Jump to content
  • Advertisement
Sign in to follow this  
ongamex92

ARM NEON

This topic is 2113 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Currently I'm extending a math library (bullet math form Bullet phyisics to be exact). Im adding Matrix4x4 Vector4 and other helper classes/methods(3d graphics friendly methods)

Tesing with SSE is easy. But there is a ARM NEON implementation in Bullet and I dont want to break the compibility.

 

Is there any way to test ARM NEON SIMD on Windows?

Share this post


Link to post
Share on other sites
Advertisement

I have regularly used qemu to emulate Cortex-A8 and later ARM devices under Linux (those devices are pretty much guaranteed to have vfs3 and NEON copros).  I know qemu is also available for Microsoft Windows™.  You could try that. or you could use the Google to find an alternative emulator of your choice.

Share this post


Link to post
Share on other sites

Right.  The whole purpose in using NEON is performance, not compatibility.  If your goal is compatibility, you are doing it wrong.  It is specifically meant to be processor-specific.  Convert the code to SSE* using macros to select which code path to follow.

 

 

L. Spiro

Share this post


Link to post
Share on other sites

Well In the most cases I'm only porting the code form btMatrix3x3 and btVector3 to 4 dimension version of the classes.

I beleve in the people who wrote bullet math. Of course I will pay attention in the situations where the performance is not good enough.

I'm not really into NEON (just can't find good papers). I'm thinking that the most optimal version using NEON should look like the SSE version of the code, is that correct?

Share this post


Link to post
Share on other sites

Well In the most cases I'm only porting the code form btMatrix3x3 and btVector3 to 4 dimension version of the classes.

I beleve in the people who wrote bullet math. Of course I will pay attention in the situations where the performance is not good enough.

I'm not really into NEON (just can't find good papers). I'm thinking that the most optimal version using NEON should look like the SSE version of the code, is that correct?

 

 

For basic ops like add/sub, then yes they should be similar. For dot/cross and other operations based around shuffles/unpack/movehl, they aren't going to look too similar at all....

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

GameDev.net is your game development community. Create an account for your GameDev Portfolio and participate in the largest developer community in the games industry.

Sign me up!