0
|
1 /*
|
710
|
2 * Copyright 1997-2009 Sun Microsystems, Inc. All Rights Reserved.
|
0
|
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
|
|
4 *
|
|
5 * This code is free software; you can redistribute it and/or modify it
|
|
6 * under the terms of the GNU General Public License version 2 only, as
|
|
7 * published by the Free Software Foundation.
|
|
8 *
|
|
9 * This code is distributed in the hope that it will be useful, but WITHOUT
|
|
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
|
12 * version 2 for more details (a copy is included in the LICENSE file that
|
|
13 * accompanied this code).
|
|
14 *
|
|
15 * You should have received a copy of the GNU General Public License version
|
|
16 * 2 along with this work; if not, write to the Free Software Foundation,
|
|
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
|
|
18 *
|
|
19 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
|
|
20 * CA 95054 USA or visit www.sun.com if you need additional information or
|
|
21 * have any questions.
|
|
22 *
|
|
23 */
|
|
24
|
|
25 #include "incls/_precompiled.incl"
|
|
26 #include "incls/_interpreter_sparc.cpp.incl"
|
|
27
|
|
28
|
|
29
|
|
30 // Generation of Interpreter
|
|
31 //
|
|
32 // The InterpreterGenerator generates the interpreter into Interpreter::_code.
|
|
33
|
|
34
|
|
35 #define __ _masm->
|
|
36
|
|
37
|
|
38 //----------------------------------------------------------------------------------------------------
|
|
39
|
|
40
|
|
41
|
|
42
|
|
43 int AbstractInterpreter::BasicType_as_index(BasicType type) {
|
|
44 int i = 0;
|
|
45 switch (type) {
|
|
46 case T_BOOLEAN: i = 0; break;
|
|
47 case T_CHAR : i = 1; break;
|
|
48 case T_BYTE : i = 2; break;
|
|
49 case T_SHORT : i = 3; break;
|
|
50 case T_INT : i = 4; break;
|
|
51 case T_LONG : i = 5; break;
|
|
52 case T_VOID : i = 6; break;
|
|
53 case T_FLOAT : i = 7; break;
|
|
54 case T_DOUBLE : i = 8; break;
|
|
55 case T_OBJECT : i = 9; break;
|
|
56 case T_ARRAY : i = 9; break;
|
|
57 default : ShouldNotReachHere();
|
|
58 }
|
|
59 assert(0 <= i && i < AbstractInterpreter::number_of_result_handlers, "index out of bounds");
|
|
60 return i;
|
|
61 }
|
|
62
|
|
63
|
|
64 #ifndef _LP64
|
|
65 address AbstractInterpreterGenerator::generate_slow_signature_handler() {
|
|
66 address entry = __ pc();
|
|
67 Argument argv(0, true);
|
|
68
|
|
69 // We are in the jni transition frame. Save the last_java_frame corresponding to the
|
|
70 // outer interpreter frame
|
|
71 //
|
|
72 __ set_last_Java_frame(FP, noreg);
|
|
73 // make sure the interpreter frame we've pushed has a valid return pc
|
|
74 __ mov(O7, I7);
|
|
75 __ mov(Lmethod, G3_scratch);
|
|
76 __ mov(Llocals, G4_scratch);
|
|
77 __ save_frame(0);
|
|
78 __ mov(G2_thread, L7_thread_cache);
|
|
79 __ add(argv.address_in_frame(), O3);
|
|
80 __ mov(G2_thread, O0);
|
|
81 __ mov(G3_scratch, O1);
|
|
82 __ call(CAST_FROM_FN_PTR(address, InterpreterRuntime::slow_signature_handler), relocInfo::runtime_call_type);
|
|
83 __ delayed()->mov(G4_scratch, O2);
|
|
84 __ mov(L7_thread_cache, G2_thread);
|
|
85 __ reset_last_Java_frame();
|
|
86
|
|
87 // load the register arguments (the C code packed them as varargs)
|
|
88 for (Argument ldarg = argv.successor(); ldarg.is_register(); ldarg = ldarg.successor()) {
|
|
89 __ ld_ptr(ldarg.address_in_frame(), ldarg.as_register());
|
|
90 }
|
|
91 __ ret();
|
|
92 __ delayed()->
|
|
93 restore(O0, 0, Lscratch); // caller's Lscratch gets the result handler
|
|
94 return entry;
|
|
95 }
|
|
96
|
|
97
|
|
98 #else
|
|
99 // LP64 passes floating point arguments in F1, F3, F5, etc. instead of
|
|
100 // O0, O1, O2 etc..
|
|
101 // Doubles are passed in D0, D2, D4
|
|
102 // We store the signature of the first 16 arguments in the first argument
|
|
103 // slot because it will be overwritten prior to calling the native
|
|
104 // function, with the pointer to the JNIEnv.
|
|
105 // If LP64 there can be up to 16 floating point arguments in registers
|
|
106 // or 6 integer registers.
|
|
107 address AbstractInterpreterGenerator::generate_slow_signature_handler() {
|
|
108
|
|
109 enum {
|
|
110 non_float = 0,
|
|
111 float_sig = 1,
|
|
112 double_sig = 2,
|
|
113 sig_mask = 3
|
|
114 };
|
|
115
|
|
116 address entry = __ pc();
|
|
117 Argument argv(0, true);
|
|
118
|
|
119 // We are in the jni transition frame. Save the last_java_frame corresponding to the
|
|
120 // outer interpreter frame
|
|
121 //
|
|
122 __ set_last_Java_frame(FP, noreg);
|
|
123 // make sure the interpreter frame we've pushed has a valid return pc
|
|
124 __ mov(O7, I7);
|
|
125 __ mov(Lmethod, G3_scratch);
|
|
126 __ mov(Llocals, G4_scratch);
|
|
127 __ save_frame(0);
|
|
128 __ mov(G2_thread, L7_thread_cache);
|
|
129 __ add(argv.address_in_frame(), O3);
|
|
130 __ mov(G2_thread, O0);
|
|
131 __ mov(G3_scratch, O1);
|
|
132 __ call(CAST_FROM_FN_PTR(address, InterpreterRuntime::slow_signature_handler), relocInfo::runtime_call_type);
|
|
133 __ delayed()->mov(G4_scratch, O2);
|
|
134 __ mov(L7_thread_cache, G2_thread);
|
|
135 __ reset_last_Java_frame();
|
|
136
|
|
137
|
|
138 // load the register arguments (the C code packed them as varargs)
|
|
139 Address Sig = argv.address_in_frame(); // Argument 0 holds the signature
|
|
140 __ ld_ptr( Sig, G3_scratch ); // Get register argument signature word into G3_scratch
|
|
141 __ mov( G3_scratch, G4_scratch);
|
|
142 __ srl( G4_scratch, 2, G4_scratch); // Skip Arg 0
|
|
143 Label done;
|
|
144 for (Argument ldarg = argv.successor(); ldarg.is_float_register(); ldarg = ldarg.successor()) {
|
|
145 Label NonFloatArg;
|
|
146 Label LoadFloatArg;
|
|
147 Label LoadDoubleArg;
|
|
148 Label NextArg;
|
|
149 Address a = ldarg.address_in_frame();
|
|
150 __ andcc(G4_scratch, sig_mask, G3_scratch);
|
|
151 __ br(Assembler::zero, false, Assembler::pt, NonFloatArg);
|
|
152 __ delayed()->nop();
|
|
153
|
|
154 __ cmp(G3_scratch, float_sig );
|
|
155 __ br(Assembler::equal, false, Assembler::pt, LoadFloatArg);
|
|
156 __ delayed()->nop();
|
|
157
|
|
158 __ cmp(G3_scratch, double_sig );
|
|
159 __ br(Assembler::equal, false, Assembler::pt, LoadDoubleArg);
|
|
160 __ delayed()->nop();
|
|
161
|
|
162 __ bind(NonFloatArg);
|
|
163 // There are only 6 integer register arguments!
|
|
164 if ( ldarg.is_register() )
|
|
165 __ ld_ptr(ldarg.address_in_frame(), ldarg.as_register());
|
|
166 else {
|
|
167 // Optimization, see if there are any more args and get out prior to checking
|
|
168 // all 16 float registers. My guess is that this is rare.
|
|
169 // If is_register is false, then we are done the first six integer args.
|
|
170 __ tst(G4_scratch);
|
|
171 __ brx(Assembler::zero, false, Assembler::pt, done);
|
|
172 __ delayed()->nop();
|
|
173
|
|
174 }
|
|
175 __ ba(false, NextArg);
|
|
176 __ delayed()->srl( G4_scratch, 2, G4_scratch );
|
|
177
|
|
178 __ bind(LoadFloatArg);
|
|
179 __ ldf( FloatRegisterImpl::S, a, ldarg.as_float_register(), 4);
|
|
180 __ ba(false, NextArg);
|
|
181 __ delayed()->srl( G4_scratch, 2, G4_scratch );
|
|
182
|
|
183 __ bind(LoadDoubleArg);
|
|
184 __ ldf( FloatRegisterImpl::D, a, ldarg.as_double_register() );
|
|
185 __ ba(false, NextArg);
|
|
186 __ delayed()->srl( G4_scratch, 2, G4_scratch );
|
|
187
|
|
188 __ bind(NextArg);
|
|
189
|
|
190 }
|
|
191
|
|
192 __ bind(done);
|
|
193 __ ret();
|
|
194 __ delayed()->
|
|
195 restore(O0, 0, Lscratch); // caller's Lscratch gets the result handler
|
|
196 return entry;
|
|
197 }
|
|
198 #endif
|
|
199
|
|
200 void InterpreterGenerator::generate_counter_overflow(Label& Lcontinue) {
|
|
201
|
|
202 // Generate code to initiate compilation on the counter overflow.
|
|
203
|
|
204 // InterpreterRuntime::frequency_counter_overflow takes two arguments,
|
|
205 // the first indicates if the counter overflow occurs at a backwards branch (NULL bcp)
|
|
206 // and the second is only used when the first is true. We pass zero for both.
|
|
207 // The call returns the address of the verified entry point for the method or NULL
|
|
208 // if the compilation did not complete (either went background or bailed out).
|
|
209 __ set((int)false, O2);
|
|
210 __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::frequency_counter_overflow), O2, O2, true);
|
|
211 // returns verified_entry_point or NULL
|
|
212 // we ignore it in any case
|
|
213 __ ba(false, Lcontinue);
|
|
214 __ delayed()->nop();
|
|
215
|
|
216 }
|
|
217
|
|
218
|
|
219 // End of helpers
|
|
220
|
|
221 // Various method entries
|
|
222
|
|
223 // Abstract method entry
|
|
224 // Attempt to execute abstract method. Throw exception
|
|
225 //
|
|
226 address InterpreterGenerator::generate_abstract_entry(void) {
|
|
227 address entry = __ pc();
|
|
228 // abstract method entry
|
|
229 // throw exception
|
|
230 __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
|
|
231 // the call_VM checks for exception, so we should never return here.
|
|
232 __ should_not_reach_here();
|
|
233 return entry;
|
|
234
|
|
235 }
|
|
236
|
|
237
|
710
|
238
|
|
239 // Method handle invoker
|
|
240 // Dispatch a method of the form java.dyn.MethodHandles::invoke(...)
|
|
241 address InterpreterGenerator::generate_method_handle_entry(void) {
|
|
242 if (!EnableMethodHandles) {
|
|
243 return generate_abstract_entry();
|
|
244 }
|
|
245 return generate_abstract_entry(); //6815692//
|
|
246 }
|
|
247
|
|
248
|
|
249
|
|
250
|
0
|
251 //----------------------------------------------------------------------------------------------------
|
|
252 // Entry points & stack frame layout
|
|
253 //
|
|
254 // Here we generate the various kind of entries into the interpreter.
|
|
255 // The two main entry type are generic bytecode methods and native call method.
|
|
256 // These both come in synchronized and non-synchronized versions but the
|
|
257 // frame layout they create is very similar. The other method entry
|
|
258 // types are really just special purpose entries that are really entry
|
|
259 // and interpretation all in one. These are for trivial methods like
|
|
260 // accessor, empty, or special math methods.
|
|
261 //
|
|
262 // When control flow reaches any of the entry types for the interpreter
|
|
263 // the following holds ->
|
|
264 //
|
|
265 // C2 Calling Conventions:
|
|
266 //
|
|
267 // The entry code below assumes that the following registers are set
|
|
268 // when coming in:
|
|
269 // G5_method: holds the methodOop of the method to call
|
|
270 // Lesp: points to the TOS of the callers expression stack
|
|
271 // after having pushed all the parameters
|
|
272 //
|
|
273 // The entry code does the following to setup an interpreter frame
|
|
274 // pop parameters from the callers stack by adjusting Lesp
|
|
275 // set O0 to Lesp
|
|
276 // compute X = (max_locals - num_parameters)
|
|
277 // bump SP up by X to accomadate the extra locals
|
|
278 // compute X = max_expression_stack
|
|
279 // + vm_local_words
|
|
280 // + 16 words of register save area
|
|
281 // save frame doing a save sp, -X, sp growing towards lower addresses
|
|
282 // set Lbcp, Lmethod, LcpoolCache
|
|
283 // set Llocals to i0
|
|
284 // set Lmonitors to FP - rounded_vm_local_words
|
|
285 // set Lesp to Lmonitors - 4
|
|
286 //
|
|
287 // The frame has now been setup to do the rest of the entry code
|
|
288
|
|
289 // Try this optimization: Most method entries could live in a
|
|
290 // "one size fits all" stack frame without all the dynamic size
|
|
291 // calculations. It might be profitable to do all this calculation
|
|
292 // statically and approximately for "small enough" methods.
|
|
293
|
|
294 //-----------------------------------------------------------------------------------------------
|
|
295
|
|
296 // C1 Calling conventions
|
|
297 //
|
|
298 // Upon method entry, the following registers are setup:
|
|
299 //
|
|
300 // g2 G2_thread: current thread
|
|
301 // g5 G5_method: method to activate
|
|
302 // g4 Gargs : pointer to last argument
|
|
303 //
|
|
304 //
|
|
305 // Stack:
|
|
306 //
|
|
307 // +---------------+ <--- sp
|
|
308 // | |
|
|
309 // : reg save area :
|
|
310 // | |
|
|
311 // +---------------+ <--- sp + 0x40
|
|
312 // | |
|
|
313 // : extra 7 slots : note: these slots are not really needed for the interpreter (fix later)
|
|
314 // | |
|
|
315 // +---------------+ <--- sp + 0x5c
|
|
316 // | |
|
|
317 // : free :
|
|
318 // | |
|
|
319 // +---------------+ <--- Gargs
|
|
320 // | |
|
|
321 // : arguments :
|
|
322 // | |
|
|
323 // +---------------+
|
|
324 // | |
|
|
325 //
|
|
326 //
|
|
327 //
|
|
328 // AFTER FRAME HAS BEEN SETUP for method interpretation the stack looks like:
|
|
329 //
|
|
330 // +---------------+ <--- sp
|
|
331 // | |
|
|
332 // : reg save area :
|
|
333 // | |
|
|
334 // +---------------+ <--- sp + 0x40
|
|
335 // | |
|
|
336 // : extra 7 slots : note: these slots are not really needed for the interpreter (fix later)
|
|
337 // | |
|
|
338 // +---------------+ <--- sp + 0x5c
|
|
339 // | |
|
|
340 // : :
|
|
341 // | | <--- Lesp
|
|
342 // +---------------+ <--- Lmonitors (fp - 0x18)
|
|
343 // | VM locals |
|
|
344 // +---------------+ <--- fp
|
|
345 // | |
|
|
346 // : reg save area :
|
|
347 // | |
|
|
348 // +---------------+ <--- fp + 0x40
|
|
349 // | |
|
|
350 // : extra 7 slots : note: these slots are not really needed for the interpreter (fix later)
|
|
351 // | |
|
|
352 // +---------------+ <--- fp + 0x5c
|
|
353 // | |
|
|
354 // : free :
|
|
355 // | |
|
|
356 // +---------------+
|
|
357 // | |
|
|
358 // : nonarg locals :
|
|
359 // | |
|
