Explore a thought-provoking conference talk that challenges conventional wisdom on parallel programming. Delve into Guy L. Steele Jr.'s argument that programmers should not be burdened with thinking about parallelism, and instead, programming languages should provide transparent parallel execution. Discover how this paradigm shift requires a new approach to language design, focusing on algorithms built on independence and build-and-conquer principles rather than linear problem decomposition. Gain insights from Steele's extensive experience as a Sun Fellow and his work in programming language research. Examine detailed examples, including IBM 1130 programming, instruction formats, and code analysis, to understand the evolution of programming concepts. Learn about key topics such as computation trees, atomic updates, and the importance of algebraic properties in algorithm design. Uncover the main points of Steele's talk and explore what makes code truly effective in the context of parallel programming.
Overview
Syllabus
Intro
The IBM 1130
IBM 1130 Console
IBM 1442 Card Read Punch
My Ugly Program
Normal Card Data Input
Program Load Card Input
IBM 1130 Instruction Format
Draw Separator Lines
Transcribe into Hexadecimal
Analyzing the Code
Preliminary Disassembly
Initialization Code
Main Program, Two Subroutines
Subroutine: Print Character
Level 4 Interrupt Address
Return (and Dismiss Interrupt)
Main Program: Counters
How the Counters Work
Print Newline and Shift to Red
Subroutine: Print Word
Counting Four Hex Digits
Clearing the Accumulator: Strange
Console Printer Control Codes
Computing Sine (HAKMEM #158)
Main Points of This Talk
What Makes Code Good?
Let's Add a Bunch of Numbers
What Does a Mathematician Say?
Sequential Computation Tree
Atomic Update Computation Tree (a)
Accumulation...
Splitting a String into Words (5)
Algebraic Properties Are Important!
Taught by
Strange Loop Conference
