Dthreads: Efficient Deterministic Multithreading
Download as PPTX, PDF0 likes1,208 views
The document discusses dthreads, an efficient deterministic multithreading framework that addresses common issues in parallel computing such as race conditions and deadlocks. It highlights the performance advantages of dthreads over traditional methods like pthreads, demonstrating its ability to enable race-free executions and faster debugging. Additionally, the document presents graphical performance analyses comparing the execution times of processes versus threads.
![12
color = ; row = 0; // globals
void nextStripe(){
for (c = 0; c < Width; c++)
drawBox (c,row,color);
color = (color == )? : ;
row++;
}
for (n = 0; n < 9; n++)
pthread_create(t[n], nextStripe);
for (n = 0; n < 9; n++)
pthread_join(t[n]);
JUST USE THREADS…](https://image.slidesharecdn.com/dthreads-sosp-modified-130620212745-phpapp01/85/Dthreads-Efficient-Deterministic-Multithreading-12-320.jpg)
Dthreads: Efficient Deterministic Multithreading
- 1. Tongping Liu, Charlie Curtsinger, Emery Berger DTHREADS: Efficient Deterministic Multithreading Insanity: Doing the same thing over and over again and expecting different results.
- 3. 3 There was the Core.
- 4. 4 And it was Good.
- 5. 5 It gave us our Daily Speed.
- 7. 7 And the Speed was no Moore.
- 8. 8 And then came a False Prophet…
- 9. 9
- 10. 10 Want speed?
- 11. 11 I BRING YOU THE GIFT OF PARALLELISM!
- 12. 12 color = ; row = 0; // globals void nextStripe(){ for (c = 0; c < Width; c++) drawBox (c,row,color); color = (color == )? : ; row++; } for (n = 0; n < 9; n++) pthread_create(t[n], nextStripe); for (n = 0; n < 9; n++) pthread_join(t[n]); JUST USE THREADS…
- 13. 13
- 14. 14
- 15. 15
- 16. 16
- 17. 17
- 19. 19 Salvation?
- 20. 20
- 21. 21 pthreads race conditions atomicity violations deadlock order violations DTHREADS deterministic race conditions atomicity violations deadlock order violations
- 22. 22 DTHREADS Enables… Race-free Executions Replay Debugging w/o Logging Replicated State Machines
- 23. 23 0 1 2 3 4 5 6 runtimerelativetopthreads CoreDet dthreads pthreads 8.47.8 DTHREADS: Efficient Determinism Usually faster than the state of the art
- 24. 24 0 1 2 3 4 5 6 runtimerelativetopthreads CoreDet dthreads pthreads 8.47.8 DTHREADS: Efficient Determinism Generally as fast or faster than pthreads
- 25. 25 % g++ myprog.cpp –l thread DTHREADS: Easy to Use p
- 26. 26 Isolation shared address space disjoint address spaces
- 27. 27 Performance: Processes vs. Threads threads processes 1 2 4 8 16 32 64 128 256 512 1024 Thread Execution Time (ms) 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 NormalizedExecutionTime
- 28. 28 Performance: Processes vs. Threads threads processes 1 2 4 8 16 32 64 128 256 512 1024 Thread Execution Time (ms) 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 NormalizedExecutionTime
- 29. 29 Performance: Processes vs. Threads threads processes 1 2 4 8 16 32 64 128 256 512 1024 Thread Execution Time (ms) 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 NormalizedExecutionTime
- 31. 31 Snapshot pages before modifications “Shared Memory”
- 32. 32 Write back diffs “Shared Memory”
- 33. 33 “Thread” 1 “Thread” 2 “Thread” 3 Parallel Serial Update in Deterministic Time & Order Para mutex_lock cond_wait pthread_create
- 35. 35 Thread 1 Main Memory Core 1 Thread 2 Core 2 Invalidate The Culprit: False Sharing
- 36. 36 Thread 1 Thread 2 Invalidate Main Memory Core 1 Core 2 The Culprit: False Sharing 20x
- 37. 37 Process 1 Process 2 Global State Core 1 Core 2 Process 2 Process 1 DTHREADS: Eliminates False Sharing!
- 38. 38 0 1 2 3 4 5 6 runtimerelativetopthreads ordering only isolation only dthreads DTHREADS: Detailed Analysis
- 39. 39 0 1 2 3 4 5 6 runtimerelativetopthreads ordering only isolation only dthreads DTHREADS: Detailed Analysis
- 40. 40 0 1 2 3 4 5 6 runtimerelativetopthreads ordering only isolation only dthreads DTHREADS: Detailed Analysis
- 41. 41 0 1 2 3 4 speedupof8coresover2cores CoreDet dthreads pthreads DTHREADS: Scalable Determinism
- 42. 42 0 1 2 3 4 speedupof8coresover2cores CoreDet dthreads pthreads DTHREADS: Scalable Determinism
- 43. 43 0 1 2 3 4 speedupof8coresover2cores CoreDet dthreads pthreads DTHREADS: Scalable Determinism
- 44. 44 DTHREADS % g++ myprog.cpp –l threadp
- 45. 45
Editor's Notes
- #2: In the beginning, there was the Core. And it was good.
- #3: Casts out the demons of nondeterminism
- #4: Highlight when same speed or faster.
- #5: Highlight when same speed or faster.
- #6: Obviously this doesn’t preserve shared memory semantics, so we need to commit changes made by one thread so they become visible to others.
- #7: ADD ANIMATIONS: threads initially on one core then migrating, vs. processes spewed across cores
- #8: ADD ANIMATIONS: threads initially on one core then migrating, vs. processes spewed across cores
- #9: ADD ANIMATIONS: threads initially on one core then migrating, vs. processes spewed across cores
- #11: It’s not *always* as fast or faster than pthreads. Slow THEN HIGHLIGHT THE FASTER PARTS.
- #36: Cache coherence protocol makes false sharing problem unpleasant performance effect
- #39: Panel 1 = what it does, panel 2 = how, panel 3 = efficient, panel 4 = easy to use