Partial Solution Set, Leon §7.1 7.1.1 Find the three-digit decimal floating point representation of each of the following numbers: (a) f l(2312) = 0.231 · 104 (b) f l(32.56) = 0.326 · 102 (c) f l(0.01277) = 0.128 · 10−1 (d) f l(82, 431) = 0.824 · 105 7.1.2 Find the absolute error and the relative error when each of the real numbers in Exercise 1 is approximated by a three-digit decimal floating-point number. Recall that absolute error is f l(x) − x and the relative error is δ = f l(x)−x . x (a) absolute error: f l(2312) − 2312 = 0.231 · 10−3 − 2312 = −2, and relative error is −2 δ = 2312 ≈ −0.865 × 10−3 . (b) 0.326 · 10−2 − 32.56 = .04, and relative error is δ = .04 32.56 ≈ 0.1229 × 10−2 . (c) 0.128 · 101 − 0.01277 = 0.00003, and relative error is δ = 0.235 × 10−2 . (d) 0.824 · 10−5 − 82, 431 = −31, and relative error is δ = −31 82,431 0.00003 0.01277 ≈ 0.00235 = ≈ −0.376 × 10−3 . 7.1.3 Represent each of the following as five-digit base 2 floating-point numbers: (a) 21 = 24 + 0 · 23 + 22 + 0 · 21 + 1 · 20 = (0.10101)2 · 25 . (b) 3 8 = (0 · 24 + 0 · 23 + 0 · 22 + 21 + 20 ) · 2−3 = (0.11000)2 · 2−1 . (c) 9.872 = 23 +1.872 = 23 +0·22 +0·21 +20 +2−1 +lower-order terms ≈ (0.10011)2 ·24 . (d) −0.1 = = = ≈ −(0.0625 + 0.0375) −(2−4 + 2−5 + 0.063) −(2−4 + 2−5 + 0 · 2−6 + 0 · 2−7 + 2−8 ) + lower-order terms (0.11001)2 · 2−3 7.1.4 Do each of the following using four-digit decimal floating-point arithmetic and calculate the absolute and relative errors in your answers. (a) 10420 + 0.0018 (b) 10424 − 10416 (c) 0.12347 − 0.12342 (d) (3626.6)(22.656) Solution: (a) The exact sum is 10420.0018, which becomes 10420 in 4-digit arithmetic. The absolute error is ε = −0.0018, and the relative error is δ= ε ≈ 0.173 × 10−6 . 10420.0018 (b) The exact difference is 10424 − 10416 = 8. The computed difference is 10420 − −8 10420 = 0. The absolute error is ε = −8, and the relative error is δ = = −1. 8 (c) The exact difference is 0.12347 − 0.12342 = 0.00005. The computed difference is 0.1235 − 0.1234 = 0.0001. The absolute error is ε = .0001 − .00005 = .00005, and .00005 = 1. the relative error is .00005 (d) The exact product (to four decimal places) is (3626.6)(22.656) = 82164.2496. The computed product, after both pre- and post-multiplication rounding, is (3627)(22.66) = 82190. The absolute error is ε = 25.7504, and the relative error is 25.7504 ≈ 0.3134 × 10−3 82164.2496 . 2