1/*
2 * Copyright (c) 2017 Lima Project
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sub license,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the
12 * next paragraph) shall be included in all copies or substantial portions
13 * of the Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
21 * DEALINGS IN THE SOFTWARE.
22 *
23 */
24
25#include <limits.h>
26
27#include "ppir.h"
28
29static int cmp_int(const void *a, const void *b)
30{
31   return (*(int*)a - *(int*)b);
32}
33
34static void ppir_schedule_calc_sched_info(ppir_instr *instr)
35{
36   int n = 0;
37   float extra_reg = 1.0;
38
39   /* update all children's sched info */
40   ppir_instr_foreach_pred(instr, dep) {
41      ppir_instr *pred = dep->pred;
42
43      if (pred->reg_pressure < 0)
44         ppir_schedule_calc_sched_info(pred);
45
46      if (instr->est < pred->est + 1)
47         instr->est = pred->est + 1;
48
49      float reg_weight = 1.0 - 1.0 / list_length(&pred->succ_list);
50      if (extra_reg > reg_weight)
51         extra_reg = reg_weight;
52
53      n++;
54   }
55
56   /* leaf instr */
57   if (!n) {
58      instr->reg_pressure = 0;
59      return;
60   }
61
62   int i = 0, reg[n];
63   ppir_instr_foreach_pred(instr, dep) {
64      ppir_instr *pred = dep->pred;
65      reg[i++] = pred->reg_pressure;
66   }
67
68   /* sort */
69   qsort(reg, n, sizeof(reg[0]), cmp_int);
70
71   for (i = 0; i < n; i++) {
72      int pressure = reg[i] + n - (i + 1);
73      if (pressure > instr->reg_pressure)
74         instr->reg_pressure = pressure;
75   }
76
77   /* If all children of this instr have multi parents, then this
78    * instr need an extra reg to store its result. For example,
79    * it's not fair for parent has the same reg pressure as child
80    * if n==1 and child's successor>1, because we need 2 reg for
81    * this.
82    *
83    * But we can't add a full reg to the reg_pressure, because the
84    * last parent of a multi-successor child doesn't need an extra
85    * reg. For example, a single child (with multi successor) instr
86    * should has less reg pressure than a two children (with single
87    * successor) instr.
88    *
89    * extra reg = min(all child)(1.0 - 1.0 / num successor)
90    */
91   instr->reg_pressure += extra_reg;
92}
93
94static void ppir_insert_ready_list(struct list_head *ready_list,
95                                   ppir_instr *insert_instr)
96{
97   struct list_head *insert_pos = ready_list;
98
99   list_for_each_entry(ppir_instr, instr, ready_list, list) {
100      if (insert_instr->parent_index < instr->parent_index ||
101          (insert_instr->parent_index == instr->parent_index &&
102           (insert_instr->reg_pressure < instr->reg_pressure ||
103            (insert_instr->reg_pressure == instr->reg_pressure &&
104             (insert_instr->est >= instr->est))))) {
105         insert_pos = &instr->list;
106         break;
107      }
108   }
109
110   list_del(&insert_instr->list);
111   list_addtail(&insert_instr->list, insert_pos);
112}
113
114static void ppir_schedule_ready_list(ppir_block *block,
115                                     struct list_head *ready_list)
116{
117   if (list_is_empty(ready_list))
118      return;
119
120   ppir_instr *instr = list_first_entry(ready_list, ppir_instr, list);
121   list_del(&instr->list);
122
123   /* schedule the instr to the block instr list */
124   list_add(&instr->list, &block->instr_list);
125   instr->scheduled = true;
126   block->sched_instr_index--;
127   instr->seq = block->sched_instr_base + block->sched_instr_index;
128
129   ppir_instr_foreach_pred(instr, dep) {
130      ppir_instr *pred = dep->pred;
131      pred->parent_index = block->sched_instr_index;
132
133      bool ready = true;
134      ppir_instr_foreach_succ(pred, dep) {
135         ppir_instr *succ = dep->succ;
136         if (!succ->scheduled) {
137            ready = false;
138            break;
139         }
140      }
141      /* all successor have been scheduled */
142      if (ready)
143         ppir_insert_ready_list(ready_list, pred);
144   }
145
146   ppir_schedule_ready_list(block, ready_list);
147}
148
149/* Register sensitive schedule algorithm from paper:
150 * "Register-Sensitive Selection, Duplication, and Sequencing of Instructions"
151 * Author: Vivek Sarkar,  Mauricio J. Serrano,  Barbara B. Simons
152 */
153static void ppir_schedule_block(ppir_block *block)
154{
155   /* move all instr to instr_list, block->instr_list will
156    * contain schedule result */
157   struct list_head instr_list;
158   list_replace(&block->instr_list, &instr_list);
159   list_inithead(&block->instr_list);
160
161   /* step 2 & 3 */
162   list_for_each_entry(ppir_instr, instr, &instr_list, list) {
163      if (ppir_instr_is_root(instr))
164         ppir_schedule_calc_sched_info(instr);
165      block->sched_instr_index++;
166   }
167   block->sched_instr_base = block->comp->sched_instr_base;
168   block->comp->sched_instr_base += block->sched_instr_index;
169
170   /* step 4 */
171   struct list_head ready_list;
172   list_inithead(&ready_list);
173
174   /* step 5 */
175   list_for_each_entry_safe(ppir_instr, instr, &instr_list, list) {
176      if (ppir_instr_is_root(instr)) {
177         instr->parent_index = INT_MAX;
178         ppir_insert_ready_list(&ready_list, instr);
179      }
180   }
181
182   /* step 6 */
183   ppir_schedule_ready_list(block, &ready_list);
184}
185
186bool ppir_schedule_prog(ppir_compiler *comp)
187{
188   list_for_each_entry(ppir_block, block, &comp->block_list, list) {
189      ppir_schedule_block(block);
190   }
191
192   return true;
193}
194