You are on page 1of 3

KALINGA INSTITUTE OF INDUSTRIAL TECHNOLOGY (KIIT)

(Deemed to be University)
SCHOOL OF COMPUTER ENGINEERING

AUTUMN SEMESTER 2022

ACTIVITY / ASSIGNMENT - 2

Date: 12/11/2022

Subject: HPC

You all have to write the answers then scasn it to upload at KIIT MOODLE class
room only in pdf format.

Last date of submission is 20th November 2022.

ACTIVITY - 2

List of Activities

1. Consider the loop on MIPS without any scheduling will execute as follows:-
Clock Cycles Issued
LOOP: LOAD F2, 0(R7) 1
STALL 2
ADD F0, F2, F4 3
STALL 4
STALL 5
SUB F6, F0, F2 6
STALL 7
STALL 8
STORE F0, 0(R7) 9
STALL 10
ADD R7, R7, #+8 11
STALL 12
BEQ F1, F5, LOOP 13
Use the benefits of instruction scheduling and loop unrolling to reduce the number of clock cycle
required to complete the execution of the above given code. [Refer to the latency table given in the test
book]

2. Assume we have 2 multipliers, 2 adders, 1 divide & 1 integer unit. The following set of MIPS
instruction is going to be executed in a pipelined system.

Consider the following instructions:-


MUL R3, R5, R4
ADD R7, R3, R5
SUB R6, R7, R5
LOAD R6, 48(F5)
LOAD R9, 48(F5)
MUL R10, R9, R3
DIV R7, R3, R2
ADD R2, R7, R3
Consider the following execution status for the above instruction set:-
 1st MUL instruction is completed in its write result phase.
 All other remaining instruction is in its execution phase.
Show the status of instruction, reservation station, & register result status using dynamic scheduling
with the Scoreboard approach.
3. Assume we have 3 multipliers, 4 adders & 3 loader units. The following set of MIPS instruction is
going to be executed in a pipelined system.

Consider the following instructions:-


LOAD R1, 23(R2)
LOAD R5, 48(F1)
MUL R3, R5, R4
LOAD R3, 48(F5)
ADD R7, R3, R5
SUB R6, R2, R5
DIV R7, R3, R2
ADD R2, R6, R3
Consider the following execution status for the above instruction set:-
 1st two LOAD instructions are completed their write result phase.
 MUL & DIV instructions are ready to go for their write result phase.
 All other remaining instruction is in its execution phase.
Show the status of instruction, reservation station, & register result status using dynamic scheduling
with tomasulo approach.

4. Assume a fully associative write back cache with many cache entries that starts empty. Below is a sequence of 7
memory operations (the address is in square brackets).
WriteMem [100]
ReadMem [300 ]
WriteMem [200]
ReadMem [100 ]
ReadMem [400 ]
WriteMem [200]
WriteMem [300]
What are the numbers of hits and misses when using no-write allocate Vs write allocate?
5. What is cache miss penalty time? Explain how to improve cache memory performance by reducing cache miss
penalty.
6. Draw and explain the states of directory-based cache protocol which implements both CPU request and BUS
request for each cache block access.

7. What is a distributed shared memory architecture? Explain cache coherence in symmetric shared-
memory multiprocessors with an example.

You might also like