minor change

test/benchmark.jl

@@ -3,29 +3,38 @@ using BenchmarkTools
 sz = (1024,1024)
 
 v = rand(sz...);
-useCuda = false
+useCuda = true
 if useCuda
     using CUDA
     CUDA.allowscalar(false);
     v = CuArray(v)
+
+    macro mytime(expr)
+        return :( @btime CUDA.@sync $expr)
+    end    
+else
+    macro mytime(expr)
+        return :( @btime $expr)
+    end
 end
 sh = (335, 444)
 
 sv = ShiftedArray(v, sh)
 cv = CircShiftedArray(v, sh)
 
+
 # timings stated for Dell Laptop XPS 15 (i7 11800) on Windows 10
-@btime q = $sv .+ 5.0 # bc version: 1.48 ms, CuArray bc: 0.016 ms, old version: 2.73 ms
+@mytime q = $sv .+ 5.0; # bc version: 1.48 ms, CuArray bc: 0.099 ms, old version: 2.73 ms
 res = sv .+ 5.0
 
-@btime res .= $sv .+ 5.0 # bc version: 0.18 ms, CuArray bc: 0.015 ms, old version: 0.37 ms
+@mytime res .= $sv .+ 5.0 # bc version: 0.18 ms, CuArray bc: 0.097 ms, old version: 0.37 ms
 
 sv = ShiftedArray(v, sh, default=0.0)
 resn = copy(v)
-@btime $resn .= $sv .+ 5.0 # bc version: 0.28 ms, CuArray bc: 0.020 ms, old version: 0.42 ms
+@mytime $resn .= $sv .+ 5.0; # bc version: 0.28 ms, CuArray bc: 0.118 ms, old version: 0.42 ms
 
-@btime $res .= $sv .+ 5.0 .* $v .*$sv # bc version: 0.727 ms, CuArray bc: 0.039 ms, old version: 1.65 ms
+@mytime $res .= $sv .+ 5.0 .* $v .*$sv; # bc version: 0.727 ms, CuArray bc: 0.264 ms, old version: 1.65 ms
 
 svi = ShiftedArrays.ifftshift(v)
-@btime $resn .= $svi .+ 5.0 .* $v .*$svi # bc version: 0.53 ms, CuArray bc: 0.029 ms, old version: 3.98 ms
-@btime $resn .= ShiftedArrays.fftshift($svi .+ 5.0 .* $v .*$svi) # bc version: 2.41 ms,  CuArray bc: 0.050 ms, old version: 3.98 ms
+@mytime $resn .= $svi .+ 5.0 .* $v .*$svi; # bc version: 0.53 ms, CuArray bc: 0.324 ms, old version: 3.98 ms
+@mytime $resn .= ShiftedArrays.fftshift($svi .+ 5.0 .* $v .*$svi); # bc version: 2.41 ms,  CuArray bc: 0.443 ms, old version: 3.98 ms