Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
PojavLauncherTeam
GitHub Repository: PojavLauncherTeam/mobile
Path: blob/master/src/hotspot/share/opto/divnode.cpp
40930 views
1
/*
2
* Copyright (c) 1997, 2021, Oracle and/or its affiliates. All rights reserved.
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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20
* or visit www.oracle.com if you need additional information or have any
21
* questions.
22
*
23
*/
24
25
#include "precompiled.hpp"
26
#include "memory/allocation.inline.hpp"
27
#include "opto/addnode.hpp"
28
#include "opto/connode.hpp"
29
#include "opto/convertnode.hpp"
30
#include "opto/divnode.hpp"
31
#include "opto/machnode.hpp"
32
#include "opto/movenode.hpp"
33
#include "opto/matcher.hpp"
34
#include "opto/mulnode.hpp"
35
#include "opto/phaseX.hpp"
36
#include "opto/subnode.hpp"
37
#include "utilities/powerOfTwo.hpp"
38
39
// Portions of code courtesy of Clifford Click
40
41
// Optimization - Graph Style
42
43
#include <math.h>
44
45
//----------------------magic_int_divide_constants-----------------------------
46
// Compute magic multiplier and shift constant for converting a 32 bit divide
47
// by constant into a multiply/shift/add series. Return false if calculations
48
// fail.
49
//
50
// Borrowed almost verbatim from Hacker's Delight by Henry S. Warren, Jr. with
51
// minor type name and parameter changes.
52
static bool magic_int_divide_constants(jint d, jint &M, jint &s) {
53
int32_t p;
54
uint32_t ad, anc, delta, q1, r1, q2, r2, t;
55
const uint32_t two31 = 0x80000000L; // 2**31.
56
57
ad = ABS(d);
58
if (d == 0 || d == 1) return false;
59
t = two31 + ((uint32_t)d >> 31);
60
anc = t - 1 - t%ad; // Absolute value of nc.
61
p = 31; // Init. p.
62
q1 = two31/anc; // Init. q1 = 2**p/|nc|.
63
r1 = two31 - q1*anc; // Init. r1 = rem(2**p, |nc|).
64
q2 = two31/ad; // Init. q2 = 2**p/|d|.
65
r2 = two31 - q2*ad; // Init. r2 = rem(2**p, |d|).
66
do {
67
p = p + 1;
68
q1 = 2*q1; // Update q1 = 2**p/|nc|.
69
r1 = 2*r1; // Update r1 = rem(2**p, |nc|).
70
if (r1 >= anc) { // (Must be an unsigned
71
q1 = q1 + 1; // comparison here).
72
r1 = r1 - anc;
73
}
74
q2 = 2*q2; // Update q2 = 2**p/|d|.
75
r2 = 2*r2; // Update r2 = rem(2**p, |d|).
76
if (r2 >= ad) { // (Must be an unsigned
77
q2 = q2 + 1; // comparison here).
78
r2 = r2 - ad;
79
}
80
delta = ad - r2;
81
} while (q1 < delta || (q1 == delta && r1 == 0));
82
83
M = q2 + 1;
84
if (d < 0) M = -M; // Magic number and
85
s = p - 32; // shift amount to return.
86
87
return true;
88
}
89
90
//--------------------------transform_int_divide-------------------------------
91
// Convert a division by constant divisor into an alternate Ideal graph.
92
// Return NULL if no transformation occurs.
93
static Node *transform_int_divide( PhaseGVN *phase, Node *dividend, jint divisor ) {
94
95
// Check for invalid divisors
96
assert( divisor != 0 && divisor != min_jint,
97
"bad divisor for transforming to long multiply" );
98
99
bool d_pos = divisor >= 0;
100
jint d = d_pos ? divisor : -divisor;
101
const int N = 32;
102
103
// Result
104
Node *q = NULL;
105
106
if (d == 1) {
107
// division by +/- 1
108
if (!d_pos) {
109
// Just negate the value
110
q = new SubINode(phase->intcon(0), dividend);
111
}
112
} else if ( is_power_of_2(d) ) {
113
// division by +/- a power of 2
114
115
// See if we can simply do a shift without rounding
116
bool needs_rounding = true;
117
const Type *dt = phase->type(dividend);
118
const TypeInt *dti = dt->isa_int();
119
if (dti && dti->_lo >= 0) {
120
// we don't need to round a positive dividend
121
needs_rounding = false;
122
} else if( dividend->Opcode() == Op_AndI ) {
123
// An AND mask of sufficient size clears the low bits and
124
// I can avoid rounding.
125
const TypeInt *andconi_t = phase->type( dividend->in(2) )->isa_int();
126
if( andconi_t && andconi_t->is_con() ) {
127
jint andconi = andconi_t->get_con();
128
if( andconi < 0 && is_power_of_2(-andconi) && (-andconi) >= d ) {
129
if( (-andconi) == d ) // Remove AND if it clears bits which will be shifted
130
dividend = dividend->in(1);
131
needs_rounding = false;
132
}
133
}
134
}
135
136
// Add rounding to the shift to handle the sign bit
137
int l = log2i_graceful(d - 1) + 1;
138
if (needs_rounding) {
139
// Divide-by-power-of-2 can be made into a shift, but you have to do
140
// more math for the rounding. You need to add 0 for positive
141
// numbers, and "i-1" for negative numbers. Example: i=4, so the
142
// shift is by 2. You need to add 3 to negative dividends and 0 to
143
// positive ones. So (-7+3)>>2 becomes -1, (-4+3)>>2 becomes -1,
144
// (-2+3)>>2 becomes 0, etc.
145
146
// Compute 0 or -1, based on sign bit
147
Node *sign = phase->transform(new RShiftINode(dividend, phase->intcon(N - 1)));
148
// Mask sign bit to the low sign bits
149
Node *round = phase->transform(new URShiftINode(sign, phase->intcon(N - l)));
150
// Round up before shifting
151
dividend = phase->transform(new AddINode(dividend, round));
152
}
153
154
// Shift for division
155
q = new RShiftINode(dividend, phase->intcon(l));
156
157
if (!d_pos) {
158
q = new SubINode(phase->intcon(0), phase->transform(q));
159
}
160
} else {
161
// Attempt the jint constant divide -> multiply transform found in
162
// "Division by Invariant Integers using Multiplication"
163
// by Granlund and Montgomery
164
// See also "Hacker's Delight", chapter 10 by Warren.
165
166
jint magic_const;
167
jint shift_const;
168
if (magic_int_divide_constants(d, magic_const, shift_const)) {
169
Node *magic = phase->longcon(magic_const);
170
Node *dividend_long = phase->transform(new ConvI2LNode(dividend));
171
172
// Compute the high half of the dividend x magic multiplication
173
Node *mul_hi = phase->transform(new MulLNode(dividend_long, magic));
174
175
if (magic_const < 0) {
176
mul_hi = phase->transform(new RShiftLNode(mul_hi, phase->intcon(N)));
177
mul_hi = phase->transform(new ConvL2INode(mul_hi));
178
179
// The magic multiplier is too large for a 32 bit constant. We've adjusted
180
// it down by 2^32, but have to add 1 dividend back in after the multiplication.
181
// This handles the "overflow" case described by Granlund and Montgomery.
182
mul_hi = phase->transform(new AddINode(dividend, mul_hi));
183
184
// Shift over the (adjusted) mulhi
185
if (shift_const != 0) {
186
mul_hi = phase->transform(new RShiftINode(mul_hi, phase->intcon(shift_const)));
187
}
188
} else {
189
// No add is required, we can merge the shifts together.
190
mul_hi = phase->transform(new RShiftLNode(mul_hi, phase->intcon(N + shift_const)));
191
mul_hi = phase->transform(new ConvL2INode(mul_hi));
192
}
193
194
// Get a 0 or -1 from the sign of the dividend.
195
Node *addend0 = mul_hi;
196
Node *addend1 = phase->transform(new RShiftINode(dividend, phase->intcon(N-1)));
197
198
// If the divisor is negative, swap the order of the input addends;
199
// this has the effect of negating the quotient.
200
if (!d_pos) {
201
Node *temp = addend0; addend0 = addend1; addend1 = temp;
202
}
203
204
// Adjust the final quotient by subtracting -1 (adding 1)
205
// from the mul_hi.
206
q = new SubINode(addend0, addend1);
207
}
208
}
209
210
return q;
211
}
212
213
//---------------------magic_long_divide_constants-----------------------------
214
// Compute magic multiplier and shift constant for converting a 64 bit divide
215
// by constant into a multiply/shift/add series. Return false if calculations
216
// fail.
217
//
218
// Borrowed almost verbatim from Hacker's Delight by Henry S. Warren, Jr. with
219
// minor type name and parameter changes. Adjusted to 64 bit word width.
220
static bool magic_long_divide_constants(jlong d, jlong &M, jint &s) {
221
int64_t p;
222
uint64_t ad, anc, delta, q1, r1, q2, r2, t;
223
const uint64_t two63 = UCONST64(0x8000000000000000); // 2**63.
224
225
ad = ABS(d);
226
if (d == 0 || d == 1) return false;
227
t = two63 + ((uint64_t)d >> 63);
228
anc = t - 1 - t%ad; // Absolute value of nc.
229
p = 63; // Init. p.
230
q1 = two63/anc; // Init. q1 = 2**p/|nc|.
231
r1 = two63 - q1*anc; // Init. r1 = rem(2**p, |nc|).
232
q2 = two63/ad; // Init. q2 = 2**p/|d|.
233
r2 = two63 - q2*ad; // Init. r2 = rem(2**p, |d|).
234
do {
235
p = p + 1;
236
q1 = 2*q1; // Update q1 = 2**p/|nc|.
237
r1 = 2*r1; // Update r1 = rem(2**p, |nc|).
238
if (r1 >= anc) { // (Must be an unsigned
239
q1 = q1 + 1; // comparison here).
240
r1 = r1 - anc;
241
}
242
q2 = 2*q2; // Update q2 = 2**p/|d|.
243
r2 = 2*r2; // Update r2 = rem(2**p, |d|).
244
if (r2 >= ad) { // (Must be an unsigned
245
q2 = q2 + 1; // comparison here).
246
r2 = r2 - ad;
247
}
248
delta = ad - r2;
249
} while (q1 < delta || (q1 == delta && r1 == 0));
250
251
M = q2 + 1;
252
if (d < 0) M = -M; // Magic number and
253
s = p - 64; // shift amount to return.
254
255
return true;
256
}
257
258
//---------------------long_by_long_mulhi--------------------------------------
259
// Generate ideal node graph for upper half of a 64 bit x 64 bit multiplication
260
static Node* long_by_long_mulhi(PhaseGVN* phase, Node* dividend, jlong magic_const) {
261
// If the architecture supports a 64x64 mulhi, there is
262
// no need to synthesize it in ideal nodes.
263
if (Matcher::has_match_rule(Op_MulHiL)) {
264
Node* v = phase->longcon(magic_const);
265
return new MulHiLNode(dividend, v);
266
}
267
268
// Taken from Hacker's Delight, Fig. 8-2. Multiply high signed.
269
// (http://www.hackersdelight.org/HDcode/mulhs.c)
270
//
271
// int mulhs(int u, int v) {
272
// unsigned u0, v0, w0;
273
// int u1, v1, w1, w2, t;
274
//
275
// u0 = u & 0xFFFF; u1 = u >> 16;
276
// v0 = v & 0xFFFF; v1 = v >> 16;
277
// w0 = u0*v0;
278
// t = u1*v0 + (w0 >> 16);
279
// w1 = t & 0xFFFF;
280
// w2 = t >> 16;
281
// w1 = u0*v1 + w1;
282
// return u1*v1 + w2 + (w1 >> 16);
283
// }
284
//
285
// Note: The version above is for 32x32 multiplications, while the
286
// following inline comments are adapted to 64x64.
287
288
const int N = 64;
289
290
// Dummy node to keep intermediate nodes alive during construction
291
Node* hook = new Node(4);
292
293
// u0 = u & 0xFFFFFFFF; u1 = u >> 32;
294
Node* u0 = phase->transform(new AndLNode(dividend, phase->longcon(0xFFFFFFFF)));
295
Node* u1 = phase->transform(new RShiftLNode(dividend, phase->intcon(N / 2)));
296
hook->init_req(0, u0);
297
hook->init_req(1, u1);
298
299
// v0 = v & 0xFFFFFFFF; v1 = v >> 32;
300
Node* v0 = phase->longcon(magic_const & 0xFFFFFFFF);
301
Node* v1 = phase->longcon(magic_const >> (N / 2));
302
303
// w0 = u0*v0;
304
Node* w0 = phase->transform(new MulLNode(u0, v0));
305
306
// t = u1*v0 + (w0 >> 32);
307
Node* u1v0 = phase->transform(new MulLNode(u1, v0));
308
Node* temp = phase->transform(new URShiftLNode(w0, phase->intcon(N / 2)));
309
Node* t = phase->transform(new AddLNode(u1v0, temp));
310
hook->init_req(2, t);
311
312
// w1 = t & 0xFFFFFFFF;
313
Node* w1 = phase->transform(new AndLNode(t, phase->longcon(0xFFFFFFFF)));
314
hook->init_req(3, w1);
315
316
// w2 = t >> 32;
317
Node* w2 = phase->transform(new RShiftLNode(t, phase->intcon(N / 2)));
318
319
// w1 = u0*v1 + w1;
320
Node* u0v1 = phase->transform(new MulLNode(u0, v1));
321
w1 = phase->transform(new AddLNode(u0v1, w1));
322
323
// return u1*v1 + w2 + (w1 >> 32);
324
Node* u1v1 = phase->transform(new MulLNode(u1, v1));
325
Node* temp1 = phase->transform(new AddLNode(u1v1, w2));
326
Node* temp2 = phase->transform(new RShiftLNode(w1, phase->intcon(N / 2)));
327
328
// Remove the bogus extra edges used to keep things alive
329
hook->destruct(phase);
330
331
return new AddLNode(temp1, temp2);
332
}
333
334
335
//--------------------------transform_long_divide------------------------------
336
// Convert a division by constant divisor into an alternate Ideal graph.
