DOC PREVIEW
UMD CMSC 411 - Lecture 2 Computer Design and Evaluation

This preview shows page 1-2 out of 7 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 7 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 7 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 7 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

CMSC 411Computer Systems ArchitectureyLecture 2Computer Design and EvaluationComputer Design and Evaluation Alan [email protected]@csu deduAdministrivia• Unit 1 Homework due Thursday, 2/12– posted on homework/project web page later today– as stated on syllabus, 2 problems will be graded• Finish reading Ch. 1 of H&PCMSC 411 - 22Manufacture of DRAM and other chipsFig. 1.12 in H&P• Chips are manufactured on wafers - circular disks containing many diesg117 AMD Opteronscontaining many dies (chips).• The wafer is tested and chopped into dieschopped into dies.CMSC 411 - 23Wafers and dies• To find the cost of a die:–Number of dies per wafer isat mostthe area ofNumber of dies per wafer is at mostthe area of the wafer divided by the area of the die.– The cost of the wafer divided by the number of working dies per wafer is the cost of each dieworking dies per wafer is the cost of each die.• The fraction of working dies is called the die yieldwhich decreases as the area ofdie yield, which decreases as the area of the die increases.• Rule of thumb (p. 20): Cost of die is (p )proportional to the square of the die areaCMSC 411 - 24Comparing performance of two machines• Definition: Performance is equal to 1 divided by execution time• Problem: How to measure execution time?CMSC 411 - 25What is time?• Unix time command example:–90.7u 12.9s 2:39 65%90.7u 12.9s 2:39 65%– The user used the CPU for 90.7 seconds (user CPU time)–The system used it for 12.9 seconds (system CPU time)–Elapsed time from the user's request toElapsed time from the user s request to completion of the task was 2 minutes, 39 seconds (159 seconds)A d (90 7 + 12 9)/159 65%–And (90.7 + 12.9)/159 = 65% » the rest of the time was spent waiting for I/O or running other programsCMSC 411 - 26Time (cont.)• Usual measurements of time:– system performance measures the elapsed time on unloaded (i l ) t(single user) system– CPU performance measures user CPU time on unloaded systemCMSC 411 - 27How to measure CPU performance• Benchmark: a program used to measure performancep– real programs - what is reality?– kernels - loops in which most of time is spent in a real programprogram– toy programs– synthetic programsFtC t ft t thi dt•Fact: Computer manufacturers tune their product to the popular benchmarks– ``Your results may vary," unless you run benchmark programs and nothing else – See Figure 1.13 listing programs in the SPEC CPU2006 benchmark suiteCMSC 411 - 28Reproducibility• Benchmarking is a laboratory experiment, and needs to be documented as fully as a well-run hit i tchemistry experiment– Identify each variable hardware component– Identify compiler flags and measure variability– Verify reproducibility and provide data for others to reproduce the benchmarking resultsCMSC 411 - 29Reporting results• Example of performance for SPEC CFP2000 benchmark is in H&P Figure 1 14benchmark is in H&P Figure 1.14– for Sun Ultra5, AMD Opteron, Intel Itanium 2– need to look on SPEC web site for all the parameters of the machines, including,g» data on hardware - CPU (how many, primary and secondary caches), memory, disk» data on software – OS, compilers, file system type, and f( )compiler flags used (very important!)• How to summarize results?CMSC 411 - 210Sample resultsFrom Figure 1.15 in H&P 3/e – which machine is fastest?Computer AComputer BComputer CProgram P1 (sec)11020Program P2 (sec)1000 100 20()Program P3 (sec)1001 110 40CMSC 411 - 211()Statistical reporting• “It is rare indeed when advertisers, politicians, pop economists, and p,pp ,drumbeaters for medical programs offer a statistical argument that is not either misleading or downright deceptive”misleading or downright deceptive. -Martin Gardner• “in mathematics you don't understand... in mathematics you don t understand things, you just get used to them.” - John von Neumann• “... if you torture your data long enough, they will tell you what you want to hear.” -James L Mills M DJames L. Mills, M.D.CMSC 411 - 212Statistical reporting (cont.)• The average execution time is the arithmetic mean:– sum the execution times for each program and divide by the number of programs n• The harmonic mean is n divided by the sum of some function of each execution timetime• Often, both of these means are modified to give more weight to more important gg pprogramsCMSC 411 - 213Statistical reporting (cont.)• Arithmetic mean is good if the weights represent your own workload - then you pyycan compare how each machine will perform for youD'ttth i ht b d f–Don'tset the weights based on performance on any particular machine; can get confusing results by doing that• Harmonic mean gives same sort of information, but is used when rates (e g instructions per second) are(e.g., instructions per second) are measured instead of timesCMSC 411 - 214Geometric mean• To measure relative performance, use the geometric mean g– Choose a reference machine, divide all execution times by the corresponding times on the reference machine multiply thoseon the reference machine, multiply those ratios together, and take the nth root of the product• Geometric means have the nice property that you get consistent results (relative performance)regardless of whichperformance) regardless of which machine is used to normalize (Figure 1.14)CMSC 411 - 215How to make computers faster• Make the common case faster!• Example: Put more effort and funds into poptimizing the hardware for addition than to optimize square root•Amdahl's law quantifies this principle:•Amdahl s law quantifies this principle:– Define speedup as the time the task took originally divided by the time the task takes after improvementCMSC 411 - 216Amdahl’s Law• Then Amdahl tells us what the speedup of a particular task is, given– fraction f of the original execution time that the task could use the improvement– the speedup s of a task that always uses the improvement• Then what is the speedup of the task?CMSC 411 - 217Amdahl’s Law (cont.)• Suppose that the original task runs for 1 second so it takes f seconds in the critical piece, and 1-fiththiin other things• Then the task on the improved machine will take only f/s seconds in the critical piece, but will still yp,take 1-f seconds in other things•speedup = =timeold _f1•speedup = =timenew_sff+− )1(CMSC 411 - 218Example 1• Suppose we work very hard improving the square root


View Full Document

UMD CMSC 411 - Lecture 2 Computer Design and Evaluation

Documents in this Course
Load more
Download Lecture 2 Computer Design and Evaluation
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Lecture 2 Computer Design and Evaluation and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Lecture 2 Computer Design and Evaluation 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?