Closed bjoo closed 4 years ago
And also in simd_common.hpp (BTW on this note, why are we setting this in two places? cuda_warp.hpp includes simd_common.hpp)
this was actually part of my PR #13
Can we (@bjoo) confirm that #13 was sufficient and that both this issue and #17 can be closed?
Having done a read of it, I think #13 will close #16 and obviate my #17. I thought I based my code on a master after #13 was merged tho… So maybe I was going through it and reverting it thinking ‘thiss stuff surely isn’t needed on the host’. So likely mea culpa there.
I am about to submit a pull that has permutes for AVX, AVX512, CUDA_WARP and HIP. I can ensure that these fixes are all in there.
Would that be a better way to go? @crtrott’s exercises are part of my testsuite.
Best, B
On May 4, 2020, at 12:49 PM, Dan Ibanez notifications@github.com wrote:
Can we (@bjoo) confirm that #13 was sufficient and that both this issue and #17 can be closed?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.
I can confirm that #13
Commit ID: f417a3872eca867de2d05eb0208010f0ecb071b5
fixes the CUDA Compilation problems. I will rebase onto that now.
Best, B
On May 4, 2020, at 12:49 PM, Dan Ibanez notifications@github.com wrote:
Can we (@bjoo) confirm that #13 was sufficient and that both this issue and #17 can be closed?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.
Ugh, I may have let a bug slip through when I added the hip-wavefront backend. I may have (in a bughunt) set SIMD_HOST_DEVICE to SIMD_DEVICE for the simd() constructors and opreator=() in cuda_warp.hpp
It is now (around line 168 in cuda_warp.hpp
These should be. SIMD_HOST_DEVICE , I just had a hard time compiling the exercise_10 complaining that
My apologies for having messed this up. I will have a PR soon for permutes which fixes this, but it may be worth hotfixing.