RLE gpu decompress kernel

Signed-off-by: Robert Kruszewski <[email protected]>

Author: robert3005

java/testfiles/Cargo.lock

@@ -12,16 +12,16 @@ __device__ void dict_take(
     ValueT *__restrict values_out
 ) {
     auto i = threadIdx.x;
-    auto block_offset = (blockIdx.x * 1024);
+    auto block_offset = blockIdx.x * 1024;
 
     auto codes = codes_array + block_offset;
     auto out = values_out + block_offset;
 
     const int thread_ops = 32;
 
     for (auto j = 0; j < thread_ops; j++) {
-            auto idx = i * thread_ops + j;
-            out[idx] = values[codes[idx]];
+        auto idx = i * thread_ops + j;
+        out[idx] = values[codes[idx]];
     }
 }
 
@@ -33,8 +33,8 @@ __device__ void dict_take_masked(
     ValueT *__restrict values_out
 ) {
     auto i = threadIdx.x;
-    auto block_offset = (blockIdx.x * 1024);
-    auto mask_block_offset = (blockIdx.x * (1024 / 32));
+    auto block_offset = blockIdx.x * 1024;
+    auto mask_block_offset = blockIdx.x * (1024 / 32);
 
     auto codes = codes_array + block_offset;
     auto mask = mask_array + mask_block_offset;