Move mapreduce fix from AK to oneAPI.jl

[michel2323]

No description provided.

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic master) to apply these changes.

Click here to view the suggested changes.

diff --git a/src/mapreduce.jl b/src/mapreduce.jl
index 809e2c6..3a9c317 100644
--- a/src/mapreduce.jl
+++ b/src/mapreduce.jl
@@ -1,13 +1,13 @@
 # On oneAPI, use the MapReduce algorithm by default as on some Intel GPUs ConcurrentWrite hangs
 # the device.
 function AK.any(
-    pred, v::AbstractArray, backend::oneAPIBackend;
+        pred, v::AbstractArray, backend::oneAPIBackend;
 
-    # Algorithm choice
-    alg::AK.PredicatesAlgorithm=AK.MapReduce(),
-    kwargs...
-)
-    AK._any_impl(
+        # Algorithm choice
+        alg::AK.PredicatesAlgorithm = AK.MapReduce(),
+        kwargs...
+    )
+    return AK._any_impl(
         pred, v, backend;
         alg,
         kwargs...
@@ -16,13 +16,13 @@ end
 
 
 function AK.all(
-    pred, v::AbstractArray, backend::oneAPIBackend;
+        pred, v::AbstractArray, backend::oneAPIBackend;
 
-    # Algorithm choice
-    alg::AK.PredicatesAlgorithm=AK.MapReduce(),
-    kwargs...
-)
-    AK._all_impl(
+        # Algorithm choice
+        alg::AK.PredicatesAlgorithm = AK.MapReduce(),
+        kwargs...
+    )
+    return AK._all_impl(
         pred, v, backend;
         alg,
         kwargs...

[michel2323]

Closing: superseded by upstream direction in AcceleratedKernels.jl. The oneAPI extension there has since gained a _decoupled_fence device override (AK #116) and JuliaGPU/AcceleratedKernels.jl#133 moves further Intel-specific workarounds (scan block size, _supports_concurrent_write) into that extension rather than out of it. Keeping the any/all default in AK's oneAPIExt is the way forward, so this PR (and its counterpart JuliaGPU/AcceleratedKernels.jl#73) is no longer needed.