Concurrent Systems II - Bad Request - Trinity College Dublin

More documents

Recommendations

Info

Problems • Voluntary ◾Mutexes ‘protect’ code. ◾Other programmers don’t have to use them to get access to the variables the code accesses. • Unfair • This is part of the tradeoff. Use processes rather than threads if you want better protection. ◾If multiple threads are blocked on a mutex, the order in which they waken up is not guaranteed to be any particular order. 54 Trinity College Dublin © Mike Brady 2007
Problem Tutorial • The transpose of a matrix M is a matrix T, where: T[i,j] = M[j,i] for all i, j. • Write a program to transpose a matrix. • How would you parallelise it? • What problems/issues do you foresee? 55 Trinity College Dublin © Mike Brady 2007–2009
Page 1 and 2:
Concurrent Systems II (CS3D4/CS3BA2
Page 3 and 4: POSIX Threads • POSIX Threads aka
Page 5 and 6: Six Separate Threads Master Thread
Page 7 and 8: Where… • ‘thread’ is the ID
Page 9 and 10: Hello World -- Creating Threads int
Page 11 and 12: Hello World -- Thread Function void
Page 13 and 14: Wilde/Stoker/Ubuntu h-3.1$ cc -lpth
Page 15 and 16: Interaction • No thread interacti
Page 17 and 18: Interactions & Dependencies • Int
Page 19 and 20: …to do Parallel Programming • H
Page 21 and 22: Hardware Main Memory Level 2 Cache
Page 23 and 24: Shared Memory Machine • Each proc
Page 25 and 26: Concurrent Program • A Concurrent
Page 27 and 28: Communication • A parallel progra
Page 29 and 30: 1 - Flow Dependence S1: write to lo
Page 31 and 32: 3 - Output Dependence S1: write to
Page 33 and 34: Dependence in Loops j k k i 1 2 j 5
Page 35 and 36: How can we transform it? • We can
Page 37 and 38: Issues: • The variable sum, as wr
Page 39 and 40: Summarising: • We could paralleli
Page 41 and 42: Example • Sample function: y = 0.
Page 43 and 44: Synchronisation • We’ve looked
Page 45 and 46: Shared Variable S: Thread 1 after T
Page 47 and 48: Mutual Exclusion in pthreads. • M
Page 49 and 50: Shared Variable S Protected by Mute
Page 51 and 52: Lock and Unlock Mutex • pthread_m
Page 53: } /********** Critical Section ****
Page 57 and 58: Producers and Consumers • Essenti
Page 59 and 60: Condition Variables (2) • Dramati
Page 61 and 62: From Thread U’s POV (2) • If C
Page 63 and 64: From Thread U’s POV (4) • Acqui
Page 65 and 66: Thread S (2) • Acquire Mutex M
Page 67 and 68: Semaphores • Computer Semaphores
Page 69 and 70: Semaphore Consumer ◾Acquire Mutex
Page 71 and 72: Semaphores (3) • Useful for count
Page 73 and 74: Sample Solution for ∏ • Calcula
Page 75 and 76: unsigned long long clock_cycles() {
Page 77 and 78: The Challenge of Concurrency • Co
Page 79 and 80: Concurrent Execution • A concurre
Page 81 and 82: Notation Trivial Concurrent Program
Page 83 and 84: Trivial Concurrent Program p1: n
Page 85 and 86: Example — Jumping Frogs • A fro
Page 87 and 88: If the frogs can only move “forwa
Page 89 and 90: State Diagrams for Processes • A
Page 91 and 92: Book Reference Principles of concur
Page 93 and 94: Correctness • (Correctness in Seq
Page 95 and 96: Using State machines • We can des
Page 97 and 98: The Critical Section Problem (2)
Page 99 and 100: Observations & Implications • Rem
Page 101 and 102: Simplify Critical Section Problem i
Page 103 and 104: State Transitions r1,s1,1 r1: await
Page 105 and 106:
Deadlock Free? • Deadlock Free: I
Page 107 and 108:
Scorecard Mutual Exclusion Deadlock
Page 109 and 110:
Attempt #2 Critical Section Problem
Page 111 and 112:
Attempt #2 Fails… • Attempt 2 f
Page 113 and 114:
Attempt #3 Critical Section Problem
Page 115 and 116:
Page 117 and 118:
Attempt #4 loop forever Critical Se
Page 119 and 120:
Page 121 and 122:
Attempt #5, derived from #4 Critica
Page 123 and 124:
Verification of concurrent programs
Page 125 and 126:
A quick overview of some mathematic
Page 127 and 128:
A quick overview of some mathematic
Page 129 and 130:
Proving correctness by induction
Page 131 and 132:
A quick overview of some temporal l
Page 133 and 134:
A quick overview of some temporal l
Page 135 and 136:
Temporal logic at work • Consider
Page 137 and 138:
Promela • Promela [Protocol/Proce
Page 139 and 140:
Process (2) • A process type (pro
Page 141 and 142:
SPIN Hello World Example SPIN = Sim
Page 143 and 144:
Promela Variables (2) • Arrays
Page 145 and 146:
Statements (2) • The skip stateme
Page 147 and 148:
If Statement if :: (n % 2 != 0) ->
Page 149 and 150:
Atomic Statement atomic { stat1; st
Page 151 and 152:
Process Execution / Evaluation Sema
Page 153 and 154:
The Closed World Assumption • The
Page 155 and 156:
Process memory management • In a
Page 157 and 158:
Virtual memory • Give each proces
Page 159 and 160:
Virtual memory 0 0 process 1 2 32 0
Page 161 and 162:
Memory management requirements and
Page 163 and 164:
E.g. Virtual memory implementation
Page 165 and 166:
Multi-level page tables Virtual add
Page 167 and 168:
Typical process memory layout memor
Page 169 and 170:
Translation look aside buffer • E
Page 171 and 172:
Page fault handling Start Access pa
Page 173 and 174:
Fetch policy • When should a page
Page 175 and 176:
Page replacement policy • A page
Page 177 and 178:
Page replacement policy • First-I
Page 179 and 180:
Clock Algorithm 179 Trinity College
Page 181 and 182:
Resident set management • Variabl
Page 183 and 184:
Resident set management Local Repla
Page 185 and 186:
Resident set management Δ Working
Page 187 and 188:
Resident set management 187 Trinity
Page 189 and 190:
Cleaning policy • Page buffering
Page 191 and 192:
Load control • We can use a few d
Page 193 and 194:
Process state transitions • Proce
Page 195 and 196:
Preemptive and nonpreemptive schedu
Page 197 and 198:
Scheduling algorithms • First-Com
Page 199 and 200:
Scheduling algorithms • Shortest-
Page 201 and 202:
Scheduling algorithms • Priority
Page 203 and 204:
Scheduling algorithms • Round-Rob
Page 205 and 206:
Example Process Arrival Time Next B
Page 207 and 208:
Scheduling algorithms • Multi-lev
Page 209 and 210:
Scheduling algorithms • Multi-lev
Page 211 and 212:
Multi-processor scheduling • Flyn
Page 213 and 214:
Layered file I/O (1) Application Di
Page 215 and 216:
Disk access • Disks are slow ◾
Page 217 and 218:
Disk scheduling • So, the order i
Page 219 and 220:
Disk scheduling algorithms (2) •
show all

Concurrent Systems II - Bad Request - Trinity College Dublin

Create successful ePaper yourself

Delete template?

Save as template?