Aor jet#690
Conversation
|
Please don't merge it just yet: I want to try adding little-endian optimization |
|
I included Looks like the speed up comes from removing |
joemfb
left a comment
There was a problem hiding this comment.
Those #define's are vestigial, and this codebase is little-endian only. So you can just use your little-endian version and drop the other one.
joemfb
left a comment
There was a problem hiding this comment.
Your naming conventions don't matching u3's suffix patterns. The name suffix should match the type of the variable, regardless of whether it's a pointer: c3_y -> foo_y, c3_w foo_w, &c.
++aorjet. 10x speed up with this input: