179:
77:
409:
36:
288:
would have an even larger loss. To ameliorate the loss, some machines implement branch target prediction: given the address of a branch, they predict the target of that branch. A refinement of the idea predicts the start of a sequential run of instructions given the address of the start of the previous sequential run of instructions.
287:
In machines where this recurrence takes two cycles, the machine loses one full cycle of fetch after every predicted taken branch. As predicted branches happen every 10 instructions or so, this can force a substantial drop in fetch bandwidth. Some machines with longer instruction cache latencies
305:
As the predictor RAM can be 5–10% of the size of the instruction cache, the fetch happens much faster than the instruction cache fetch, and so this recurrence is much faster. If it were not fast enough, it could be parallelized, by predicting target addresses of target branches.
94:
49:
450:
333:
327:
321:
141:
217:
160:
113:
63:
474:
469:
120:
98:
238:
is the part of a processor that predicts the target, i.e. the address of the instruction that is executed next, of a taken
264:
latency grows longer and the fetch width grows wider, branch target extraction becomes a bottleneck. The recurrence is:
188:
127:
55:
443:
109:
87:
315:
243:
436:
231:
134:
257:
239:
370:
253:
which attempts to guess whether a conditional branch will be taken or not-taken (i.e., binary).
280:
261:
250:
420:
298:
Fetch the prediction for the addresses of the targets of branches in that run of instructions
416:
195:
Please help update this article to reflect recent events or newly available information.
383:
463:
17:
76:
357:
The
Behavior of Efficient Virtual Machine Interpreters on Modern Architectures
379:
408:
301:
Select the address corresponding to the branch predicted taken
172:
70:
29:
424:
271:
Instructions in block are scanned to identify branches
246:
is computed by the execution unit of the processor.
101:. Unsourced material may be challenged and removed.
384:"The microarchitecture of Intel, AMD and VIA CPUs"
295:Hash the address of the first instruction in a run
242:or an unconditional branch instruction before the
191:may be compromised due to out-of-date information
291:This predictor reduces the recurrence above to:
268:Instruction cache fetches block of instructions
444:
8:
249:Branch target prediction is not the same as
64:Learn how and when to remove these messages
451:
437:
274:First predicted taken branch is identified
218:Learn how and when to remove this message
161:Learn how and when to remove this message
334:Single thread indirect branch predictor
328:Indirect branch restricted speculation
7:
405:
403:
99:adding citations to reliable sources
349:Accurate Indirect Branch Prediction
423:. You can help Knowledge (XXG) by
322:Indirect branch prediction barrier
25:
277:Target of that branch is computed
45:This article has multiple issues.
407:
244:target of the branch instruction
177:
75:
34:
86:needs additional citations for
53:or discuss these issues on the
1:
27:Part of a computer processor
491:
402:
283:restarts at branch target
110:"Branch target predictor"
347:Driesen; Hölzle (1992),
371:"Branch Target Buffers"
316:Indirect branch control
236:branch target predictor
475:Computer science stubs
470:Instruction processing
232:computer architecture
355:Ertl; Gregg (2001),
95:improve this article
18:Branch Target Buffer
258:parallel processor
240:conditional branch
432:
431:
281:Instruction fetch
262:instruction cache
251:branch prediction
228:
227:
220:
210:
209:
171:
170:
163:
145:
68:
16:(Redirected from
482:
453:
446:
439:
417:computer science
411:
404:
396:
394:
393:
388:
374:
359:
351:
260:designs, as the
223:
216:
205:
202:
196:
189:factual accuracy
181:
180:
173:
166:
159:
155:
152:
146:
144:
103:
79:
71:
60:
38:
37:
30:
21:
490:
489:
485:
484:
483:
481:
480:
479:
460:
459:
458:
457:
400:
391:
389:
386:
378:
369:
366:
354:
346:
343:
341:Further reading
312:
224:
213:
212:
211:
206:
200:
197:
194:
186:This article's
182:
178:
167:
156:
150:
147:
104:
102:
92:
80:
39:
35:
28:
23:
22:
15:
12:
11:
5:
488:
486:
478:
477:
472:
462:
461:
456:
455:
448:
441:
433:
430:
429:
412:
398:
397:
376:
365:
364:External links
362:
361:
360:
352:
342:
339:
338:
337:
331:
325:
319:
311:
308:
303:
302:
299:
296:
285:
284:
278:
275:
272:
269:
226:
225:
208:
207:
185:
183:
176:
169:
168:
83:
81:
74:
69:
43:
42:
40:
33:
26:
24:
14:
13:
10:
9:
6:
4:
3:
2:
487:
476:
473:
471:
468:
467:
465:
454:
449:
447:
442:
440:
435:
434:
428:
426:
422:
419:article is a
418:
413:
410:
406:
401:
385:
381:
377:
372:
368:
367:
363:
358:
353:
350:
345:
344:
340:
335:
332:
329:
326:
323:
320:
317:
314:
313:
309:
307:
300:
297:
294:
293:
292:
289:
282:
279:
276:
273:
270:
267:
266:
265:
263:
259:
254:
252:
247:
245:
241:
237:
233:
222:
219:
204:
192:
190:
184:
175:
174:
165:
162:
154:
143:
140:
136:
133:
129:
126:
122:
119:
115:
112: –
111:
107:
106:Find sources:
100:
96:
90:
89:
84:This article
82:
78:
73:
72:
67:
65:
58:
57:
52:
51:
46:
41:
32:
31:
19:
425:expanding it
414:
399:
390:. Retrieved
356:
348:
304:
290:
286:
255:
248:
235:
229:
214:
198:
187:
157:
151:October 2007
148:
138:
131:
124:
117:
105:
93:Please help
88:verification
85:
61:
54:
48:
47:Please help
44:
464:Categories
392:2017-08-09
380:Fog, Agner
201:March 2017
121:newspapers
50:improve it
56:talk page
310:See also
256:In more
375:(EE461)
336:(STIBP)
135:scholar
330:(IBRS)
324:(IBPB)
137:
130:
123:
116:
108:
415:This
387:(PDF)
318:(IBC)
142:JSTOR
128:books
421:stub
234:, a
114:news
230:In
97:by
466::
382:.
59:.
452:e
445:t
438:v
427:.
395:.
373:.
221:)
215:(
203:)
199:(
193:.
164:)
158:(
153:)
149:(
139:·
132:·
125:·
118:·
91:.
66:)
62:(
20:)
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.
