From 1f91edaca6948b1dd7e0087f5b7dffcd08127222 Mon Sep 17 00:00:00 2001
From: Valentin Churavy <v.churavy@gmail.com>
Date: Thu, 20 Jun 2024 13:30:55 -0400
Subject: [PATCH 01/20] Use preferences to switch between SIMD and
 KernelAbstractions

---
 Project.toml     |  3 ++-
 src/WaterLily.jl | 21 ++++++++++++++-------
 src/util.jl      | 19 ++++++++++++++++---
 3 files changed, 32 insertions(+), 11 deletions(-)

diff --git a/Project.toml b/Project.toml
index 10fae339..627eac49 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,7 +1,7 @@
 name = "WaterLily"
 uuid = "ed894a53-35f9-47f1-b17f-85db9237eebd"
 authors = ["Gabriel Weymouth <gabriel.weymouth@gmail.com>"]
-version = "1.1"
+version = "1.1.0"
 
 [deps]
 DocStringExtensions = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
@@ -9,6 +9,7 @@ EllipsisNotation = "da5c29d0-fa7d-589e-88eb-ea29b0a81949"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+Preferences = "21216c6a-2e73-6563-6e65-726566657250"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
 Requires = "ae029012-a4dd-5104-9daa-d747884805df"
diff --git a/src/WaterLily.jl b/src/WaterLily.jl
index bdb39de6..471804d5 100644
--- a/src/WaterLily.jl
+++ b/src/WaterLily.jl
@@ -134,15 +134,22 @@ export restart_sim!
 
 # Check number of threads when loading WaterLily
 """
-    check_nthreads(::Val{1})
+    check_nthreads()
 
 Check the number of threads available for the Julia session that loads WaterLily.
-A warning is shown when running in serial (JULIA_NUM_THREADS=1).
+A warning is shown when running in serial (JULIA_NUM_THREADS=1) with KernelAbstractions enabled.
 """
-check_nthreads(::Val{1}) = @warn("\nUsing WaterLily in serial (ie. JULIA_NUM_THREADS=1) is not recommended because \
-    it disables the GPU backend and defaults to serial CPU."*
-    "\nUse JULIA_NUM_THREADS=auto, or any number of threads greater than 1, to allow multi-threading in CPU or GPU backends.")
-check_nthreads(_) = nothing
+function check_nthreads()
+    if backend == :KernelAbstractions && Threads.nthreads() == 1
+        @warn """
+            Using WaterLily in serial (ie. JULIA_NUM_THREADS=1) is not recommended because
+            it defaults to serial CPU execution."*
+            Use JULIA_NUM_THREADS=auto, or any number of threads greater than 1, to allow multi-threading in CPU backends.
+
+            For a low-overhead single-threaded CPU only backend set: `WaterLily.set_backend("SIMD")`
+        """
+    end
+end
 
 # Backward compatibility for extensions
 if !isdefined(Base, :get_extension)
@@ -155,7 +162,7 @@ function __init__()
         @require WriteVTK = "64499a7a-5c06-52f2-abe2-ccb03c286192" include("../ext/WaterLilyWriteVTKExt.jl")
         @require ReadVTK = "dc215faf-f008-4882-a9f7-a79a826fadc3" include("../ext/WaterLilyReadVTKExt.jl")
     end
-    check_nthreads(Val{Threads.nthreads()}())
+    check_nthreads()
 end
 
 end # module
diff --git a/src/util.jl b/src/util.jl
index 5edc8be6..55c4ca6d 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -65,6 +65,19 @@ macro inside(ex)
     end |> esc
 end
 
+# Could also use ScopedValues in Julia 1.11+
+const backend = Symbol(@load_preference("backend", "KernelAbstractions"))
+function set_backend(new_backend::String)
+    if !(new_backend in ("SIMD", "KernelAbstractions"))
+        throw(ArgumentError("Invalid backend: \"$(new_backend)\""))
+    end
+
+    # Set it in our runtime values, as well as saving it to disk
+    @set_preferences!("backend" => new_backend)
+    @info("New backend set; restart your Julia session for this change to take effect!")
+end
+
+
 """
     @loop <expr> over <I ∈ R>
 
