CUDA/HIP: Refactor the use of _sync warp primitives #31617
IMbackK
force pushed
from
c8ef12a7
to
10e72647
13 days ago
IMbackK
force pushed
from
10e72647
to
d2cef041
13 days ago
IMbackK
force pushed
from
d2cef041
to
83a8eb2f
12 days ago
IMbackK
force pushed
from
83a8eb2f
to
5a5203bc
7 days ago
IMbackK
force pushed
from
5e408cbb
to
af507251
7 days ago
IMbackK
force pushed
from
af507251
to
61cba3af
7 days ago
IMbackK
marked this pull request as ready for review 7 days ago
IMbackK
requested a review
7 days ago
IMbackK
requested a review
7 days ago
IMbackK
requested a review
7 days ago
am17an
approved these changes
on 2026-11-09
IMbackK
force pushed
from
498e3b73
to
3fac570d
6 days ago
CISC
approved these changes
on 2026-11-10
IMbackK
changed the title HIP: Bump required rocm version to 6.3 CUDA/HIP: Refactor the use of _sync warp primitives 5 days ago
IMbackK
force pushed
from
f8cd3b4d
to
18a8d9f1
4 days ago
CISC
approved these changes
on 2026-11-12
HIP: Bump required rocm version to 6.3
c075b4ac
CUDA/HIP: dehardcode topk warp size
28c9d17e
CUDA/HIP: Ensure all vendor headers are included with defines in def…
d84e452a
lane -> thread
d95a9210
add __device__ for cuda constexpr
a77cf1f9
Add wrappers for _sync intrinsics
9d3e4a6f
Remove redundant code in top-k
90f577eb
Fix WARP_SIZE introduced by rebase
5b394b41
IMbackK
force pushed
from
18a8d9f1
to
5b394b41
1 day ago
IMbackK
merged
278a89cb
into master 1 day ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub