IntrodParallel

Question 1

Find dependencies in the following code sequence:

St. 1. Loop: 	L.D  		F0, 0(R1)
St. 2.		ADD.D 	F4, F0, F2
St. 3.		S.D 		F4,0(R1)
St. 4. 		DADDUI 	R1,R1,-8
St. 5. 		BNE 		R1,R2,LOOP

Accepted Answer

There are true dependence between St1 and St2, output dependence between St2 and St3 and true dependence between St4 and St5.

Answer

There are true dependence between St1 and St2, output dependence between St2 and St3 and output dependence between St4 and St5

Answer

There are three true dependences: between St1 and St2, between St2 and St3 and between St4 and St5

Answer

No correct answer

Question 2

Consider a loop branch whose behavior is taken nine times in a row, then not taken once. What is the prediction accuracy for this branch, assuming that we use a 2-bit prediction scheme?

Accepted Answer

No correct answer

Answer

80%

Answer

95%

Answer

83%

Question 3

Consider the following code: 
for ( l=1; l<=100; l++)
{ 
X[2*I+3]=x[2*l]*5.0;
} 
The GCD test gives you the following answer:

Accepted Answer

The dependence doesn't exist

Answer

The dependence exists

Answer

You can not use GCD test for above example

Answer

No correct answer

Question 4

What kind of data dependences can be eliminated using register renaming technique?

Accepted Answer

Output and antidependences

Answer

True and output dependences

Answer

All types of dependences

Answer

Antidependence and true dependences

Question 5

Suppose that you execute parallel program using 15 processors and that sequential fraction of your program is 20%. What is speedup which can be achieved?

Accepted Answer

~5

Answer

~15

Answer

~10

Answer

No correct answer

Question 6

What does it mean that network has blocking property?

Accepted Answer

It means that when one processor is accessing some memory block other processor cannot access different memory block.

Answer

It means that if one processor is accessing some memory block, other processor can not access the same memory block

Answer

It means that two different processors can not access the same memory location at the same time

Answer

No correct answer

Question 7

When we have a loop-carried dependence?

Accepted Answer

Whether data accesses in later interactions are dependent on the data values produces in earlier iterations

Answer

Whether data accesses in earlier iterations are dependent on the data values produces in later iterations.

Answer

When we have data dependence between different loop instructions.

Answer

No correct answer

Question 8

What is a difference between store and forward and cut-through routing mechanisms?

Accepted Answer

In the store and forward routing the message is sending between processors and each intermediate processor store it in the local memory until received the whole message, when in cut-through routing the message is divided on parts which are sending between processors without waiting for the whole message.

Answer

In the cut-through routing the message is sending between processors and each intermediate processor store it in the local memory until received the whole message, when in store and forward routing the message is divided on parts which are sending between processors without waiting for the whole message.

Answer

There are no any difference, it depends on hardware features.

Answer

Cut-through routing is used for distributed memory computers, when store and forward for shared memory computers

Question 9

When we use store and forward routing, total communication time can be expressed by the following equation (l – the number of communication lines used, m – the size of message)

Accepted Answer

tcomm = ts + (mtw + th)l

Answer

tcomm = ts + mtw + lth

Answer

tcomm = (ts + mtw + th)l

Answer

No correct answer

Question 10

for (I=1; I<=100; I=I+1)
{ X[2*I+3] = X[2*I-1]*5.0; }

The GCD test gives you the following answer:

Accepted Answer

The dependence exists

Answer

The dependence doesn’t exist.

Answer

You can not use GCD test for above example

Answer

No correct answer

Question 11

Consider the mesh network with p processors as in figure below (no wraparound connections). The communication time for one_to_all broadcast can be expressed by the following equation (m- message size):

Accepted Answer

2(ts+twm)*sqrt(p)-1

Answer

2(ts +twm)*sqrt(p)

Answer

2(ts+twm)*sqrt(p)/2

Answer

No correct answer

Question 12

RAW hazard is a consequence of:

Accepted Answer

Data dependence

Answer

Antidependence

Answer

Output dependence

Answer

Control dependence

Question 13

When we use the reservation station and Tomasulo’s scheme, instruction execution is divided onto:

Accepted Answer

No correct answer

Answer

Five steps like using standard pipeline

Answer

Four steps: Issue, Execute, Write Result, Commit

Answer

Four steps: Fetch, Decode, Execute, Write result

Question 14

Hardware-based speculation combines the following key ideas:

Accepted Answer

Dynamic branch prediction, dynamic scheduling and speculations to allow the execution of instructions before the control dependence are solved.

Answer

Static branch predictions, dynamic scheduling and speculation to allow the execution of instructions before the control dependence are solved.

Answer

Static branch predictions and register renaming

Answer

No correct answer

Question 15

When using hardware based speculation we need to add additional step during instruction execution called commit. The are following types of commits:

Accepted Answer

Normal commit, store and forward commit and incorrect branch prediction commit.

Answer

We have only one type of commit

Answer

Normal and incorrect branch prediction commits.

Answer

Normal and store commits

Question 16

What kind of partitioning should be used in parallel matrix multiplication algorithms.

Accepted Answer

No correct answer

Answer

domain partitioning

Answer

functional partitioning

Answer

domain or functional partitioning

Question 17

Duplication of data and / or computation takes place

Accepted Answer

We can not duplicate data or calculations, we have to avoid this.

Answer

It does not make sense, it increase the cost in the sense of the total processing time.