@@ -99,7 +112,7 @@ macro loop(args...)
     setdiff!(sym,[I]) # don't want to pass I as an argument
     @gensym(kern, kern_) # generate unique kernel function names for serial and KA execution
     return quote
-        function $kern($(rep.(sym)...),::Val{1})
+        function $kern($(rep.(sym)...),::Val{:SIMD})
             @simd for $I ∈ $R
                 @fastmath @inbounds $ex
             end
@@ -109,10 +122,10 @@ macro loop(args...)
             $I += I0
             @fastmath @inbounds $ex
         end
-        function $kern($(rep.(sym)...),_)
+        function $kern($(rep.(sym)...),::Val{:KernelAbstractions})
             $kern_(get_backend($(sym[1])),64)($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
         end
-        $kern($(sym...),Val{Threads.nthreads()}()) # dispatch to SIMD for -t 1, or KA otherwise
+        $kern($(sym...),Val{$backend}()) # dispatch to SIMD or KA otherwise
     end |> esc
 end
 function grab!(sym,ex::Expr)

From 4b7a8468a2ac44eca03c02c4bf21e2acc62e6671 Mon Sep 17 00:00:00 2001
From: Valentin Churavy <v.churavy@gmail.com>
Date: Thu, 20 Jun 2024 13:35:49 -0400
Subject: [PATCH 02/20] fixup! Use preferences to switch between SIMD and
 KernelAbstractions

---
 src/util.jl | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/util.jl b/src/util.jl
index 55c4ca6d..6988fd0c 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -66,6 +66,7 @@ macro inside(ex)
 end
 
 # Could also use ScopedValues in Julia 1.11+
+using Preferences
 const backend = Symbol(@load_preference("backend", "KernelAbstractions"))
 function set_backend(new_backend::String)
     if !(new_backend in ("SIMD", "KernelAbstractions"))

From e77bca4e6a9d393e040cbe980ea17295f41bc9d3 Mon Sep 17 00:00:00 2001
From: Valentin Churavy <v.churavy@gmail.com>
Date: Thu, 20 Jun 2024 13:39:22 -0400
Subject: [PATCH 03/20] fixup! Use preferences to switch between SIMD and
 KernelAbstractions

---
 src/util.jl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/util.jl b/src/util.jl
index 6988fd0c..96780c5f 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -126,7 +126,7 @@ macro loop(args...)
         function $kern($(rep.(sym)...),::Val{:KernelAbstractions})
             $kern_(get_backend($(sym[1])),64)($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
         end
-        $kern($(sym...),Val{$backend}()) # dispatch to SIMD or KA otherwise
+        $kern($(sym...),Val{$(QuoteNode(backend))}()) # dispatch to SIMD or KA otherwise
     end |> esc
 end
 function grab!(sym,ex::Expr)

From b01cdce0e00d85331b2a63cf885c6f394c5148b5 Mon Sep 17 00:00:00 2001
From: Valentin Churavy <v.churavy@gmail.com>
Date: Thu, 20 Jun 2024 14:17:07 -0400
Subject: [PATCH 04/20] don't limit workgroups to 64

---
 src/util.jl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/util.jl b/src/util.jl
index 96780c5f..80efd415 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -124,7 +124,7 @@ macro loop(args...)
             @fastmath @inbounds $ex
         end
         function $kern($(rep.(sym)...),::Val{:KernelAbstractions})
-            $kern_(get_backend($(sym[1])),64)($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
+            $kern_(get_backend($(sym[1])))($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
         end
         $kern($(sym...),Val{$(QuoteNode(backend))}()) # dispatch to SIMD or KA otherwise
     end |> esc

From f0b351f6c5b1e6003ab6895869fc796146d02342 Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 01:14:39 +0200
Subject: [PATCH 05/20] Updated loop macro and cleaned up Preferences.jl
 routines. When the SIMD backend is selected, the loop macro generates only
 the for loop, without a function wrapper. Also, dispatch based on number of
 threads has been removed, and now only the backend-specific kernel is
 compiled. The CI needs to be fixed for the allocations tests, which first
 needs to set the SIMD backend, and then re-run the tests.

