Class ME703, Fridtjof Siebert Challenges for the RealTime

1Dr. Fridtjof Siebert, aicas

5th Annual

ME703: Challenges for RealTime SW Development for Multicore Systems

Class ME703, Fridtjof Siebert

Challenges for the RealTimeSoftware Development

for Multicore Systems

Dr. Fridtjof Siebert CTOaicas

29 April 2010

http://esc-sv09.techinsightsevents.com/


5th Annual


Typical Problems on Multicore

typical code sequence (C/C++ or Java)

int counter;

void increment(){ counter++; }



5th Annual




int counter;

void increment(){ counter++; } r1 = counter;

r2 = r1 + 1; counter = r2;



5th Annual




int counter;


r2 = r1 + 1; counter = r2;

r1 = counter; r2 = r1 + 1; counter = r2;

Thread 1 Thread 2



5th Annual




int counter;


r2 = r1 + 1; counter = r2;

r1 = counter; r2 = r1 + 1; counter = r2;

Thread 1 Thread 2

One increment() can get lost!



5th Annual




int counter;




5th Annual




int counter;


this code misses synchronization

but on a single core, it practically always works!

on a multicore, chances for failure explode!



5th Annual


Synchronization

solution: synchronize

int counter;

synchronized void increment(){ counter++; }

easy, problem solved.

Or? See later.



5th Annual


Atomic Operations

What is the result of

int a, b; /* 32 bit, initially 0 */

Thread 1 Thread 2 b = a; a = 1;

?



5th Annual


Atomic Operations


int a, b; /* 32 bit, initially 0 */


?

b == 0b == 1



5th Annual


Atomic Operations


long a, b; /* 64 bit, initially 0 */


?



5th Annual


Atomic Operations


long a, b; /* 64 bit, initially 0 */


?

b == 0b == 1b == 4294967296b == 4294967295



5th Annual



polling update

long counter;[..]do { doSomething(); }while (counter < MAX);

counter is incremented by parallel thread

on multicore, changes to counter may not become visible!



5th Annual



polling update

long counter;[..]do { doSomething(); }while (counter < MAX);

counter is incremented by parallel thread

on multicore, changes to counter may not become visible!



5th Annual


Solution: volatile?

polling update

volatile long counter;[..]do { doSomething(); }while (counter < MAX);

works for Java

does not work for C!



5th Annual


Solution: volatile?

polling update

volatile long counter;[..]do { doSomething(); }while (counter < MAX);

works for Java

does not work for C!



5th Annual


We must understand the memory model!

Memory model specifies what optimisations are permitted by the compiler or underlying hardware

C/C++ programs have undefined semantics in case of race conditions

Java defines a strict memory model



5th Annual


Java's memory model:

ordering operations aresynchronized block

accessing a volatile variable

The presence of an ordering operation determines the visible state in shared memory



5th Annual


Java's memory model: Enforcing Order

all reads are completed before entering synchronized block, or

reading a volatile variable

read fence

all writes are completed before exiting a synchronized block, or

writing a volatile var

write fence



5th Annual


Java's memory model: Date Races

data races are not forbidden in Java you can use shared memory variables

your code has to tolerate optimizations

examples collecting debugging / profiling information

useful if occasional errors due to data reaces are tolerable



5th Annual


Example use of Java's memory model:

Shared memory communicationPtr p;boolean p_valid;

Thread 1 p = new Ptr(); p_valid = true;



5th Annual




Thread 1 Thread 2 p = new Ptr(); p_valid = true;



5th Annual




Thread 1 Thread 2 p = new Ptr(); if (p_valid)p_valid = true; { p.call(); }



5th Annual




Thread 1 Thread 2 t1 = new Ptr(); if (p_valid)t2 = true; {p_valid = t2; p.call();p = t1; }

Writes reordered!



5th Annual



Thread 1 Thread 2 t1 = new Ptr(); t3 = p;t2 = true; if (p_valid) p_valid = t2; { p = t1; t3.call(); }

Writes reordered! Reads reordered!




5th Annual



Thread 1 Thread 2 t1 = new Ptr(); t3 = p;t2 = true; if (p_valid) p_valid = t2; { p = t1; t3.call(); }

Writes reordered! Reads reordered!




5th Annual


Shared memory communicationvolatile Ptr p;volatile boolean p_valid;


in Java




5th Annual




in Java




5th Annual




in Java




5th Annual


Shared memory communicationvolatile Obj *p;volatile boolean p_valid;

Thread 1 Thread 2 p = malloc(..); if (p_valid)p_valid = TRUE; { p>f = ..; }

in C?

Example use of C's memory model:



5th Annual




in C? CPU may still reorder memory accesses!




5th Annual








5th Annual








5th Annual




How to fix it? Add memory fences!




5th Annual



Thread 1 Thread 2 p = malloc(..); if (p_valid)asm volatile( { "sfence":::"memory"); p>f = ..;p_valid = TRUE; }





5th Annual



Thread 1 Thread 2 p = malloc(..); if (p_valid)asm volatile( { "sfence":::"memory"); asm volatile(p_valid = TRUE; "lfence":::"memory"); p>f = ..; }





5th Annual



Thread 1 Thread 2 p = malloc(..); if (p_valid)asm volatile( { "sfence":::"memory"); asm volatile(p_valid = TRUE; "lfence":::"memory"); p>f = ..; }





5th Annual


OutofthinAir

imagine this codeint x = 0, n = 0;

Thread 1 Thread 2 for(i=0;i<n;i++) x = 42; x += f(n); print(x);



5th Annual


OutofthinAir

imagine this codeint x = 0, n = 0;

Thread 1 Thread 2 for(i=0;i<n;i++) x = 42; x += f(n); print(x);

can only print 42 in Java



5th Annual


OutofthinAir: Introduction of Writes

loop optimization in C/C++int x = 0, n = 0;

Thread 1 Thread 2 tmp = x; for(i=0;i<n;i++) x = 42; tmp += f(n);x = tmp; print(x);



5th Annual


OutofthinAir: Introduction of Writes

loop optimization in C/C++int x = 0, n = 0;

Thread 1 Thread 2 tmp = x; for(i=0;i<n;i++) x = 42; tmp += f(n);x = tmp; print(x);

can print 0 in C/C++



5th Annual


OutofthinAir

imagine this codeint x = 0, y = 0;

Thread 1 Thread 2 r1 = x; r2 = y;y = r1; x = r2;



5th Annual


OutofthinAir


Thread 1 Thread 2 r1 = x; r2 = y;y = r1; x = r2;

Expected result x == 0; y == 0;

Only possible result in Java



5th Annual


OutofthinAir: Optimization in C/C++


Thread 1 Thread 2 y = 42; r2 = y;r1 = x; x = r2;if (r1 != 42) y = r1;

Possible in upcoming C++ MM. Results in x == 42; y == 42;



5th Annual


Performance on a Multicore

example: singlecore app, 3 threads

all threads synchronize frequently on the same lock



5th Annual






5th Annual



on a multicore




5th Annual


Lockfree Algorithms

typical code sequence

do { x = counter; result = CAS(counter,x,x+1); }while (result != x);



5th Annual


CompareAndSwap Issues



what is the WCET? ∞?



5th Annual


On dual core:

# iterations

fre

qu

en

cy



what is the WCET? ∞?

CompareAndSwap Issues



5th Annual


CompareAndSwap Solution

introduce long enough code sections in between 2 compareandswap loops

then, if a retry is required, one other CPU was successful

after n1 conflicts, we can be sure that all other CPUs are outside the CAS loop



5th Annual


Example: On an 8 CPU system: # of tries limited

CompareAndSwap Solution



5th Annual


Lockfree library code

use of libraries helpsAtomicInteger counter = new AtomicInteger();void increment(){ (void) counter.incrementAndGet();}

Code is easier and safer

Handmade lockfree algorithms are not for everyday development



5th Annual


Conclusion

Code that runs well on single CPU suddenly fails on a multicore

Clear semantics of concurrent code is required for safe applications

Concurrency at a high level is most beneficial

Thanks: This work was partially funded by the European Commission's 7th framework program's JEOPARD project, #216682.


Class ME703, Fridtjof Siebert Challenges for the RealTime

Documents