You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Its needed here because we are unifying the ranges API and main API handling of __parallel_copy_if to enable the performance improvements to be felt in both cases, and also to remove redundant code.
This is to avoid an issue with making a zip_view from a non const lvalue reference to a zip view. There may be changes we can make to our internal tuple class to resolve this, but I think that lies outside the scope of this PR.
An issue arises in reduce_by_key_ranges_sycl.pass when using make_zip_view and forwarding instead of using zip_view constructor here with a copy of the range. This seems wrong, and may imply changes with zip_view or our internal tuple.
The text was updated successfully, but these errors were encountered:
I don't like that this change is required, but it mirrors how it is currently handled in the ranges implementation.
oneDPL/include/oneapi/dpl/pstl/hetero/algorithm_ranges_impl_hetero.h
Line 364 in aec12a4
Its needed here because we are unifying the ranges API and main API handling of
__parallel_copy_if
to enable the performance improvements to be felt in both cases, and also to remove redundant code.This is to avoid an issue with making a
zip_view
from a non const lvalue reference to a zip view. There may be changes we can make to our internal tuple class to resolve this, but I think that lies outside the scope of this PR.Originally posted by @danhoeflinger in #1763 (comment)
An issue arises in reduce_by_key_ranges_sycl.pass when using
make_zip_view
and forwarding instead of usingzip_view
constructor here with a copy of the range. This seems wrong, and may imply changes with zip_view or our internal tuple.The text was updated successfully, but these errors were encountered: