1. Consider a floating-point format that uses 8 bits of memory to store numbers: • 1 bit for the sign; 3 bits for the exponent; • 4 bits for the significand including the leading 1, i.e., significand is of the form (1.d₁d2d3)2. (a) Compute fl(23.375), the floating point approximation of 23.375 using rounding. (b) To how many significant digits is fl(23.375) accurate? (c) When would you get numerical underflow and overflow errors with this format? Note that we are not dealing with a shifted exponent.

1. Consider a floating-point format that uses 8 bits of memory to store numbers: • 1 bit for the sign; 3 bits for the exponent; • 4 bits for the significand including the leading 1, i.e., significand is of the form (1.d₁d2d3)2. (a) Compute fl(23.375), the floating point approximation of 23.375 using rounding. (b) To how many significant digits is fl(23.375) accurate? (c) When would you get numerical underflow and overflow errors with this format? Note that we are not dealing with a shifted exponent.

Database System Concepts

7th Edition

ISBN:9780078022159

Author:Abraham Silberschatz Professor, Henry F. Korth, S. Sudarshan

Publisher:Abraham Silberschatz Professor, Henry F. Korth, S. Sudarshan

Chapter1: Introduction

Section: Chapter Questions

Problem 1PE

See similar textbooks

Similar questions

Consider a 16-bit binary floating point number representation system: + | SE E E E E E m m т m m m m m m The first bit of the exponent is dedicated to its sign. Assume that the mantissa must start with a '1'. Use this system to answer the following question What is the smallest (magnitude) number that can be represented with this system?
Consider a floating point format that uses 4 digits in the significand including the leading 1. In other words, the significand is given by (1.d₁d2d3)2. (a) Assuming you have enough bits in the exponent, compute fl(25.75) for this floating-point format. (b) How many significant digits is fl(25.75) accurate to? (c) How many bits are needed in the exponent such that computing fl(25.75) will not result in nu- merical overflow?
Consider a 16-bit binary floating point number representation system: + SE E EEEE m m m m The first bit of the exponent is dedicated to its sign. Assume that the mantissa must start with a '1'. Use this system to answer the following questions: 1. What is the smallest (magnitude) number that can be represented with this system? 2. Calculate the relative round-off error introduced when this system tries to represent e = 2.718281828459045235360287 .. using rounding.
5. Consider the binary numbers s= (0.101)2, r= (10.101) 2. (a) For fl(r) the floating point value using the Rounding to Nearest rule, write out bits 47-52 of the mantissa. (b) Converts and a to base 10 (without rounding). Your answers should be expressed as fractions.
We are using the simple model for floating point representation which is a 14-bit format, 5 bits for the exponent with a bias of 15, a normalized mantissa of 8 bits, and a single sign bit. Show step by step how the computer would represent the following numbers using this floating- point format | a. -6.47510 i. Convert it into a binary number ii. Normalize this binary number iii. Fill up the fields
Consider the half-precision floating-point format, which has five bits for the exponent and ten bits for the significand. Find the decimal (i.e., base 10) number represented by the following bits in the half-precision floating- point format: (a) 0100110001110000. (b) 1000001111 1111. (c) 0111110000000000.
Given the following 32-bit number in IEEE754 floating-point number format,B = Ox4OBC0000 b) Determine the hexadecimal representation in IEEE754 double-precision floating-point number format of the following two cases,i. \sqrt{-1} and, ii. Negative infinity Remark: need clearly show all bit fields in hexadecimal format and clear step
Scenario: In a biased N-bit binary number system with bias B, positive and negative numbers are represented as their value plus the bias B. For example, for 5-bit numbers with a bias of 15, the number 0 is represented as 01111, 1 as 10000, and so forth. Biased number systems are sometimes used in floating point mathematics. Consider a biased 8-bit binary number system with a bias of 12710 Question: What is the representation and value of the most negative number?
Consider an 8-bit floating point format. In this representation 4 bits are reserved for the mantissa and 3 bits are reserved for the exponent and 1 bit is reserved for the sign of the number. The exponent of the number is stored with a bias = 3. Express the decimal number 4.75 in this 8 bit floating point format. Write down the next bigger representable number in decimal base. B If you would be conducting the following operation using this 8 bit floating point format what would the result of the following calculation be ? 4.75 + 0.05 - 4.75 Express the decimal number 4.75 in this 8 bit floating point format. [10] b) Write down the next bigger representable number in decimal base. [7.5] c) If you would be conducting the following operation using this 8 bit floating point format what would the result of the following calculation be ? 4.75 + 0.05 - 4.75
4. The bfloat16 "brain floating point" format is a 16 bit format used in Google's machine learning and AI software. It is a binary floating point format which is very similar to the single precision IEEE-754 format: 1 bit is allocated for the sign, 8 bits for the exponent with a bias of 127, but only 7 bits are allocated for the fraction (the exponent is always chosen so that the first digit of the mantissa is 1, and then only the fraction is stored in memory). (a) What is the approximate decimal precision of a brain floating point? (b) If the bits are stored in the order: sign, exponent, fraction, and 0 corresponds to a positive sign, then calculate the decimal representation of the number stored as 1 00000110 0100010 (c) Given that the largest exponent actually used for numbers is 11111110, what is the largest number that can be expressed as a bfloat 16? (d) Supposing that numbers are truncated (rounded down), what is the maximum absolute rounding error in bfloat16?
Given the following 32-bit number in IEEE754 floating-point number format, B = Ox4OBC0000 c) Consider the following subtraction, Z = B - 9.625 Convert the result Z to 32-bit IEEE754 floating number format. Express it in hexadecimal.
E10] Consider a floating point representation format where 8 bits are reserved for the mantissa. How many real numbers between 1 and 2 can be represented in this format ? Answer: