Testing block_size=16 for AMDGPU

Author: anicusan

README.md

@@ -139,8 +139,6 @@ Julia v1.11
 
 [Metal](https://github.com/JuliaGPU/Metal.jl)
 
-[Known Issue with `accumulate` Only](https://github.com/JuliaGPU/AcceleratedKernels.jl/issues/10) 
-
 </td>
 <td>
 

prototype/accumulate_benchmark.jl

@@ -7,7 +7,7 @@ Random.seed!(0)
 
 
 function akacc(v)
-    va = AK.accumulate(+, v, init=zero(eltype(v)), block_size=512)
+    va = AK.accumulate(+, v, init=zero(eltype(v)), block_size=1024)
     Metal.synchronize()
     va
 end

test/runtests.jl

@@ -1064,7 +1064,7 @@ end
         num_elems = rand(1:100_000)
         x = array_from_host(rand(1:1000, num_elems), Int32)
         y = copy(x)
-        AK.accumulate!(+, y; init=0, block_size=8)
+        AK.accumulate!(+, y; init=0, block_size=16)
         @test all(Array(y) .== accumulate(+, Array(x)))
     end