Chapter 5, Problem 5.13HW

A.

Program Plan Intro

Cycles per element (CPE):

• The CPE denotes performance of program that helps in improving code.
• It helps to understand detailed level loop performance for an iterative program.
• It is suitable for programs that use a repetitive calculation.
• The processor’s activity sequencing is measured by a clock that provides signal of some frequency.

Explanation of Solution

Diagram for instruction sequence:

Explanation:

• The data dependencies between instructions are been depicted in diagram.
• The given instruction sequence is been decoded into operations.
• It creates a critical path of operations.
• The data flow between instructions is been shown in diagram.

B.

Program Plan Intro

Cycles per element (CPE):

• The CPE denotes performance of program that helps in improving code.
• It helps to understand detailed level loop performance for an iterative program.
• It is suitable for programs that use a repetitive calculation.
• The processor’s activity sequencing is measured by a clock that provides signal of some frequency.

Explanation of Solution

Lower bound on CPE:

• The lower bound on CPE is been determined by critical path.
• For data type “double”, it denotes the float add cell.
• The lower bound on CPE is 3.0 based on the architecture.

C.

Program Plan Intro

Cycles per element (CPE):

• The CPE denotes performance of program that helps in improving code.
• It helps to understand detailed level loop performance for an iterative program.
• It is suitable for programs that use a repetitive calculation.
• The processor’s activity sequencing is measured by a clock that provides signal of some frequency.

Explanation of Solution

Lower bound on CPE:

• The lower bound on CPE is been determined by critical path.
• For data type “integer”, it denotes the long add cell.
• The lower bound on CPE is 1.0 based on the architecture.

D.

Program Plan Intro

Cycles per element (CPE):

• The CPE denotes performance of program that helps in improving code.
• It helps to understand detailed level loop performance for an iterative program.
• It is suitable for programs that use a repetitive calculation.
• The processor’s activity sequencing is measured by a clock that provides signal of some frequency.

Explanation of Solution

Given C Code:

// Define method inner4

void inner4(vec_ptr u, vec_ptr v, data_t *dest)

{

// Declare variable

long i;

//Compute length of vector

long length = vec_length(u);

//Get first vector

data_t *udata = get_vec_start(u);

//Get second vector

data_t *vdata = get_vec_start(v);

//Initialize variable

data_t sum = (data_t) 0;

//Loop

for (i = 0; i < length; i++)

{

//Compute product and add

sum = sum + udata[i] * vdata[i];

}

//Store result

*dest = sum;

}

CPE value for floating-point versions:

• The inner product computed is been accumulated in temporary.
• The float add operation is only on key path.
• The multiplication operation takes 5 clock cycles.
• The overall operation takes 3 cycles to complete on average.
• Hence, CPE value for floating-point versions is 3.0.