---
 .gitignore        |  1 +
 src/WaterLily.jl  |  2 +-
 src/util.jl       | 41 ++++++++++++++++++++++-------------------
 test/alloctest.jl |  2 ++
 test/maintests.jl |  2 ++
 5 files changed, 28 insertions(+), 20 deletions(-)

diff --git a/.gitignore b/.gitignore
index 8cd2947b..b7d3610f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -2,6 +2,7 @@
 .vscode/
 .DS_Store
 Manifest.toml
+LocalPreferences.toml
 /docs/build/
 
 # gfx
diff --git a/src/WaterLily.jl b/src/WaterLily.jl
index 67ec14fb..99528196 100644
--- a/src/WaterLily.jl
+++ b/src/WaterLily.jl
@@ -6,7 +6,7 @@ module WaterLily
 using DocStringExtensions
 
 include("util.jl")
-export L₂,BC!,@inside,inside,δ,apply!,loc,@log
+export L₂,BC!,@inside,inside,δ,apply!,loc,@log,set_backend,backend
 
 using Reexport
 @reexport using KernelAbstractions: @kernel,@index,get_backend
diff --git a/src/util.jl b/src/util.jl
index 960d1186..0119abb6 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -91,7 +91,7 @@ end
 
 # Could also use ScopedValues in Julia 1.11+
 using Preferences
-const backend = Symbol(@load_preference("backend", "KernelAbstractions"))
+const backend = @load_preference("backend", "KernelAbstractions")
 function set_backend(new_backend::String)
     if !(new_backend in ("SIMD", "KernelAbstractions"))
         throw(ArgumentError("Invalid backend: \"$(new_backend)\""))
@@ -102,7 +102,6 @@ function set_backend(new_backend::String)
     @info("New backend set; restart your Julia session for this change to take effect!")
 end
 
-
 """
     @loop <expr> over <I ∈ R>
 
@@ -132,26 +131,30 @@ Note that `get_backend` is used on the _first_ variable in `expr` (`a` in this e
 """
 macro loop(args...)
     ex,_,itr = args
-    _,I,R = itr.args; sym = []
-    grab!(sym,ex)     # get arguments and replace composites in `ex`
-    setdiff!(sym,[I]) # don't want to pass I as an argument
-    @gensym(kern, kern_) # generate unique kernel function names for serial and KA execution
-    return quote
-        function $kern($(rep.(sym)...),::Val{:SIMD})
+    _,I,R = itr.args
+    @static if backend == "KernelAbstractions"
+        sym = []
+        grab!(sym,ex)     # get arguments and replace composites in `ex`
+        setdiff!(sym,[I]) # don't want to pass I as an argument
+        @gensym(kern, kern_) # generate unique kernel function names for serial and KA execution
+        return quote
+            @kernel function $kern_($(rep.(sym)...),@Const(I0)) # replace composite arguments
+                $I = @index(Global,Cartesian)
+                $I += I0
+                @fastmath @inbounds $ex
+            end
+            function $kern($(rep.(sym)...))
+                $kern_(get_backend($(sym[1])))($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
+            end
+            $kern($(sym...))
+        end |> esc
+    else # backend == "SIMD"
+        return quote
             @simd for $I ∈ $R
                 @fastmath @inbounds $ex
             end
-        end
-        @kernel function $kern_($(rep.(sym)...),@Const(I0)) # replace composite arguments
-            $I = @index(Global,Cartesian)
-            $I += I0
-            @fastmath @inbounds $ex
-        end
-        function $kern($(rep.(sym)...),::Val{:KernelAbstractions})
-            $kern_(get_backend($(sym[1])))($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
-        end
-        $kern($(sym...),Val{$(QuoteNode(backend))}()) # dispatch to SIMD or KA otherwise
-    end |> esc
+        end |> esc
+    end
 end
 function grab!(sym,ex::Expr)
     ex.head == :. && return union!(sym,[ex])      # grab composite name and return
diff --git a/test/alloctest.jl b/test/alloctest.jl
index 5e860d2c..a6e41d10 100644
--- a/test/alloctest.jl
+++ b/test/alloctest.jl
@@ -1,5 +1,7 @@
 using BenchmarkTools, Printf
 
+backend == "KernelAbstractions" && (set_backend("SIMD"); exit())
+
 @testset "mom_step! allocations" begin
     function Sim(θ;L=32,U=1,Re=100,perdir=())
         function map(x,t)
diff --git a/test/maintests.jl b/test/maintests.jl
index 10e79a5d..59136e2f 100644
--- a/test/maintests.jl
+++ b/test/maintests.jl
@@ -1,6 +1,8 @@
 using GPUArrays
 using ReadVTK, WriteVTK
 
+backend == "SIMD" && (set_backend("KernelAbstractions"); exit())
+
 @info "Test backends: $(join(arrays,", "))"
 @testset "util.jl" begin
     I = CartesianIndex(1,2,3,4)

From 2369279bc9de4ce56a6e20ce96c20b8c943dca9a Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 01:21:38 +0200
Subject: [PATCH 06/20] Small bug fix.

