Skip to content

Clarify the correlation output layout - #333

Open
sylvesterkaczmarek wants to merge 1 commit into
ARM-software:mainfrom
sylvesterkaczmarek:docs/correlation-output-layout-120
Open

Clarify the correlation output layout#333
sylvesterkaczmarek wants to merge 1 commit into
ARM-software:mainfrom
sylvesterkaczmarek:docs/correlation-output-layout-120

Conversation

@sylvesterkaczmarek

Copy link
Copy Markdown

Summary

  • distinguish the srcALen + srcBLen - 1 correlation values from the larger zero-padded destination buffer
  • document whether the padding appears before or after the correlation values for unequal input lengths
  • retain the existing destination allocation and zero-initialisation requirements

Context

The correlation implementations intentionally use a destination length of 2 * max(srcALen, srcBLen) - 1. The existing group documentation describes that entire buffer as the correlation result, but does not show where the actual correlation values are located when the input lengths differ.

This change documents the existing layout without changing the implementation or API.

Fixes #120.

Validation

  • git diff --check
  • compiled the scalar arm_correlate_f32.c path with GCC 13.3.0 and -Wall -Wextra -Werror, using minimal test-only stubs for the external CMSIS Core headers
  • ran focused C assertions for both unequal-length orders and confirmed the documented leading and trailing zero padding

Documentation-only change. The repository test suite was not run.

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Correlation function arm_correlate_f32 has wrong output array length expectation for NEON

1 participant