337
// Return NULL if no transformation occurs.
338
static Node *transform_long_divide( PhaseGVN *phase, Node *dividend, jlong divisor ) {
339
// Check for invalid divisors
340
assert( divisor != 0L && divisor != min_jlong,
341
"bad divisor for transforming to long multiply" );
342
343
bool d_pos = divisor >= 0;
344
jlong d = d_pos ? divisor : -divisor;
345
const int N = 64;
346
347
// Result
348
Node *q = NULL;
349
350
if (d == 1) {
351
// division by +/- 1
352
if (!d_pos) {
353
// Just negate the value
354
q = new SubLNode(phase->longcon(0), dividend);
355
}
356
} else if ( is_power_of_2(d) ) {
357
358
// division by +/- a power of 2
359
360
// See if we can simply do a shift without rounding
361
bool needs_rounding = true;
362
const Type *dt = phase->type(dividend);
363
const TypeLong *dtl = dt->isa_long();
364
365
if (dtl && dtl->_lo > 0) {
366
// we don't need to round a positive dividend
367
needs_rounding = false;
368
} else if( dividend->Opcode() == Op_AndL ) {
369
// An AND mask of sufficient size clears the low bits and
370
// I can avoid rounding.
371
const TypeLong *andconl_t = phase->type( dividend->in(2) )->isa_long();
372
if( andconl_t && andconl_t->is_con() ) {
373
jlong andconl = andconl_t->get_con();
374
if( andconl < 0 && is_power_of_2(-andconl) && (-andconl) >= d ) {
375
if( (-andconl) == d ) // Remove AND if it clears bits which will be shifted
376
dividend = dividend->in(1);
377
needs_rounding = false;
378
}
379
}
380
}
381
382
// Add rounding to the shift to handle the sign bit
383
int l = log2i_graceful(d - 1) + 1;
384
if (needs_rounding) {
385
// Divide-by-power-of-2 can be made into a shift, but you have to do
386
// more math for the rounding. You need to add 0 for positive
387
// numbers, and "i-1" for negative numbers. Example: i=4, so the
388
// shift is by 2. You need to add 3 to negative dividends and 0 to
389
// positive ones. So (-7+3)>>2 becomes -1, (-4+3)>>2 becomes -1,
390
// (-2+3)>>2 becomes 0, etc.
391
392
// Compute 0 or -1, based on sign bit
393
Node *sign = phase->transform(new RShiftLNode(dividend, phase->intcon(N - 1)));
394
// Mask sign bit to the low sign bits
395
Node *round = phase->transform(new URShiftLNode(sign, phase->intcon(N - l)));
396
// Round up before shifting
397
dividend = phase->transform(new AddLNode(dividend, round));
398
}
399
400
// Shift for division
401
q = new RShiftLNode(dividend, phase->intcon(l));
402
403
if (!d_pos) {
404
q = new SubLNode(phase->longcon(0), phase->transform(q));
405
}
406
} else if ( !Matcher::use_asm_for_ldiv_by_con(d) ) { // Use hardware DIV instruction when
407
// it is faster than code generated below.
408
// Attempt the jlong constant divide -> multiply transform found in
409
// "Division by Invariant Integers using Multiplication"
410
// by Granlund and Montgomery
411
// See also "Hacker's Delight", chapter 10 by Warren.
412
413
jlong magic_const;
414
jint shift_const;
415
if (magic_long_divide_constants(d, magic_const, shift_const)) {
416
// Compute the high half of the dividend x magic multiplication
417
Node *mul_hi = phase->transform(long_by_long_mulhi(phase, dividend, magic_const));
418
419
// The high half of the 128-bit multiply is computed.
420
if (magic_const < 0) {
421
// The magic multiplier is too large for a 64 bit constant. We've adjusted
422
// it down by 2^64, but have to add 1 dividend back in after the multiplication.
423
// This handles the "overflow" case described by Granlund and Montgomery.
424
mul_hi = phase->transform(new AddLNode(dividend, mul_hi));
425
}
426
427
// Shift over the (adjusted) mulhi
428
if (shift_const != 0) {
429
mul_hi = phase->transform(new RShiftLNode(mul_hi, phase->intcon(shift_const)));
430
}
431
432
// Get a 0 or -1 from the sign of the dividend.
433
Node *addend0 = mul_hi;
434
Node *addend1 = phase->transform(new RShiftLNode(dividend, phase->intcon(N-1)));
435
436
// If the divisor is negative, swap the order of the input addends;
437
// this has the effect of negating the quotient.
438
if (!d_pos) {
439
Node *temp = addend0; addend0 = addend1; addend1 = temp;
440
}
441
442
// Adjust the final quotient by subtracting -1 (adding 1)
443
// from the mul_hi.
444
q = new SubLNode(addend0, addend1);
445
}
446
}
447
448
return q;
449
}
450
451
//=============================================================================
452
//------------------------------Identity---------------------------------------
453
// If the divisor is 1, we are an identity on the dividend.
454
Node* DivINode::Identity(PhaseGVN* phase) {
455
return (phase->type( in(2) )->higher_equal(TypeInt::ONE)) ? in(1) : this;
456
}
457
458
//------------------------------Idealize---------------------------------------
459
// Divides can be changed to multiplies and/or shifts
460
Node *DivINode::Ideal(PhaseGVN *phase, bool can_reshape) {
461
if (in(0) && remove_dead_region(phase, can_reshape)) return this;
462
// Don't bother trying to transform a dead node
463
if( in(0) && in(0)->is_top() ) return NULL;
464
465
const Type *t = phase->type( in(2) );
466
if( t == TypeInt::ONE ) // Identity?
467
return NULL; // Skip it
468
469
const TypeInt *ti = t->isa_int();
470
if( !ti ) return NULL;
471
472
// Check for useless control input
473
// Check for excluding div-zero case
474
if (in(0) && (ti->_hi < 0 || ti->_lo > 0)) {
475
set_req(0, NULL); // Yank control input
476
return this;
477
}
478
479
if( !ti->is_con() ) return NULL;
480
jint i = ti->get_con(); // Get divisor
481
482
if (i == 0) return NULL; // Dividing by zero constant does not idealize
483
484
// Dividing by MININT does not optimize as a power-of-2 shift.
485
if( i == min_jint ) return NULL;
486
487
return transform_int_divide( phase, in(1), i );
488
}
489
490
//------------------------------Value------------------------------------------
491
// A DivINode divides its inputs. The third input is a Control input, used to
492
// prevent hoisting the divide above an unsafe test.
493
const Type* DivINode::Value(PhaseGVN* phase) const {
494
// Either input is TOP ==> the result is TOP
495
const Type *t1 = phase->type( in(1) );
496
const Type *t2 = phase->type( in(2) );
497
if( t1 == Type::TOP ) return Type::TOP;
498
if( t2 == Type::TOP ) return Type::TOP;
499
500
// x/x == 1 since we always generate the dynamic divisor check for 0.
501
if (in(1) == in(2)) {
502
return TypeInt::ONE;
503
}
504
505
// Either input is BOTTOM ==> the result is the local BOTTOM
506
const Type *bot = bottom_type();
507
if( (t1 == bot) || (t2 == bot) ||
508
(t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
509
return bot;
510
511
// Divide the two numbers. We approximate.
512
// If divisor is a constant and not zero
513
const TypeInt *i1 = t1->is_int();
514
const TypeInt *i2 = t2->is_int();
515
int widen = MAX2(i1->_widen, i2->_widen);
516
517
if( i2->is_con() && i2->get_con() != 0 ) {
518
int32_t d = i2->get_con(); // Divisor
519
jint lo, hi;
520
if( d >= 0 ) {
521
lo = i1->_lo/d;
522
hi = i1->_hi/d;
523
} else {
524
if( d == -1 && i1->_lo == min_jint ) {
525
// 'min_jint/-1' throws arithmetic exception during compilation
526
lo = min_jint;
527
// do not support holes, 'hi' must go to either min_jint or max_jint:
528
// [min_jint, -10]/[-1,-1] ==> [min_jint] UNION [10,max_jint]
529
hi = i1->_hi == min_jint ? min_jint : max_jint;
530
} else {
531
lo = i1->_hi/d;
532
hi = i1->_lo/d;
533
}
534
}
535
return TypeInt::make(lo, hi, widen);
536
}
537
538
// If the dividend is a constant
539
if( i1->is_con() ) {
540
int32_t d = i1->get_con();
541
if( d < 0 ) {
542
if( d == min_jint ) {
543
// (-min_jint) == min_jint == (min_jint / -1)
544
return TypeInt::make(min_jint, max_jint/2 + 1, widen);
545
} else {
546
return TypeInt::make(d, -d, widen);
547
}
548
}
549
return TypeInt::make(-d, d, widen);
550
}
551
552
// Otherwise we give up all hope
553
return TypeInt::INT;
554
}
555
556
557
//=============================================================================
558
//------------------------------Identity---------------------------------------
559
// If the divisor is 1, we are an identity on the dividend.
560
Node* DivLNode::Identity(PhaseGVN* phase) {
561
return (phase->type( in(2) )->higher_equal(TypeLong::ONE)) ? in(1) : this;
562
}
563
564
//------------------------------Idealize---------------------------------------
565
// Dividing by a power of 2 is a shift.
566
Node *DivLNode::Ideal( PhaseGVN *phase, bool can_reshape) {
567
if (in(0) && remove_dead_region(phase, can_reshape)) return this;
568
// Don't bother trying to transform a dead node
569
if( in(0) && in(0)->is_top() ) return NULL;
570
571
const Type *t = phase->type( in(2) );
572
if( t == TypeLong::ONE ) // Identity?
573
return NULL; // Skip it
574
575
const TypeLong *tl = t->isa_long();
576
if( !tl ) return NULL;
577
578
// Check for useless control input
579
// Check for excluding div-zero case
580
if (in(0) && (tl->_hi < 0 || tl->_lo > 0)) {
581
set_req(0, NULL); // Yank control input
582
return this;
583
}
584
585
if( !tl->is_con() ) return NULL;
586
jlong l = tl->get_con(); // Get divisor
587
588
if (l == 0) return NULL; // Dividing by zero constant does not idealize
589
590
// Dividing by MINLONG does not optimize as a power-of-2 shift.
591
if( l == min_jlong ) return NULL;
592
593
return transform_long_divide( phase, in(1), l );
594
}
595
596
//------------------------------Value------------------------------------------
597
// A DivLNode divides its inputs. The third input is a Control input, used to
598
// prevent hoisting the divide above an unsafe test.
599
const Type* DivLNode::Value(PhaseGVN* phase) const {
600
// Either input is TOP ==> the result is TOP
601
const Type *t1 = phase->type( in(1) );
602
const Type *t2 = phase->type( in(2) );
603
if( t1 == Type::TOP ) return Type::TOP;
604
if( t2 == Type::TOP ) return Type::TOP;
605
606
// x/x == 1 since we always generate the dynamic divisor check for 0.
607
if (in(1) == in(2)) {
608
return TypeLong::ONE;
609
}
610
611
// Either input is BOTTOM ==> the result is the local BOTTOM
612
const Type *bot = bottom_type();
613
if( (t1 == bot) || (t2 == bot) ||
614
(t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
615
return bot;
616
617
// Divide the two numbers. We approximate.
618
// If divisor is a constant and not zero
619
const TypeLong *i1 = t1->is_long();
620
const TypeLong *i2 = t2->is_long();
621
int widen = MAX2(i1->_widen, i2->_widen);
622
623
if( i2->is_con() && i2->get_con() != 0 ) {
624
jlong d = i2->get_con(); // Divisor
625
jlong lo, hi;
626
if( d >= 0 ) {
627
lo = i1->_lo/d;
628
hi = i1->_hi/d;
629
} else {
630
if( d == CONST64(-1) && i1->_lo == min_jlong ) {
631
// 'min_jlong/-1' throws arithmetic exception during compilation
632
lo = min_jlong;
633
// do not support holes, 'hi' must go to either min_jlong or max_jlong:
634
// [min_jlong, -10]/[-1,-1] ==> [min_jlong] UNION [10,max_jlong]
635
hi = i1->_hi == min_jlong ? min_jlong : max_jlong;
636
} else {
637
lo = i1->_hi/d;
638
hi = i1->_lo/d;
639
}
640
}
641
return TypeLong::make(lo, hi, widen);
642
}
643
644
// If the dividend is a constant
645
if( i1->is_con() ) {
646
jlong d = i1->get_con();
647
if( d < 0 ) {
648
if( d == min_jlong ) {
649
// (-min_jlong) == min_jlong == (min_jlong / -1)
650
return TypeLong::make(min_jlong, max_jlong/2 + 1, widen);
651
} else {
652
return TypeLong::make(d, -d, widen);
653
}
654
}
655
return TypeLong::make(-d, d, widen);
656
}
657
658
// Otherwise we give up all hope
659
return TypeLong::LONG;
660
}
661
662
663
//=============================================================================
664
//------------------------------Value------------------------------------------
665
// An DivFNode divides its inputs. The third input is a Control input, used to
666
// prevent hoisting the divide above an unsafe test.
667
const Type* DivFNode::Value(PhaseGVN* phase) const {
668
// Either input is TOP ==> the result is TOP
669
const Type *t1 = phase->type( in(1) );
670
const Type *t2 = phase->type( in(2) );
671
if( t1 == Type::TOP ) return Type::TOP;
672
if( t2 == Type::TOP ) return Type::TOP;
673
674
// Either input is BOTTOM ==> the result is the local BOTTOM
675
const Type *bot = bottom_type();
676
if( (t1 == bot) || (t2 == bot) ||
677
(t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
678
return bot;
679
680
// x/x == 1, we ignore 0/0.