Answer

During the agglomeration and aims to reduce the cost of communication

Answer

no correct answer

Question 18

What kind of communication, and for which architecture is implemented by the following program fragment

Accepted Answer

Communication all_to_all for ring architecture

Answer

Communication all_to_all for mesh architecture

Answer

Communication one_to_all for hypercube architecture

Answer

No correct answer

Question 19

Using geometric means of normalized execution times…

Accepted Answer

we can easily predict the execution time.

Answer

we cannot predict the execution time.

Answer

depending on the workload, sometimes we can predict the execution time

Answer

no correct answer.

Question 20

Consider the following omega network. Is it blocking network?

Accepted Answer

Yes, it is blocking network, for example when processor 000 refers to memory 010 & processor 110 refers to memory 011

Answer

No, "omega" network is completely non-blocking network

Answer

Yes, it is blocking network, for example processor 000 refers to memory 010 & processor 011 refers to memory 010

Answer

No correct answer

IntrodParallel

✨ Parecidos

Introducción a Microsof

Ingles

Guia Taller Informàtica 2

Retroalimentación: Ontología

redes de computadoras complexivo II

Tema 2 Derecho

🔥 Quizzes populares

TEST DE MILIPILI

Tincho

TEST DE TURRA

¿Cuánto sábes de "The Amazing Digital Circus"?

Cinturon Blanco punta amarillo

Adivina la raza de PERRO!

Comentarios · 0

Find dependencies in the following code sequence: St. 1. Loop: L.D F0, 0(R1) St. 2. ADD.D F4, F0, F2 St. 3. S.D F4,0(R1) St. 4. DADDUI R1,R1,-8 St. 5. BNE R1,R2,LOOP

Consider a loop branch whose behavior is taken nine times in a row, then not taken once. What is the prediction accuracy for this branch, assuming that we use a 2-bit prediction scheme?

Consider the following code: for ( l=1; l<=100; l++) { X[2I+3]=x[2l]*5.0; } The GCD test gives you the following answer:

What kind of data dependences can be eliminated using register renaming technique?

Suppose that you execute parallel program using 15 processors and that sequential fraction of your program is 20%. What is speedup which can be achieved?

What does it mean that network has blocking property?

When we have a loop-carried dependence?

What is a difference between store and forward and cut-through routing mechanisms?

When we use store and forward routing, total communication time can be expressed by the following equation (l – the number of communication lines used, m – the size of message)

for (I=1; I<=100; I=I+1) { X[2I+3] = X[2I-1]*5.0; } The GCD test gives you the following answer:

Consider the mesh network with p processors as in figure below (no wraparound connections). The communication time for one_to_all broadcast can be expressed by the following equation (m- message size):

RAW hazard is a consequence of:

When we use the reservation station and Tomasulo’s scheme, instruction execution is divided onto:

Hardware-based speculation combines the following key ideas:

When using hardware based speculation we need to add additional step during instruction execution called commit. The are following types of commits:

What kind of partitioning should be used in parallel matrix multiplication algorithms.

Duplication of data and / or computation takes place

What kind of communication, and for which architecture is implemented by the following program fragment

Using geometric means of normalized execution times…

Consider the following omega network. Is it blocking network?

IntrodParallel

✨ Parecidos

Introducción a Microsof

Ingles

Guia Taller Informàtica 2

Retroalimentación: Ontología

redes de computadoras complexivo II

Tema 2 Derecho

🔥 Quizzes populares

TEST DE MILIPILI

Tincho

TEST DE TURRA

¿Cuánto sábes de "The Amazing Digital Circus"?

Cinturon Blanco punta amarillo

Adivina la raza de PERRO!

Comentarios · 0

Find dependencies in the following code sequence: St. 1. Loop: L.D F0, 0(R1) St. 2. ADD.D F4, F0, F2 St. 3. S.D F4,0(R1) St. 4. DADDUI R1,R1,-8 St. 5. BNE R1,R2,LOOP

Consider a loop branch whose behavior is taken nine times in a row, then not taken once. What is the prediction accuracy for this branch, assuming that we use a 2-bit prediction scheme?

Consider the following code: for ( l=1; l<=100; l++) { X[2*I+3]=x[2*l]*5.0; } The GCD test gives you the following answer:

What kind of data dependences can be eliminated using register renaming technique?

Suppose that you execute parallel program using 15 processors and that sequential fraction of your program is 20%. What is speedup which can be achieved?

What does it mean that network has blocking property?

When we have a loop-carried dependence?

What is a difference between store and forward and cut-through routing mechanisms?

When we use store and forward routing, total communication time can be expressed by the following equation (l – the number of communication lines used, m – the size of message)

for (I=1; I<=100; I=I+1) { X[2*I+3] = X[2*I-1]*5.0; } The GCD test gives you the following answer:

Consider the mesh network with p processors as in figure below (no wraparound connections). The communication time for one_to_all broadcast can be expressed by the following equation (m- message size):

RAW hazard is a consequence of:

When we use the reservation station and Tomasulo’s scheme, instruction execution is divided onto:

Hardware-based speculation combines the following key ideas:

When using hardware based speculation we need to add additional step during instruction execution called commit. The are following types of commits:

What kind of partitioning should be used in parallel matrix multiplication algorithms.

Duplication of data and / or computation takes place

What kind of communication, and for which architecture is implemented by the following program fragment

Using geometric means of normalized execution times…

Consider the following omega network. Is it blocking network?

Consider the following code: for ( l=1; l<=100; l++) { X[2I+3]=x[2l]*5.0; } The GCD test gives you the following answer:

for (I=1; I<=100; I=I+1) { X[2I+3] = X[2I-1]*5.0; } The GCD test gives you the following answer: