MiniProject on Algorithms for Multicore Systems 20214181

1
Mini-Project on Algorithms for Multicore
Systems(20214181)
http//www.cs.bgu.ac.il/mpam092/Main

• Instructor Danny Hendler

2
Mini-project Plan
An introductory lecture on shared-memory
algorithms for multi-core systems
20/4/09

• Team formulation paper selection
• Teams of 1, 2, or 3 students
• Each team should give (at least!) 3 priorities
  for papers they prefer

25/4/09
Paper allocation published
27/4/09
2
3
From the New York Times
3
4
The Future of Computing

• Speeding up uni-processors is harder and harder
• Intel, Sun, AMD, IBM now focusing on multi-core
  architectures
• Already, most computers are multiprocessors

How can we write correct and scalable algorithms
for multiprocessors?
4
5
A fundamental problem of thread-level parallelism
Thread A
Thread B
.
. Accounti Accounti-X Accountj
AccountjX .
. .
.
. Accounti Accounti-X Accountj
AccountjX .
. .
But what if execution is concurrent?
Must avoid race conditions
6
Inter-thread synch. alternatives

• Coarse-grained locks
• Fine-grained locks
• Lock-free implementations
• Transactional memory

7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
More on lock-free synchronization

• The use of locks can sometimes be avoided if the
  hardware supports stronger Read-Modify-Write
  operations and not just read and write
• Test-and-set
• Fetch-and-add
• Compare-and-swap

Test-and-set(w) atomically v read from
w w 1 return v
Fetch-and-add(w, delta) atomically v
read from w w vdelta return v
11
The compare-and-swap (CAS) operation
Comareswap(w, expected, new) atomically v
read from w if (v expected) w
new return success else return
failure
Motorola 680x0 IBM 370 Sun SPARC 80X86, Pentium
MIPS PowerPC DECAlpha
12
An example CAS usageTreibers stack algorithm

val
val
val
Top
next
next
next

• Push(int v, Stack S)
• n new NODE create node for new stack item
• n.val v write item value
• do forever repeat until success
• node top S.top
• n.next top next points to current
  top (LIFO order)
• if compareswap(S.top, top, n) try to add
  new item
• return return
  if succeeded
• od

13
Treibers stack algorithm (contd)

val
val
val
Top
next
next
next

• Pop(Stack S)
• do forever
• top S.top
• if top null
• return empty
• if compareswap(S.top, top, top.next)
• return-valtop.val
• free top
• return return-val
• od

14
More on lock-free synchronization (contd)
Lock-free synchronization algorithms are often
complicated
15
Transactional Memory

• A transaction is a sequence of memory reads and
  writes, executed by a single thread, that either
  commits or aborts
• If a transaction commits, all the reads and
  writes appear to have executed atomically
• If a transaction aborts, none of its stores take
  effect
• Transaction operations aren't visible until they
  commit (if they do)

16
Transactional Memory Goals

• A new multiprocessor architecture
• The goal Implementing lock-free synchronization
  that is
• efficient
• easy to use compared with conventional
  techniques based on mutual exclusion
• Implemented by hardware support (such as
  straightforward extensions to multiprocessor
  cache-coherence protocols) and / or by software
  mechanisms

17
A Usage Example
Accounti Accounti-X Accountj
AccountjX
Locks Lock(Li) Lock(Lj) Accounti
Accounti X Accountj Accountj X
Unlock(Lj) Unlock(Li)

• Transactional Memory
• atomic
• Accounti Accounti X
• Accountj Accountj X

18
Back to Lock-free and lock-based algorithms
19
Progress Conditions

• Wait-freedom Each thread terminates its
  operation in a finite number of its steps
• Nonblocking (a.k.a. lock-freedom) After a finite
  number of steps, some thread terminates its
  operation
• Obstruction-freedom If a thread runs by itself
  long enough, it will finish its operation

20
Correctness condition linearizability
Linearizability (an intuitive definition) Can
find a point within the time-interval of each
operation, where the operation took place, such
that the operations order is legal.
21
Example A queue
linearizable
q.enq(x)
q.deq(y)
q.enq(y)
q.deq(x)
time
(6)
22
Example
not linearizable
q.enq(x)
q.deq(y)
q.enq(y)
(5)
23
Example
linearizable
q.enq(x)
q.deq(x)
time
(4)
24
Example
multiple orders OK
linearizable
q.enq(x)
q.deq(y)
q.enq(y)
q.deq(x)
time
(8)
25
Lin. points in Treibers algorithm

• Push(int v, Stack S)
• n new NODE create node for new stack item
• n.val v write item value
• do forever repeat until success
• node top S.top
• n.next top next points to current
  top (LIFO order)
• if compareswap(S.top, top, n) try to add
  new item
• return return
  if succeeded
• od

• Pop(Stack S)
• do forever
• top S.top
• if top null
• return empty
• if compareswap(S.top, top, top.next)
• return-valtop.val
• free top
• return return-val
• od

26
Development/deployment platforms

• You will test your algorithm on an 8-core Xeon
  machine, running Gentoo Linux (kernel 2.6.22)
• We will set up accounts for you
• You can also develop (and do initial tests) on
  any PC (obviously, a multi-core if preferable).
  Easiest way for that seems to be
• Install VMware
• Run a compatible Linux on top of VMWare(e.g.,
  Ubuntu 8.1, kernel Linux 2.6.27 should work)

27
Suggested Projects

• A lock-free extensible hash table
• A lock-free open addressing hash table
• Yet another lock-free hash table
• A lock-free elimination-based stack
• A lock-based skip-list
• A lock-based heap
• A lock-based linked list
• A lock-free double-ended queue