---
 src/WaterLily.jl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/WaterLily.jl b/src/WaterLily.jl
index 99528196..4c4f0e8e 100644
--- a/src/WaterLily.jl
+++ b/src/WaterLily.jl
@@ -169,7 +169,7 @@ Check the number of threads available for the Julia session that loads WaterLily
 A warning is shown when running in serial (JULIA_NUM_THREADS=1) with KernelAbstractions enabled.
 """
 function check_nthreads()
-    if backend == :KernelAbstractions && Threads.nthreads() == 1
+    if backend == "KernelAbstractions" && Threads.nthreads() == 1
         @warn """
             Using WaterLily in serial (ie. JULIA_NUM_THREADS=1) is not recommended because
             it defaults to serial CPU execution."*

From 2c629c20020ddf67d2acc9291052a33dd499e0bc Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 01:24:25 +0200
Subject: [PATCH 07/20] Cleaned up -t 1 warning message.

---
 src/WaterLily.jl | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/src/WaterLily.jl b/src/WaterLily.jl
index 4c4f0e8e..08b8c47f 100644
--- a/src/WaterLily.jl
+++ b/src/WaterLily.jl
@@ -171,11 +171,9 @@ A warning is shown when running in serial (JULIA_NUM_THREADS=1) with KernelAbstr
 function check_nthreads()
     if backend == "KernelAbstractions" && Threads.nthreads() == 1
         @warn """
-            Using WaterLily in serial (ie. JULIA_NUM_THREADS=1) is not recommended because
-            it defaults to serial CPU execution."*
-            Use JULIA_NUM_THREADS=auto, or any number of threads greater than 1, to allow multi-threading in CPU backends.
-
-            For a low-overhead single-threaded CPU only backend set: `WaterLily.set_backend("SIMD")`
+        Using WaterLily in serial (ie. JULIA_NUM_THREADS=1) is not recommended because it defaults to serial CPU execution.
+        Use JULIA_NUM_THREADS=auto, or any number of threads greater than 1, to allow multi-threading in CPU backends.
+        For a low-overhead single-threaded CPU only backend set: WaterLily.set_backend("SIMD")
         """
     end
 end

From 5f17edef71a55cbe0ebb26206d07905e62e4f767 Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 01:42:52 +0200
Subject: [PATCH 08/20] testing 64,64 workgroup size

---
 src/util.jl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/util.jl b/src/util.jl
index 0119abb6..f8829591 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -144,7 +144,7 @@ macro loop(args...)
                 @fastmath @inbounds $ex
             end
             function $kern($(rep.(sym)...))
-                $kern_(get_backend($(sym[1])))($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
+                $kern_(get_backend($(sym[1])),(64,64))($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
             end
             $kern($(sym...))
         end |> esc

From f1621be73421d6bde4b0f5c65de63a46f3eb76fc Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 02:26:36 +0200
Subject: [PATCH 09/20] Workgroup size is 64, like master.

---
 src/util.jl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/util.jl b/src/util.jl
index f8829591..a2b24a4f 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -144,7 +144,7 @@ macro loop(args...)
                 @fastmath @inbounds $ex
             end
             function $kern($(rep.(sym)...))
-                $kern_(get_backend($(sym[1])),(64,64))($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
+                $kern_(get_backend($(sym[1])),64)($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
             end
             $kern($(sym...))
         end |> esc

From 6f9aa992dd18ce62d72881d48cf2483a583e19c0 Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 16:57:25 +0200
Subject: [PATCH 10/20] Wrapped SIMD kernels in a fuction too, as
 single-threaded test were 40pc slower than master...

---
 src/util.jl | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/src/util.jl b/src/util.jl
index a2b24a4f..c6172d6e 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -132,11 +132,11 @@ Note that `get_backend` is used on the _first_ variable in `expr` (`a` in this e
 macro loop(args...)
     ex,_,itr = args
     _,I,R = itr.args
+    sym = []
+    grab!(sym,ex)     # get arguments and replace composites in `ex`
+    setdiff!(sym,[I]) # don't want to pass I as an argument
+    @gensym(kern, kern_) # generate unique kernel function names for serial and KA execution
     @static if backend == "KernelAbstractions"
-        sym = []
-        grab!(sym,ex)     # get arguments and replace composites in `ex`
-        setdiff!(sym,[I]) # don't want to pass I as an argument
-        @gensym(kern, kern_) # generate unique kernel function names for serial and KA execution
         return quote
             @kernel function $kern_($(rep.(sym)...),@Const(I0)) # replace composite arguments
                 $I = @index(Global,Cartesian)
@@ -150,9 +150,12 @@ macro loop(args...)
         end |> esc
     else # backend == "SIMD"
         return quote
-            @simd for $I ∈ $R
-                @fastmath @inbounds $ex
+            function $kern($(rep.(sym)...))
+                @simd for $I ∈ $R
+                    @fastmath @inbounds $ex
+                end
             end
+            $kern($(sym...))
         end |> esc
     end
 end

