diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index f53f890d..4d055fbb 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -9,13 +9,10 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        version:
-          - '1.6'
-          - '1' # Latest 1.X
-        os:
-          - ubuntu-latest
-        arch:
-          - x64
+        include:
+          - {version: '1', os: ubuntu-latest,   arch: x64}
+          - {version: '1.6', os: ubuntu-latest, arch: x64}
+          - {version: '1', os: macos-14,        arch: aarch64}
     steps:
       - uses: actions/checkout@v2
       - uses: julia-actions/setup-julia@v1
diff --git a/CHANGELOG.md b/CHANGELOG.md
index bcad921d..0e60d69b 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,6 +5,12 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.3.5] - 2025-07-31
+
+### Fixed
+
+- Bugfix: Use standard operations for reduce to be compatible with aarch64 architecture (see https://github.com/PartitionedArrays/PartitionedArrays.jl/pull/209)
+
 ## [0.3.4] - 2023-09-06
 
 ### Added 
diff --git a/Project.toml b/Project.toml
index 35e5e577..1470fcf7 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,7 +1,7 @@
 name = "PartitionedArrays"
 uuid = "5a9dfac6-5c52-46f7-8278-5e2210713be9"
 authors = ["Francesc Verdugo <f.verdugo.rojano@vu.nl> and contributors"]
-version = "0.3.4"
+version = "0.3.5"
 
 [deps]
 CircularArrays = "7a955b69-7140-5f4e-a0ed-f168c5e2e749"
diff --git a/src/p_vector.jl b/src/p_vector.jl
index 8cebf37a..f32478c4 100644
--- a/src/p_vector.jl
+++ b/src/p_vector.jl
@@ -1,6 +1,30 @@
 
 function local_values end
 
+function reduce_for_distance(d, partials, init)
+    reduce((i, j) -> Distances.eval_reduce(d, i, j), partials; init)
+end
+
+reduce_for_distance(::Distances.Chebyshev, partials, init) =
+    reduce(max, partials; init)
+
+for T in (Distances.Euclidean,
+          Distances.SqEuclidean,
+          Distances.PeriodicEuclidean,
+          Distances.Cityblock,
+          Distances.TotalVariation,
+          Distances.Minkowski,
+          Distances.Hamming,
+          Distances.ChiSqDist,
+          Distances.KLDivergence,
+          Distances.GenKLDivergence)
+    # all Distances.metrics except Jaccard, RogersTanimoto, CosineDist,
+    # RenyiDivergence, BrayCurtis, SpanNormDist, since these either operate with
+    # tuples or have a complex reduction logic.
+    @eval reduce_for_distance(::$(T), partials, init) =
+        reduce(+, partials; init)
+end
+
 function own_values end
 
 function ghost_values end
@@ -834,9 +858,8 @@ for M in Distances.metrics
                     return s
                 end
             end
-            s = reduce((i,j)->Distances.eval_reduce(d,i,j),
-                       partials,
-                       init=Distances.eval_start(d, a, b))
+            init_val = Distances.eval_start(d, a, b)
+            s = reduce_for_distance(d, partials, init_val)
             Distances.eval_end(d,s)
         end
     end
diff --git a/src/primitives.jl b/src/primitives.jl
index c0868867..349bd785 100644
--- a/src/primitives.jl
+++ b/src/primitives.jl
@@ -601,14 +601,11 @@ function ExchangeGraph_impl_with_neighbors(snd_ids,neighbors::ExchangeGraph)
         is_included_all=map(snd_ids_a, snd_ids_b) do snd_ids_a, snd_ids_b
             all(i->i in snd_ids_b,snd_ids_a)
         end
-        result=false
-        and(a,b)=a && b
-        is_included_all=reduction(and,is_included_all,destination=:all,init=one(eltype(is_included_all)))
-        map(is_included_all) do is_included_all
-            result = is_included_all
-        end
-        result
-    end 
+        # perform a global AND using built-in & operator (static MPI-op)
+        is_included_all = reduction(&, is_included_all; destination=:all, init=one(eltype(is_included_all)))
+        # extract the single boolean result
+        collect(is_included_all)[1]
+    end
     @boundscheck is_included(snd_ids,neighbors.snd) || error("snd_ids must be a subset of neighbors.snd")
     rank = linear_indices(snd_ids)
     # Tell the neighbors whether I want to send to them