|
360 // +---------------+
|
|
361 // | |
|
|
362 // : arguments :
|
|
363 // | | <--- Llocals
|
|
364 // +---------------+ <--- Gargs
|
|
365 // | |
|
|
366
|
|
367 address AbstractInterpreterGenerator::generate_method_entry(AbstractInterpreter::MethodKind kind) {
|
|
368 // determine code generation flags
|
|
369 bool synchronized = false;
|
|
370 address entry_point = NULL;
|
|
371
|
|
372 switch (kind) {
|
|
373 case Interpreter::zerolocals : break;
|
|
374 case Interpreter::zerolocals_synchronized: synchronized = true; break;
|
|
375 case Interpreter::native : entry_point = ((InterpreterGenerator*)this)->generate_native_entry(false); break;
|
|
376 case Interpreter::native_synchronized : entry_point = ((InterpreterGenerator*)this)->generate_native_entry(true); break;
|
|
377 case Interpreter::empty : entry_point = ((InterpreterGenerator*)this)->generate_empty_entry(); break;
|
|
378 case Interpreter::accessor : entry_point = ((InterpreterGenerator*)this)->generate_accessor_entry(); break;
|
|
379 case Interpreter::abstract : entry_point = ((InterpreterGenerator*)this)->generate_abstract_entry(); break;
|
710
|
380 case Interpreter::method_handle : entry_point = ((InterpreterGenerator*)this)->generate_method_handle_entry(); break;
|
0
|
381 case Interpreter::java_lang_math_sin : break;
|
|
382 case Interpreter::java_lang_math_cos : break;
|
|
383 case Interpreter::java_lang_math_tan : break;
|
|
384 case Interpreter::java_lang_math_sqrt : break;
|
|
385 case Interpreter::java_lang_math_abs : break;
|
|
386 case Interpreter::java_lang_math_log : break;
|
|
387 case Interpreter::java_lang_math_log10 : break;
|
|
388 default : ShouldNotReachHere(); break;
|
|
389 }
|
|
390
|
|
391 if (entry_point) return entry_point;
|
|
392
|
|
393 return ((InterpreterGenerator*)this)->generate_normal_entry(synchronized);
|
|
394 }
|
|
395
|
|
396
|
|
397 // This method tells the deoptimizer how big an interpreted frame must be:
|
|
398 int AbstractInterpreter::size_activation(methodOop method,
|
|
399 int tempcount,
|
|
400 int popframe_extra_args,
|
|
401 int moncount,
|
|
402 int callee_param_count,
|
|
403 int callee_locals,
|
|
404 bool is_top_frame) {
|
|
405 return layout_activation(method,
|
|
406 tempcount,
|
|
407 popframe_extra_args,
|
|
408 moncount,
|
|
409 callee_param_count,
|
|
410 callee_locals,
|
|
411 (frame*)NULL,
|
|
412 (frame*)NULL,
|
|
413 is_top_frame);
|
|
414 }
|
|
415
|
|
416 void Deoptimization::unwind_callee_save_values(frame* f, vframeArray* vframe_array) {
|
|
417
|
|
418 // This code is sort of the equivalent of C2IAdapter::setup_stack_frame back in
|
|
419 // the days we had adapter frames. When we deoptimize a situation where a
|
|
420 // compiled caller calls a compiled caller will have registers it expects
|
|
421 // to survive the call to the callee. If we deoptimize the callee the only
|
|
422 // way we can restore these registers is to have the oldest interpreter
|
|
423 // frame that we create restore these values. That is what this routine
|
|
424 // will accomplish.
|
|
425
|
|
426 // At the moment we have modified c2 to not have any callee save registers
|
|
427 // so this problem does not exist and this routine is just a place holder.
|
|
428
|
|
429 assert(f->is_interpreted_frame(), "must be interpreted");
|
|
430 }
|
|
431
|
|
432
|
|
433 //----------------------------------------------------------------------------------------------------
|
|
434 // Exceptions
|