Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

fix: Treat equivalent pooling qinfo as non-requantized#1293

Open
morgolock wants to merge 1 commit into
main from
pr/fix_pool_layer_perf
Open

fix: Treat equivalent pooling qinfo as non-requantized #1293
morgolock wants to merge 1 commit into
main from
pr/fix_pool_layer_perf

Conversation

@morgolock

@morgolock morgolock commented Jun 11, 2026

Copy link
Copy Markdown
Contributor

CpuPool2dAssemblyWrapperKernel::configure() decided whether to use the requantized pooling path by comparing the raw QuantizationInfo objects, while validate() compared their normalized uniform() values.

That made numerically equivalent qinfo such as QuantizationInfo(1.0f) and QuantizationInfo(1.0f, 0) take the slower differing-qinfo pooling path even though they represent the same scale and zero point.

Use normalized qinfo equality for pooling dispatch as well, and add NEON pooling coverage for the equivalent-qinfo NHWC QASYMM8 padded MAX case.

Change-Id: I54ab2cc9b2fae810b9d2767676097858eb8621c8

CpuPool2dAssemblyWrapperKernel::configure() decided whether to use the requantized pooling path by comparing the raw QuantizationInfo objects, while validate() compared their normalized uniform() values.
That made numerically equivalent qinfo such as QuantizationInfo(1.0f) and QuantizationInfo(1.0f, 0) take the slower differing-qinfo pooling path even though they represent the same scale and zero point.
Use normalized qinfo equality for pooling dispatch as well, and add NEON pooling coverage for the equivalent-qinfo NHWC QASYMM8 padded MAX case.
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Change-Id: I54ab2cc9b2fae810b9d2767676097858eb8621c8
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@gunes-arm gunes-arm Awaiting requested review from gunes-arm

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

AltStyle によって変換されたページ (->オリジナル) /