681
// Note: if t1 and t2 are zero then result is NaN (JVMS page 213)
682
// Does not work for variables because of NaN's
683
if (in(1) == in(2) && t1->base() == Type::FloatCon &&
684
!g_isnan(t1->getf()) && g_isfinite(t1->getf()) && t1->getf() != 0.0) { // could be negative ZERO or NaN
685
return TypeF::ONE;
686
}
687
688
if( t2 == TypeF::ONE )
689
return t1;
690
691
// If divisor is a constant and not zero, divide them numbers
692
if( t1->base() == Type::FloatCon &&
693
t2->base() == Type::FloatCon &&
694
t2->getf() != 0.0 ) // could be negative zero
695
return TypeF::make( t1->getf()/t2->getf() );
696
697
// If the dividend is a constant zero
698
// Note: if t1 and t2 are zero then result is NaN (JVMS page 213)
699
// Test TypeF::ZERO is not sufficient as it could be negative zero
700
701
if( t1 == TypeF::ZERO && !g_isnan(t2->getf()) && t2->getf() != 0.0 )
702
return TypeF::ZERO;
703
704
// Otherwise we give up all hope
705
return Type::FLOAT;
706
}
707
708
//------------------------------isA_Copy---------------------------------------
709
// Dividing by self is 1.
710
// If the divisor is 1, we are an identity on the dividend.
711
Node* DivFNode::Identity(PhaseGVN* phase) {
712
return (phase->type( in(2) ) == TypeF::ONE) ? in(1) : this;
713
}
714
715
716
//------------------------------Idealize---------------------------------------
717
Node *DivFNode::Ideal(PhaseGVN *phase, bool can_reshape) {
718
if (in(0) && remove_dead_region(phase, can_reshape)) return this;
719
// Don't bother trying to transform a dead node
720
if( in(0) && in(0)->is_top() ) return NULL;
721
722
const Type *t2 = phase->type( in(2) );
723
if( t2 == TypeF::ONE ) // Identity?
724
return NULL; // Skip it
725
726
const TypeF *tf = t2->isa_float_constant();
727
if( !tf ) return NULL;
728
if( tf->base() != Type::FloatCon ) return NULL;
729
730
// Check for out of range values
731
if( tf->is_nan() || !tf->is_finite() ) return NULL;
732
733
// Get the value
734
float f = tf->getf();
735
int exp;
736
737
// Only for special case of dividing by a power of 2
738
if( frexp((double)f, &exp) != 0.5 ) return NULL;
739
740
// Limit the range of acceptable exponents
741
if( exp < -126 || exp > 126 ) return NULL;
742
743
// Compute the reciprocal
744
float reciprocal = ((float)1.0) / f;
745
746
assert( frexp((double)reciprocal, &exp) == 0.5, "reciprocal should be power of 2" );
747
748
// return multiplication by the reciprocal
749
return (new MulFNode(in(1), phase->makecon(TypeF::make(reciprocal))));
750
}
751
752
//=============================================================================
753
//------------------------------Value------------------------------------------
754
// An DivDNode divides its inputs. The third input is a Control input, used to
755
// prevent hoisting the divide above an unsafe test.
756
const Type* DivDNode::Value(PhaseGVN* phase) const {
757
// Either input is TOP ==> the result is TOP
758
const Type *t1 = phase->type( in(1) );
759
const Type *t2 = phase->type( in(2) );
760
if( t1 == Type::TOP ) return Type::TOP;
761
if( t2 == Type::TOP ) return Type::TOP;
762
763
// Either input is BOTTOM ==> the result is the local BOTTOM
764
const Type *bot = bottom_type();
765
if( (t1 == bot) || (t2 == bot) ||
766
(t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
767
return bot;
768
769
// x/x == 1, we ignore 0/0.
770
// Note: if t1 and t2 are zero then result is NaN (JVMS page 213)
771
// Does not work for variables because of NaN's
772
if (in(1) == in(2) && t1->base() == Type::DoubleCon &&
773
!g_isnan(t1->getd()) && g_isfinite(t1->getd()) && t1->getd() != 0.0) { // could be negative ZERO or NaN
774
return TypeD::ONE;
775
}
776
777
if( t2 == TypeD::ONE )
778
return t1;
779
780
// IA32 would only execute this for non-strict FP, which is never the
781
// case now.
782
#if ! defined(IA32)
783
// If divisor is a constant and not zero, divide them numbers
784
if( t1->base() == Type::DoubleCon &&
785
t2->base() == Type::DoubleCon &&
786
t2->getd() != 0.0 ) // could be negative zero
787
return TypeD::make( t1->getd()/t2->getd() );
788
#endif
789
790
// If the dividend is a constant zero
791
// Note: if t1 and t2 are zero then result is NaN (JVMS page 213)
792
// Test TypeF::ZERO is not sufficient as it could be negative zero
793
if( t1 == TypeD::ZERO && !g_isnan(t2->getd()) && t2->getd() != 0.0 )
794
return TypeD::ZERO;
795
796
// Otherwise we give up all hope
797
return Type::DOUBLE;
798
}
799
800
801
//------------------------------isA_Copy---------------------------------------
802
// Dividing by self is 1.
803
// If the divisor is 1, we are an identity on the dividend.
804
Node* DivDNode::Identity(PhaseGVN* phase) {
805
return (phase->type( in(2) ) == TypeD::ONE) ? in(1) : this;
806
}
807
808
//------------------------------Idealize---------------------------------------
809
Node *DivDNode::Ideal(PhaseGVN *phase, bool can_reshape) {
810
if (in(0) && remove_dead_region(phase, can_reshape)) return this;
811
// Don't bother trying to transform a dead node
812
if( in(0) && in(0)->is_top() ) return NULL;
813
814
const Type *t2 = phase->type( in(2) );
815
if( t2 == TypeD::ONE ) // Identity?
816
return NULL; // Skip it
817
818
const TypeD *td = t2->isa_double_constant();
819
if( !td ) return NULL;
820
if( td->base() != Type::DoubleCon ) return NULL;
821
822
// Check for out of range values
823
if( td->is_nan() || !td->is_finite() ) return NULL;
824
825
// Get the value
826
double d = td->getd();
827
int exp;
828
829
// Only for special case of dividing by a power of 2
830
if( frexp(d, &exp) != 0.5 ) return NULL;
831
832
// Limit the range of acceptable exponents
833
if( exp < -1021 || exp > 1022 ) return NULL;
834
835
// Compute the reciprocal
836
double reciprocal = 1.0 / d;
837
838
assert( frexp(reciprocal, &exp) == 0.5, "reciprocal should be power of 2" );
839
840
// return multiplication by the reciprocal
841
return (new MulDNode(in(1), phase->makecon(TypeD::make(reciprocal))));
842
}
843
844
//=============================================================================
845
//------------------------------Idealize---------------------------------------
846
Node *ModINode::Ideal(PhaseGVN *phase, bool can_reshape) {
847
// Check for dead control input
848
if( in(0) && remove_dead_region(phase, can_reshape) ) return this;
849
// Don't bother trying to transform a dead node
850
if( in(0) && in(0)->is_top() ) return NULL;
851
852
// Get the modulus
853
const Type *t = phase->type( in(2) );
854
if( t == Type::TOP ) return NULL;
855
const TypeInt *ti = t->is_int();
856
857
// Check for useless control input
858
// Check for excluding mod-zero case
859
if (in(0) && (ti->_hi < 0 || ti->_lo > 0)) {
860
set_req(0, NULL); // Yank control input
861
return this;
862
}
863
864
// See if we are MOD'ing by 2^k or 2^k-1.
