the
// first C-argument slot later on, it is OK to occupy this slot
// temporarilly. Then we copy the argument list on the java
// expression stack into native varargs format on the native stack
// and load arguments into argument registers. Integer arguments in
// the varargs vector will be sign-extended to 8 bytes.
//
// On entry:
// R3_ARG1 - intptr_t* Address of java argument list in memory.
// R15_prev_state - BytecodeInterpreter* Address of interpreter state for
// this method
// R19_method
//
// On exit (just before return instruction):
// R3_RET - contains the address of the result_handler.
// R4_ARG2 - is not updated for static methods and contains "this" otherwise.
// R5_ARG3-R10_ARG8: - When the (i-2)th Java argument is not of type float or double,
// ARGi contains this argument. Otherwise, ARGi is not updated.
// F1_ARG1-F13_ARG13 - contain the first 13 arguments of type float or double.
const int LogSizeOfTwoInstructions = 3;
// FIXME: use Argument:: GL: Argument names different numbers!
const int max_fp_register_arguments = 13;
const int max_int_register_arguments = 6; // first 2 are reserved
const Register arg_java = R21_tmp1;
const Register arg_c = R22_tmp2;
const Register signature = R23_tmp3; // is string
const Register sig_byte = R24_tmp4;
const Register fpcnt = R25_tmp5;
const Register argcnt = R26_tmp6;
const Register intSlot = R27_tmp7;
const Register target_sp = R28_tmp8;
const FloatRegister floatSlot = F0;
address entry = __ function_entry();
__ save_LR_CR(R0);
__ save_nonvolatile_gprs(R1_SP, _spill_nonvolatiles_neg(r14));
// We use target_sp for storing arguments in the C frame.
__ mr(target_sp, R1_SP);
__ push_frame_reg_args_nonvolatiles(0, R11_scratch1);
__ mr(arg_java, R3_ARG1);
__ call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::get_signature), R16_thread, R19_method);
// Signature is in R3_RET. Signature is callee saved.
__ mr(signature, R3_RET);
// Get the result handler.
__ call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::get_result_handler), R16_thread, R19_method);
{
Label L;
// test if static
// _access_flags._flags must be at offset 0.
// TODO PPC port: requires change in shared code.
//assert(in_bytes(AccessFlags::flags_offset()) == 0,
// "MethodDesc._access_flags == MethodDesc._access_flags._flags");
// _access_flags must be a 32 bit value.
assert(sizeof(AccessFlags) == 4, "wrong size");
__ lwa(R11_scratch1/*access_flags*/, method_(access_flags));
// testbit with condition register.
__ testbitdi(CCR0, R0, R11_scratch1/*access_flags*/, JVM_ACC_STATIC_BIT);
__ btrue(CCR0, L);
// For non-static functions, pass "this" in R4_ARG2 and copy it
// to 2nd C-arg slot.
// We need to box the Java object here, so we use arg_java
// (address of current Java stack slot) as argument and don't
// dereference it as in case of ints, floats, etc.
__ mr(R4_ARG2, arg_java);
__ addi(arg_java, arg_java, -BytesPerWord);
__ std(R4_ARG2, _abi(carg_2), target_sp);
__ bind(L);
}
// Will be incremented directly after loop_start. argcnt=0
// corresponds to 3rd C argument.
__ li(argcnt, -1);
// arg_c points to 3rd C argument
__ addi(arg_c, target_sp, _abi(carg_3));
// no floating-point args parsed so far
__ li(fpcnt, 0);
Label move_intSlot_to_ARG, move_floatSlot_to_FARG;
Label loop_start, loop_end;
Label do_int, do_long, do_float, do_double, do_dontreachhere, do_object, do_array, do_boxed;
// signature points to '(' at entry
#ifdef ASSERT
__ lbz(sig_byte, 0, signature);
__ cmplwi(CCR0, sig_byte, '(');
__ bne(CCR0, do_dontreachhere);
#endif
__ bind(loop_start);
__ addi(argcnt, argcnt, 1);
__ lbzu(sig_byte, 1, signature);
__ cmplwi(CCR0, sig_byte, ')'); // end of signature
__ beq(CCR0, loop_end);
__ cmplwi(CCR0, sig_byte, 'B'); // byte
__ beq(CCR0, do_int);
__ cmplwi(CCR0, sig_byte, 'C'); // char
__ beq(CCR0, do_int);
__ cmplwi(CCR0, sig_byte, 'D'); // double
__ beq(CCR0, do_double);
__ cmplwi(CCR0, sig_byte, 'F'); // float
__ beq(CCR0, do_float);
__ cmplwi(CCR0, sig_byte, 'I'); // int
__ beq(CCR0, do_int);
__ cmplwi(CCR0, sig_byte, 'J'); // long
__ beq(CCR0, do_long);
__ cmplwi(CCR0, sig_byte, 'S'); // short
__ beq(CCR0, do_int);
__ cmplwi(CCR0, sig_byte, 'Z'); // boolean
__ beq(CCR0, do_int);
__ cmplwi(CCR0, sig_byte, 'L'); // object
__ beq(CCR0, do_object);
__ cmplwi(CCR0, sig_byte, '['); // array
__ beq(CCR0, do_array);
// __ cmplwi(CCR0, sig_byte, 'V'); // void cannot appear since we do not parse the return type
// __ beq(CCR0, do_void);
__ bind(do_dontreachhere);
__ unimplemented("ShouldNotReachHere in slow_signature_handler", 120);
__ bind(do_array);
{
Label start_skip, end_skip;
__ bind(start_skip);
__ lbzu(sig_byte, 1, signature);
__ cmplwi(CCR0, sig_byte, '[');
__ beq(CCR0, start_skip); // skip further brackets
__ cmplwi(CCR0, sig_byte, '9');
__ bgt(CCR0, end_skip); // no optional size
__ cmplwi(CCR0, sig_byte, '0');
__ bge(CCR0, start_skip); // skip optional size
__ bind(end_skip);
__ cmplwi(CCR0, sig_byte, 'L');
__ beq(CCR0, do_object); // for arrays of objects, the name of the object must be skipped
__ b(do_boxed); // otherwise, go directly to do_boxed
}
__ bind(do_object);
{
Label L;
__ bind(L);
__ lbzu(sig_byte, 1, signature);
__ cmplwi(CCR0, sig_byte, ';');
__ bne(CCR0, L);
}
// Need to box the Java object here, so we use arg_java (address of
// current Java stack slot) as argument and don't dereference it as
// in case of ints, floats, etc.
Label do_null;
__ bind(do_boxed);
__ ld(R0,0, arg_java);
__ cmpdi(CCR0, R0, 0);
__ li(intSlot,0);
__ beq(CCR0, do_null);
__ mr(intSlot, arg_java);
__ bind(do_null);
__ std(intSlot, 0, arg_c);
__ addi(arg_java, arg_java, -BytesPerWord);
__ addi(arg_c, arg_c, BytesPerWord);
__ cmplwi(CCR0, argcnt, max_int_register_arguments);
__ blt(CCR0, move_intSlot_to_ARG);
__ b(loop_start);
__ bind(do_int);
__ lwa(intSlot, 0, arg_java);
__ std(intSlot, 0, arg_c);
__ addi(arg_java, arg_java, -BytesPerWord);
__ addi(arg_c, arg_c, BytesPerWord);
__ cmplwi(CCR0, argcnt, max_int_register_arguments);
__ blt(CCR0, move_intSlot_to_ARG);
__ b(loop_start);
__ bind(do_long);
__ ld(intSlot, -BytesPerWord, arg_java);
__ std(intSlot, 0, arg_c);
__ addi(arg_java, arg_java, - 2 * BytesPerWord);
__ addi(arg_c, arg_c, BytesPerWord);
__ cmplwi(CCR0, argcnt, max_int_register_arguments);
__ blt(CCR0, move_intSlot_to_ARG);
__ b(loop_start);
__ bind(do_float);
__ lfs(floatSlot, 0, arg_java);
#if defined(LINUX)
// Linux uses ELF ABI. Both original ELF and ELFv2 ABIs have float
// in the least significant word of an argument slot.
#if defined(VM_LITTLE_ENDIAN)
__ stfs(floatSlot, 0, arg_c);
#else
__ stfs(floatSlot, 4, arg_c);
#endif
#elif defined(AIX)
// Although AIX runs on big endian CPU, float is in most significant
// word of an argument slot.
__ stfs(floatSlot, 0, arg_c);
#else
#error "unknown OS"
#endif
__ addi(arg_java, arg_java, -BytesPerWord);
__ addi(arg_c, arg_c, BytesPerWord);
__ cmplwi(CCR0, fpcnt, max_fp_register_arguments);
__ blt(CCR0, move_floatSlot_to_FARG);
__ b(loop_start);
__ bind(do_double);
__ lfd(floatSlot, - BytesPerWord, arg_java);
__ stfd(floatSlot, 0, arg_c);
__ addi(arg_java, arg_java, - 2 * BytesPerWord);
__ addi(arg_c, arg_c, BytesPerWord);
__ cmplwi(CCR0, fpcnt, max_fp_register_arguments);
__ blt(CCR0, move_floatSlot_to_FARG);
__ b(loop_start);
__ bind(loop_end);
__ pop_frame();
__ restore_nonvolatile_gprs(R1_SP, _spill_nonvolatiles_neg(r14));
__ restore_LR_CR(R0);
__ blr();
Label move_int_arg, move_float_arg;
__ bind(move_int_arg); // each case must consist of 2 instructions (otherwise adapt LogSizeOfTwoInstructions)
__ mr(R5_ARG3, intSlot); __ b(loop_start);
__ mr(R6_ARG4, intSlot); __ b(loop_start);
__ mr(R7_ARG5, intSlot); __ b(loop_start);
__ mr(R8_ARG6, intSlot); __ b(loop_start);
__ mr(R9_ARG7, intSlot); __ b(loop_start);
__ mr(R10_ARG8, intSlot); __ b(loop_start);
__ bind(move_float_arg); // each case must consist of 2 instructions (otherwise adapt LogSizeOfTwoInstructions)
__ fmr(F1_ARG1, floatSlot); __ b(loop_start);
__ fmr(F2_ARG2, floatSlot); __ b(loop_start);
__ fmr(F3_ARG3, floatSlot); __ b(loop_start);
__ fmr(F4_ARG4, floatSlot); __ b(loop_start);
__ fmr(F5_ARG5, floatSlot); __ b(loop_start);
__ fmr(F6_ARG6, floatSlot); __ b(loop_start);
__ fmr(F7_ARG7, floatSlot); __ b(loop_start);
__ fmr(F8_ARG8, floatSlot); __ b(loop_start);
__ fmr(F9_ARG9, floatSlot); __ b(loop_start);
__ fmr(F10_ARG10, floatSlot); __ b(loop_start);
__ fmr(F11_ARG11, floatSlot); __ b(loop_start);
__ fmr(F12_ARG12, floatSlot); __ b(loop_start);
__ fmr(F13_ARG13, floatSlot); __ b(loop_start);
__ bind(move_intSlot_to_ARG);
__ sldi(R0, argcnt, LogSizeOfTwoInstructions);
__ load_const(R11_scratch1, move_int_arg); // Label must be bound here.
__ add(R11_scratch1, R0, R11_scratch1);
__ mtctr(R11_scratch1/*branch_target*/);
__ bctr();
__ bind(move_floatSlot_to_FARG);
__ sldi(R0, fpcnt, LogSizeOfTwoInstructions);
__ addi(fpcnt, fpcnt, 1);
__ load_const(R11_scratch1, move_float_arg); // Label must be bound here.
__ add(R11_scratch1, R0, R11_scratch1);
__ mtctr(R11_scratch1/*branch_target*/);
__ bctr();
return entry;
}
int main( void )
{
ntree res = br(snoc( snoc( NULL, lf(1) ), lfs( 3, 2, 3, 4 )));
nprint( res );
return 0;
}
