-
Notifications
You must be signed in to change notification settings - Fork 0
/
branch_perdictor_carousel.cpp
601 lines (422 loc) · 18.1 KB
/
branch_perdictor_carousel.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
#include <bits/stdc++.h>
#include <fstream>
// include <iostream>
using namespace std;
//THIS FILE IS UNTESTED
//Things I have learned while writing this: I do not know c++. We need an algorithim that is Unsupervised!!!
//Squeezing has been implemented, however this file remains untested.
// complete branch perdictors
int* bimodal(int m, string tracefile) {
vector<int> table(1 << m, 4);
bool actual_taken;
bool pred_taken;
// File
ifstream InFile(tracefile);
string line;
// Stats
int predictions = 0;
int mispredictions = 0;
int pc = 0;
// Find and predict at each branch
while (getline(InFile, line)) {
actual_taken = (line[7] == 't');
// Get PC
pc = stoi(line.substr(0,6), 0, 16);
int index = (pc >> 2) % (1 << m);
// Predict and update
int value = table[index];
pred_taken = (value >= 4);
if (actual_taken && value < 7)
table[index]++;
else if (!actual_taken && value > 0)
table[index]--;
predictions++;
if (actual_taken != pred_taken) mispredictions++;
}
InFile.close();
int *ret = (int *)malloc(sizeof(int) * 2);
ret[0] = predictions;
ret[1] = mispredictions;
return ret;
}
int* gshare(int m, int n, string tracefile) {
int bhr_largest_bit = 1 << (n - 1);
vector<int> table(1 << m, 4);
bool actual_taken;
bool pred_taken;
// File
ifstream InFile(tracefile);
string line;
// Stats
int predictions = 0;
int mispredictions = 0;
int bhr = 0;
int pc = 0;
// Find and predict at each branch
while (getline(InFile, line)) {
actual_taken = (line[7] == 't');
// Get PC
pc = stoi(line.substr(0,6), 0, 16);
// Remove right 2 bits
pc >>= 2;
// Removing last 2 bits, and using m left
// Adding N-M 0's to bhr
// XOR
// n from bhr
// n from PC
// m-n from PC
// shift all of the above until they all have m bits
// M to N bits of shifted PC
// n = 3, m = 6
// 11111111111111111 Original
// 00011111111111111 Shift Right by N
// 11111111111111000 Shift Left By N (now right N bits are 0)
// 1000000 2^M
// 00000000000111000 Mod by 2^M (bits before Mth place are 0)
// ^^^
int index_A = (((pc >> n) << n) % (1 << m));
// Rightmost N bits of shifted PC (XOR with bhr)
int index_B = (pc % (1 << n)) ^ bhr;
int index = index_A + index_B;
// Predict and update
int value = table[index];
pred_taken = (value >= 4);
if (actual_taken && value < 7)
table[index]++;
else if (!actual_taken && value > 0)
table[index]--;
bhr >>= 1;
if (actual_taken)
bhr += bhr_largest_bit;
predictions++;
if (actual_taken != pred_taken) mispredictions++;
}
InFile.close();
int *ret = (int *)malloc(sizeof(int) * 2);
ret[0] = predictions;
ret[1] = mispredictions;
return ret;
}
int* smith(int num_bits, string tracefile) {
int largest_bit = 1 << (num_bits - 1);
int counter = largest_bit;
bool actual_taken;
bool pred_taken;
// File
ifstream InFile(tracefile);
string line;
// Stats
int predictions = 0;
int mispredictions = 0;
// Find and predict at each branch
while (getline(InFile, line)) {
actual_taken = (line[7] == 't');
pred_taken = (counter >= largest_bit);
if (actual_taken && counter < 7)
counter++;
else if (!actual_taken && counter > 0)
counter--;
predictions++;
if (actual_taken != pred_taken) mispredictions++;
}
int *ret = (int *)malloc(sizeof(int) * 2);
ret[0] = predictions;
ret[1] = mispredictions;
return ret;
}
// segmented branch perdictors
int* bimodal_segmented(int m, string tracefile, int seg_len, int seg_offset) {
vector<int> table(1 << m, 4);
bool actual_taken;
bool pred_taken;
// File
ifstream InFile(tracefile);
string line;
// Stats
int predictions = 0;
int mispredictions = 0;
int pc = 0;
int linecounter = 0;
// Find and predict at each branch
while (getline(InFile, line)) {
linecounter++;
if ((seg_offset <= linecounter) && (linecounter < (seg_offset + seg_len))){
actual_taken = (line[7] == 't');
// Get PC
pc = stoi(line.substr(0,6), 0, 16);
int index = (pc >> 2) % (1 << m);
// Predict and update
int value = table[index];
pred_taken = (value >= 4);
if (actual_taken && value < 7)
table[index]++;
else if (!actual_taken && value > 0)
table[index]--;
predictions++;
if (actual_taken != pred_taken) mispredictions++;
}
}
InFile.close();
int *ret = (int *)malloc(sizeof(int) * 4);
ret[0] = predictions;
ret[1] = mispredictions;
ret[2] = seg_offset;
ret[3] = seg_len + seg_offset;
return ret;
}
int* gshare_segmented(int m, int n, string tracefile, int seg_len, int seg_offset) {
int bhr_largest_bit = 1 << (n - 1);
vector<int> table(1 << m, 4);
bool actual_taken;
bool pred_taken;
// File
ifstream InFile(tracefile);
string line;
// Stats
int predictions = 0;
int mispredictions = 0;
int bhr = 0;
int pc = 0;
int linecounter = 0;
// Find and predict at each branch
while (getline(InFile, line)) {
linecounter ++;
if ((seg_offset <= linecounter) && (linecounter < (seg_offset + seg_len))){
actual_taken = (line[7] == 't');
// Get PC
pc = stoi(line.substr(0,6), 0, 16);
// Remove right 2 bits
pc >>= 2;
// Removing last 2 bits, and using m left
// Adding N-M 0's to bhr
// XOR
// n from bhr
// n from PC
// m-n from PC
// shift all of the above until they all have m bits
// M to N bits of shifted PC
// n = 3, m = 6
// 11111111111111111 Original
// 00011111111111111 Shift Right by N
// 11111111111111000 Shift Left By N (now right N bits are 0)
// 1000000 2^M
// 00000000000111000 Mod by 2^M (bits before Mth place are 0)
// ^^^
int index_A = (((pc >> n) << n) % (1 << m));
// Rightmost N bits of shifted PC (XOR with bhr)
int index_B = (pc % (1 << n)) ^ bhr;
int index = index_A + index_B;
// Predict and update
int value = table[index];
pred_taken = (value >= 4);
if (actual_taken && value < 7)
table[index]++;
else if (!actual_taken && value > 0)
table[index]--;
bhr >>= 1;
if (actual_taken)
bhr += bhr_largest_bit;
predictions++;
if (actual_taken != pred_taken) mispredictions++;
}
}
InFile.close();
int *ret = (int *)malloc(sizeof(int) * 4);
ret[0] = predictions;
ret[1] = mispredictions;
ret[2] = seg_offset;
ret[3] = seg_len + seg_offset;
return ret;
}
int* smith_segmented(int num_bits, string tracefile, int seg_len, int seg_offset) {
int largest_bit = 1 << (num_bits - 1);
int counter = largest_bit;
bool actual_taken;
bool pred_taken;
// File
ifstream InFile(tracefile);
string line;
// Stats
int predictions = 0;
int mispredictions = 0;
int linecounter = 0;
// Find and predict at each branch
while (getline(InFile, line)) {
linecounter ++;
if ((seg_offset <= linecounter) && (linecounter < (seg_offset + seg_len))){
actual_taken = (line[7] == 't');
pred_taken = (counter >= largest_bit);
if (actual_taken && counter < 7)
counter++;
else if (!actual_taken && counter > 0)
counter--;
predictions++;
if (actual_taken != pred_taken) mispredictions++;
}
}
int *ret = (int *)malloc(sizeof(int) * 2);
ret[0] = predictions;
ret[1] = mispredictions;
ret[2] = seg_offset;
ret[3] = seg_len + seg_offset;
return ret;
}
void autosqueeze(int segment_length, int iterations, int amount_traces, string path_to_trace){
//basic branch perdictor fitting below
int perdictor_amount = 3; //amount of perdictors we're using
int segement_amount = amount_traces / segment_length; //amount of segments trace is divided into
// configure branch perdictor defualts
int bimodal_m = 6;
int gshare_m = 9;
int gshare_n = 3;
int smith_bits = 3;
// intialize biases for branch perdictors
float perdictor_bias[segement_amount][perdictor_amount];
cout << segement_amount;
// calculate bias for each section
int offset = 0; //inital offset for determining where in trace this segment begins
/*
for(int segment = 0; segment <= segement_amount; segment++){
int* bimodal_inital_bias = bimodal_segmented(bimodal_m, path_to_trace, segment_length, offset);
int* gshare_inital_bias = gshare_segmented(gshare_m, gshare_n, path_to_trace, segment_length, offset);
int* smith_inital_bias = smith_segmented(smith_bits, path_to_trace, segment_length, offset);
perdictor_bias[segment][0] = 1 - (static_cast<float>(bimodal_inital_bias[1]) / static_cast<float>(bimodal_inital_bias[0]));
perdictor_bias[segment][1] = 1 - (static_cast<float>(gshare_inital_bias[1]) / static_cast<float>(gshare_inital_bias[0]));
perdictor_bias[segment][2] = 1 - (static_cast<float>(smith_inital_bias[1]) / static_cast<float>(smith_inital_bias[0]));
// this is to visualize the parameters in the actual result array
cout << "biases for Segment " << segment;
cout << " bimodal, gshare, smith. ";
cout << "\n\n";
cout << perdictor_bias[segment][0];
cout << "\n";
cout << perdictor_bias[segment][1];
cout << "\n";
cout << perdictor_bias[segment][2];
cout << "\n";
offset = segment + offset;
}
*/
// this threshold value is the value by determining if the segment or 'window' will be expanded
float threshold = 0.80;
// this is the amount to expand current segment, deincrement by
int segment_step = segment_length/10;
// this is the segment/window to start on
int segment = 0;
// this is the value by which to squeeze the next segment (i.e. if we're expanding segment "3" for example by 140, we need to subtract
// that amount from the next segment, in this case segment 4)
int segment_squeeze = 0;
// this value stores the original segment length for use in calulating the segment length of new segments
int original_segment_length = segment_length;
// this value tells you if you've actually jumped to a new segment or not (flag variable)
bool new_segment = false;
// this method implements squeezing rudimentairily
while(segment <= segement_amount){
// if we're on a new segment, we don't wan't to use the old value of length that may have been modified for one of the older perdictors
if (new_segment){
//also if we're on a new segment, we want to make sure it's appropiatly squeezed, that is we don't want to overlap the first part of this
//trace if it's already been analyzed by the last perdictor. (if the previous segment expanded onto this one)
segment_length = original_segment_length - segment_squeeze;
//If this segment has been squeezed, we want to make sure we're starting in the right area in the trace.
offset = offset + segment_squeeze;
}
// implementation of the for loop with autosqueezing - squeeze forward
int* bimodal_inital_bias = bimodal_segmented(bimodal_m, path_to_trace, segment_length, offset);
int* gshare_inital_bias = gshare_segmented(gshare_m, gshare_n, path_to_trace, segment_length, offset);
int* smith_inital_bias = smith_segmented(smith_bits, path_to_trace, segment_length, offset);
perdictor_bias[segment][0] = 1 - (static_cast<float>(bimodal_inital_bias[1]) / static_cast<float>(bimodal_inital_bias[0]));
perdictor_bias[segment][1] = 1 - (static_cast<float>(gshare_inital_bias[1]) / static_cast<float>(gshare_inital_bias[0]));
perdictor_bias[segment][2] = 1 - (static_cast<float>(smith_inital_bias[1]) / static_cast<float>(smith_inital_bias[0]));
// this is where 'squeezing' is implemented -
// let's take the best value of these perdictors - that is the one with the highest accuracy
float best_perdictor = std::max({perdictor_bias[segment][0], perdictor_bias[segment][1], perdictor_bias[segment][2] });
// if it's soo good that we've expanded it 10 times, (as in it basically consumed the next segment, we want to make sure we don't
// try to analyze that one again)
if (segment_length = 2 * original_segment_length){
segment = segment + 2;
// don't forget to update the offset
offset = segment * offset;
// at this point we can do no more expanding or squeezing, so just move to the next segment
new_segment = true;
}
// if one of the perdictors is above the threshold value, expand by [segment step] and recalculate the percentage
else if (best_perdictor > threshold){
segment_squeeze = segment_squeeze + segment_step;
segment_length = segment_length + segment_squeeze;
new_segment = false;
}
// if the best perdictor is not above the threshold value, just move on.
else{
segment++;
offset = segment * offset;
new_segment = true;
}
// this is to visualize the parameters in the actual result array
cout << "biases for Segment " << segment;
cout << " bimodal, gshare, smith. ";
cout << "\n\n";
cout << perdictor_bias[segment][0];
cout << "\n";
cout << perdictor_bias[segment][1];
cout << "\n";
cout << perdictor_bias[segment][2];
cout << "\n";
}
}
void mv_autosqueeze(int segment_length, int iterations, int amount_traces, string paths_to_traces){
// to implement squeezing
//for(int iteration = 0; iteration < iterations; iteration++)
//SQUEEZING HAS NOT YET BEEN IMPLEMENTED YET!!!!!
//This version is diffrent!!! it takes in an array of string traces, and attempts to average biases across all traces!!
//This is basically the auto squeeze function implemented over a list of strings, averaging out biases over a bunch of trace files.
//To use this one we need multiple trace files, with their paths in a list of strings. Perferably with them all having the same amount of
//perdictions in them.
//Initialization stuff
int perdictor_amount = 3; //amount of perdictors we're using
int segement_amount = amount_traces / segment_length; //amount of segments trace is divided into
// configure branch perdictor defualts
int bimodal_m = 6;
int gshare_m = 9;
int gshare_n = 3;
int smith_bits = 3;
// intialize biases for branch perdictors
float perdictor_bias[segement_amount][perdictor_amount];
cout << segement_amount;
// calculate bias for each section
for(int trace = 0; trace <= sizeof(paths_to_traces); trace++){
int offset = 0; //inital offset for determining where in trace this segment begins
for(int segment = 0; segment <= segement_amount; segment++){
int* bimodal_inital_bias = bimodal_segmented(bimodal_m, paths_to_traces, segment_length, offset);
int* gshare_inital_bias = gshare_segmented(gshare_m, gshare_n, paths_to_traces, segment_length, offset);
int* smith_inital_bias = smith_segmented(smith_bits, paths_to_traces, segment_length, offset);
if(trace == 0){
perdictor_bias[segment][0] = 1 - (static_cast<float>(bimodal_inital_bias[1]) / static_cast<float>(bimodal_inital_bias[0]));
perdictor_bias[segment][1] = 1 - (static_cast<float>(gshare_inital_bias[1]) / static_cast<float>(gshare_inital_bias[0]));
perdictor_bias[segment][2] = 1 - (static_cast<float>(smith_inital_bias[1]) / static_cast<float>(smith_inital_bias[0]));
}
else{
perdictor_bias[segment][0] = (perdictor_bias[segment][0] + (1 - (static_cast<float>(bimodal_inital_bias[1]) / static_cast<float>(bimodal_inital_bias[0]))))/2;
perdictor_bias[segment][1] = (perdictor_bias[segment][1] + (1 - (static_cast<float>(gshare_inital_bias[1]) / static_cast<float>(gshare_inital_bias[0]))))/2;
perdictor_bias[segment][2] = (perdictor_bias[segment][2] + (1 - (static_cast<float>(smith_inital_bias[1]) / static_cast<float>(smith_inital_bias[0]))))/2;
}
//this is to visualize the parameters in the actual result array
cout << "biases for Segment " << segment;
cout << " bimodal, gshare, smith. ";
cout << "\n\n";
cout << perdictor_bias[segment][0];
cout << "\n";
cout << perdictor_bias[segment][1];
cout << "\n";
cout << perdictor_bias[segment][2];
cout << "\n";
offset = segment + offset;
}
}
}
int main(string path_to_tracefile){
// For this to work, we need to know the aproximate total amount of traces (branches) in the program
string tracefile = path_to_tracefile;
int amount_traces = 2000000; //originally used gcc1
cout << "testing \n";
autosqueeze(100000, 1, amount_traces, tracefile);
return 0;
}