Try using unified memory.

Author: maleadt

test/base/random.jl

@@ -200,12 +200,8 @@ end
 
 @testset "counter overflow" begin
     rng = CUDA.RNG()
-    # we may not be able to allocate over 4GB on the GPU, so use CPU memory
-    #c = CUDA.zeros(Float16, (64, 32, 512, 32, 64))
-    c = Array{Float16}(undef, 64, 32, 512, 32, 64)
-    GC.@preserve c begin
-        dc = unsafe_wrap(CuArray, c)
-        rand!(rng, dc)
-        randn!(rng, dc)
-    end
+    # we may not be able to allocate over 4GB on the GPU, so use unified memory
+    c = CuArray{Float16, 5, CUDA.UnifiedMemory}(undef, 64, 32, 512, 32, 64)
+    rand!(rng, c)
+    randn!(rng, c)
 end