1 /*
2  * Copyright (c) 1997, 2012, 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/divnode.hpp"
30 #include "opto/machnode.hpp"
31 #include "opto/matcher.hpp"
32 #include "opto/mulnode.hpp"
33 #include "opto/phaseX.hpp"
34 #include "opto/subnode.hpp"
35 
36 // Portions of code courtesy of Clifford Click
37 
38 // Optimization - Graph Style
39 
40 #include <math.h>
41 
42 //----------------------magic_int_divide_constants-----------------------------
43 // Compute magic multiplier and shift constant for converting a 32 bit divide
44 // by constant into a multiply/shift/add series. Return false if calculations
45 // fail.
46 //
47 // Borrowed almost verbatim from Hacker's Delight by Henry S. Warren, Jr. with
48 // minor type name and parameter changes.
magic_int_divide_constants(jint d,jint & M,jint & s)49 static bool magic_int_divide_constants(jint d, jint &M, jint &s) {
50   int32_t p;
51   uint32_t ad, anc, delta, q1, r1, q2, r2, t;
52   const uint32_t two31 = 0x80000000L;     // 2**31.
53 
54   ad = ABS(d);
55   if (d == 0 || d == 1) return false;
56   t = two31 + ((uint32_t)d >> 31);
57   anc = t - 1 - t%ad;     // Absolute value of nc.
58   p = 31;                 // Init. p.
59   q1 = two31/anc;         // Init. q1 = 2**p/|nc|.
60   r1 = two31 - q1*anc;    // Init. r1 = rem(2**p, |nc|).
61   q2 = two31/ad;          // Init. q2 = 2**p/|d|.
62   r2 = two31 - q2*ad;     // Init. r2 = rem(2**p, |d|).
63   do {
64     p = p + 1;
65     q1 = 2*q1;            // Update q1 = 2**p/|nc|.
66     r1 = 2*r1;            // Update r1 = rem(2**p, |nc|).
67     if (r1 >= anc) {      // (Must be an unsigned
68       q1 = q1 + 1;        // comparison here).
69       r1 = r1 - anc;
70     }
71     q2 = 2*q2;            // Update q2 = 2**p/|d|.
72     r2 = 2*r2;            // Update r2 = rem(2**p, |d|).
73     if (r2 >= ad) {       // (Must be an unsigned
74       q2 = q2 + 1;        // comparison here).
75       r2 = r2 - ad;
76     }
77     delta = ad - r2;
78   } while (q1 < delta || (q1 == delta && r1 == 0));
79 
80   M = q2 + 1;
81   if (d < 0) M = -M;      // Magic number and
82   s = p - 32;             // shift amount to return.
83 
84   return true;
85 }
86 
87 //--------------------------transform_int_divide-------------------------------
88 // Convert a division by constant divisor into an alternate Ideal graph.
89 // Return NULL if no transformation occurs.
transform_int_divide(PhaseGVN * phase,Node * dividend,jint divisor)90 static Node *transform_int_divide( PhaseGVN *phase, Node *dividend, jint divisor ) {
91 
92   // Check for invalid divisors
93   assert( divisor != 0 && divisor != min_jint,
94           "bad divisor for transforming to long multiply" );
95 
96   bool d_pos = divisor >= 0;
97   jint d = d_pos ? divisor : -divisor;
98   const int N = 32;
99 
100   // Result
101   Node *q = NULL;
102 
103   if (d == 1) {
104     // division by +/- 1
105     if (!d_pos) {
106       // Just negate the value
107       q = new (phase->C) SubINode(phase->intcon(0), dividend);
108     }
109   } else if ( is_power_of_2(d) ) {
110     // division by +/- a power of 2
111 
112     // See if we can simply do a shift without rounding
113     bool needs_rounding = true;
114     const Type *dt = phase->type(dividend);
115     const TypeInt *dti = dt->isa_int();
116     if (dti && dti->_lo >= 0) {
117       // we don't need to round a positive dividend
118       needs_rounding = false;
119     } else if( dividend->Opcode() == Op_AndI ) {
120       // An AND mask of sufficient size clears the low bits and
121       // I can avoid rounding.
122       const TypeInt *andconi_t = phase->type( dividend->in(2) )->isa_int();
123       if( andconi_t && andconi_t->is_con() ) {
124         jint andconi = andconi_t->get_con();
125         if( andconi < 0 && is_power_of_2(-andconi) && (-andconi) >= d ) {
126           if( (-andconi) == d ) // Remove AND if it clears bits which will be shifted
127             dividend = dividend->in(1);
128           needs_rounding = false;
129         }
130       }
131     }
132 
133     // Add rounding to the shift to handle the sign bit
134     int l = log2_jint(d-1)+1;
135     if (needs_rounding) {
136       // Divide-by-power-of-2 can be made into a shift, but you have to do
137       // more math for the rounding.  You need to add 0 for positive
138       // numbers, and "i-1" for negative numbers.  Example: i=4, so the
139       // shift is by 2.  You need to add 3 to negative dividends and 0 to
140       // positive ones.  So (-7+3)>>2 becomes -1, (-4+3)>>2 becomes -1,
141       // (-2+3)>>2 becomes 0, etc.
142 
143       // Compute 0 or -1, based on sign bit
144       Node *sign = phase->transform(new (phase->C) RShiftINode(dividend, phase->intcon(N - 1)));
145       // Mask sign bit to the low sign bits
146       Node *round = phase->transform(new (phase->C) URShiftINode(sign, phase->intcon(N - l)));
147       // Round up before shifting
148       dividend = phase->transform(new (phase->C) AddINode(dividend, round));
149     }
150 
151     // Shift for division
152     q = new (phase->C) RShiftINode(dividend, phase->intcon(l));
153 
154     if (!d_pos) {
155       q = new (phase->C) SubINode(phase->intcon(0), phase->transform(q));
156     }
157   } else {
158     // Attempt the jint constant divide -> multiply transform found in
159     //   "Division by Invariant Integers using Multiplication"
160     //     by Granlund and Montgomery
161     // See also "Hacker's Delight", chapter 10 by Warren.
162 
163     jint magic_const;
164     jint shift_const;
165     if (magic_int_divide_constants(d, magic_const, shift_const)) {
166       Node *magic = phase->longcon(magic_const);
167       Node *dividend_long = phase->transform(new (phase->C) ConvI2LNode(dividend));
168 
169       // Compute the high half of the dividend x magic multiplication
170       Node *mul_hi = phase->transform(new (phase->C) MulLNode(dividend_long, magic));
171 
172       if (magic_const < 0) {
173         mul_hi = phase->transform(new (phase->C) RShiftLNode(mul_hi, phase->intcon(N)));
174         mul_hi = phase->transform(new (phase->C) ConvL2INode(mul_hi));
175 
176         // The magic multiplier is too large for a 32 bit constant. We've adjusted
177         // it down by 2^32, but have to add 1 dividend back in after the multiplication.
178         // This handles the "overflow" case described by Granlund and Montgomery.
179         mul_hi = phase->transform(new (phase->C) AddINode(dividend, mul_hi));
180 
181         // Shift over the (adjusted) mulhi
182         if (shift_const != 0) {
183           mul_hi = phase->transform(new (phase->C) RShiftINode(mul_hi, phase->intcon(shift_const)));
184         }
185       } else {
186         // No add is required, we can merge the shifts together.
187         mul_hi = phase->transform(new (phase->C) RShiftLNode(mul_hi, phase->intcon(N + shift_const)));
188         mul_hi = phase->transform(new (phase->C) ConvL2INode(mul_hi));
189       }
190 
191       // Get a 0 or -1 from the sign of the dividend.
192       Node *addend0 = mul_hi;
193       Node *addend1 = phase->transform(new (phase->C) RShiftINode(dividend, phase->intcon(N-1)));
194 
195       // If the divisor is negative, swap the order of the input addends;
196       // this has the effect of negating the quotient.
197       if (!d_pos) {
198         Node *temp = addend0; addend0 = addend1; addend1 = temp;
199       }
200 
201       // Adjust the final quotient by subtracting -1 (adding 1)
202       // from the mul_hi.
203       q = new (phase->C) SubINode(addend0, addend1);
204     }
205   }
206 
207   return q;
208 }
209 
210 //---------------------magic_long_divide_constants-----------------------------
211 // Compute magic multiplier and shift constant for converting a 64 bit divide
212 // by constant into a multiply/shift/add series. Return false if calculations
213 // fail.
214 //
215 // Borrowed almost verbatim from Hacker's Delight by Henry S. Warren, Jr. with
216 // minor type name and parameter changes.  Adjusted to 64 bit word width.
magic_long_divide_constants(jlong d,jlong & M,jint & s)217 static bool magic_long_divide_constants(jlong d, jlong &M, jint &s) {
218   int64_t p;
219   uint64_t ad, anc, delta, q1, r1, q2, r2, t;
220   const uint64_t two63 = 0x8000000000000000LL;     // 2**63.
221 
222   ad = ABS(d);
223   if (d == 0 || d == 1) return false;
224   t = two63 + ((uint64_t)d >> 63);
225   anc = t - 1 - t%ad;     // Absolute value of nc.
226   p = 63;                 // Init. p.
227   q1 = two63/anc;         // Init. q1 = 2**p/|nc|.
228   r1 = two63 - q1*anc;    // Init. r1 = rem(2**p, |nc|).
229   q2 = two63/ad;          // Init. q2 = 2**p/|d|.
230   r2 = two63 - q2*ad;     // Init. r2 = rem(2**p, |d|).
231   do {
232     p = p + 1;
233     q1 = 2*q1;            // Update q1 = 2**p/|nc|.
234     r1 = 2*r1;            // Update r1 = rem(2**p, |nc|).
235     if (r1 >= anc) {      // (Must be an unsigned
236       q1 = q1 + 1;        // comparison here).
237       r1 = r1 - anc;
238     }
239     q2 = 2*q2;            // Update q2 = 2**p/|d|.
240     r2 = 2*r2;            // Update r2 = rem(2**p, |d|).
241     if (r2 >= ad) {       // (Must be an unsigned
242       q2 = q2 + 1;        // comparison here).
243       r2 = r2 - ad;
244     }
245     delta = ad - r2;
246   } while (q1 < delta || (q1 == delta && r1 == 0));
247 
248   M = q2 + 1;
249   if (d < 0) M = -M;      // Magic number and
250   s = p - 64;             // shift amount to return.
251 
252   return true;
253 }
254 
255 //---------------------long_by_long_mulhi--------------------------------------
256 // Generate ideal node graph for upper half of a 64 bit x 64 bit multiplication
long_by_long_mulhi(PhaseGVN * phase,Node * dividend,jlong magic_const)257 static Node* long_by_long_mulhi(PhaseGVN* phase, Node* dividend, jlong magic_const) {
258   // If the architecture supports a 64x64 mulhi, there is
259   // no need to synthesize it in ideal nodes.
260   if (Matcher::has_match_rule(Op_MulHiL)) {
261     Node* v = phase->longcon(magic_const);
262     return new (phase->C) MulHiLNode(dividend, v);
263   }
264 
265   // Taken from Hacker's Delight, Fig. 8-2. Multiply high signed.
266   // (http://www.hackersdelight.org/HDcode/mulhs.c)
267   //
268   // int mulhs(int u, int v) {
269   //    unsigned u0, v0, w0;
270   //    int u1, v1, w1, w2, t;
271   //
272   //    u0 = u & 0xFFFF;  u1 = u >> 16;
273   //    v0 = v & 0xFFFF;  v1 = v >> 16;
274   //    w0 = u0*v0;
275   //    t  = u1*v0 + (w0 >> 16);
276   //    w1 = t & 0xFFFF;
277   //    w2 = t >> 16;
278   //    w1 = u0*v1 + w1;
279   //    return u1*v1 + w2 + (w1 >> 16);
280   // }
281   //
282   // Note: The version above is for 32x32 multiplications, while the
283   // following inline comments are adapted to 64x64.
284 
285   const int N = 64;
286 
287   // Dummy node to keep intermediate nodes alive during construction
288   Node* hook = new (phase->C) Node(4);
289 
290   // u0 = u & 0xFFFFFFFF;  u1 = u >> 32;
291   Node* u0 = phase->transform(new (phase->C) AndLNode(dividend, phase->longcon(0xFFFFFFFF)));
292   Node* u1 = phase->transform(new (phase->C) RShiftLNode(dividend, phase->intcon(N / 2)));
293   hook->init_req(0, u0);
294   hook->init_req(1, u1);
295 
296   // v0 = v & 0xFFFFFFFF;  v1 = v >> 32;
297   Node* v0 = phase->longcon(magic_const & 0xFFFFFFFF);
298   Node* v1 = phase->longcon(magic_const >> (N / 2));
299 
300   // w0 = u0*v0;
301   Node* w0 = phase->transform(new (phase->C) MulLNode(u0, v0));
302 
303   // t = u1*v0 + (w0 >> 32);
304   Node* u1v0 = phase->transform(new (phase->C) MulLNode(u1, v0));
305   Node* temp = phase->transform(new (phase->C) URShiftLNode(w0, phase->intcon(N / 2)));
306   Node* t    = phase->transform(new (phase->C) AddLNode(u1v0, temp));
307   hook->init_req(2, t);
308 
309   // w1 = t & 0xFFFFFFFF;
310   Node* w1 = phase->transform(new (phase->C) AndLNode(t, phase->longcon(0xFFFFFFFF)));
311   hook->init_req(3, w1);
312 
313   // w2 = t >> 32;
314   Node* w2 = phase->transform(new (phase->C) RShiftLNode(t, phase->intcon(N / 2)));
315 
316   // w1 = u0*v1 + w1;
317   Node* u0v1 = phase->transform(new (phase->C) MulLNode(u0, v1));
318   w1         = phase->transform(new (phase->C) AddLNode(u0v1, w1));
319 
320   // return u1*v1 + w2 + (w1 >> 32);
321   Node* u1v1  = phase->transform(new (phase->C) MulLNode(u1, v1));
322   Node* temp1 = phase->transform(new (phase->C) AddLNode(u1v1, w2));
323   Node* temp2 = phase->transform(new (phase->C) RShiftLNode(w1, phase->intcon(N / 2)));
324 
325   // Remove the bogus extra edges used to keep things alive
326   PhaseIterGVN* igvn = phase->is_IterGVN();
327   if (igvn != NULL) {
328     igvn->remove_dead_node(hook);
329   } else {
330     for (int i = 0; i < 4; i++) {
331       hook->set_req(i, NULL);
332     }
333   }
334 
335   return new (phase->C) AddLNode(temp1, temp2);
336 }
337 
338 
339 //--------------------------transform_long_divide------------------------------
340 // Convert a division by constant divisor into an alternate Ideal graph.
341 // Return NULL if no transformation occurs.
transform_long_divide(PhaseGVN * phase,Node * dividend,jlong divisor)342 static Node *transform_long_divide( PhaseGVN *phase, Node *dividend, jlong divisor ) {
343   // Check for invalid divisors
344   assert( divisor != 0L && divisor != min_jlong,
345           "bad divisor for transforming to long multiply" );
346 
347   bool d_pos = divisor >= 0;
348   jlong d = d_pos ? divisor : -divisor;
349   const int N = 64;
350 
351   // Result
352   Node *q = NULL;
353 
354   if (d == 1) {
355     // division by +/- 1
356     if (!d_pos) {
357       // Just negate the value
358       q = new (phase->C) SubLNode(phase->longcon(0), dividend);
359     }
360   } else if ( is_power_of_2_long(d) ) {
361 
362     // division by +/- a power of 2
363 
364     // See if we can simply do a shift without rounding
365     bool needs_rounding = true;
366     const Type *dt = phase->type(dividend);
367     const TypeLong *dtl = dt->isa_long();
368 
369     if (dtl && dtl->_lo > 0) {
370       // we don't need to round a positive dividend
371       needs_rounding = false;
372     } else if( dividend->Opcode() == Op_AndL ) {
373       // An AND mask of sufficient size clears the low bits and
374       // I can avoid rounding.
375       const TypeLong *andconl_t = phase->type( dividend->in(2) )->isa_long();
376       if( andconl_t && andconl_t->is_con() ) {
377         jlong andconl = andconl_t->get_con();
378         if( andconl < 0 && is_power_of_2_long(-andconl) && (-andconl) >= d ) {
379           if( (-andconl) == d ) // Remove AND if it clears bits which will be shifted
380             dividend = dividend->in(1);
381           needs_rounding = false;
382         }
383       }
384     }
385 
386     // Add rounding to the shift to handle the sign bit
387     int l = log2_long(d-1)+1;
388     if (needs_rounding) {
389       // Divide-by-power-of-2 can be made into a shift, but you have to do
390       // more math for the rounding.  You need to add 0 for positive
391       // numbers, and "i-1" for negative numbers.  Example: i=4, so the
392       // shift is by 2.  You need to add 3 to negative dividends and 0 to
393       // positive ones.  So (-7+3)>>2 becomes -1, (-4+3)>>2 becomes -1,
394       // (-2+3)>>2 becomes 0, etc.
395 
396       // Compute 0 or -1, based on sign bit
397       Node *sign = phase->transform(new (phase->C) RShiftLNode(dividend, phase->intcon(N - 1)));
398       // Mask sign bit to the low sign bits
399       Node *round = phase->transform(new (phase->C) URShiftLNode(sign, phase->intcon(N - l)));
400       // Round up before shifting
401       dividend = phase->transform(new (phase->C) AddLNode(dividend, round));
402     }
403 
404     // Shift for division
405     q = new (phase->C) RShiftLNode(dividend, phase->intcon(l));
406 
407     if (!d_pos) {
408       q = new (phase->C) SubLNode(phase->longcon(0), phase->transform(q));
409     }
410   } else if ( !Matcher::use_asm_for_ldiv_by_con(d) ) { // Use hardware DIV instruction when
411                                                        // it is faster than code generated below.
412     // Attempt the jlong constant divide -> multiply transform found in
413     //   "Division by Invariant Integers using Multiplication"
414     //     by Granlund and Montgomery
415     // See also "Hacker's Delight", chapter 10 by Warren.
416 
417     jlong magic_const;
418     jint shift_const;
419     if (magic_long_divide_constants(d, magic_const, shift_const)) {
420       // Compute the high half of the dividend x magic multiplication
421       Node *mul_hi = phase->transform(long_by_long_mulhi(phase, dividend, magic_const));
422 
423       // The high half of the 128-bit multiply is computed.
424       if (magic_const < 0) {
425         // The magic multiplier is too large for a 64 bit constant. We've adjusted
426         // it down by 2^64, but have to add 1 dividend back in after the multiplication.
427         // This handles the "overflow" case described by Granlund and Montgomery.
428         mul_hi = phase->transform(new (phase->C) AddLNode(dividend, mul_hi));
429       }
430 
431       // Shift over the (adjusted) mulhi
432       if (shift_const != 0) {
433         mul_hi = phase->transform(new (phase->C) RShiftLNode(mul_hi, phase->intcon(shift_const)));
434       }
435 
436       // Get a 0 or -1 from the sign of the dividend.
437       Node *addend0 = mul_hi;
438       Node *addend1 = phase->transform(new (phase->C) RShiftLNode(dividend, phase->intcon(N-1)));
439 
440       // If the divisor is negative, swap the order of the input addends;
441       // this has the effect of negating the quotient.
442       if (!d_pos) {
443         Node *temp = addend0; addend0 = addend1; addend1 = temp;
444       }
445 
446       // Adjust the final quotient by subtracting -1 (adding 1)
447       // from the mul_hi.
448       q = new (phase->C) SubLNode(addend0, addend1);
449     }
450   }
451 
452   return q;
453 }
454 
455 //=============================================================================
456 //------------------------------Identity---------------------------------------
457 // If the divisor is 1, we are an identity on the dividend.
Identity(PhaseTransform * phase)458 Node *DivINode::Identity( PhaseTransform *phase ) {
459   return (phase->type( in(2) )->higher_equal(TypeInt::ONE)) ? in(1) : this;
460 }
461 
462 //------------------------------Idealize---------------------------------------
463 // Divides can be changed to multiplies and/or shifts
Ideal(PhaseGVN * phase,bool can_reshape)464 Node *DivINode::Ideal(PhaseGVN *phase, bool can_reshape) {
465   if (in(0) && remove_dead_region(phase, can_reshape))  return this;
466   // Don't bother trying to transform a dead node
467   if( in(0) && in(0)->is_top() )  return NULL;
468 
469   const Type *t = phase->type( in(2) );
470   if( t == TypeInt::ONE )       // Identity?
471     return NULL;                // Skip it
472 
473   const TypeInt *ti = t->isa_int();
474   if( !ti ) return NULL;
475   if( !ti->is_con() ) return NULL;
476   jint i = ti->get_con();       // Get divisor
477 
478   if (i == 0) return NULL;      // Dividing by zero constant does not idealize
479 
480   set_req(0,NULL);              // Dividing by a not-zero constant; no faulting
481 
482   // Dividing by MININT does not optimize as a power-of-2 shift.
483   if( i == min_jint ) return NULL;
484 
485   return transform_int_divide( phase, in(1), i );
486 }
487 
488 //------------------------------Value------------------------------------------
489 // A DivINode divides its inputs.  The third input is a Control input, used to
490 // prevent hoisting the divide above an unsafe test.
Value(PhaseTransform * phase) const491 const Type *DivINode::Value( PhaseTransform *phase ) const {
492   // Either input is TOP ==> the result is TOP
493   const Type *t1 = phase->type( in(1) );
494   const Type *t2 = phase->type( in(2) );
495   if( t1 == Type::TOP ) return Type::TOP;
496   if( t2 == Type::TOP ) return Type::TOP;
497 
498   // x/x == 1 since we always generate the dynamic divisor check for 0.
499   if( phase->eqv( in(1), in(2) ) )
500     return TypeInt::ONE;
501 
502   // Either input is BOTTOM ==> the result is the local BOTTOM
503   const Type *bot = bottom_type();
504   if( (t1 == bot) || (t2 == bot) ||
505       (t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
506     return bot;
507 
508   // Divide the two numbers.  We approximate.
509   // If divisor is a constant and not zero
510   const TypeInt *i1 = t1->is_int();
511   const TypeInt *i2 = t2->is_int();
512   int widen = MAX2(i1->_widen, i2->_widen);
513 
514   if( i2->is_con() && i2->get_con() != 0 ) {
515     int32 d = i2->get_con(); // Divisor
516     jint lo, hi;
517     if( d >= 0 ) {
518       lo = i1->_lo/d;
519       hi = i1->_hi/d;
520     } else {
521       if( d == -1 && i1->_lo == min_jint ) {
522         // 'min_jint/-1' throws arithmetic exception during compilation
523         lo = min_jint;
524         // do not support holes, 'hi' must go to either min_jint or max_jint:
525         // [min_jint, -10]/[-1,-1] ==> [min_jint] UNION [10,max_jint]
526         hi = i1->_hi == min_jint ? min_jint : max_jint;
527       } else {
528         lo = i1->_hi/d;
529         hi = i1->_lo/d;
530       }
531     }
532     return TypeInt::make(lo, hi, widen);
533   }
534 
535   // If the dividend is a constant
536   if( i1->is_con() ) {
537     int32 d = i1->get_con();
538     if( d < 0 ) {
539       if( d == min_jint ) {
540         //  (-min_jint) == min_jint == (min_jint / -1)
541         return TypeInt::make(min_jint, max_jint/2 + 1, widen);
542       } else {
543         return TypeInt::make(d, -d, widen);
544       }
545     }
546     return TypeInt::make(-d, d, widen);
547   }
548 
549   // Otherwise we give up all hope
550   return TypeInt::INT;
551 }
552 
553 
554 //=============================================================================
555 //------------------------------Identity---------------------------------------
556 // If the divisor is 1, we are an identity on the dividend.
Identity(PhaseTransform * phase)557 Node *DivLNode::Identity( PhaseTransform *phase ) {
558   return (phase->type( in(2) )->higher_equal(TypeLong::ONE)) ? in(1) : this;
559 }
560 
561 //------------------------------Idealize---------------------------------------
562 // Dividing by a power of 2 is a shift.
Ideal(PhaseGVN * phase,bool can_reshape)563 Node *DivLNode::Ideal( PhaseGVN *phase, bool can_reshape) {
564   if (in(0) && remove_dead_region(phase, can_reshape))  return this;
565   // Don't bother trying to transform a dead node
566   if( in(0) && in(0)->is_top() )  return NULL;
567 
568   const Type *t = phase->type( in(2) );
569   if( t == TypeLong::ONE )      // Identity?
570     return NULL;                // Skip it
571 
572   const TypeLong *tl = t->isa_long();
573   if( !tl ) return NULL;
574   if( !tl->is_con() ) return NULL;
575   jlong l = tl->get_con();      // Get divisor
576 
577   if (l == 0) return NULL;      // Dividing by zero constant does not idealize
578 
579   set_req(0,NULL);              // Dividing by a not-zero constant; no faulting
580 
581   // Dividing by MINLONG does not optimize as a power-of-2 shift.
582   if( l == min_jlong ) return NULL;
583 
584   return transform_long_divide( phase, in(1), l );
585 }
586 
587 //------------------------------Value------------------------------------------
588 // A DivLNode divides its inputs.  The third input is a Control input, used to
589 // prevent hoisting the divide above an unsafe test.
Value(PhaseTransform * phase) const590 const Type *DivLNode::Value( PhaseTransform *phase ) const {
591   // Either input is TOP ==> the result is TOP
592   const Type *t1 = phase->type( in(1) );
593   const Type *t2 = phase->type( in(2) );
594   if( t1 == Type::TOP ) return Type::TOP;
595   if( t2 == Type::TOP ) return Type::TOP;
596 
597   // x/x == 1 since we always generate the dynamic divisor check for 0.
598   if( phase->eqv( in(1), in(2) ) )
599     return TypeLong::ONE;
600 
601   // Either input is BOTTOM ==> the result is the local BOTTOM
602   const Type *bot = bottom_type();
603   if( (t1 == bot) || (t2 == bot) ||
604       (t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
605     return bot;
606 
607   // Divide the two numbers.  We approximate.
608   // If divisor is a constant and not zero
609   const TypeLong *i1 = t1->is_long();
610   const TypeLong *i2 = t2->is_long();
611   int widen = MAX2(i1->_widen, i2->_widen);
612 
613   if( i2->is_con() && i2->get_con() != 0 ) {
614     jlong d = i2->get_con();    // Divisor
615     jlong lo, hi;
616     if( d >= 0 ) {
617       lo = i1->_lo/d;
618       hi = i1->_hi/d;
619     } else {
620       if( d == CONST64(-1) && i1->_lo == min_jlong ) {
621         // 'min_jlong/-1' throws arithmetic exception during compilation
622         lo = min_jlong;
623         // do not support holes, 'hi' must go to either min_jlong or max_jlong:
624         // [min_jlong, -10]/[-1,-1] ==> [min_jlong] UNION [10,max_jlong]
625         hi = i1->_hi == min_jlong ? min_jlong : max_jlong;
626       } else {
627         lo = i1->_hi/d;
628         hi = i1->_lo/d;
629       }
630     }
631     return TypeLong::make(lo, hi, widen);
632   }
633 
634   // If the dividend is a constant
635   if( i1->is_con() ) {
636     jlong d = i1->get_con();
637     if( d < 0 ) {
638       if( d == min_jlong ) {
639         //  (-min_jlong) == min_jlong == (min_jlong / -1)
640         return TypeLong::make(min_jlong, max_jlong/2 + 1, widen);
641       } else {
642         return TypeLong::make(d, -d, widen);
643       }
644     }
645     return TypeLong::make(-d, d, widen);
646   }
647 
648   // Otherwise we give up all hope
649   return TypeLong::LONG;
650 }
651 
652 
653 //=============================================================================
654 //------------------------------Value------------------------------------------
655 // An DivFNode divides its inputs.  The third input is a Control input, used to
656 // prevent hoisting the divide above an unsafe test.
Value(PhaseTransform * phase) const657 const Type *DivFNode::Value( PhaseTransform *phase ) const {
658   // Either input is TOP ==> the result is TOP
659   const Type *t1 = phase->type( in(1) );
660   const Type *t2 = phase->type( in(2) );
661   if( t1 == Type::TOP ) return Type::TOP;
662   if( t2 == Type::TOP ) return Type::TOP;
663 
664   // Either input is BOTTOM ==> the result is the local BOTTOM
665   const Type *bot = bottom_type();
666   if( (t1 == bot) || (t2 == bot) ||
667       (t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
668     return bot;
669 
670   // x/x == 1, we ignore 0/0.
