From f3b6ca5659bd0e1be0cb0c0ec5113ee7cc910725 Mon Sep 17 00:00:00 2001
From: Tim Besard <tim.besard@gmail.com>
Date: Tue, 27 Oct 2026 12:39:43 +0100
Subject: [PATCH] Define the KernelAbstractions methods in oneAPI rather than
 an extension

KernelAbstractions is always loaded anyway, since GPUArrays depends on it, so
make it a dependency and move the extension's methods into oneAPIKernels.

Here the extension did not invalidate anything, and oneAPI does not precompile
KernelAbstractions launches, but the other back-ends need this change to keep
their precompiled launches valid. Doing the same keeps them uniform, and makes
sure the KernelAbstractions methods exist whenever oneAPI's precompiled code is
compiled.
---
 Project.toml                 |  7 +------
 ext/KernelAbstractionsExt.jl | 22 ----------------------
 src/oneAPIKernels.jl         | 14 ++++++++++++++
 3 files changed, 15 insertions(+), 28 deletions(-)
 delete mode 100644 ext/KernelAbstractionsExt.jl

diff --git a/Project.toml b/Project.toml
index c819a1036..8b2f986c7 100644
--- a/Project.toml
+++ b/Project.toml
@@ -13,6 +13,7 @@ GPUArrays = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
 GPUCompiler = "61eb1bfa-7361-4325-ad38-22787b887f55"
 GPUToolbox = "096a3bc2-3ced-46d0-87f4-dd12716f4bfc"
 IntelITT = "c9b2f978-7543-4802-ae44-75068f23ee64"
+KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
 KernelInterface = "4ee993da-d684-4d17-a7dd-4e58e78d92bf"
 LLVM = "929cbde3-209d-540e-8aea-75f648917ca0"
 Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
@@ -35,12 +36,6 @@ oneAPI_Level_Zero_Headers_jll = "f4bc562b-d309-54f8-9efb-476e56f0410d"
 oneAPI_Level_Zero_Loader_jll = "13eca655-d68d-5b81-8367-6d99d727ab01"
 oneAPI_Support_jll = "b049733a-a71d-5ed3-8eba-7d323ac00b36"
 
-[weakdeps]
-KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
-
-[extensions]
-KernelAbstractionsExt = "KernelAbstractions"
-
 [compat]
 AbstractFFTs = "1.5.0"
 Adapt = "4"
diff --git a/ext/KernelAbstractionsExt.jl b/ext/KernelAbstractionsExt.jl
deleted file mode 100644
index f9a0d9fcd..000000000
--- a/ext/KernelAbstractionsExt.jl
+++ /dev/null
@@ -1,22 +0,0 @@
-module KernelAbstractionsExt
-
-using oneAPI
-
-import KernelAbstractions as KA
-
-import Adapt
-
-Adapt.adapt_storage(::KA.CPU, a::oneArray) = convert(Array, a)
-
-# sparse arrays (oneMKL is only available on Linux)
-@static if Sys.islinux()
-    import SparseArrays
-    Adapt.adapt_storage(::KA.CPU, a::oneAPI.oneMKL.oneAbstractSparseMatrix) = SparseArrays.SparseMatrixCSC(a)
-end
-
-
-## other
-
-Adapt.adapt_storage(to::KA.ConstAdaptor, a::oneDeviceArray) = Base.Experimental.Const(a)
-
-end
diff --git a/src/oneAPIKernels.jl b/src/oneAPIKernels.jl
index 6923e14b7..a2c7d8869 100644
--- a/src/oneAPIKernels.jl
+++ b/src/oneAPIKernels.jl
@@ -3,6 +3,7 @@ module oneAPIKernels
 using ..oneAPI
 using ..oneAPI: @device_override, SPIRVIntrinsics, method_table, kernel_convert, zefunction
 
+import KernelAbstractions as KA
 import KernelInterface as KI
 
 import Adapt
@@ -213,6 +214,19 @@ function KI.multiprocessor_count(::oneAPIBackend)::Int
 end
 
 
+## KernelAbstractions
+
+Adapt.adapt_storage(::KA.CPU, a::oneArray) = convert(Array, a)
+
+# sparse arrays (oneMKL is only available on Linux)
+@static if Sys.islinux()
+    import SparseArrays
+    Adapt.adapt_storage(::KA.CPU, a::oneAPI.oneMKL.oneAbstractSparseMatrix) = SparseArrays.SparseMatrixCSC(a)
+end
+
+Adapt.adapt_storage(to::KA.ConstAdaptor, a::oneDeviceArray) = Base.Experimental.Const(a)
+
+
 ## Indexing Functions
 ## COV_EXCL_START