865
if( !ti->is_con() ) return NULL;
866
jint con = ti->get_con();
867
868
Node *hook = new Node(1);
869
870
// First, special check for modulo 2^k-1
871
if( con >= 0 && con < max_jint && is_power_of_2(con+1) ) {
872
uint k = exact_log2(con+1); // Extract k
873
874
// Basic algorithm by David Detlefs. See fastmod_int.java for gory details.
875
static int unroll_factor[] = { 999, 999, 29, 14, 9, 7, 5, 4, 4, 3, 3, 2, 2, 2, 2, 2, 1 /*past here we assume 1 forever*/};
876
int trip_count = 1;
877
if( k < ARRAY_SIZE(unroll_factor)) trip_count = unroll_factor[k];
878
879
// If the unroll factor is not too large, and if conditional moves are
880
// ok, then use this case
881
if( trip_count <= 5 && ConditionalMoveLimit != 0 ) {
882
Node *x = in(1); // Value being mod'd
883
Node *divisor = in(2); // Also is mask
884
885
hook->init_req(0, x); // Add a use to x to prevent him from dying
886
// Generate code to reduce X rapidly to nearly 2^k-1.
887
for( int i = 0; i < trip_count; i++ ) {
888
Node *xl = phase->transform( new AndINode(x,divisor) );
889
Node *xh = phase->transform( new RShiftINode(x,phase->intcon(k)) ); // Must be signed
890
x = phase->transform( new AddINode(xh,xl) );
891
hook->set_req(0, x);
892
}
893
894
// Generate sign-fixup code. Was original value positive?
895
// int hack_res = (i >= 0) ? divisor : 1;
896
Node *cmp1 = phase->transform( new CmpINode( in(1), phase->intcon(0) ) );
897
Node *bol1 = phase->transform( new BoolNode( cmp1, BoolTest::ge ) );
898
Node *cmov1= phase->transform( new CMoveINode(bol1, phase->intcon(1), divisor, TypeInt::POS) );
899
// if( x >= hack_res ) x -= divisor;
900
Node *sub = phase->transform( new SubINode( x, divisor ) );
901
Node *cmp2 = phase->transform( new CmpINode( x, cmov1 ) );
902
Node *bol2 = phase->transform( new BoolNode( cmp2, BoolTest::ge ) );
903
// Convention is to not transform the return value of an Ideal
904
// since Ideal is expected to return a modified 'this' or a new node.
905
Node *cmov2= new CMoveINode(bol2, x, sub, TypeInt::INT);
906
// cmov2 is now the mod
907
908
// Now remove the bogus extra edges used to keep things alive
909
hook->destruct(phase);
910
return cmov2;
911
}
912
}
913
914
// Fell thru, the unroll case is not appropriate. Transform the modulo
915
// into a long multiply/int multiply/subtract case
916
917
// Cannot handle mod 0, and min_jint isn't handled by the transform
918
if( con == 0 || con == min_jint ) return NULL;
919
920
// Get the absolute value of the constant; at this point, we can use this
921
jint pos_con = (con >= 0) ? con : -con;
922
923
// integer Mod 1 is always 0
924
if( pos_con == 1 ) return new ConINode(TypeInt::ZERO);
925
926
int log2_con = -1;
927
928
// If this is a power of two, they maybe we can mask it
929
if (is_power_of_2(pos_con)) {
930
log2_con = log2i_exact(pos_con);
931
932
const Type *dt = phase->type(in(1));
933
const TypeInt *dti = dt->isa_int();
934
935
// See if this can be masked, if the dividend is non-negative
936
if( dti && dti->_lo >= 0 )
937
return ( new AndINode( in(1), phase->intcon( pos_con-1 ) ) );
938
}
939
940
// Save in(1) so that it cannot be changed or deleted
941
hook->init_req(0, in(1));
942
943
// Divide using the transform from DivI to MulL
944
Node *result = transform_int_divide( phase, in(1), pos_con );
945
if (result != NULL) {
946
Node *divide = phase->transform(result);
947
948
// Re-multiply, using a shift if this is a power of two
949
Node *mult = NULL;
950
951
if( log2_con >= 0 )
952
mult = phase->transform( new LShiftINode( divide, phase->intcon( log2_con ) ) );
953
else
954
mult = phase->transform( new MulINode( divide, phase->intcon( pos_con ) ) );
955
956
// Finally, subtract the multiplied divided value from the original
957
result = new SubINode( in(1), mult );
958
}
959
960
// Now remove the bogus extra edges used to keep things alive
961
hook->destruct(phase);
962
963
// return the value
964
return result;
965
}
966
967
//------------------------------Value------------------------------------------
968
const Type* ModINode::Value(PhaseGVN* phase) const {
969
// Either input is TOP ==> the result is TOP
970
const Type *t1 = phase->type( in(1) );
971
const Type *t2 = phase->type( in(2) );
972
if( t1 == Type::TOP ) return Type::TOP;
973
if( t2 == Type::TOP ) return Type::TOP;
974
975
// We always generate the dynamic check for 0.
976
// 0 MOD X is 0
977
if( t1 == TypeInt::ZERO ) return TypeInt::ZERO;
978
// X MOD X is 0
979
if (in(1) == in(2)) {
980
return TypeInt::ZERO;
981
}
982
983
// Either input is BOTTOM ==> the result is the local BOTTOM
984
const Type *bot = bottom_type();
985
if( (t1 == bot) || (t2 == bot) ||
986
(t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
987
return bot;
988
989
const TypeInt *i1 = t1->is_int();
990
const TypeInt *i2 = t2->is_int();
991
if( !i1->is_con() || !i2->is_con() ) {
992
if( i1->_lo >= 0 && i2->_lo >= 0 )
993
return TypeInt::POS;
994
// If both numbers are not constants, we know little.
995
return TypeInt::INT;
996
}
997
// Mod by zero? Throw exception at runtime!
998
if( !i2->get_con() ) return TypeInt::POS;
999
1000
// We must be modulo'ing 2 float constants.
1001
// Check for min_jint % '-1', result is defined to be '0'.