671   // Note: if t1 and t2 are zero then result is NaN (JVMS page 213)
672   // Does not work for variables because of NaN's
673   if( phase->eqv( in(1), in(2) ) && t1->base() == Type::FloatCon)
674     if (!g_isnan(t1->getf()) && g_isfinite(t1->getf()) && t1->getf() != 0.0) // could be negative ZERO or NaN
675       return TypeF::ONE;
676 
677   if( t2 == TypeF::ONE )
678     return t1;
679 
680   // If divisor is a constant and not zero, divide them numbers
681   if( t1->base() == Type::FloatCon &&
682       t2->base() == Type::FloatCon &&
683       t2->getf() != 0.0 ) // could be negative zero
684     return TypeF::make( t1->getf()/t2->getf() );
685 
686   // If the dividend is a constant zero
687   // Note: if t1 and t2 are zero then result is NaN (JVMS page 213)
688   // Test TypeF::ZERO is not sufficient as it could be negative zero
689 
690   if( t1 == TypeF::ZERO && !g_isnan(t2->getf()) && t2->getf() != 0.0 )
691     return TypeF::ZERO;
692 
693   // Otherwise we give up all hope
694   return Type::FLOAT;
695 }
696 
697 //------------------------------isA_Copy---------------------------------------
698 // Dividing by self is 1.
699 // If the divisor is 1, we are an identity on the dividend.
Identity(PhaseTransform * phase)700 Node *DivFNode::Identity( PhaseTransform *phase ) {
701   return (phase->type( in(2) ) == TypeF::ONE) ? in(1) : this;
702 }
703 
704 
705 //------------------------------Idealize---------------------------------------
Ideal(PhaseGVN * phase,bool can_reshape)706 Node *DivFNode::Ideal(PhaseGVN *phase, bool can_reshape) {
707   if (in(0) && remove_dead_region(phase, can_reshape))  return this;
708   // Don't bother trying to transform a dead node
709   if( in(0) && in(0)->is_top() )  return NULL;
710 
711   const Type *t2 = phase->type( in(2) );
712   if( t2 == TypeF::ONE )         // Identity?
713     return NULL;                // Skip it
714 
715   const TypeF *tf = t2->isa_float_constant();
716   if( !tf ) return NULL;
717   if( tf->base() != Type::FloatCon ) return NULL;
718 
719   // Check for out of range values
720   if( tf->is_nan() || !tf->is_finite() ) return NULL;
721 
722   // Get the value
723   float f = tf->getf();
724   int exp;
725 
726   // Only for special case of dividing by a power of 2
727   if( frexp((double)f, &exp) != 0.5 ) return NULL;
728 
729   // Limit the range of acceptable exponents
730   if( exp < -126 || exp > 126 ) return NULL;
731 
732   // Compute the reciprocal
733   float reciprocal = ((float)1.0) / f;
734 
735   assert( frexp((double)reciprocal, &exp) == 0.5, "reciprocal should be power of 2" );
736 
737   // return multiplication by the reciprocal
738   return (new (phase->C) MulFNode(in(1), phase->makecon(TypeF::make(reciprocal))));
739 }
740 
741 //=============================================================================
742 //------------------------------Value------------------------------------------
743 // An DivDNode divides its inputs.  The third input is a Control input, used to
744 // prevent hoisting the divide above an unsafe test.
Value(PhaseTransform * phase) const745 const Type *DivDNode::Value( PhaseTransform *phase ) const {
746   // Either input is TOP ==> the result is TOP
747   const Type *t1 = phase->type( in(1) );
748   const Type *t2 = phase->type( in(2) );
749   if( t1 == Type::TOP ) return Type::TOP;
750   if( t2 == Type::TOP ) return Type::TOP;
751 
752   // Either input is BOTTOM ==> the result is the local BOTTOM
753   const Type *bot = bottom_type();
754   if( (t1 == bot) || (t2 == bot) ||
755       (t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
756     return bot;
757 
758   // x/x == 1, we ignore 0/0.
759   // Note: if t1 and t2 are zero then result is NaN (JVMS page 213)
760   // Does not work for variables because of NaN's
761   if( phase->eqv( in(1), in(2) ) && t1->base() == Type::DoubleCon)
762     if (!g_isnan(t1->getd()) && g_isfinite(t1->getd()) && t1->getd() != 0.0) // could be negative ZERO or NaN
763       return TypeD::ONE;
764 
765   if( t2 == TypeD::ONE )
766     return t1;
767 
768 #if defined(IA32)
769   if (!phase->C->method()->is_strict())
770     // Can't trust native compilers to properly fold strict double
771     // division with round-to-zero on this platform.
772 #endif
773     {
774       // If divisor is a constant and not zero, divide them numbers
775       if( t1->base() == Type::DoubleCon &&
776           t2->base() == Type::DoubleCon &&
777           t2->getd() != 0.0 ) // could be negative zero
778         return TypeD::make( t1->getd()/t2->getd() );
779     }
780 
781   // If the dividend is a constant zero
782   // Note: if t1 and t2 are zero then result is NaN (JVMS page 213)
783   // Test TypeF::ZERO is not sufficient as it could be negative zero
784   if( t1 == TypeD::ZERO && !g_isnan(t2->getd()) && t2->getd() != 0.0 )
785     return TypeD::ZERO;
786 
787   // Otherwise we give up all hope
788   return Type::DOUBLE;
789 }
790 
791 
792 //------------------------------isA_Copy---------------------------------------
793 // Dividing by self is 1.
794 // If the divisor is 1, we are an identity on the dividend.
Identity(PhaseTransform * phase)795 Node *DivDNode::Identity( PhaseTransform *phase ) {
796   return (phase->type( in(2) ) == TypeD::ONE) ? in(1) : this;
797 }
798 
799 //------------------------------Idealize---------------------------------------
Ideal(PhaseGVN * phase,bool can_reshape)800 Node *DivDNode::Ideal(PhaseGVN *phase, bool can_reshape) {
801   if (in(0) && remove_dead_region(phase, can_reshape))  return this;
802   // Don't bother trying to transform a dead node
803   if( in(0) && in(0)->is_top() )  return NULL;
804 
805   const Type *t2 = phase->type( in(2) );
806   if( t2 == TypeD::ONE )         // Identity?
807     return NULL;                // Skip it
808 
809   const TypeD *td = t2->isa_double_constant();
810   if( !td ) return NULL;
811   if( td->base() != Type::DoubleCon ) return NULL;
812 
813   // Check for out of range values
814   if( td->is_nan() || !td->is_finite() ) return NULL;
815 
816   // Get the value
817   double d = td->getd();
818   int exp;
819 
820   // Only for special case of dividing by a power of 2
821   if( frexp(d, &exp) != 0.5 ) return NULL;
822 
823   // Limit the range of acceptable exponents
824   if( exp < -1021 || exp > 1022 ) return NULL;
825 
826   // Compute the reciprocal
827   double reciprocal = 1.0 / d;
828 
829   assert( frexp(reciprocal, &exp) == 0.5, "reciprocal should be power of 2" );
830 
831   // return multiplication by the reciprocal
832   return (new (phase->C) MulDNode(in(1), phase->makecon(TypeD::make(reciprocal))));
833 }
834 
835 //=============================================================================
836 //------------------------------Idealize---------------------------------------
Ideal(PhaseGVN * phase,bool can_reshape)837 Node *ModINode::Ideal(PhaseGVN *phase, bool can_reshape) {
838   // Check for dead control input
839   if( in(0) && remove_dead_region(phase, can_reshape) )  return this;
840   // Don't bother trying to transform a dead node
841   if( in(0) && in(0)->is_top() )  return NULL;
842 
843   // Get the modulus
844   const Type *t = phase->type( in(2) );
845   if( t == Type::TOP ) return NULL;
846   const TypeInt *ti = t->is_int();
847 
848   // Check for useless control input
849   // Check for excluding mod-zero case
850   if( in(0) && (ti->_hi < 0 || ti->_lo > 0) ) {
851     set_req(0, NULL);        // Yank control input
852     return this;
853   }
854 
855   // See if we are MOD'ing by 2^k or 2^k-1.