From fb85fd13cb71accf555e6c0cd22179bd9b56947f Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 20:27:31 +0200
Subject: [PATCH 11/20] First attempt at adjusting CI to work with
 LocalPreferences

---
 .github/workflows/ci.yml | 18 ++++++++++++++++--
 test/alloctest.jl        |  3 +--
 test/maintests.jl        |  3 +--
 3 files changed, 18 insertions(+), 6 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 502084cb..aa616cb4 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -56,10 +56,24 @@ jobs:
             ${{ runner.os }}-test-${{ env.cache-name }}-
             ${{ runner.os }}-test-
             ${{ runner.os }}-
-      - uses: julia-actions/julia-buildpkg@v1
-      - uses: julia-actions/julia-runtest@v1
+      # - uses: julia-actions/julia-buildpkg@v1
+      # - uses: julia-actions/julia-runtest@v1
+      - name: Setup and run WaterLily tests
         env:
           JULIA_NUM_THREADS: ${{ matrix.nthreads }}
+        shell: bash
+        run: |
+          if [ "${{ matrix.nthreads }}" = "auto" ]
+          then
+            printf "[WaterLily]\nbackend = \"KernelAbstractions\"" > LocalPreferences.toml
+          else
+            printf "[WaterLily]\nbackend = \"SIMD\"" > LocalPreferences.toml
+          fi
+        shell: julia --color=yes {0}
+        run: |
+          using Pkg
+          Pkg.activate(".")
+          Pkg.test(; coverage=true)
       - uses: julia-actions/julia-processcoverage@v1
       - uses: codecov/codecov-action@v5
         with:
diff --git a/test/alloctest.jl b/test/alloctest.jl
index a6e41d10..f0fa3c76 100644
--- a/test/alloctest.jl
+++ b/test/alloctest.jl
@@ -1,7 +1,6 @@
 using BenchmarkTools, Printf
 
-backend == "KernelAbstractions" && (set_backend("SIMD"); exit())
-
+backend != "SIMD" && throw(ArgumentError("KernelAbstractions backend not allowed to run allocations tests, use SIMD backend"))
 @testset "mom_step! allocations" begin
     function Sim(θ;L=32,U=1,Re=100,perdir=())
         function map(x,t)
diff --git a/test/maintests.jl b/test/maintests.jl
index 59136e2f..1268c857 100644
--- a/test/maintests.jl
+++ b/test/maintests.jl
@@ -1,8 +1,7 @@
 using GPUArrays
 using ReadVTK, WriteVTK
 
-backend == "SIMD" && (set_backend("KernelAbstractions"); exit())
-
+backend != "KernelAbstractions" && throw(ArgumentError("SIMD backend not allowed to run main tests, use KernelAbstractions backend"))
 @info "Test backends: $(join(arrays,", "))"
 @testset "util.jl" begin
     I = CartesianIndex(1,2,3,4)

From 2c7c41eea400f075f03e993a6ff421903fda03b8 Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 20:37:47 +0200
Subject: [PATCH 12/20] CI fix.

---
 .github/workflows/ci.yml | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index aa616cb4..af492dff 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -69,11 +69,7 @@ jobs:
           else
             printf "[WaterLily]\nbackend = \"SIMD\"" > LocalPreferences.toml
           fi
-        shell: julia --color=yes {0}
-        run: |
-          using Pkg
-          Pkg.activate(".")
-          Pkg.test(; coverage=true)
+          julia --color=yes -e "using Pkg; Pkg.activate(\".\"); Pkg.test(; coverage=true);"
       - uses: julia-actions/julia-processcoverage@v1
       - uses: codecov/codecov-action@v5
         with:

From 4ea8e73fbf654ccde067621887e12cad7e6717ce Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 23:08:11 +0200
Subject: [PATCH 13/20] Added automatic specialization for kernel functions
 generated with loop macro.