1002
if( i1->get_con() == min_jint && i2->get_con() == -1 )
1003
return TypeInt::ZERO;
1004
1005
return TypeInt::make( i1->get_con() % i2->get_con() );
1006
}
1007
1008
1009
//=============================================================================
1010
//------------------------------Idealize---------------------------------------
1011
Node *ModLNode::Ideal(PhaseGVN *phase, bool can_reshape) {
1012
// Check for dead control input
1013
if( in(0) && remove_dead_region(phase, can_reshape) ) return this;
1014
// Don't bother trying to transform a dead node
1015
if( in(0) && in(0)->is_top() ) return NULL;
1016
1017
// Get the modulus
1018
const Type *t = phase->type( in(2) );
1019
if( t == Type::TOP ) return NULL;
1020
const TypeLong *tl = t->is_long();
1021
1022
// Check for useless control input
1023
// Check for excluding mod-zero case
1024
if (in(0) && (tl->_hi < 0 || tl->_lo > 0)) {
1025
set_req(0, NULL); // Yank control input
1026
return this;
1027
}
1028
1029
// See if we are MOD'ing by 2^k or 2^k-1.
1030
if( !tl->is_con() ) return NULL;
1031
jlong con = tl->get_con();
1032
1033
Node *hook = new Node(1);
1034
1035
// Expand mod
1036
if(con >= 0 && con < max_jlong && is_power_of_2(con + 1)) {
1037
uint k = log2i_exact(con + 1); // Extract k
1038
1039
// Basic algorithm by David Detlefs. See fastmod_long.java for gory details.
1040
// Used to help a popular random number generator which does a long-mod
1041
// of 2^31-1 and shows up in SpecJBB and SciMark.
1042
static int unroll_factor[] = { 999, 999, 61, 30, 20, 15, 12, 10, 8, 7, 6, 6, 5, 5, 4, 4, 4, 3, 3, 3, 3, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1 /*past here we assume 1 forever*/};
1043
int trip_count = 1;
1044
if( k < ARRAY_SIZE(unroll_factor)) trip_count = unroll_factor[k];
1045
1046
// If the unroll factor is not too large, and if conditional moves are
1047
// ok, then use this case
1048
if( trip_count <= 5 && ConditionalMoveLimit != 0 ) {
1049
Node *x = in(1); // Value being mod'd
1050
Node *divisor = in(2); // Also is mask
1051
1052
hook->init_req(0, x); // Add a use to x to prevent him from dying
1053
// Generate code to reduce X rapidly to nearly 2^k-1.
1054
for( int i = 0; i < trip_count; i++ ) {
1055
Node *xl = phase->transform( new AndLNode(x,divisor) );
1056
Node *xh = phase->transform( new RShiftLNode(x,phase->intcon(k)) ); // Must be signed
1057
x = phase->transform( new AddLNode(xh,xl) );
1058
hook->set_req(0, x); // Add a use to x to prevent him from dying
1059
}
1060
1061
// Generate sign-fixup code. Was original value positive?
1062
// long hack_res = (i >= 0) ? divisor : CONST64(1);
1063
Node *cmp1 = phase->transform( new CmpLNode( in(1), phase->longcon(0) ) );
1064
Node *bol1 = phase->transform( new BoolNode( cmp1, BoolTest::ge ) );
1065
Node *cmov1= phase->transform( new CMoveLNode(bol1, phase->longcon(1), divisor, TypeLong::LONG) );
1066
// if( x >= hack_res ) x -= divisor;
1067
Node *sub = phase->transform( new SubLNode( x, divisor ) );
1068
Node *cmp2 = phase->transform( new CmpLNode( x, cmov1 ) );
1069
Node *bol2 = phase->transform( new BoolNode( cmp2, BoolTest::ge ) );
1070
// Convention is to not transform the return value of an Ideal
1071
// since Ideal is expected to return a modified 'this' or a new node.
1072
Node *cmov2= new CMoveLNode(bol2, x, sub, TypeLong::LONG);
1073
// cmov2 is now the mod
1074
1075
// Now remove the bogus extra edges used to keep things alive
1076
hook->destruct(phase);
1077
return cmov2;
1078
}
1079
}
1080
1081
// Fell thru, the unroll case is not appropriate. Transform the modulo
1082
// into a long multiply/int multiply/subtract case
1083
1084
// Cannot handle mod 0, and min_jlong isn't handled by the transform
1085
if( con == 0 || con == min_jlong ) return NULL;
1086
1087
// Get the absolute value of the constant; at this point, we can use this
1088
jlong pos_con = (con >= 0) ? con : -con;
1089
1090
// integer Mod 1 is always 0
1091
if( pos_con == 1 ) return new ConLNode(TypeLong::ZERO);
1092
1093
int log2_con = -1;
1094
1095
// If this is a power of two, then maybe we can mask it
1096
if (is_power_of_2(pos_con)) {
1097
log2_con = log2i_exact(pos_con);
1098
1099
const Type *dt = phase->type(in(1));
1100
const TypeLong *dtl = dt->isa_long();
1101
1102
// See if this can be masked, if the dividend is non-negative
1103
if( dtl && dtl->_lo >= 0 )
1104
return ( new AndLNode( in(1), phase->longcon( pos_con-1 ) ) );
1105
}
1106
1107
// Save in(1) so that it cannot be changed or deleted
1108
hook->init_req(0, in(1));
1109
1110
// Divide using the transform from DivL to MulL
1111
Node *result = transform_long_divide( phase, in(1), pos_con );
1112
if (result != NULL) {
1113
Node *divide = phase->transform(result);
1114
1115
// Re-multiply, using a shift if this is a power of two
1116
Node *mult = NULL;
1117
1118
if( log2_con >= 0 )
1119
mult = phase->transform( new LShiftLNode( divide, phase->intcon( log2_con ) ) );
1120
else
1121
mult = phase->transform( new MulLNode( divide, phase->longcon( pos_con ) ) );
1122
1123
// Finally, subtract the multiplied divided value from the original
1124
result = new SubLNode( in(1), mult );
1125
}
1126
1127
// Now remove the bogus extra edges used to keep things alive
1128
hook->destruct(phase);
1129
1130
// return the value
1131
return result;
1132
}
1133
1134
//------------------------------Value------------------------------------------
1135
const Type* ModLNode::Value(PhaseGVN* phase) const {
1136
// Either input is TOP ==> the result is TOP
1137
const Type *t1 = phase->type( in(1) );
1138
const Type *t2 = phase->type( in(2) );
1139
if( t1 == Type::TOP ) return Type::TOP;
1140
if( t2 == Type::TOP ) return Type::TOP;
1141
1142
// We always generate the dynamic check for 0.
1143
// 0 MOD X is 0
1144
if( t1 == TypeLong::ZERO ) return TypeLong::ZERO;
1145
// X MOD X is 0
1146
if (in(1) == in(2)) {
1147
return TypeLong::ZERO;
1148
}
1149
1150
// Either input is BOTTOM ==> the result is the local BOTTOM
1151
const Type *bot = bottom_type();
1152
if( (t1 == bot) || (t2 == bot) ||
1153
(t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
1154
return bot;
1155
1156
const TypeLong *i1 = t1->is_long();
1157
const TypeLong *i2 = t2->is_long();
1158
if( !i1->is_con() || !i2->is_con() ) {
1159
if( i1->_lo >= CONST64(0) && i2->_lo >= CONST64(0) )
1160
return TypeLong::POS;
1161
// If both numbers are not constants, we know little.