856   if( !ti->is_con() ) return NULL;
857   jint con = ti->get_con();
858 
859   Node *hook = new (phase->C) Node(1);
860 
861   // First, special check for modulo 2^k-1
862   if( con >= 0 && con < max_jint && is_power_of_2(con+1) ) {
863     uint k = exact_log2(con+1);  // Extract k
864 
865     // Basic algorithm by David Detlefs.  See fastmod_int.java for gory details.
866     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*/};
867     int trip_count = 1;
868     if( k < ARRAY_SIZE(unroll_factor))  trip_count = unroll_factor[k];
869 
870     // If the unroll factor is not too large, and if conditional moves are
871     // ok, then use this case
872     if( trip_count <= 5 && ConditionalMoveLimit != 0 ) {
873       Node *x = in(1);            // Value being mod'd
874       Node *divisor = in(2);      // Also is mask
875 
876       hook->init_req(0, x);       // Add a use to x to prevent him from dying
877       // Generate code to reduce X rapidly to nearly 2^k-1.
878       for( int i = 0; i < trip_count; i++ ) {
879         Node *xl = phase->transform( new (phase->C) AndINode(x,divisor) );
880         Node *xh = phase->transform( new (phase->C) RShiftINode(x,phase->intcon(k)) ); // Must be signed
881         x = phase->transform( new (phase->C) AddINode(xh,xl) );
882         hook->set_req(0, x);
883       }
884 
885       // Generate sign-fixup code.  Was original value positive?
886       // int hack_res = (i >= 0) ? divisor : 1;
887       Node *cmp1 = phase->transform( new (phase->C) CmpINode( in(1), phase->intcon(0) ) );
888       Node *bol1 = phase->transform( new (phase->C) BoolNode( cmp1, BoolTest::ge ) );
889       Node *cmov1= phase->transform( new (phase->C) CMoveINode(bol1, phase->intcon(1), divisor, TypeInt::POS) );
890       // if( x >= hack_res ) x -= divisor;
891       Node *sub  = phase->transform( new (phase->C) SubINode( x, divisor ) );
892       Node *cmp2 = phase->transform( new (phase->C) CmpINode( x, cmov1 ) );
893       Node *bol2 = phase->transform( new (phase->C) BoolNode( cmp2, BoolTest::ge ) );
894       // Convention is to not transform the return value of an Ideal
895       // since Ideal is expected to return a modified 'this' or a new node.
896       Node *cmov2= new (phase->C) CMoveINode(bol2, x, sub, TypeInt::INT);
897       // cmov2 is now the mod
898 
899       // Now remove the bogus extra edges used to keep things alive
900       if (can_reshape) {
901         phase->is_IterGVN()->remove_dead_node(hook);
902       } else {
903         hook->set_req(0, NULL);   // Just yank bogus edge during Parse phase
904       }
905       return cmov2;
906     }
907   }
908 
909   // Fell thru, the unroll case is not appropriate. Transform the modulo
910   // into a long multiply/int multiply/subtract case
911 
912   // Cannot handle mod 0, and min_jint isn't handled by the transform
913   if( con == 0 || con == min_jint ) return NULL;
914 
915   // Get the absolute value of the constant; at this point, we can use this
916   jint pos_con = (con >= 0) ? con : -con;
917 
918   // integer Mod 1 is always 0
919   if( pos_con == 1 ) return new (phase->C) ConINode(TypeInt::ZERO);
920 
921   int log2_con = -1;
922 
923   // If this is a power of two, they maybe we can mask it
924   if( is_power_of_2(pos_con) ) {
925     log2_con = log2_intptr((intptr_t)pos_con);
926 
927     const Type *dt = phase->type(in(1));
928     const TypeInt *dti = dt->isa_int();
929 
930     // See if this can be masked, if the dividend is non-negative
931     if( dti && dti->_lo >= 0 )
932       return ( new (phase->C) AndINode( in(1), phase->intcon( pos_con-1 ) ) );
933   }
934 
935   // Save in(1) so that it cannot be changed or deleted
936   hook->init_req(0, in(1));
937 
938   // Divide using the transform from DivI to MulL
939   Node *result = transform_int_divide( phase, in(1), pos_con );
940   if (result != NULL) {
941     Node *divide = phase->transform(result);
942 
943     // Re-multiply, using a shift if this is a power of two
944     Node *mult = NULL;
945 
946     if( log2_con >= 0 )
947       mult = phase->transform( new (phase->C) LShiftINode( divide, phase->intcon( log2_con ) ) );
948     else
949       mult = phase->transform( new (phase->C) MulINode( divide, phase->intcon( pos_con ) ) );
950 
951     // Finally, subtract the multiplied divided value from the original
952     result = new (phase->C) SubINode( in(1), mult );
953   }
954 
955   // Now remove the bogus extra edges used to keep things alive
956   if (can_reshape) {
957     phase->is_IterGVN()->remove_dead_node(hook);
958   } else {
959     hook->set_req(0, NULL);       // Just yank bogus edge during Parse phase
960   }
961 
962   // return the value
963   return result;
964 }
965 
966 //------------------------------Value------------------------------------------
Value(PhaseTransform * phase) const967 const Type *ModINode::Value( PhaseTransform *phase ) const {
968   // Either input is TOP ==> the result is TOP
969   const Type *t1 = phase->type( in(1) );
970   const Type *t2 = phase->type( in(2) );
971   if( t1 == Type::TOP ) return Type::TOP;
972   if( t2 == Type::TOP ) return Type::TOP;
973 
974   // We always generate the dynamic check for 0.
975   // 0 MOD X is 0
976   if( t1 == TypeInt::ZERO ) return TypeInt::ZERO;
977   // X MOD X is 0
978   if( phase->eqv( in(1), in(2) ) ) return TypeInt::ZERO;
979 
980   // Either input is BOTTOM ==> the result is the local BOTTOM
981   const Type *bot = bottom_type();
982   if( (t1 == bot) || (t2 == bot) ||
983       (t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
984     return bot;
985 
986   const TypeInt *i1 = t1->is_int();
987   const TypeInt *i2 = t2->is_int();
988   if( !i1->is_con() || !i2->is_con() ) {
989     if( i1->_lo >= 0 && i2->_lo >= 0 )
990       return TypeInt::POS;
991     // If both numbers are not constants, we know little.
992     return TypeInt::INT;
993   }
994   // Mod by zero?  Throw exception at runtime!
995   if( !i2->get_con() ) return TypeInt::POS;
996 
997   // We must be modulo'ing 2 float constants.
998   // Check for min_jint % '-1', result is defined to be '0'.
999   if( i1->get_con() == min_jint && i2->get_con() == -1 )
1000     return TypeInt::ZERO;
1001 
1002   return TypeInt::make( i1->get_con() % i2->get_con() );
1003 }
1004 
1005 
1006 //=============================================================================
1007 //------------------------------Idealize---------------------------------------
Ideal(PhaseGVN * phase,bool can_reshape)1008 Node *ModLNode::Ideal(PhaseGVN *phase, bool can_reshape) {
1009   // Check for dead control input
1010   if( in(0) && remove_dead_region(phase, can_reshape) )  return this;
1011   // Don't bother trying to transform a dead node
1012   if( in(0) && in(0)->is_top() )  return NULL;
1013 
1014   // Get the modulus
1015   const Type *t = phase->type( in(2) );
1016   if( t == Type::TOP ) return NULL;
1017   const TypeLong *tl = t->is_long();
1018 
1019   // Check for useless control input
1020   // Check for excluding mod-zero case
1021   if( in(0) && (tl->_hi < 0 || tl->_lo > 0) ) {
1022     set_req(0, NULL);        // Yank control input
1023     return this;
1024   }
1025 
1026   // See if we are MOD'ing by 2^k or 2^k-1.
1027   if( !tl->is_con() ) return NULL;
1028   jlong con = tl->get_con();
1029 
1030   Node *hook = new (phase->C) Node(1);
1031 
1032   // Expand mod
1033   if( con >= 0 && con < max_jlong && is_power_of_2_long(con+1) ) {
1034     uint k = exact_log2_long(con+1);  // Extract k
1035 
1036     // Basic algorithm by David Detlefs.  See fastmod_long.java for gory details.
1037     // Used to help a popular random number generator which does a long-mod
1038     // of 2^31-1 and shows up in SpecJBB and SciMark.
1039     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*/};
1040     int trip_count = 1;
1041     if( k < ARRAY_SIZE(unroll_factor)) trip_count = unroll_factor[k];
1042 
1043     // If the unroll factor is not too large, and if conditional moves are
1044     // ok, then use this case
1045     if( trip_count <= 5 && ConditionalMoveLimit != 0 ) {
1046       Node *x = in(1);            // Value being mod'd
1047       Node *divisor = in(2);      // Also is mask
1048 
1049       hook->init_req(0, x);       // Add a use to x to prevent him from dying
1050       // Generate code to reduce X rapidly to nearly 2^k-1.
1051       for( int i = 0; i < trip_count; i++ ) {
1052         Node *xl = phase->transform( new (phase->C) AndLNode(x,divisor) );
1053         Node *xh = phase->transform( new (phase->C) RShiftLNode(x,phase->intcon(k)) ); // Must be signed
1054         x = phase->transform( new (phase->C) AddLNode(xh,xl) );
1055         hook->set_req(0, x);    // Add a use to x to prevent him from dying
1056       }
1057 
1058       // Generate sign-fixup code.  Was original value positive?
1059       // long hack_res = (i >= 0) ? divisor : CONST64(1);
1060       Node *cmp1 = phase->transform( new (phase->C) CmpLNode( in(1), phase->longcon(0) ) );
1061       Node *bol1 = phase->transform( new (phase->C) BoolNode( cmp1, BoolTest::ge ) );
1062       Node *cmov1= phase->transform( new (phase->C) CMoveLNode(bol1, phase->longcon(1), divisor, TypeLong::LONG) );
1063       // if( x >= hack_res ) x -= divisor;
1064       Node *sub  = phase->transform( new (phase->C) SubLNode( x, divisor ) );
1065       Node *cmp2 = phase->transform( new (phase->C) CmpLNode( x, cmov1 ) );
1066       Node *bol2 = phase->transform( new (phase->C) BoolNode( cmp2, BoolTest::ge ) );
1067       // Convention is to not transform the return value of an Ideal
1068       // since Ideal is expected to return a modified 'this' or a new node.
1069       Node *cmov2= new (phase->C) CMoveLNode(bol2, x, sub, TypeLong::LONG);
1070       // cmov2 is now the mod
1071 
1072       // Now remove the bogus extra edges used to keep things alive
1073       if (can_reshape) {
1074         phase->is_IterGVN()->remove_dead_node(hook);
1075       } else {
1076         hook->set_req(0, NULL);   // Just yank bogus edge during Parse phase
1077       }
1078       return cmov2;
1079     }
1080   }
1081 
1082   // Fell thru, the unroll case is not appropriate. Transform the modulo
1083   // into a long multiply/int multiply/subtract case
1084 
1085   // Cannot handle mod 0, and min_jlong isn't handled by the transform
1086   if( con == 0 || con == min_jlong ) return NULL;
1087 
1088   // Get the absolute value of the constant; at this point, we can use this
1089   jlong pos_con = (con >= 0) ? con : -con;
1090 
1091   // integer Mod 1 is always 0
1092   if( pos_con == 1 ) return new (phase->C) ConLNode(TypeLong::ZERO);
1093 
1094   int log2_con = -1;
1095 
1096   // If this is a power of two, then maybe we can mask it
1097   if( is_power_of_2_long(pos_con) ) {
1098     log2_con = exact_log2_long(pos_con);
1099 
1100     const Type *dt = phase->type(in(1));
1101     const TypeLong *dtl = dt->isa_long();
1102 
1103     // See if this can be masked, if the dividend is non-negative
1104     if( dtl && dtl->_lo >= 0 )
1105       return ( new (phase->C) AndLNode( in(1), phase->longcon( pos_con-1 ) ) );
1106   }
1107 
1108   // Save in(1) so that it cannot be changed or deleted
1109   hook->init_req(0, in(1));
1110 
1111   // Divide using the transform from DivL to MulL
1112   Node *result = transform_long_divide( phase, in(1), pos_con );
1113   if (result != NULL) {
1114     Node *divide = phase->transform(result);
1115 
1116     // Re-multiply, using a shift if this is a power of two
1117     Node *mult = NULL;
1118 
1119     if( log2_con >= 0 )
1120       mult = phase->transform( new (phase->C) LShiftLNode( divide, phase->intcon( log2_con ) ) );
1121     else
1122       mult = phase->transform( new (phase->C) MulLNode( divide, phase->longcon( pos_con ) ) );
1123 
1124     // Finally, subtract the multiplied divided value from the original
1125     result = new (phase->C) SubLNode( in(1), mult );
1126   }
1127 
1128   // Now remove the bogus extra edges used to keep things alive
1129   if (can_reshape) {
1130     phase->is_IterGVN()->remove_dead_node(hook);
1131   } else {
1132     hook->set_req(0, NULL);       // Just yank bogus edge during Parse phase
1133   }
1134 
1135   // return the value
1136   return result;
1137 }
1138 
1139 //------------------------------Value------------------------------------------
Value(PhaseTransform * phase) const1140 const Type *ModLNode::Value( PhaseTransform *phase ) const {
1141   // Either input is TOP ==> the result is TOP
1142   const Type *t1 = phase->type( in(1) );
1143   const Type *t2 = phase->type( in(2) );
1144   if( t1 == Type::TOP ) return Type::TOP;
1145   if( t2 == Type::TOP ) return Type::TOP;
1146 
1147   // We always generate the dynamic check for 0.
1148   // 0 MOD X is 0
1149   if( t1 == TypeLong::ZERO ) return TypeLong::ZERO;
1150   // X MOD X is 0
1151   if( phase->eqv( in(1), in(2) ) ) return TypeLong::ZERO;
1152 
1153   // Either input is BOTTOM ==> the result is the local BOTTOM
1154   const Type *bot = bottom_type();
1155   if( (t1 == bot) || (t2 == bot) ||
1156       (t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
1157     return bot;
1158 
1159   const TypeLong *i1 = t1->is_long();
1160   const TypeLong *i2 = t2->is_long();
1161   if( !i1->is_con() || !i2->is_con() ) {
1162     if( i1->_lo >= CONST64(0) && i2->_lo >= CONST64(0) )
1163       return TypeLong::POS;
1164     // If both numbers are not constants, we know little.
1165     return TypeLong::LONG;
1166   }
1167   // Mod by zero?  Throw exception at runtime!
1168   if( !i2->get_con() ) return TypeLong::POS;
1169 
1170   // We must be modulo'ing 2 float constants.
1171   // Check for min_jint % '-1', result is defined to be '0'.
1172   if( i1->get_con() == min_jlong && i2->get_con() == -1 )
1173     return TypeLong::ZERO;
1174 
1175   return TypeLong::make( i1->get_con() % i2->get_con() );
1176 }
1177 
1178 
1179 //=============================================================================
1180 //------------------------------Value------------------------------------------
Value(PhaseTransform * phase) const1181 const Type *ModFNode::Value( PhaseTransform *phase ) const {
1182   // Either input is TOP ==> the result is TOP
1183   const Type *t1 = phase->type( in(1) );
1184   const Type *t2 = phase->type( in(2) );
1185   if( t1 == Type::TOP ) return Type::TOP;
1186   if( t2 == Type::TOP ) return Type::TOP;
1187 
1188   // Either input is BOTTOM ==> the result is the local BOTTOM
1189   const Type *bot = bottom_type();
1190   if( (t1 == bot) || (t2 == bot) ||
1191       (t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
1192     return bot;
1193 
1194   // If either number is not a constant, we know nothing.
1195   if ((t1->base() != Type::FloatCon) || (t2->base() != Type::FloatCon)) {
1196     return Type::FLOAT;         // note: x%x can be either NaN or 0
1197   }
1198 
1199   float f1 = t1->getf();
1200   float f2 = t2->getf();
1201   jint  x1 = jint_cast(f1);     // note:  *(int*)&f1, not just (int)f1
1202   jint  x2 = jint_cast(f2);
1203 
1204   // If either is a NaN, return an input NaN
1205   if (g_isnan(f1))    return t1;
1206   if (g_isnan(f2))    return t2;
1207 
1208   // If an operand is infinity or the divisor is +/- zero, punt.
1209   if (!g_isfinite(f1) || !g_isfinite(f2) || x2 == 0 || x2 == min_jint)
1210     return Type::FLOAT;
1211 
1212   // We must be modulo'ing 2 float constants.
1213   // Make sure that the sign of the fmod is equal to the sign of the dividend
1214   jint xr = jint_cast(fmod(f1, f2));
1215   if ((x1 ^ xr) < 0) {
1216     xr ^= min_jint;
1217   }
1218 
1219   return TypeF::make(jfloat_cast(xr));
1220 }
1221 
1222 
1223 //=============================================================================
1224 //------------------------------Value------------------------------------------
Value(PhaseTransform * phase) const1225 const Type *ModDNode::Value( PhaseTransform *phase ) const {
1226   // Either input is TOP ==> the result is TOP
1227   const Type *t1 = phase->type( in(1) );
1228   const Type *t2 = phase->type( in(2) );
1229   if( t1 == Type::TOP ) return Type::TOP;
1230   if( t2 == Type::TOP ) return Type::TOP;
1231 
1232   // Either input is BOTTOM ==> the result is the local BOTTOM
1233   const Type *bot = bottom_type();
1234   if( (t1 == bot) || (t2 == bot) ||
1235       (t1 == Type::BOTTOM) || (t2 == Type::BOTTOM) )
1236     return bot;
1237 
1238   // If either number is not a constant, we know nothing.
1239   if ((t1->base() != Type::DoubleCon) || (t2->base() != Type::DoubleCon)) {
1240     return Type::DOUBLE;        // note: x%x can be either NaN or 0
1241   }
1242 
1243   double f1 = t1->getd();
1244   double f2 = t2->getd();
1245   jlong  x1 = jlong_cast(f1);   // note:  *(long*)&f1, not just (long)f1
1246   jlong  x2 = jlong_cast(f2);
1247 
1248   // If either is a NaN, return an input NaN
1249   if (g_isnan(f1))    return t1;
1250   if (g_isnan(f2))    return t2;
1251 
1252   // If an operand is infinity or the divisor is +/- zero, punt.
1253   if (!g_isfinite(f1) || !g_isfinite(f2) || x2 == 0 || x2 == min_jlong)
1254     return Type::DOUBLE;
1255 
1256   // We must be modulo'ing 2 double constants.
1257   // Make sure that the sign of the fmod is equal to the sign of the dividend
1258   jlong xr = jlong_cast(fmod(f1, f2));
1259   if ((x1 ^ xr) < 0) {
1260     xr ^= min_jlong;
1261   }
1262 
1263   return TypeD::make(jdouble_cast(xr));
1264 }
1265 
1266 //=============================================================================
1267 
DivModNode(Node * c,Node * dividend,Node * divisor)1268 DivModNode::DivModNode( Node *c, Node *dividend, Node *divisor ) : MultiNode(3) {
1269   init_req(0, c);
1270   init_req(1, dividend);
1271   init_req(2, divisor);
1272 }
1273 
1274 //------------------------------make------------------------------------------
make(Compile * C,Node * div_or_mod)1275 DivModINode* DivModINode::make(Compile* C, Node* div_or_mod) {
1276   Node* n = div_or_mod;
1277   assert(n->Opcode() == Op_DivI || n->Opcode() == Op_ModI,
1278          "only div or mod input pattern accepted");
1279 
1280   DivModINode* divmod = new (C) DivModINode(n->in(0), n->in(1), n->in(2));
1281   Node*        dproj  = new (C) ProjNode(divmod, DivModNode::div_proj_num);
1282   Node*        mproj  = new (C) ProjNode(divmod, DivModNode::mod_proj_num);
1283   return divmod;
1284 }
1285 
1286 //------------------------------make------------------------------------------
make(Compile * C,Node * div_or_mod)1287 DivModLNode* DivModLNode::make(Compile* C, Node* div_or_mod) {
1288   Node* n = div_or_mod;
1289   assert(n->Opcode() == Op_DivL || n->Opcode() == Op_ModL,
1290          "only div or mod input pattern accepted");
1291 
1292   DivModLNode* divmod = new (C) DivModLNode(n->in(0), n->in(1), n->in(2));
1293   Node*        dproj  = new (C) ProjNode(divmod, DivModNode::div_proj_num);
1294   Node*        mproj  = new (C) ProjNode(divmod, DivModNode::mod_proj_num);
1295   return divmod;
1296 }
1297 
1298 //------------------------------match------------------------------------------
1299 // return result(s) along with their RegMask info
match(const ProjNode * proj,const Matcher * match)1300 Node *DivModINode::match( const ProjNode *proj, const Matcher *match ) {
1301   uint ideal_reg = proj->ideal_reg();
1302   RegMask rm;
1303   if (proj->_con == div_proj_num) {
1304     rm = match->divI_proj_mask();
1305   } else {
1306     assert(proj->_con == mod_proj_num, "must be div or mod projection");
1307     rm = match->modI_proj_mask();
1308   }
1309   return new (match->C)MachProjNode(this, proj->_con, rm, ideal_reg);
1310 }
1311 
1312 
1313 //------------------------------match------------------------------------------
1314 // return result(s) along with their RegMask info
match(const ProjNode * proj,const Matcher * match)1315 Node *DivModLNode::match( const ProjNode *proj, const Matcher *match ) {
1316   uint ideal_reg = proj->ideal_reg();
1317   RegMask rm;
1318   if (proj->_con == div_proj_num) {
1319     rm = match->divL_proj_mask();
1320   } else {
1321     assert(proj->_con == mod_proj_num, "must be div or mod projection");
1322     rm = match->modL_proj_mask();
1323   }
1324   return new (match->C)MachProjNode(this, proj->_con, rm, ideal_reg);
1325 }
1326