---
 Project.toml | 6 ++++--
 src/util.jl  | 9 +++++++--
 2 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/Project.toml b/Project.toml
index 752bc8d9..bacbdece 100644
--- a/Project.toml
+++ b/Project.toml
@@ -9,9 +9,10 @@ EllipsisNotation = "da5c29d0-fa7d-589e-88eb-ea29b0a81949"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
-Preferences = "21216c6a-2e73-6563-6e65-726566657250"
 LoggingExtras = "e6f89c97-d47a-5376-807f-9c37f3926c36"
+Preferences = "21216c6a-2e73-6563-6e65-726566657250"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
+Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
 Requires = "ae029012-a4dd-5104-9daa-d747884805df"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
@@ -20,8 +21,8 @@ TOML = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
 [weakdeps]
 AMDGPU = "21141c5a-9bdb-4563-92ae-f87d6854732e"
 CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
-GeometryBasics = "5c1252a2-5f33-56bf-86c9-59e7332b4326"
 GLMakie = "e9467ef8-e4e7-5192-8a1a-b1aee30e663a"
+GeometryBasics = "5c1252a2-5f33-56bf-86c9-59e7332b4326"
 Meshing = "e6723b4c-ebff-59f1-b4b7-d97aa5274f73"
 Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
 ReadVTK = "dc215faf-f008-4882-a9f7-a79a826fadc3"
@@ -42,6 +43,7 @@ EllipsisNotation = "1.8"
 ForwardDiff = "^0.10.18"
 KernelAbstractions = "0.9.1"
 LoggingExtras = "1.1"
+Random = "1.11.0"
 Reexport = "^1.2.2"
 Requires = "1.3"
 StaticArrays = "^1.1.0"
diff --git a/src/util.jl b/src/util.jl
index c6172d6e..00a815fe 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -1,5 +1,6 @@
 using KernelAbstractions: get_backend, @index, @kernel
 using LoggingExtras
+using Random
 
 # custom log macro
 _psolver = Logging.LogLevel(-123) # custom log level for pressure solver, needs the negative sign
@@ -135,6 +136,7 @@ macro loop(args...)
     sym = []
     grab!(sym,ex)     # get arguments and replace composites in `ex`
     setdiff!(sym,[I]) # don't want to pass I as an argument
+    symT = symtypes(sym) # generate a list of types for each symbol
     @gensym(kern, kern_) # generate unique kernel function names for serial and KA execution
     @static if backend == "KernelAbstractions"
         return quote
@@ -143,14 +145,14 @@ macro loop(args...)
                 $I += I0
                 @fastmath @inbounds $ex
             end
-            function $kern($(rep.(sym)...))
+            function $kern($(joinsymtype(rep.(sym),symT)...)) where {$(symT...)}
                 $kern_(get_backend($(sym[1])),64)($(sym...),$R[1]-oneunit($R[1]),ndrange=size($R))
             end
             $kern($(sym...))
         end |> esc
     else # backend == "SIMD"
         return quote
-            function $kern($(rep.(sym)...))
+            function $kern($(joinsymtype(rep.(sym),symT)...)) where {$(symT...)}
                 @simd for $I ∈ $R
                     @fastmath @inbounds $ex
                 end
@@ -169,6 +171,9 @@ grab!(sym,ex::Symbol) = union!(sym,[ex])          # grab symbol name
 grab!(sym,ex) = nothing
 rep(ex) = ex
 rep(ex::Expr) = ex.head == :. ? Symbol(ex.args[2].value) : ex
+symtypes(sym) = [Symbol.(randstring('A':'Z',4)) for _ in 1:length(sym)]
+joinsymtype(sym::Symbol,symT::Symbol) = Expr(:(::), sym, symT)
+joinsymtype(sym,symT) = zip(sym,symT) .|> x->joinsymtype(x...)
 
 using StaticArrays
 """

From eea319b17ad9df0c3ca00ca91546ede61ff358ed Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 23:41:24 +0200
Subject: [PATCH 14/20] Cleaned up CI.

---
 .github/workflows/ci.yml | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index af492dff..f0edd0a1 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -56,9 +56,7 @@ jobs:
             ${{ runner.os }}-test-${{ env.cache-name }}-
             ${{ runner.os }}-test-
             ${{ runner.os }}-
-      # - uses: julia-actions/julia-buildpkg@v1
-      # - uses: julia-actions/julia-runtest@v1
-      - name: Setup and run WaterLily tests
+      - name: WaterLily tests
         env:
           JULIA_NUM_THREADS: ${{ matrix.nthreads }}
         shell: bash
@@ -69,7 +67,7 @@ jobs:
           else
             printf "[WaterLily]\nbackend = \"SIMD\"" > LocalPreferences.toml
           fi
-          julia --color=yes -e "using Pkg; Pkg.activate(\".\"); Pkg.test(; coverage=true);"
+          julia --proj --color=yes -e "using Pkg; Pkg.test(; coverage=true);"
       - uses: julia-actions/julia-processcoverage@v1
       - uses: codecov/codecov-action@v5
         with:

From 12396106fbab6adf950da601f6057c011c3651a8 Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 23:45:14 +0200
Subject: [PATCH 15/20] Adding pkg Random in tests.

---
 Project.toml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/Project.toml b/Project.toml
index d8126407..f798b5b9 100644
--- a/Project.toml
+++ b/Project.toml
@@ -64,10 +64,11 @@ Makie = "537997a7-5e4e-5d89-9595-2241ea00577e"
 Meshing = "e6723b4c-ebff-59f1-b4b7-d97aa5274f73"
 PerformanceTestTools = "dc46b164-d16f-48ec-a853-60448fc869fe"
 Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
+Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 ReadVTK = "dc215faf-f008-4882-a9f7-a79a826fadc3"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 UnicodePlots = "b8865327-cd53-5732-bb35-84acbb429228"
 WriteVTK = "64499a7a-5c06-52f2-abe2-ccb03c286192"
 
 [targets]
-test = ["Test", "BenchmarkTools", "CUDA", "AMDGPU", "GPUArrays", "WriteVTK", "ReadVTK", "JLD2"]
+test = ["Test", "BenchmarkTools", "CUDA", "AMDGPU", "GPUArrays", "WriteVTK", "ReadVTK", "JLD2", "Random"]

From a50deb1eca5472625de530e5acb151e780110ef7 Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 23:52:34 +0200
Subject: [PATCH 16/20] Trying to fix Random pkg error in CI.

---
 Project.toml | 5 ++---
 src/util.jl  | 4 ++--
 2 files changed, 4 insertions(+), 5 deletions(-)

diff --git a/Project.toml b/Project.toml
index f798b5b9..e8d7172e 100644
--- a/Project.toml
+++ b/Project.toml
@@ -21,8 +21,8 @@ TOML = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
 [weakdeps]
 AMDGPU = "21141c5a-9bdb-4563-92ae-f87d6854732e"
 CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
-GeometryBasics = "5c1252a2-5f33-56bf-86c9-59e7332b4326"
 GLMakie = "e9467ef8-e4e7-5192-8a1a-b1aee30e663a"
+GeometryBasics = "5c1252a2-5f33-56bf-86c9-59e7332b4326"
 JLD2 = "033835bb-8acc-5ee8-8aae-3f567f8a3819"
 Meshing = "e6723b4c-ebff-59f1-b4b7-d97aa5274f73"
 Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
@@ -64,11 +64,10 @@ Makie = "537997a7-5e4e-5d89-9595-2241ea00577e"
 Meshing = "e6723b4c-ebff-59f1-b4b7-d97aa5274f73"
 PerformanceTestTools = "dc46b164-d16f-48ec-a853-60448fc869fe"
 Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
-Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 ReadVTK = "dc215faf-f008-4882-a9f7-a79a826fadc3"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 UnicodePlots = "b8865327-cd53-5732-bb35-84acbb429228"
 WriteVTK = "64499a7a-5c06-52f2-abe2-ccb03c286192"
 
 [targets]
-test = ["Test", "BenchmarkTools", "CUDA", "AMDGPU", "GPUArrays", "WriteVTK", "ReadVTK", "JLD2", "Random"]
+test = ["Test", "BenchmarkTools", "CUDA", "AMDGPU", "GPUArrays", "WriteVTK", "ReadVTK", "JLD2"]
diff --git a/src/util.jl b/src/util.jl
index 00a815fe..14fbf672 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -1,6 +1,5 @@
 using KernelAbstractions: get_backend, @index, @kernel
 using LoggingExtras
-using Random
 
 # custom log macro
 _psolver = Logging.LogLevel(-123) # custom log level for pressure solver, needs the negative sign
@@ -171,7 +170,8 @@ grab!(sym,ex::Symbol) = union!(sym,[ex])          # grab symbol name
 grab!(sym,ex) = nothing
 rep(ex) = ex
 rep(ex::Expr) = ex.head == :. ? Symbol(ex.args[2].value) : ex
-symtypes(sym) = [Symbol.(randstring('A':'Z',4)) for _ in 1:length(sym)]
+using Random
+symtypes(sym) = [Symbol.(Random.randstring('A':'Z',4)) for _ in 1:length(sym)]
 joinsymtype(sym::Symbol,symT::Symbol) = Expr(:(::), sym, symT)
 joinsymtype(sym,symT) = zip(sym,symT) .|> x->joinsymtype(x...)
 

