Skip to content

Conversation

@Anipik
Copy link
Contributor

@Anipik Anipik commented Oct 10, 2018

Fixes #1245
For inputs that are not naturally aligned (the alignment is not a multiple of 4), it does exclusively unaligned loads
For all other inputs, it will do at most two unaligned loads (one each for any leading/trailing unaligned elements) and all other loads will be aligned.

cc @tannergooding @eerhardt @danmosemsft @TomFinley

@Anipik
Copy link
Contributor Author

Anipik commented Oct 10, 2018

Before
Method Avx Sse Native
MatMulX 93.80 us 105.50 us 101.2 us
MatMulTranX 94.60 us 112.50 us 106.0 us
After
Method Avx Sse Native
MatMulX 93.40 us 100.70 us 105.3 us
MatMulTranX 94.60 us 111.80 us 115.5 us

// update with changes

@Anipik
Copy link
Contributor Author

Anipik commented Oct 11, 2018

@tannergooding are the performance tests fine ?

@Anipik
Copy link
Contributor Author

Anipik commented Oct 12, 2018

@eerhardt @tannergooding can you take another look ?

@Anipik
Copy link
Contributor Author

Anipik commented Oct 16, 2018

@tannergooding @eerhardt cany more feedback ?

Copy link
Member

@eerhardt eerhardt left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

:shipit:

Copy link
Member

@tannergooding tannergooding left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Overall, LGTM

@Anipik Anipik merged commit 7d9660a into dotnet:master Oct 16, 2018
@Anipik Anipik deleted the matMul branch October 16, 2018 19:01
@ghost ghost locked as resolved and limited conversation to collaborators Mar 28, 2022
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants