Skip to content

docs(matmul): recompute FLOP table cells with decimal n#119

Closed
boskodev790 wants to merge 1 commit into
zeokin:mainfrom
boskodev790:fix/matmul-readme-flop-decimal-102
Closed

docs(matmul): recompute FLOP table cells with decimal n#119
boskodev790 wants to merge 1 commit into
zeokin:mainfrom
boskodev790:fix/matmul-readme-flop-decimal-102

Conversation

@boskodev790

Copy link
Copy Markdown

Lane

fix / bug — docs-only, no code change. CPU-safe.

Problem

matmul/README.md's "The core problem: it doesn't fit" table mixes unit bases. The memory columns and the 128k compute cell use decimal-k n (12000, 128000), but the 1k/16k/32k cells of the compute (2n³) column were computed with binary-k n (1024/16384/32768) — overstating 2n³ by ~7% and disagreeing with the decimal memory cells in their own rows.

Fix

Recompute the three cells with the same decimal n:

n was now (2·decimal-n³)
1k (1000) 2.1 GFLOP 2.0 GFLOP
16k (16000) 8.8 TFLOP 8.2 TFLOP
32k (32000) 70 TFLOP 66 TFLOP

128k (4.2 PFLOP) was already decimal and is unchanged. Docs-only; touches only matmul/README.md.

Fixes #102

…statement)

fix / bug lane — docs-only, no code change.

The `matmul/README.md` "it doesn't fit" table mixes unit bases: the memory
columns and the 128k compute cell use decimal-k n (12000, 128000), but the
1k/16k/32k compute cells were computed with binary-k n (1024/16384/32768),
overstating 2n³ by ~7% and disagreeing with the decimal memory cells in their
own rows. Recomputed those three cells with decimal n to match:

- 1k  (n=1000):  2·1000³  = 2.0 GFLOP  (was 2.1)
- 16k (n=16000): 2·16000³ = 8.2 TFLOP  (was 8.8)
- 32k (n=32000): 2·32000³ = 66 TFLOP   (was 70)

128k (4.2 PFLOP) was already decimal and is unchanged. Docs-only; touches only
matmul/README.md.

Fixes zeokin#102
@github-actions github-actions Bot added area:matmul Exact engine and tile schedules (matmul/) area:docs README, CONTRIBUTING, BENCHMARKS and other docs status:needs-pr-kind PR must declare whether it is a fix/bug PR or a feat/strategy PR status:needs-review Awaiting maintainer review copycat Blocked: substantially reproduces an earlier PR (see .github/COPYCATS.md) labels Jul 8, 2026
@github-actions

github-actions Bot commented Jul 8, 2026

Copy link
Copy Markdown

Closed as a copycat submission: 100% of added lines already appear in an earlier PR (matches an earlier PR by luciferlive112116). See .github/COPYCATS.md.

@github-actions

github-actions Bot commented Jul 8, 2026

Copy link
Copy Markdown

copycat

@github-actions github-actions Bot closed this Jul 8, 2026
@github-actions

github-actions Bot commented Jul 8, 2026

Copy link
Copy Markdown

Closed as a copycat submission: matches an earlier PR by luciferlive112116: 100% of added lines already appear in an earlier PR

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area:docs README, CONTRIBUTING, BENCHMARKS and other docs area:matmul Exact engine and tile schedules (matmul/) copycat Blocked: substantially reproduces an earlier PR (see .github/COPYCATS.md) status:needs-pr-kind PR must declare whether it is a fix/bug PR or a feat/strategy PR status:needs-review Awaiting maintainer review

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[docs] matmul README FLOP table mixes decimal and binary n, overstating 2n³ by ~7%

1 participant