team-level std algos: part 2 #6205
Conversation
fnrizzi
force-pushed
the
team_level_nonmod_seq_p2
branch
from
3af43a3
to
9013aa6
Compare
fnrizzi
changed the title
fnrizzi
changed the title
fnrizzi
force-pushed
the
team_level_nonmod_seq_p2
branch
from
29b7043
to
7ec06a2
Compare
|
CI failure unrelated |
This comment was marked as outdated.
This comment was marked as outdated.
|
ci acting up but was fine before |
|
retest this please |
|
retest this please |
|
all CI failures are unrelated |
|
only hvhpc fails with compiler error opc |
| const auto s_count = KE::distance(s_first, s_last); | ||
| namespace KE = ::Kokkos::Experimental; | ||
| const auto num_elements = KE::distance(first, last); | ||
| [[maybe_unused]] const auto s_count = KE::distance(s_first, s_last); |
There was a problem hiding this comment.
Isn't this always used (in line 115)?
There was a problem hiding this comment.
the complain comes from line 106, it sees this possibility of existing earlier
There was a problem hiding this comment.
the same should be true for num_elements though , but don't recall seeing that warning
There was a problem hiding this comment.
I will try to remove them again
| // the target sequence should not be larger than the range [first, last) | ||
| namespace KE = ::Kokkos::Experimental; | ||
| const auto num_elements = KE::distance(first, last); | ||
| [[maybe_unused]] const auto s_count = KE::distance(s_first, s_last); |
There was a problem hiding this comment.
Why is this unused? Which compiler complains?
There was a problem hiding this comment.
the complain comes from line 189 and 193, it sees this possibility of existing earlier. I don't remember which compiler is doing this but definitely was getting warnings. I can dig it back up if you really want to know
There was a problem hiding this comment.
the same should be true for num_elements though
|
@masterleinad I removed all the |
| template <class ExecutionSpace, class IteratorType1, class IteratorType2> | ||
| template < | ||
| typename ExecutionSpace, typename IteratorType1, typename IteratorType2, | ||
| std::enable_if_t<::Kokkos::is_execution_space_v<ExecutionSpace>, int> = 0> |
There was a problem hiding this comment.
Is there a reason these guys are not constrained on are_iterators like the equal algorithm?
There was a problem hiding this comment.
the different # of params does not cause the same ambiguous error from compiler
Co-authored-by: Cezary Skrzyński
Co-authored-by: Jakub Strzebonski
Content
Team-level implementation and tests for:
Kokkos_Equal.hppKokkos_Search.hppKokkos_FindEnd.hppKokkos_FindFirstOf.hppTechnical comments
the change
class->typenameand the enable if defaulted in the template parameters comes from this requestthe code duplication is intentional and agreed upon after the long discussion in here
tests are run for both contiguous views and strided views
Other comments
all team-level PRs are tracked in: Team-level std algorithms PRs: issue tracker #6352
this comes from breaking apart team impl of std algorithms #5150
Possible conflicts
Implementation-wise, this has not conflict with other team-level PRs so it could be merged independently.
However, a trivial conflict in the CMakeLists inside the unit test occurs if another of the team-level PRs is merged before this because each team-level PR adds test using a specific "label". Resolving this is trivial.
Associated wiki PRs
Kokkos_Equal.hpp: see wiki PR 366Kokkos_Search.hpp: see wiki PR 371Kokkos_FindEnd.hpp: see wiki PR 366Kokkos_FindFirstOf.hpp: see wiki PR 366IMPORTANT: wiki PRs are not complete yet but 99% there
The issue tracking all wiki PRs is here