address generate_d2i_wrapper( address fcn ) {
StubCodeMark mark(this, "StubRoutines", "d2i_wrapper");
address start = __ pc();
// Capture info about frame layout
enum layout { FPUState_off = 0,
ebp_off = FPUStateSizeInWords,
edi_off,
esi_off,
ecx_off,
ebx_off,
saved_argument_off,
saved_argument_off2, // 2nd half of double
framesize
};
assert(FPUStateSizeInWords == 27, "update stack layout");
// Save outgoing argument to stack across push_FPU_state()
__ subl(esp, wordSize * 2);
__ fstp_d(Address(esp));
// Save CPU & FPU state
__ pushl(ebx);
__ pushl(ecx);
__ pushl(esi);
__ pushl(edi);
__ pushl(ebp);
__ push_FPU_state();
// push_FPU_state() resets the FP top of stack
// Load original double into FP top of stack
__ fld_d(Address(esp, saved_argument_off * wordSize));
// Store double into stack as outgoing argument
__ subl(esp, wordSize*2);
__ fst_d(Address(esp));
// Prepare FPU for doing math in C-land
__ empty_FPU_stack();
// Call the C code to massage the double. Result in EAX
__ call_VM_leaf( fcn, 2 );
// Restore CPU & FPU state
__ pop_FPU_state();
__ popl(ebp);
__ popl(edi);
__ popl(esi);
__ popl(ecx);
__ popl(ebx);
__ addl(esp, wordSize * 2);
__ ret(0);
return start;
}
address TemplateInterpreterGenerator::generate_math_entry(AbstractInterpreter::MethodKind kind) {
// rbx,: Method*
// rcx: scratrch
// r13: sender sp
if (!InlineIntrinsics) return NULL; // Generate a vanilla entry
address entry_point = __ pc();
// These don't need a safepoint check because they aren't virtually
// callable. We won't enter these intrinsics from compiled code.
// If in the future we added an intrinsic which was virtually callable
// we'd have to worry about how to safepoint so that this code is used.
// mathematical functions inlined by compiler
// (interpreter must provide identical implementation
// in order to avoid monotonicity bugs when switching
// from interpreter to compiler in the middle of some
// computation)
//
// stack: [ ret adr ] <-- rsp
// [ lo(arg) ]
// [ hi(arg) ]
//
if (kind == Interpreter::java_lang_math_fmaD) {
__ movdbl(xmm0, Address(rsp, wordSize));
__ movdbl(xmm1, Address(rsp, 3 * wordSize));
__ movdbl(xmm2, Address(rsp, 5 * wordSize));
__ fmad(xmm0, xmm1, xmm2, xmm0);
} else if (kind == Interpreter::java_lang_math_fmaF) {
__ movflt(xmm0, Address(rsp, wordSize));
__ movflt(xmm1, Address(rsp, 2 * wordSize));
__ movflt(xmm2, Address(rsp, 3 * wordSize));
__ fmaf(xmm0, xmm1, xmm2, xmm0);
} else if (kind == Interpreter::java_lang_math_sqrt) {
__ sqrtsd(xmm0, Address(rsp, wordSize));
} else if (kind == Interpreter::java_lang_math_exp) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dexp() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dexp())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dexp));
}
} else if (kind == Interpreter::java_lang_math_log) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dlog() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dlog())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dlog));
}
} else if (kind == Interpreter::java_lang_math_log10) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dlog10() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dlog10())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dlog10));
}
} else if (kind == Interpreter::java_lang_math_sin) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dsin() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dsin())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dsin));
}
} else if (kind == Interpreter::java_lang_math_cos) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dcos() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dcos())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dcos));
}
} else if (kind == Interpreter::java_lang_math_pow) {
__ movdbl(xmm1, Address(rsp, wordSize));
__ movdbl(xmm0, Address(rsp, 3 * wordSize));
if (StubRoutines::dpow() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dpow())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dpow));
}
} else if (kind == Interpreter::java_lang_math_tan) {
__ movdbl(xmm0, Address(rsp, wordSize));
if (StubRoutines::dtan() != NULL) {
__ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::dtan())));
} else {
__ call_VM_leaf0(CAST_FROM_FN_PTR(address, SharedRuntime::dtan));
}
} else {
__ fld_d(Address(rsp, wordSize));
switch (kind) {
case Interpreter::java_lang_math_abs:
__ fabs();
break;
default:
ShouldNotReachHere();
}
// return double result in xmm0 for interpreter and compilers.
__ subptr(rsp, 2*wordSize);
// Round to 64bit precision
__ fstp_d(Address(rsp, 0));
__ movdbl(xmm0, Address(rsp, 0));
__ addptr(rsp, 2*wordSize);
}
__ pop(rax);
__ mov(rsp, r13);
__ jmp(rax);
return entry_point;
}
address JNI_FastGetField::generate_fast_get_float_field0(BasicType type) {
const char *name;
switch (type) {
case T_FLOAT: name = "jni_fast_GetFloatField"; break;
case T_DOUBLE: name = "jni_fast_GetDoubleField"; break;
default: ShouldNotReachHere();
}
ResourceMark rm;
BufferBlob* b = BufferBlob::create(name, BUFFER_SIZE*wordSize);
address fast_entry = b->instructions_begin();
CodeBuffer cbuf(fast_entry, b->instructions_size());
MacroAssembler* masm = new MacroAssembler(&cbuf);
Label slow_with_pop, slow;
// stack layout: offset from rsp (in words):
// return pc 0
// jni env 1
// obj 2
// jfieldID 3
ExternalAddress counter(SafepointSynchronize::safepoint_counter_addr());
__ mov32 (rcx, counter);
__ testb (rcx, 1);
__ jcc (Assembler::notZero, slow);
if (os::is_MP()) {
__ mov(rax, rcx);
__ andptr(rax, 1); // rax, must end up 0
__ movptr(rdx, Address(rsp, rax, Address::times_1, 2*wordSize));
// obj, notice rax, is 0.
// rdx is data dependent on rcx.
} else {
__ movptr(rdx, Address(rsp, 2*wordSize)); // obj
}
__ movptr(rax, Address(rsp, 3*wordSize)); // jfieldID
__ movptr(rdx, Address(rdx, 0)); // *obj
__ shrptr(rax, 2); // offset
assert(count < LIST_CAPACITY, "LIST_CAPACITY too small");
speculative_load_pclist[count] = __ pc();
switch (type) {
#ifndef _LP64
case T_FLOAT: __ fld_s (Address(rdx, rax, Address::times_1)); break;
case T_DOUBLE: __ fld_d (Address(rdx, rax, Address::times_1)); break;
#else
case T_FLOAT: __ movflt (xmm0, Address(robj, roffset, Address::times_1)); break;
case T_DOUBLE: __ movdbl (xmm0, Address(robj, roffset, Address::times_1)); break;
#endif // _LP64
default: ShouldNotReachHere();
}
Address ca1;
if (os::is_MP()) {
__ fst_s (Address(rsp, -4));
__ lea(rdx, counter);
__ movl (rax, Address(rsp, -4));
// garbage hi-order bits on 64bit are harmless.
__ xorptr(rdx, rax);
__ xorptr(rdx, rax);
__ cmp32(rcx, Address(rdx, 0));
// rax, ^ counter_addr ^ rax, = address
// ca1 is data dependent on the field
// access.
} else {
__ cmp32(rcx, counter);
}
__ jcc (Assembler::notEqual, slow_with_pop);
#ifndef _WINDOWS
__ ret (0);
#else
// __stdcall calling convention
__ ret (3*wordSize);
#endif
__ bind (slow_with_pop);
// invalid load. pop FPU stack.
__ fstp_d (0);
slowcase_entry_pclist[count++] = __ pc();
__ bind (slow);
address slow_case_addr;
switch (type) {
case T_FLOAT: slow_case_addr = jni_GetFloatField_addr(); break;
case T_DOUBLE: slow_case_addr = jni_GetDoubleField_addr(); break;
default: ShouldNotReachHere();
}
// tail call
__ jump (ExternalAddress(slow_case_addr));
__ flush ();
#ifndef _WINDOWS
return fast_entry;
#else
switch (type) {
case T_FLOAT: jni_fast_GetFloatField_fp = (GetFloatField_t)fast_entry; break;
case T_DOUBLE: jni_fast_GetDoubleField_fp = (GetDoubleField_t)fast_entry;
}
return os::win32::fast_jni_accessor_wrapper(type);
#endif
}
address InterpreterGenerator::generate_math_entry(AbstractInterpreter::MethodKind kind) {
// rbx,: methodOop
// rcx: scratrch
// rsi: sender sp
if (!InlineIntrinsics) return NULL; // Generate a vanilla entry
address entry_point = __ pc();
// These don't need a safepoint check because they aren't virtually
// callable. We won't enter these intrinsics from compiled code.
// If in the future we added an intrinsic which was virtually callable
// we'd have to worry about how to safepoint so that this code is used.
// mathematical functions inlined by compiler
// (interpreter must provide identical implementation
// in order to avoid monotonicity bugs when switching
// from interpreter to compiler in the middle of some
// computation)
//
// stack: [ ret adr ] <-- rsp
// [ lo(arg) ]
// [ hi(arg) ]
//
// Note: For JDK 1.2 StrictMath doesn't exist and Math.sin/cos/sqrt are
// native methods. Interpreter::method_kind(...) does a check for
// native methods first before checking for intrinsic methods and
// thus will never select this entry point. Make sure it is not
// called accidentally since the SharedRuntime entry points will
// not work for JDK 1.2.
//
// We no longer need to check for JDK 1.2 since it's EOL'ed.
// The following check existed in pre 1.6 implementation,
// if (Universe::is_jdk12x_version()) {
// __ should_not_reach_here();
// }
// Universe::is_jdk12x_version() always returns false since
// the JDK version is not yet determined when this method is called.
// This method is called during interpreter_init() whereas
// JDK version is only determined when universe2_init() is called.
// Note: For JDK 1.3 StrictMath exists and Math.sin/cos/sqrt are
// java methods. Interpreter::method_kind(...) will select
// this entry point for the corresponding methods in JDK 1.3.
// get argument
__ fld_d(Address(rsp, 1*wordSize));
switch (kind) {
case Interpreter::java_lang_math_sin :
__ trigfunc('s');
break;
case Interpreter::java_lang_math_cos :
__ trigfunc('c');
break;
case Interpreter::java_lang_math_tan :
__ trigfunc('t');
break;
case Interpreter::java_lang_math_sqrt:
__ fsqrt();
break;
case Interpreter::java_lang_math_abs:
__ fabs();
break;
case Interpreter::java_lang_math_log:
__ flog();
// Store to stack to convert 80bit precision back to 64bits
__ push_fTOS();
__ pop_fTOS();
break;
case Interpreter::java_lang_math_log10:
__ flog10();
// Store to stack to convert 80bit precision back to 64bits
__ push_fTOS();
__ pop_fTOS();
break;
default :
ShouldNotReachHere();
}
// return double result in xmm0 for interpreter and compilers.
if (UseSSE >= 2) {
__ subptr(rsp, 2*wordSize);
__ fstp_d(Address(rsp, 0));
__ movdbl(xmm0, Address(rsp, 0));
__ addptr(rsp, 2*wordSize);
}
// done, result in FPU ST(0) or XMM0
__ pop(rdi); // get return address
__ mov(rsp, rsi); // set sp to sender sp
__ jmp(rdi);
return entry_point;
}
address generate_call_stub(address& return_address) {
StubCodeMark mark(this, "StubRoutines", "call_stub");
address start = __ pc();
// stub code parameters / addresses
assert(frame::entry_frame_call_wrapper_offset == 2, "adjust this code");
bool sse_save = false;
const Address esp_after_call(ebp, -4 * wordSize); // same as in generate_catch_exception()!
const Address mxcsr_save (ebp, -4 * wordSize);
const Address result (ebp, 3 * wordSize);
const Address result_type (ebp, 4 * wordSize);
const Address method (ebp, 5 * wordSize);
const Address entry_point (ebp, 6 * wordSize);
const Address parameters (ebp, 7 * wordSize);
const Address parameter_size(ebp, 8 * wordSize);
const Address thread (ebp, 9 * wordSize); // same as in generate_catch_exception()!
#ifdef COMPILER2
sse_save = VM_Version::supports_sse();
#endif
// stub code
__ enter();
// save edi, esi, & ebx, according to C calling conventions
__ pushl(edi);
__ pushl(esi);
__ pushl(ebx);
__ subl(esp, wordSize); // space for %mxcsr save
// save and initialize %mxcsr
if (sse_save) {
__ stmxcsr(mxcsr_save);
__ ldmxcsr(Address((int) StubRoutines::addr_mxcsr_std(), relocInfo::none));
}
#ifdef ASSERT
// make sure we have no pending exceptions
{ Label L;
__ movl(ecx, thread);
__ cmpl(Address(ecx, Thread::pending_exception_offset()), (int)NULL);
__ jcc(Assembler::equal, L);
__ stop("StubRoutines::call_stub: entered with pending exception");
__ bind(L);
}
#endif
// pass parameters if any
Label parameters_done;
__ movl(ecx, parameter_size); // parameter counter
__ testl(ecx, ecx);
__ jcc(Assembler::zero, parameters_done);
// parameter passing loop
Label loop;
__ movl(edx, parameters); // parameter pointer
__ movl(esi, ecx); // parameter counter is in esi now
__ movl(ecx, Address(edx)); // get first parameter in case it is a receiver
__ bind(loop);
__ movl(eax, Address(edx)); // get parameter
__ addl(edx, wordSize); // advance to next parameter
__ decl(esi); // decrement counter
__ pushl(eax); // pass parameter
__ jcc(Assembler::notZero, loop);
// call Java function
__ bind(parameters_done);
__ movl(ebx, method); // get methodOop
__ movl(esi, entry_point); // get entry_point
__ call(esi, relocInfo::none);
return_address = __ pc();
// store result depending on type
// (everything that is not T_LONG, T_FLOAT or T_DOUBLE is treated as T_INT)
__ movl(edi, result);
Label is_long, is_float, is_double, exit;
__ movl(esi, result_type);
__ cmpl(esi, T_LONG);
__ jcc(Assembler::equal, is_long);
__ cmpl(esi, T_FLOAT);
__ jcc(Assembler::equal, is_float);
__ cmpl(esi, T_DOUBLE);
__ jcc(Assembler::equal, is_double);
// handle T_INT case
__ movl(Address(edi), eax);
__ bind(exit);
// pop parameters
__ movl(ecx, parameter_size);
__ leal(esp, Address(esp, ecx, Address::times_4));
// check if parameters have been popped correctly
#ifdef ASSERT
Label esp_wrong;
__ leal(edi, esp_after_call);
__ cmpl(esp, edi);
__ jcc(Assembler::notEqual, esp_wrong);
#endif
// restore %mxcsr
if (sse_save) {
__ ldmxcsr(mxcsr_save);
}
// restore edi & esi
__ addl(esp, wordSize); // remove %mxcsr save area
__ popl(ebx);
__ popl(esi);
__ popl(edi);
// return
__ popl(ebp);
__ ret(0);
// handle return types different from T_INT
__ bind(is_long);
__ movl(Address(edi, 0 * wordSize), eax);
__ movl(Address(edi, 1 * wordSize), edx);
__ jmp(exit);
__ bind(is_float);
__ fstp_s(Address(edi));
__ jmp(exit);
__ bind(is_double);
__ fstp_d(Address(edi));
__ jmp(exit);
#ifdef ASSERT
// stack pointer misadjusted
__ bind(esp_wrong);
__ stop("esp wrong after Java call");
#endif
return start;
}
void MacroAssembler::fast_exp(XMMRegister xmm0, XMMRegister xmm1, XMMRegister xmm2, XMMRegister xmm3, XMMRegister xmm4, XMMRegister xmm5, XMMRegister xmm6, XMMRegister xmm7, Register eax, Register ecx, Register edx, Register tmp) {
Label L_2TAG_PACKET_0_0_2, L_2TAG_PACKET_1_0_2, L_2TAG_PACKET_2_0_2, L_2TAG_PACKET_3_0_2;
Label L_2TAG_PACKET_4_0_2, L_2TAG_PACKET_5_0_2, L_2TAG_PACKET_6_0_2, L_2TAG_PACKET_7_0_2;
Label L_2TAG_PACKET_8_0_2, L_2TAG_PACKET_9_0_2, L_2TAG_PACKET_10_0_2, L_2TAG_PACKET_11_0_2;
Label L_2TAG_PACKET_12_0_2, L_2TAG_PACKET_13_0_2, B1_3, B1_5, start;
assert_different_registers(tmp, eax, ecx, edx);
jmp(start);
address static_const_table = (address)_static_const_table;
bind(start);
subl(rsp, 120);
movl(Address(rsp, 64), tmp);
lea(tmp, ExternalAddress(static_const_table));
movdqu(xmm0, Address(rsp, 128));
unpcklpd(xmm0, xmm0);
movdqu(xmm1, Address(tmp, 64)); // 0x652b82feUL, 0x40571547UL, 0x652b82feUL, 0x40571547UL
movdqu(xmm6, Address(tmp, 48)); // 0x00000000UL, 0x43380000UL, 0x00000000UL, 0x43380000UL
movdqu(xmm2, Address(tmp, 80)); // 0xfefa0000UL, 0x3f862e42UL, 0xfefa0000UL, 0x3f862e42UL
movdqu(xmm3, Address(tmp, 96)); // 0xbc9e3b3aUL, 0x3d1cf79aUL, 0xbc9e3b3aUL, 0x3d1cf79aUL
pextrw(eax, xmm0, 3);
andl(eax, 32767);
movl(edx, 16527);
subl(edx, eax);
subl(eax, 15504);
orl(edx, eax);
cmpl(edx, INT_MIN);
jcc(Assembler::aboveEqual, L_2TAG_PACKET_0_0_2);
mulpd(xmm1, xmm0);
addpd(xmm1, xmm6);
movapd(xmm7, xmm1);
subpd(xmm1, xmm6);
mulpd(xmm2, xmm1);
movdqu(xmm4, Address(tmp, 128)); // 0xe3289860UL, 0x3f56c15cUL, 0x555b9e25UL, 0x3fa55555UL
mulpd(xmm3, xmm1);
movdqu(xmm5, Address(tmp, 144)); // 0xc090cf0fUL, 0x3f811115UL, 0x55548ba1UL, 0x3fc55555UL
subpd(xmm0, xmm2);
movdl(eax, xmm7);
movl(ecx, eax);
andl(ecx, 63);
shll(ecx, 4);
sarl(eax, 6);
movl(edx, eax);
movdqu(xmm6, Address(tmp, 16)); // 0xffffffc0UL, 0x00000000UL, 0xffffffc0UL, 0x00000000UL
pand(xmm7, xmm6);
movdqu(xmm6, Address(tmp, 32)); // 0x0000ffc0UL, 0x00000000UL, 0x0000ffc0UL, 0x00000000UL
paddq(xmm7, xmm6);
psllq(xmm7, 46);
subpd(xmm0, xmm3);
movdqu(xmm2, Address(tmp, ecx, Address::times_1, 160));
mulpd(xmm4, xmm0);
movapd(xmm6, xmm0);
movapd(xmm1, xmm0);
mulpd(xmm6, xmm6);
mulpd(xmm0, xmm6);
addpd(xmm5, xmm4);
mulsd(xmm0, xmm6);
mulpd(xmm6, Address(tmp, 112)); // 0xfffffffeUL, 0x3fdfffffUL, 0xfffffffeUL, 0x3fdfffffUL
addsd(xmm1, xmm2);
unpckhpd(xmm2, xmm2);
mulpd(xmm0, xmm5);
addsd(xmm1, xmm0);
por(xmm2, xmm7);
unpckhpd(xmm0, xmm0);
addsd(xmm0, xmm1);
addsd(xmm0, xmm6);
addl(edx, 894);
cmpl(edx, 1916);
jcc (Assembler::above, L_2TAG_PACKET_1_0_2);
mulsd(xmm0, xmm2);
addsd(xmm0, xmm2);
jmp(L_2TAG_PACKET_2_0_2);
bind(L_2TAG_PACKET_1_0_2);
fnstcw(Address(rsp, 24));
movzwl(edx, Address(rsp, 24));
orl(edx, 768);
movw(Address(rsp, 28), edx);
fldcw(Address(rsp, 28));
movl(edx, eax);
sarl(eax, 1);
subl(edx, eax);
movdqu(xmm6, Address(tmp, 0)); // 0x00000000UL, 0xfff00000UL, 0x00000000UL, 0xfff00000UL
pandn(xmm6, xmm2);
addl(eax, 1023);
movdl(xmm3, eax);
psllq(xmm3, 52);
por(xmm6, xmm3);
addl(edx, 1023);
movdl(xmm4, edx);
psllq(xmm4, 52);
movsd(Address(rsp, 8), xmm0);
fld_d(Address(rsp, 8));
movsd(Address(rsp, 16), xmm6);
fld_d(Address(rsp, 16));
fmula(1);
faddp(1);
movsd(Address(rsp, 8), xmm4);
fld_d(Address(rsp, 8));
fmulp(1);
fstp_d(Address(rsp, 8));
movsd(xmm0,Address(rsp, 8));
fldcw(Address(rsp, 24));
pextrw(ecx, xmm0, 3);
andl(ecx, 32752);
cmpl(ecx, 32752);
jcc(Assembler::greaterEqual, L_2TAG_PACKET_3_0_2);
cmpl(ecx, 0);
jcc(Assembler::equal, L_2TAG_PACKET_4_0_2);
jmp(L_2TAG_PACKET_2_0_2);
cmpl(ecx, INT_MIN);
jcc(Assembler::less, L_2TAG_PACKET_3_0_2);
cmpl(ecx, -1064950997);
jcc(Assembler::less, L_2TAG_PACKET_2_0_2);
jcc(Assembler::greater, L_2TAG_PACKET_4_0_2);
movl(edx, Address(rsp, 128));
cmpl(edx ,-17155601);
jcc(Assembler::less, L_2TAG_PACKET_2_0_2);
jmp(L_2TAG_PACKET_4_0_2);
bind(L_2TAG_PACKET_3_0_2);
movl(edx, 14);
jmp(L_2TAG_PACKET_5_0_2);
bind(L_2TAG_PACKET_4_0_2);
movl(edx, 15);
bind(L_2TAG_PACKET_5_0_2);
movsd(Address(rsp, 0), xmm0);
movsd(xmm0, Address(rsp, 128));
fld_d(Address(rsp, 0));
jmp(L_2TAG_PACKET_6_0_2);
bind(L_2TAG_PACKET_7_0_2);
cmpl(eax, 2146435072);
jcc(Assembler::greaterEqual, L_2TAG_PACKET_8_0_2);
movl(eax, Address(rsp, 132));
cmpl(eax, INT_MIN);
jcc(Assembler::greaterEqual, L_2TAG_PACKET_9_0_2);
movsd(xmm0, Address(tmp, 1208)); // 0xffffffffUL, 0x7fefffffUL
mulsd(xmm0, xmm0);
movl(edx, 14);
jmp(L_2TAG_PACKET_5_0_2);
bind(L_2TAG_PACKET_9_0_2);
movsd(xmm0, Address(tmp, 1216));
mulsd(xmm0, xmm0);
movl(edx, 15);
jmp(L_2TAG_PACKET_5_0_2);
bind(L_2TAG_PACKET_8_0_2);
movl(edx, Address(rsp, 128));
cmpl(eax, 2146435072);
jcc(Assembler::above, L_2TAG_PACKET_10_0_2);
cmpl(edx, 0);
jcc(Assembler::notEqual, L_2TAG_PACKET_10_0_2);
movl(eax, Address(rsp, 132));
cmpl(eax, 2146435072);
jcc(Assembler::notEqual, L_2TAG_PACKET_11_0_2);
movsd(xmm0, Address(tmp, 1192)); // 0x00000000UL, 0x7ff00000UL
jmp(L_2TAG_PACKET_2_0_2);
bind(L_2TAG_PACKET_11_0_2);
movsd(xmm0, Address(tmp, 1200)); // 0x00000000UL, 0x00000000UL
jmp(L_2TAG_PACKET_2_0_2);
bind(L_2TAG_PACKET_10_0_2);
movsd(xmm0, Address(rsp, 128));
addsd(xmm0, xmm0);
jmp(L_2TAG_PACKET_2_0_2);
bind(L_2TAG_PACKET_0_0_2);
movl(eax, Address(rsp, 132));
andl(eax, 2147483647);
cmpl(eax, 1083179008);
jcc(Assembler::aboveEqual, L_2TAG_PACKET_7_0_2);
movsd(xmm0, Address(rsp, 128));
addsd(xmm0, Address(tmp, 1184)); // 0x00000000UL, 0x3ff00000UL
jmp(L_2TAG_PACKET_2_0_2);
bind(L_2TAG_PACKET_2_0_2);
movsd(Address(rsp, 48), xmm0);
fld_d(Address(rsp, 48));
bind(L_2TAG_PACKET_6_0_2);
movl(tmp, Address(rsp, 64));
}
