kernels: add 8i_x2_saturated_sum_8i #29
Conversation
…2, SSE2 and NEON along the generic kernel.
| @@ -0,0 +1,277 @@ | |||
| /* -*- c++ -*- */ | |||
| /* | |||
| * Copyright 2015 Free Software Foundation, Inc. | |||
|
|
||
| #endif /* LV_HAVE_AVX2 for unaligned */ | ||
|
|
||
| #endif /* INCLUDED_VOLK_8s_SATURATED_SUM_8s_UNALIGNED8_H */ |
There was a problem hiding this comment.
This #endif needs to be removed.
|
|
||
|
|
||
| #ifndef INCLUDED_volk_8i_x2_saturated_sum_8i_a_H | ||
| #define INCLUDED_volk_8i_x2_saturated_sum_8i_a_H |
There was a problem hiding this comment.
Uh-oh- multiple include guard defines! That means that from line 171-end, things should never be compiled, since line 72 defines that macro!
There was a problem hiding this comment.
These 2 lines as Marcus notes need to be removed.
|
pinging @evilynux here: You still up for fixing this code in a spare minute? |
…1-add-avx2-to-fm_detect-conv_k7_r2 to next * commit '028652df04a978772b9aa61ac2f5b2ed351dc98c': added avx2 to volk 8u x4 conv k7 r2 kernel Added AVX2 to fm_detect, still working on adding to conv_k7_r2
|
@evilynux are you still interested in getting the changes merged? |
|
@noc0lour My apologies, I'm completely overwhelmed and have been unable to find the required time to fix this. I don't see this changing any time soon... :-/ |
|
No problem, let's see if I can look at this anytime soon. Thanks for your work until now! |
|
@evilynux otherwise can you hit [email protected] an email to do a quick CLA so I can pick your work up in a spare minute? |
There was a problem hiding this comment.
I like this PR's addition, but the macro protection needs to be cleaned up! And, this PR needs to be rebased to the current GIT master. And I believe we need a CLA (license assignment) in place for large of an addition of code, per a prior comment about contacting Ben Hilburn who handles CLAs.
|
|
||
| #endif /* LV_HAVE_AVX2 for unaligned */ | ||
|
|
||
| #endif /* INCLUDED_VOLK_8s_SATURATED_SUM_8s_UNALIGNED8_H */ |
There was a problem hiding this comment.
This #endif needs to be removed.
|
|
||
|
|
||
| #ifndef INCLUDED_volk_8i_x2_saturated_sum_8i_a_H | ||
| #define INCLUDED_volk_8i_x2_saturated_sum_8i_a_H |
There was a problem hiding this comment.
These 2 lines as Marcus notes need to be removed.
|
|
||
| #endif /* LV_HAVE_NEON for aligned */ | ||
|
|
||
| #endif /* INCLUDED_VOLK_8s_SATURATED_SUM_8s_ALIGNED8_H */ |
There was a problem hiding this comment.
this #endif should be INCLUDED_volk_8i_x2_saturated_sum_8i_u_H ... same as that starting the header. case matters!
|
@evilynux Is there any update on the CLA? |
|
@evilynux I assume there's still no update on your CLA? Anyways, we adopted a new system: DCO. Thus, you only need to sign off your commits. I got this approach to sign previous commits from stackoverflow I updated the option |
Includes implementations for AVX2, SSE2 and NEON along the generic kernel.
Beware, the NEON kernel is untested (don't have the HW handy).