You are on page 1of 4

702

Printed Pages : 4

lD and Roll No. to be filled in your Answer Book

B. Tech.
2009-10 M.Vil) (ODDSEM.)EXAMTNATTON, ARCHITECTURE )VANCED COMPUTER

Time: 3 Hoursl
Note : Answer all questions.

F-otalMarks : 100

Attempt any two of the following ',

2xL0=20

(a) Explain how instruction set, compiler and control, technology, CPU implementation aftectthe CPU andcacheandmemoryhierarchy performance andjustrfy the effectsin terms of programlength,clock rate, and effectiveCPI.

and (bfnistinguish betweenmultiprocessors mufticomputers based on their structures, "f


resource sharing and interprocesser communications, with A workstationusesa 15-MFIzprocessor rating to executea given a claimed1O-MIPs programmix. Assumeone-cycledelayfor each memoryaccess. (i) What is the effective CPI (Cycles per of this computer? instructions) lContd...

(c)

EE-10811

(ii)

is beingupgraded the computer Suppose with a 30-MHz clock. However, the remains speedof the memorysubsystem two clock and consequently unchanged, If per memoryaccess. cyclesare needed 30% of the instructionsrequire one require and another50lo memoryaccess what per instruction, two memoryaccess is the performanceof the uPgraded instruction with a compatible processor set and equal instructioncounts in the given programmix? 2xl0=20

Attempt any two of the following :

(a) (b) (c)

involvedinto writing code What are the issues for multi-corearchitectures? techniques' program optimization List thevarious Explain.thefollowing : (i) (ii) Threads Mutex

(iii) OpenMP
Attempt any two of the following :

2x10:20

(a) (b)

and the problem Explaingeneralarchitectures of cachecoherence. Compare instructionlevel parallelismwith thread level Parallelism.
lContd...

EE-l0811

(c)

Write short note on : (i) (ii) Atomic primitives Sharedmemorymultiprocessor

(iii) Communication latency. Attempt any two of the following : (a) Considerthe following loop nest: D O J=l ,N D O I-1 ,N S1 : A (I,J+l) : B(I,J) + (I,D 52:D End do (i) (I,J): A (J,J)1 2 \ End do Show how to compile the code for vectorization in the IJoop, assuming Fortran column-major storageorder. Showhowto compile the loop to perform the JJoop in vector mode, while using for the the Doall and Endall commands outerIJoop. 2xl0=20

(ii)

(b)

Explain any two memoryconsistenry model. );- *.'r

(iii) Alias analysis

EE-10E11

lContd...

Attempt any two of the following :

2xl0=20

(a)

processes, A file is to be shared among different eachof which has a uniquenumber.The file can be accessedsimultaneously by several processes, subjectto the following constraint. The sumof all uniquenumbers associated with all the processes currently accessing the file must be less than n. Write a monitor to coordinateaccess to the file.

(b)"/Define the difference pre-emptive between and non-preemptive scheduling. Whatarethevarious operatingsystemissuesfor multi processing. (c) Whatarevariousimplementation techniques on multi-cores?

EE-l0811

t 1100 l

You might also like