Skip to content

Test sub-group shuffles of a mixed-width tuple #732

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
maleadt merged 1 commit into from
Oct 23, 2026
Merged

Test sub-group shuffles of a mixed-width tuple #732

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 21 additions & 0 deletions test/execution.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1345,3 +1345,24 @@ end
@oneapi count_until_wrap(out, flags, typemax(UInt64) - 3)
@test Array(out) == [typemax(UInt64) - 2]
end

@testset "sub-group shuffle of a mixed-width tuple" begin
# The shuffle packs the Bool, Int8 and Float16 into one word and unpacks them again. On
# Gen12LP, IGC could place the byte that receives the Int8 at byte 3 of a dword, which
# the hardware doesn't write, returning 0 for the Int8 (intel/intel-graphics-compiler#456,
# fixed in libigc_jll 2.41.5+2). The pointers avoid bounds checks, whose throw paths
# change the register allocation, and thus whether the bug shows.
function shuffle_kernel(out, a, lane)
i = get_local_id()
val = oneAPI.KernelInterface.shfl(unsafe_load(pointer(a), i), lane)
unsafe_store!(pointer(out), val, i)
return
end
T = Tuple{Bool, Int8, Float16, UInt16}
for n in (1, 8)
a = T[(isodd(i), -i % Int8, Float16(i), UInt16(1000i)) for i in 1:n]
out = oneArray{T}(undef, n)
@oneapi items=n shuffle_kernel(out, oneArray(a), 1)
@test Array(out) == fill(a[1], n)
end
end
Loading