Download presentation
Presentation is loading. Please wait.
Published byHester Allison Modified over 9 years ago
1
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 HIGH PERFORMANCE COMPUTING: MODELS, METHODS, & MEANS FINAL EXAM REVIEW Daniel Kogler, Chirag Dekate & Timur Gilmanov Department of Computer Science Louisiana State University May 5th, 2011
2
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 2 : Parallel Computer Architecture HPC System Stack (5) Performance Factors (7, 8, 9, 10) Scalability (21, 22) MIMD, SIMD, Vector Processing (Pipelineing – Important), Shared Memory 2
3
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 3: Commodity Clusters Commodity clusters vs Constellations (8) Key parameters for cluster computing (24) Where is the parallelism (25) Constituent hardware elements (27) Decoupled Work Queue Model (46) 3
4
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 4 : Benchmarking Basic Performance metrics (4) Benchmarking Definition (5) Purpose of Benchmarking (6) Linpack and HPL 4
5
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 5: Capacity Computing Speedup & Efficiency (5) Capacity, Capability, Cooperative – Important (7,8,11) Ideal Speedup (18,19) Granularities in parallelism (20) Overhead (21, 22, 23, 24) Condor Class Ads (30, 31) Condor MatchMaker(32) Condor commands (37) Capacity Computing Performance issues (53) 5
6
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 6: Communicating Sequential Processes (CSP) Scalability, Strong Scaling, Weak Scaling (7, 8, 9, 10) Cooperative Computing (12) Cooperative Computing Highlights : –Data decomposition Goals in CSP (13) –Distributed Concurrent Processes (14) –Data Exchange (15) –Synchronization(16) Performance issues in CSP (64, 65) 6
7
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 7: MPI Point to point communication in-depth Deadlock & how to resolve deadlocks Be able to Understand MPI programs and detect conceptual flaws in the programs and correct the errors etc. 7
8
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 8: MPI Collective Calls Be able to understand MPI programs and detect conceptual flaws in the program and correct the errors 8
9
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 9: SMP IMPORTANT - AMDAHLS LAW(9,10,11,12, 13) Levels of Memory Hierarchy (30) Cache Measures (31, 32) IMPORTANT - CACHE PERFORMANCE (33) refer to lecture 10 that has more complete and comprehensive coverage of this topic. 9
10
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 10: Enabling Technologies Logic technology metrics (12) What is bisection bandwidth (67) 10
11
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture11: Pthreads IMPORTANT – CPI – Cycles Per Instructions (7, 8, 9, 10) Race Conditions, Critical Sections (15, 16) Thread Synchronization Mechanisms(17 – 22) Important - Deadlock, Livelock, Starvation(29) Priority inversion (30) 11
12
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 12: OpenMP Where are we? (4) – use as a guide for material to study HPC Modelities (5) OpenMP data environment (28) OpenMP work-sharing directives (29) OpenMP thread synchronizations (40, 41, 42, 43) OpenMP reduction Be able to read and understand OpenMP C source code, and detect anomalies, and correct the errors. 12
13
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 13: Performance Performance Counters (11, 12, 13, 14) Performance Analysis Tools (15) Gprof (20) PerfSuite (24,25) PAPI (28, 29, 30, 31, 32) TAU (49, 50, 51, 78, 79) SMP to MPP (60, 62, 63) 13
14
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 14: Visualization Basic gnuplot commands (48 – 56) 14
15
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 15: Parallel Algorithms 1 OpenMP and MPI Matrix Multiplication (54 – 70) Review source code 15
16
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 16: Parallel Algorithms 2 Parallel Matrix Processing and Locality (5, 6, 7, 8, 9) Matrix Transpose (18-19, 21-23) 16
17
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 17: Parallel Algorithms 3 Review Parallel Sorting Algorithms 17
18
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 18: Parallel Algorithms 4 Not Testable 18
19
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 19: Parallel File I/O 1 RAID (9, 10,11, 12, 13, 14) Distributed File Systems: NFS (16, 17, 18) Parallel File Systems (20, 21, 22, 23) 19
20
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 20: Parallel File I/O 2 Not Testable 20
21
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 21: Operating System 1 Operating System (7, 12, 13, 14, 15, 16) Process Management (18, 19, 20, 23, 24, 25) Threads (27, 28) Memory Management (31, 32) Storage Management (35, 36) OS Kernel (39) Modern Operating Systems (41, 42, 44, 45) Unix and Linux (57, 67, 68) 21
22
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 22 Libraries 1 Lecture 19: –Static & Dynamic libraries (27, 28) 22
23
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 23 PFIO3 + Libraries 2 Not testable 23
24
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 24: Operating Systems 2 Linux and Unix Concurrency mechanisms (24, 25) Linux and Unix Scheduling (26, 27,28) Unix & Linux I/O (32, 33) Lightweight Kernels (49, 50) Compute Node Kernels (55, 56, 57) 24
25
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 25: Scheduling Job Scheduling (4, 5) CPU Scheduling algorithms (9 – 15) Workload Management Systems (20 – 27) Scheduling Algorithms for WMS FIFO, FIFO with Backfill, EASY, Conservative (29 – 46) 25
26
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 26 : Checkpointing and System Administration Not Testable 26
27
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 Lecture 27: Beyond and Beyond Not Testable 27
28
CSC 7600 Lecture 28 : Final Exam Review Spring 2010 28
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.