Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Stream functions should be optimized for SSE/ARM-NEON #25

Closed
walbourn opened this issue Jun 9, 2016 · 4 comments
Closed

Stream functions should be optimized for SSE/ARM-NEON #25

walbourn opened this issue Jun 9, 2016 · 4 comments
Assignees
@walbourn
Labels
optimization

Comments

@walbourn
Copy link
Member

walbourn commented Jun 9, 2016

The existing Stream implementations in DirectXMath for SSE/SSE2 are basic loops that use the non-stream version.

XMVector2TransformStream
XMVector2TransformCoordStream
XMVector2TransformNormalStream
XMVector3TransformStream
XMVector3TransformCoordStream
XMVector3TransformNormalStream
XMVector3ProjectStream
XMVector3UnprojectStream
XMVector4TransformStream
XMPlaneTransformStream
@walbourn
Copy link
Member Author

walbourn commented Jun 9, 2016

Fixed for DirectXMath 3.04

@walbourn walbourn closed this as completed Jun 9, 2016
@walbourn
Copy link
Member Author

walbourn commented Jun 9, 2016

For SSE, these make use of _mm_stream_ps where possible to use 'non-temporal' stores to avoid polluting the cache. You can disable this behavior by using:

#define _XM_NO_MOVNT_

@walbourn
Copy link
Member Author

walbourn commented Jun 9, 2016

The original implementation failed to include a call to _mm_sfence at the end of the Stream methods. This was fixed in DirectXMath 3.08.

@walbourn walbourn self-assigned this Jun 9, 2016
@walbourn
Copy link
Member Author

walbourn commented Jun 9, 2016

Additional optimizations made in DirectXMath 3.05

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

@walbourn walbourn

Labels
optimization
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@walbourn