From d2b3ade22b1901001a6a41e22bd9955a3987bd43 Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Thu, 22 May 2025 23:57:48 +0200
Subject: [PATCH 17/20] Bumping CI to Julia 1.11, trying to fix Random pkg
 error during tests

---
 .github/workflows/ci.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index f0edd0a1..d131a8a0 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -26,7 +26,7 @@ jobs:
       fail-fast: false
       matrix:
         version:
-          - '1.10'
+          - '1.11'
         os:
           - ubuntu-latest
           - macOS-latest
@@ -67,7 +67,7 @@ jobs:
           else
             printf "[WaterLily]\nbackend = \"SIMD\"" > LocalPreferences.toml
           fi
-          julia --proj --color=yes -e "using Pkg; Pkg.test(; coverage=true);"
+          julia --proj --color=yes -e "using Pkg; Pkg.instantiate(); Pkg.test(; coverage=true);"
       - uses: julia-actions/julia-processcoverage@v1
       - uses: codecov/codecov-action@v5
         with:

From 87cddb8ae70cc78928c019564451bfd18b26a70e Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Fri, 23 May 2025 00:08:52 +0200
Subject: [PATCH 18/20] Removed x86 architecture from CI.

---
 .github/workflows/ci.yml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index d131a8a0..3993e4d5 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -33,7 +33,6 @@ jobs:
           - windows-latest
         arch:
           - x64
-          - x86
         nthreads:
           - '1'
           - 'auto'

From f8633ea3b82aa8a8abb8a9d21573cc017da5331d Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Sat, 24 May 2025 16:02:28 +0200
Subject: [PATCH 19/20] Moved nthreads check outside of __init__ so it is only
 checked once and not for every ext module when loading.

---
 src/WaterLily.jl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/WaterLily.jl b/src/WaterLily.jl
index 1d91fca8..13b202f6 100644
--- a/src/WaterLily.jl
+++ b/src/WaterLily.jl
@@ -180,6 +180,7 @@ function check_nthreads()
         """
     end
 end
+check_nthreads()
 
 # Backward compatibility for extensions
 if !isdefined(Base, :get_extension)
@@ -196,7 +197,6 @@ function __init__()
         @require Meshing = "e6723b4c-ebff-59f1-b4b7-d97aa5274f73" include("../ext/WaterLilyMeshingExt.jl")
         @require JLD2 = "033835bb-8acc-5ee8-8aae-3f567f8a3819" include("../ext/WaterLilyJLD2Ext.jl")
     end
-    check_nthreads()
 end
 
 end # module

From 7da05269fc7bfca0807a5c9bb631005b343c6bd5 Mon Sep 17 00:00:00 2001
From: Bernat Font <bernatfontgarcia@gmail.com>
Date: Sat, 24 May 2025 16:23:51 +0200
Subject: [PATCH 20/20] Added test for new routines used in automatic kernel
 generation.

---
 test/maintests.jl | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/test/maintests.jl b/test/maintests.jl
index d3ca2085..20973d13 100644
--- a/test/maintests.jl
+++ b/test/maintests.jl
@@ -1,5 +1,5 @@
 using GPUArrays
-using ReadVTK, WriteVTK, JLD2
+using ReadVTK, WriteVTK, JLD2, Random
 
 backend != "KernelAbstractions" && throw(ArgumentError("SIMD backend not allowed to run main tests, use KernelAbstractions backend"))
 @info "Test backends: $(join(arrays,", "))"
@@ -18,6 +18,10 @@ backend != "KernelAbstractions" && throw(ArgumentError("SIMD backend not allowed
     WaterLily.grab!(sym,ex)
     @test ex == :(a[I, i] = Math.add(b[I], func(I, q)))
     @test sym == [:a, :I, :i, :(p.b), :q]
+    sym = [:a,:b,:c]
+    Random.seed!(99)
+    symT = WaterLily.symtypes(sym)
+    @test WaterLily.joinsymtype(sym,symT) == Expr[:(a::MDFZ), :(b::AXYU), :(c::RFIB)]
 
     for f ∈ arrays
         p = zeros(4,5) |> f