Homework 8 Solution

$35.00 $29.05

You'll get a: . zip file solution, download link after Payment

Description

  1. Short answer questions

    1. (3 points) What are three specific benefits of using virtual memory?

    1. (2 points) Do you think the virtual and physical page number must be the same size? what about the page size?

  1. (2 points) Do different processes have (typically) access to each other’s memory (Yes/ No). What is the name of the technique that is responsible for this?

  1. (3 points) What does TLB stand for and what does it do?

  1. (3 points) What should happen to the TLB when a new value is loaded into the page table address register?(only need to consider the valid bits of TLB)

  1. Memory Access

Consider a 32-bit physical memory space and a 32 KiB 4-way associative cache with LRU replacement. You are told that the hit rate of loop two is 9/16.

1

2

3

4

5

6

7

int ARRAY_SIZE = 32 * 1024;

int arr [ARRAY_SIZE]; // *arr is aligned to a cache block

/* loop one */

for (int i = 0; i < ARRAY_SIZE; i += 4) arr[i] = i; /* loop two */

for (int i = ARRAY_SIZE – 4; i >= 0; i -= 4) arr[i+1] = arr[i]-1;

(a)

(4 points)

Fill the number of bits in the tag, index and offset fields in the figure

below.

Tag

Index

Offset

(b)

It’s Not My Fault

1

2

3

4

5

6

7

8

9

Consider the following OpenMP snippet:

int values [ size ];

#pragma omp parallel

{

int i = omp_get_thread_num();

int n = omp_get_num_threads();

for(int j = i * ( size / n); j < (i + 1) * ( size / n); j++){ values [ j ] = j;

}

}

All cores share the same physical memory and we are running 2 threads. This is the sole process running. Each page is 1 KiB, and you have 2 pages of physical memory. The code snippet above starts at virtual address 0x400, and the values array starts at 0x800. The size, n, i, and j variables are all stored in registers. The functions omp_get_thread_num and omp_get_num_threads are stored in virtual addresses 0x440 to 0x480. The replacement policy for the page table is Least Recently Used. At the start of the pragma omp parallel call the page table looks as follows:

Virtual

Valid

Dirty

Physical

Page

Page

Number

Number

0

0

0

0

1

1

0

0

2

1

1

1

3

0

0

1

  1. (3 points) How many page faults will occur if size=0x080?

    • 0

(5 points) What is the minimum number of page faults that will occur if size=0x200? And what is the maximum?

  1. (3 points) How could you reduce the maximum page faults for part (b) (updated to (d))? Choose the valid options amongst the following ( – 1 pt for every incorrect answer):

    1. increase virtual address space

    1. decrease number of threads

    1. increase number of threads

    1. use SIMD instructions

    1. change page table replacement policy to Random

    1. increase physical address space

    1. add a 4-entry TLB

  1. Meltdown

    1. (4 points) Shortly explain (on a high level – we are mainly looking for the correct keywords) how meltdown works.

DMA, Dependency and RAID

  1. (2 points) Please briefly explain what DMA is in 2 sentences.

  1. (2 points) During the DMA procedure, which of them may raise an interrupt when handling the incoming data? circle them out.

      1. CPU

      1. ALU

  1. (2 points) Which of the following will decrease the availability? Circle them.

      1. increase MTTF

      1. decrease MTBF

      1. increase MTBF

      1. decrease MTTR

  1. (4 points) True or False Questions, please circle the correct answer.

T / F : RAID 1 will result in slower read but it has very high availability T / F: RAID 1 is the most expensive architecture among RAID 1,3,4,5 T / F : RAID 3 uses hamming ECC to check and correct the data

T / F: RAID 4 is still slow on small writes

lu66049ced2 tmp d827708e7877339 lu66049ced2 tmp 9ca0117eaf062321 lu66049ced2 tmp 6dd8fdcea8404abe lu66049ced2 tmp d897ecb0e64835f4

Page 4