Skip to content

Fix AVX-512 reduce_max/reduce_min missing bytes 32-39 for 8-bit types #1431

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
DiamonDinoia merged 1 commit into from
Oct 23, 2026
+22 −12
Merged

Fix AVX-512 reduce_max/reduce_min missing bytes 32-39 for 8-bit types #1431

Show file tree
Hide file tree
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
18 changes: 6 additions & 12 deletions include/xsimd/arch/xsimd_avx512f.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1962,25 +1962,19 @@ namespace xsimd
}

// reduce_max
template <class A, class T, class = std::enable_if_t<(sizeof(T) == 1)>>
template <class A, class T, class = std::enable_if_t<(sizeof(T) <= 2)>>
XSIMD_INLINE T reduce_max(batch<T, A> const& self, requires_arch<avx512f>) noexcept
{
constexpr batch_constant<uint64_t, A, 5, 6, 7, 8, 0, 0, 0, 0> mask;
batch<T, A> step = _mm512_permutexvar_epi64(mask.as_batch(), self);
batch<T, A> acc = max(self, step);
__m256i low = _mm512_castsi512_si256(acc);
return reduce_max(batch<T, avx2>(low));
batch<T, avx2> low(detail::lower_half(self)), high(detail::upper_half(self));
return reduce_max(max(low, high));
}

// reduce_min
template <class A, class T, class = std::enable_if_t<(sizeof(T) == 1)>>
template <class A, class T, class = std::enable_if_t<(sizeof(T) <= 2)>>
XSIMD_INLINE T reduce_min(batch<T, A> const& self, requires_arch<avx512f>) noexcept
{
constexpr batch_constant<uint64_t, A, 5, 6, 7, 8, 0, 0, 0, 0> mask;
batch<T, A> step = _mm512_permutexvar_epi64(mask.as_batch(), self);
batch<T, A> acc = min(self, step);
__m256i low = _mm512_castsi512_si256(acc);
return reduce_min(batch<T, avx2>(low));
batch<T, avx2> low(detail::lower_half(self)), high(detail::upper_half(self));
return reduce_min(min(low, high));
}

// reduce_mul
Expand Down
16 changes: 16 additions & 0 deletions test/test_batch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1070,6 +1070,22 @@ struct batch_test
INFO("reduce_min");
CHECK_SCALAR_EQ(res, expected);
}
// reduce_max / reduce_min with the extremum at each position
for (std::size_t i = 0; i < size; ++i)
{
array_type data;
data.fill(value_type(1));
{
data[i] = std::numeric_limits<value_type>::max();
INFO("reduce_max, extremum at index ", i);
CHECK_SCALAR_EQ(reduce_max(batch_type::load_unaligned(data.data())), data[i]);
}
{
data[i] = std::numeric_limits<value_type>::lowest();
INFO("reduce_min, extremum at index ", i);
CHECK_SCALAR_EQ(reduce_min(batch_type::load_unaligned(data.data())), data[i]);
}
}
// reduce_mul
{
value_type expected = std::accumulate(lhs.cbegin(), lhs.cend(), value_type(1), std::multiplies<value_type>());
Expand Down
Loading