Efficient Data-Race Detection 4/26/2011 Efficient

conchita-marotz

Published 2025-05-12 · 28 slides · 0 views

↓ Download

Efficient Data-Race Detection 4/26/2011 Efficient

Efficient Data-Race Detection 4/26/2011 Efficient - slide 1 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 2 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 3 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 4 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 5 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 6 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 7 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 8 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 9 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 10 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 11 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 12 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 13 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 14 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 15 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 16 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 17 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 18 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 19 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 20 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 21 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 22 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 23 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 24 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 25 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 26 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 27 of 28

Efficient Data-Race Detection 4/26/2011 Efficient - slide 28 of 28

Description

Efficient Data-Race Detection 4262011 Efficient Data-Race Detection 1 Recap Happens-before based data-race detection Shazs guest lecture Vector clock algorithm Lock-set based data-race detection Eraser paper discussion 4262011

Download this presentation From Below

"Efficient Data-Race Detection 4/26/2011 Efficient" is the property of its rightful owner. Permission is granted to download and print the materials on this website for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.

Presentation Transcript

Efficient Data-Race Detection 4/26/2011 Efficient Data-Race Detection 1

Recap Happens-before based data-race detection
Shaz’s guest lecture
Vector clock algorithm

Lock-set based data-race detection
Eraser paper discussion 4/26/2011 Efficient Data-Race Detection 2

New Definition of a Data Race Two instruction conflict if
They access the same memory location
At least one of them is a write

A program contains a data race if
One can schedule the program on a multiprocessor
Such that conflicting instructions execute simultaneously 4/26/2011 Efficient Data-Race Detection 3

New Definition vs Happens-Before 4/26/2011 Efficient Data-Race Detection 4 Lock L X = 1 X = 3 Unlock L Lock L X = 2 Unlock L

New Definition vs Happens-Before 4/26/2011 Efficient Data-Race Detection 5 Lock L X = 1 X = 3 Unlock L Lock L X = 2 Unlock L

New Definition vs Happens-Before True or False ?

If a program has a data race as per the new definition
Then it has a data race as per the happens-before definition 4/26/2011 Efficient Data-Race Detection 6

New Definition vs Happens-Before True or False ?

If a program has a data race as per the new definition
Then it has a happens-before data race

If a program has a happens-before data race
Then it has a data race as per the new definition 4/26/2011 Efficient Data-Race Detection 7

New Definition vs Lock Set True or False ?

If a program has a data race as per the new definition
Then it has a lock-set based data race

If a program has a lock-set based data race
Then it has a data race as per the new definition 4/26/2011 Efficient Data-Race Detection 8

Challenges For Dynamic Data Race Detection Dynamically monitor memory accesses
Need instrumentation infrastructure
Performance overhead 4/26/2011 Efficient Data-Race Detection 9

Challenges For Dynamic Data Race Detection Dynamically monitor memory accesses
Need instrumentation infrastructure
Performance overhead

Understand synchronization mechanisms
Handle home-grown locks
Need to annotate the happens-before relationship 4/26/2011 Efficient Data-Race Detection 10

DataCollider [OSDI ‘10] Lightweight
< 5% overhead

Effective
Found data races in all applications we have run

Easy to implement
The algorithm can be described in ~ 10 lines 4/26/2011 Efficient Data-Race Detection 12

Design Choice #1 Sample memory accesses
Detect data races on a small, random sample of memory accesses 4/26/2011 Efficient Data-Race Detection 13

Design Choice #1 Sample memory accesses
Detect data races on a small, random sample of memory accesses

It is acceptable to miss races
Dynamic techniques cannot find all data races anyway

A lightweight tool will have more coverage
Likely to be run on a lot more tests

But data races are ‘rare’ events
Need to be careful when sampling 4/26/2011 Efficient Data-Race Detection 14

Design Choice #2 Perturb thread schedules
By inserting delays
Force the data race to happen, rather than infer from past memory/synchronization accesses 4/26/2011 Efficient Data-Race Detection 15

Design Choice #2 Perturb thread schedules 4/26/2011 Efficient Data-Race Detection 16 Lock L X = 1 X = 3 Unlock L Lock L X = 2 Unlock L

Design Choice #2 Perturb thread schedules 4/26/2011 Efficient Data-Race Detection 17 Lock L X = 1 X = 3 Unlock L Lock L X = 2 Unlock L X = 1 Insert
Delay

The Algorithm 4/26/2011 Efficient Data-Race Detection 18 At_Every_Memory_Access ( x ) {
if (not sampled) return;

delay for some time;

if another thread performed
a conflicting operation
during the delay

report data race;
}

Use Data Breakpoints for Detecting Conflicting Operations 4/26/2011 Efficient Data-Race Detection 19 At_Every_Memory_Access ( x ) {

if (not sampled) return;

rw = accessIsWrite ? RW : W;

SetDataBreakpoint ( x, rw );

Delay();

ClearDataBreakpoint ( x );

if (breakpoint fired)
ReportDataRace();
}

Sampling Memory Accesses Need to execute the sampling function at every access
Performance overhead

Need a good sampling algorithm
Data races are ‘rare’ events 4/26/2011 Efficient Data-Race Detection 20

Dynamic Sampling Toss a biased coin at each memory access
Sample if coin turns up heads (with small probability) 4/26/2011 Efficient Data-Race Detection 21

Dynamic Sampling Toss a biased coin at each memory access
Sample if coin turns up heads (with small probability)

Hot (frequent) instructions are sampled overwhelmingly more that cold instructions

Bugs (data races) are likely to be present on cold instructions 4/26/2011 Efficient Data-Race Detection 22

Code Breakpoint Sampling Pick a random instruction X
Set a code breakpoint at X
Overwrite the first byte of X with a ‘0xcc’
Sample X when the breakpoint fires and goto Step 1 4/26/2011 Efficient Data-Race Detection 23

Sampling Using Code Breakpoints Samples instructions independent of their execution frequency
Hot and cold instructions are sampled uniformly 4/26/2011 Efficient Data-Race Detection 24

Sampling Using Code Breakpoints Samples instructions independent of their execution frequency
Hot and code instructions are sampled uniformly

Sampling distribution is determined by you (fair_coin_toss)
Sampling rate is determined by the program (unfair_coin_toss) repeat {

t = fair_coin_toss();

while( t != unfair_coin_toss() );

print( t );

} 4/26/2011 Efficient Data-Race Detection 25

The DataCollider Algorithm 4/26/2011 Efficient Data-Race Detection 27 Init:
SetRandomCodeBreakpoints( n );

AtCodeBreakpoint( x ) {

SetDataBreakpoint ( x );

Delay();

if (breakpoint fired)
ReportDataRace();

ClearDataBreakPoint ( x );

SetRandomCodeBreakpoints( 1 );
}

DataCollider Summary Tunable runtime overhead
Tune code breakpoint rate so that we get k samples per second

Finds lots of data races in practice

The fastest implementation
Someone hacked it up during the talk

Pruning intended data races is a problem

Sampling algorithm improvements? 4/26/2011 Efficient Data-Race Detection 28

Efficient Data-Race Detection 4/26/2011 Efficient

Related Topics

Share

Embed code

Download this presentation From Below

Presentation Transcript