The x here is an exponent. What needs to happen to solve this is that both the 2 and the 9
must become exponents themselves so the one can gauge what the 2 is missing to become
a 9.
A logarithm is a base ten 'exponent'.
log 2 = .3010
log 9 = .9542
log9/log2 = 3.169
That is the value of x.
Note: the calculator will output a long series of numbers. One never rounds a log, as that would
loose the value!
* * *
Computing; how does a computer come to output and use a logarithm; is it a look-up, an algorithm...??
Here’s the fun part about logarithms in computing: computers don’t “think” in logs the way humans do, but they do compute them—very efficiently. And no, it’s not magic, and it’s not just a giant lookup table either. It’s a clever blend of math tricks, approximations, and hardware-level optimizations.
Let me walk you through it in a way that respects your love of physical plausibility and layered mechanisms.
🧮 How a Computer Actually Computes a Logarithm
Computers typically use three complementary strategies, depending on the precision needed and the hardware available:
🔹 1. Range Reduction (the essential first step)
Before doing anything fancy, the computer rewrites the number into a “nice” range where approximations behave well.
🔹 2. Polynomial or Rational Approximations (the real workhorse)
Once the number is in a stable range, the computer uses a fast approximation such as:
Taylor series (rarely used directly)
Chebyshev polynomials
Minimax polynomials (very common)
Padé approximants (rational functions)
These are pre‑computed formulas that give extremely accurate results with very few operations.
🔹 3. Lookup Tables (tiny ones)
Some implementations use very small lookup tables (maybe 16–128 entries) to:
speed up range reduction
improve initial approximations
reduce polynomial degree
But they never store every possible log value. That would be absurd—there are infinitely many real numbers.
Think of lookup tables as “hints,” not the main computation.
🧩 What About Hardware?
Some CPUs have dedicated instructions (like x87’s FYL2X) that compute logs using microcoded algorithms similar to the above.
Modern math libraries (glibc, Intel SVML, Apple Accelerate) use highly optimized versions of the same principles.
🧠 Why Not Just Use Taylor Series?
Because Taylor series:
converge slowly unless the input is extremely close to 1
require many terms for high precision
are numerically unstable at the edges of the interval
Minimax polynomials or rational approximations are far superior.
🔍 If you want, we can go deeper
We can explore:
how IEEE‑754 floating‑point format makes range reduction trivial
why minimax polynomials are optimal
how error bounds are guaranteed
how CPUs pipeline these operations
how log algorithms differ for base‑10 vs base‑2
Or we can map this to physical intuition—how you’d approximate logs with only a slide rule and a few clever tricks.
Where would you like to take it next?
Tell me how slide rules are used for logarithms...
Slide rules are one of those beautiful pieces of analog engineering where logarithms become geometry. They don’t compute logs the way a computer does—they embody them. Once you see the mechanism, it’s almost shocking how elegant it is.
Let me walk you through it in a way that respects your love of physical plausibility and layered mechanisms.
🧩 1. The Scales Are Logarithmic, Not Linear
On a slide rule, the numbers 1–10 are not spaced evenly.
The distance from 1 to 2 is larger than the distance from 8 to 9.
Why?
🔄 2. Sliding = Adding Logarithms
When you slide one logarithmic scale against another, you are physically adding distances, which means you are adding logarithms.
Adding logs corresponds to multiplying numbers.
So to multiply
𝑎⋅𝑏:
Put the “1” on the sliding scale over “a” on the fixed scale.
Move along the sliding scale to “b”.
The number on the fixed scale under “b” is the product.
Why does this work?
🔍 5. Why This Works So Smoothly
A slide rule is essentially a mechanical analog computer that:
stores the logarithm function in its scale geometry
performs addition/subtraction by sliding
converts back to multiplication/division by reading the scale
It’s the same conceptual trick computers use—range reduction + log identities—but implemented with wood, plastic, and human eyes.
🎨 A Physical Intuition You’ll Appreciate
If you took a strip of paper and marked positions according to
log(𝑥), then slid it over another identical strip, you would have built a primitive slide rule.
It’s that physically grounded.
If you want, we can go deeper
We can explore:
how the different scales (C, D, A, B, K, L, S, T) encode different functions
how slide rules handle trigonometry without electronics
why the log scale compresses large numbers so elegantly
how engineers used slide rules to design bridges, rockets, and circuits
Or we can compare slide-rule computation to floating‑point computation in modern CPUs—there’s a surprising structural parallel.
Where would you like to take it next?
No comments:
Post a Comment