[GH-ISSUE #1009] Amdahl's Law Calculation Error (Page 11, Chapter 1) #1664

New Issue

GiteaMirror · 2026-04-11T08:01:02-05:00

GiteaMirror commented

2026-04-11 08:01:02 -05:00

Originally created by @Jahnic-kb on GitHub (Oct 31, 2025).
Original GitHub issue: https://github.com/harvard-edge/cs249r_book/issues/1009

Summary

The example calculation provided for Amdahl's Law on Page 11 of Chapter 1 seems to incorrectly match the required sequential execution time with the resulting maximum speedup. The text states:

"if data movement takes 80% of execution time, even infinite compute capacity provides only 5x speedup."

This statement is inconsistent. If 80% of a task cannot be sped up (it's sequential), the maximum speedup you can ever achieve is only 1.25x. This error likely stems from confusing the sequential fraction (data movement) with the parallelizable fraction, as a 5x speedup requires 80% to be parallelizable.

Detailed Explanation

Amdahl's Law is a principle that states the overall speedup of a system is limited by the fraction of the task that must be performed sequentially (non-parallelized).

What the book claims (80% sequential / 20% parallelized):
If the sequential portion is 80%, the maximum possible speedup is 1 / (1 - 0.2) = 1.25 times. The claimed speedup of 5x is therefore incorrect for an 80% sequential task.

Suggested Fix

To resolve this inconsistency, the sequential fraction should be changed to match the resulting speedup:

"if data movement takes 20% of execution time, even infinite compute capacity provides only 5x speedup."

OR alternatively the speed up should be changed:

"if data movement takes 80% of execution time, even infinite compute capacity provides only 1.25x speedup."

Originally created by @Jahnic-kb on GitHub (Oct 31, 2025). Original GitHub issue: https://github.com/harvard-edge/cs249r_book/issues/1009 **Summary** The example calculation provided for Amdahl's Law on Page 11 of Chapter 1 seems to incorrectly match the required sequential execution time with the resulting maximum speedup. The text states: "if data movement takes 80% of execution time, even infinite compute capacity provides only 5x speedup." This statement is inconsistent. If 80% of a task cannot be sped up (it's sequential), the maximum speedup you can ever achieve is only 1.25x. This error likely stems from confusing the sequential fraction (data movement) with the parallelizable fraction, as a 5x speedup requires 80% to be parallelizable. **Detailed Explanation** Amdahl's Law is a principle that states the overall speedup of a system is limited by the fraction of the task that must be performed sequentially (non-parallelized). What the book claims (80% sequential / 20% parallelized): If the sequential portion is 80%, the maximum possible speedup is $1 / (1 - 0.2) = 1.25$ times. The claimed speedup of 5x is therefore incorrect for an 80% sequential task. **Suggested Fix** To resolve this inconsistency, the sequential fraction should be changed to match the resulting speedup: _"if data movement takes 20% of execution time, even infinite compute capacity provides only 5x speedup."_ OR alternatively the speed up should be changed: _"if data movement takes 80% of execution time, even infinite compute capacity provides only 1.25x speedup."_

GiteaMirror added the area: book label 2026-04-11 08:01:02 -05:00

GiteaMirror closed this issue

2026-04-11 08:01:03 -05:00

GiteaMirror commented

2026-04-11 08:01:04 -05:00

@profvjreddi commented on GitHub (Oct 31, 2025):

I fixed it. If you don't mind, could you share the GitHub email address you used with GitHub so I can tag you as a contributor? I like to recognize everyone who's contributed: https://www.mlsysbook.ai/contents/frontmatter/acknowledgements/acknowledgements.html#sec-acknowledgements-contributors-fe00

@profvjreddi commented on GitHub (Oct 31, 2025): I fixed it. If you don't mind, could you share the GitHub email address you used with GitHub so I can tag you as a contributor? I like to recognize everyone who's contributed: https://www.mlsysbook.ai/contents/frontmatter/acknowledgements/acknowledgements.html#sec-acknowledgements-contributors-fe00

GiteaMirror commented

2026-04-11 08:01:04 -05:00

@profvjreddi commented on GitHub (Oct 31, 2025):

Hi @Jahnic-kb,

Great catch, and thank you for taking the time to explain it so clearly.

You are right that I mixed up the sequential and parallelizable fractions. Your breakdown makes this obvious.

I’ve fixed it and will push a release later today or over the weekend, ensuring that you are credited for the correction. These kinds of sharp observations enhance the book for everyone, so thank you again for catching it and explaining it so well.

Could you drop me your email that is associated with your GitHub, so that I can credit you 😄 contributors

Best,
VJ

@profvjreddi commented on GitHub (Oct 31, 2025): Hi @Jahnic-kb, Great catch, and thank you for taking the time to explain it so clearly. You are right that I mixed up the sequential and parallelizable fractions. Your breakdown makes this obvious. I’ve fixed it and will push a release later today or over the weekend, ensuring that you are credited for the correction. These kinds of sharp observations enhance the book for everyone, so thank you again for catching it and explaining it so well. Could you drop me your email that is associated with your GitHub, so that I can credit you 😄 [contributors](https://www.mlsysbook.ai/contents/frontmatter/acknowledgements/acknowledgements.html#sec-acknowledgements-contributors-fe00) Best, VJ

GiteaMirror commented

2026-04-11 08:01:04 -05:00

@Jahnic-kb commented on GitHub (Nov 1, 2025):

Hi @profvjreddi,
That's great news!

Glad I could help. My email associated with this account is jahnic.beck@kienbaum.de.

I appreciate the quick and positive response. Thanks! 🙂

@Jahnic-kb commented on GitHub (Nov 1, 2025): Hi @profvjreddi, That's great news! Glad I could help. My email associated with this account is jahnic.beck@kienbaum.de. I appreciate the quick and positive response. Thanks! 🙂

GiteaMirror commented

2026-04-11 08:01:05 -05:00

@profvjreddi commented on GitHub (Nov 1, 2025):

@all-contributors, please add @Jahnic-kb for helping with the Amdahl's law error fix

@profvjreddi commented on GitHub (Nov 1, 2025): @all-contributors, please add @Jahnic-kb for helping with the Amdahl's law error fix

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: github-starred/cs249r_book#1664