1162
return TypeLong::LONG;
1163
}
1164
// Mod by zero? Throw exception at runtime!
1165
if( !i2->get_con() ) return TypeLong::POS;
1166
1167
// We must be modulo'ing 2 float constants.
1168
// Check for min_jint % '-1', result is defined to be '0'.
1169
if( i1->get_con() == min_jlong && i2->get_con() == -1 )
1170
return TypeLong::ZERO;
1171
1172
return TypeLong::make( i1->get_con() % i2->get_con() );
1173
}
1174
1175
1176
//=============================================================================
1177
//------------------------------Value------------------------------------------
1178
const Type* ModFNode::Value(PhaseGVN* phase) const {
1179
// Either input is TOP ==> the result is TOP
1180
const Type *t1 = phase->type( in(1) );
1181
const Type *t2 = phase->type( in(2) );
1182
if( t1 == Type::TOP ) return Type::TOP;
1183
if( t2 == Type::TOP ) return Type::TOP;
1184
1185
// Either input is BOTTOM ==> the result is the local BOTTOM
1186
const Type *bot = bottom_type();
1187
if( (t1 == bot) || (t2 == bot) ||
1188
(t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
1189
return bot;
1190
1191
// If either number is not a constant, we know nothing.
1192
if ((t1->base() != Type::FloatCon) || (t2->base() != Type::FloatCon)) {
1193
return Type::FLOAT; // note: x%x can be either NaN or 0
1194
}
1195
1196
float f1 = t1->getf();
1197
float f2 = t2->getf();
1198
jint x1 = jint_cast(f1); // note: *(int*)&f1, not just (int)f1
1199
jint x2 = jint_cast(f2);
1200
1201
// If either is a NaN, return an input NaN
1202
if (g_isnan(f1)) return t1;
1203
if (g_isnan(f2)) return t2;
1204
1205
// If an operand is infinity or the divisor is +/- zero, punt.
1206
if (!g_isfinite(f1) || !g_isfinite(f2) || x2 == 0 || x2 == min_jint)
1207
return Type::FLOAT;
1208
1209
// We must be modulo'ing 2 float constants.
1210
// Make sure that the sign of the fmod is equal to the sign of the dividend
1211
jint xr = jint_cast(fmod(f1, f2));
1212
if ((x1 ^ xr) < 0) {
1213
xr ^= min_jint;
1214
}
1215
1216
return TypeF::make(jfloat_cast(xr));
1217
}
1218
1219
1220
//=============================================================================
1221
//------------------------------Value------------------------------------------
1222
const Type* ModDNode::Value(PhaseGVN* phase) const {
1223
// Either input is TOP ==> the result is TOP
1224
const Type *t1 = phase->type( in(1) );
1225
const Type *t2 = phase->type( in(2) );
1226
if( t1 == Type::TOP ) return Type::TOP;
1227
if( t2 == Type::TOP ) return Type::TOP;
1228
1229
// Either input is BOTTOM ==> the result is the local BOTTOM
1230
const Type *bot = bottom_type();
1231
if( (t1 == bot) || (t2 == bot) ||
1232
(t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
1233
return bot;
1234
1235
// If either number is not a constant, we know nothing.
1236
if ((t1->base() != Type::DoubleCon) || (t2->base() != Type::DoubleCon)) {
1237
return Type::DOUBLE; // note: x%x can be either NaN or 0
1238
}
1239
1240
double f1 = t1->getd();
1241
double f2 = t2->getd();
1242
jlong x1 = jlong_cast(f1); // note: *(long*)&f1, not just (long)f1
1243
jlong x2 = jlong_cast(f2);
1244
1245
// If either is a NaN, return an input NaN
1246
if (g_isnan(f1)) return t1;
1247
if (g_isnan(f2)) return t2;
1248
1249
// If an operand is infinity or the divisor is +/- zero, punt.
1250
if (!g_isfinite(f1) || !g_isfinite(f2) || x2 == 0 || x2 == min_jlong)
1251
return Type::DOUBLE;
1252
1253
// We must be modulo'ing 2 double constants.
1254
// Make sure that the sign of the fmod is equal to the sign of the dividend
1255
jlong xr = jlong_cast(fmod(f1, f2));
1256
if ((x1 ^ xr) < 0) {
1257
xr ^= min_jlong;
1258
}
1259
1260
return TypeD::make(jdouble_cast(xr));
1261
}
1262
1263
//=============================================================================
1264
1265
DivModNode::DivModNode( Node *c, Node *dividend, Node *divisor ) : MultiNode(3) {
1266
init_req(0, c);
1267
init_req(1, dividend);
1268
init_req(2, divisor);
1269
}
1270
1271
//------------------------------make------------------------------------------
1272
DivModINode* DivModINode::make(Node* div_or_mod) {
1273
Node* n = div_or_mod;
1274
assert(n->Opcode() == Op_DivI || n->Opcode() == Op_ModI,
1275
"only div or mod input pattern accepted");
1276
1277
DivModINode* divmod = new DivModINode(n->in(0), n->in(1), n->in(2));
1278
Node* dproj = new ProjNode(divmod, DivModNode::div_proj_num);
1279
Node* mproj = new ProjNode(divmod, DivModNode::mod_proj_num);
1280
return divmod;
1281
}
1282
1283
//------------------------------make------------------------------------------
1284
DivModLNode* DivModLNode::make(Node* div_or_mod) {
1285
Node* n = div_or_mod;
1286
assert(n->Opcode() == Op_DivL || n->Opcode() == Op_ModL,
1287
"only div or mod input pattern accepted");
1288
1289
DivModLNode* divmod = new DivModLNode(n->in(0), n->in(1), n->in(2));
1290
Node* dproj = new ProjNode(divmod, DivModNode::div_proj_num);
1291
Node* mproj = new ProjNode(divmod, DivModNode::mod_proj_num);
1292
return divmod;
1293
}
1294
1295
//------------------------------match------------------------------------------
1296
// return result(s) along with their RegMask info
1297
Node *DivModINode::match( const ProjNode *proj, const Matcher *match ) {
1298
uint ideal_reg = proj->ideal_reg();
1299
RegMask rm;
1300
if (proj->_con == div_proj_num) {
1301
rm = match->divI_proj_mask();
1302
} else {
1303
assert(proj->_con == mod_proj_num, "must be div or mod projection");
1304
rm = match->modI_proj_mask();
1305
}
1306
return new MachProjNode(this, proj->_con, rm, ideal_reg);
1307
}
1308
1309
1310
//------------------------------match------------------------------------------
1311
// return result(s) along with their RegMask info
1312
Node *DivModLNode::match( const ProjNode *proj, const Matcher *match ) {
1313
uint ideal_reg = proj->ideal_reg();
1314
RegMask rm;
1315
if (proj->_con == div_proj_num) {
1316
rm = match->divL_proj_mask();
1317
} else {
1318
assert(proj->_con == mod_proj_num, "must be div or mod projection");
1319
rm = match->modL_proj_mask();
1320
}
1321
return new MachProjNode(this, proj->_con, rm, ideal_reg);
1322
}
1323
1324