Fix fill!

[michel2323]

There was a test failure on a Max 1100 GPU in

oneAPI.jl/test/level-zero.jl

Lines 274 to 281 in 010bd13

	execute!(queue) do list
	pattern = [42]
	append_fill!(list, pointer(dst), pointer(pattern), sizeof(pattern), sizeof(src))
	append_barrier!(list)
	append_copy!(list, pointer(chk), pointer(dst), sizeof(src))
	end
	synchronize(queue)
	@test all(isequal(42), chk)

caused by passing a pointer to a standard Julia array (pattern = [42]) to zeCommandListAppendMemoryFill. AFAIK, on discrete Intel GPUs (unlike integrated ones), standard host memory is often not directly accessible by the device command processor. I also fixed fill! to address the same issue.

In that vein, I will also add a GitHub Actions runner for that GPU.

[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/array.jl b/src/array.jl
index 4d621cf..8ab4ba8 100644
--- a/src/array.jl
+++ b/src/array.jl
@@ -505,17 +505,17 @@ fill(v, dims...) = fill!(oneArray{typeof(v)}(undef, dims...), v)
 fill(v, dims::Dims) = fill!(oneArray{typeof(v)}(undef, dims...), v)
 
 function Base.fill!(A::oneDenseArray{T}, val) where T
-  length(A) == 0 && return A
-  val = convert(T, val)
-  sizeof(T) == 0 && return A
-
-  # execute! is async, so we need to allocate the pattern in USM memory
-  # and keep it alive until the operation completes.
-  buf = oneL0.host_alloc(context(A), sizeof(T), Base.datatype_alignment(T))
-  unsafe_store!(convert(Ptr{T}, buf), val)
-  unsafe_fill!(context(A), device(), pointer(A), convert(ZePtr{T}, buf), length(A))
-  synchronize(global_queue(context(A), device()))
-  oneL0.free(buf)
+    length(A) == 0 && return A
+    val = convert(T, val)
+    sizeof(T) == 0 && return A
+
+    # execute! is async, so we need to allocate the pattern in USM memory
+    # and keep it alive until the operation completes.
+    buf = oneL0.host_alloc(context(A), sizeof(T), Base.datatype_alignment(T))
+    unsafe_store!(convert(Ptr{T}, buf), val)
+    unsafe_fill!(context(A), device(), pointer(A), convert(ZePtr{T}, buf), length(A))
+    synchronize(global_queue(context(A), device()))
+    oneL0.free(buf)
   A
 end
 
diff --git a/test/level-zero.jl b/test/level-zero.jl
index ed7b283..3b13f34 100644
--- a/test/level-zero.jl
+++ b/test/level-zero.jl
@@ -271,22 +271,22 @@ let src = rand(Int, 1024)
     synchronize(queue)
     @test chk == src
 
-    # FIX: Allocate pattern in USM Host Memory
-    # Standard Host memory (stack/heap) is not accessible by discrete GPUs for fill patterns.
-    # We must use USM Host Memory.
-    pattern_val = 42
-    pattern_buf = oneL0.host_alloc(ctx, sizeof(Int), Base.datatype_alignment(Int))
-    unsafe_store!(convert(Ptr{Int}, pattern_buf), pattern_val)
+        # FIX: Allocate pattern in USM Host Memory
+        # Standard Host memory (stack/heap) is not accessible by discrete GPUs for fill patterns.
+        # We must use USM Host Memory.
+        pattern_val = 42
+        pattern_buf = oneL0.host_alloc(ctx, sizeof(Int), Base.datatype_alignment(Int))
+        unsafe_store!(convert(Ptr{Int}, pattern_buf), pattern_val)
 
     execute!(queue) do list
-        # Use the USM pointer (converted to ZePtr)
-        append_fill!(list, pointer(dst), convert(ZePtr{Int}, pattern_buf), sizeof(Int), sizeof(src))
+            # Use the USM pointer (converted to ZePtr)
+            append_fill!(list, pointer(dst), convert(ZePtr{Int}, pattern_buf), sizeof(Int), sizeof(src))
         append_barrier!(list)
         append_copy!(list, pointer(chk), pointer(dst), sizeof(src))
     end
     synchronize(queue)
 
-    oneL0.free(pattern_buf)
+        oneL0.free(pattern_buf)
 
     @test all(isequal(42), chk)
 

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.28%. Comparing base (a00fad6) to head (a498edc).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #555      +/-   ##
==========================================
+ Coverage   79.24%   79.28%   +0.04%     
==========================================
  Files          46       46              
  Lines        3064     3070       +6     
==========================================
+ Hits         2428     2434       +6     
  Misses        636      636              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[maleadt]

That's curious; I developed this on an A770 discrete GPU where it worked fine.