diff --git a/source/adapters/cuda/enqueue.cpp b/source/adapters/cuda/enqueue.cpp
index 5046f4c865..f21f5b00ee 100644
--- a/source/adapters/cuda/enqueue.cpp
+++ b/source/adapters/cuda/enqueue.cpp
@@ -169,18 +169,10 @@ void guessLocalWorkSize(ur_device_handle_t Device, size_t *ThreadsPerBlock,
   ThreadsPerBlock[0] = std::min(
       MaxThreadsPerBlock[0], std::min(GlobalSizeNormalized[0], MaxBlockDim[0]));
 
-  static auto IsPowerOf2 = [](size_t Value) -> bool {
-    return Value && !(Value & (Value - 1));
-  };
-
   // Find a local work group size that is a divisor of the global
   // work group size to produce uniform work groups.
-  // Additionally, for best compute utilisation, the local size has
-  // to be a power of two.
-  while (0u != (GlobalSizeNormalized[0] % ThreadsPerBlock[0]) ||
-         !IsPowerOf2(ThreadsPerBlock[0])) {
+  while (GlobalSizeNormalized[0] % ThreadsPerBlock[0])
     --ThreadsPerBlock[0];
-  }
 }
 
 // Helper to verify out-of-registers case (exceeded block max registers).
diff --git a/source/adapters/hip/enqueue.cpp b/source/adapters/hip/enqueue.cpp
index 80ec41f984..ce048d05d0 100644
--- a/source/adapters/hip/enqueue.cpp
+++ b/source/adapters/hip/enqueue.cpp
@@ -51,13 +51,9 @@ ur_result_t enqueueEventsWait(ur_queue_handle_t, hipStream_t Stream,
 
 void simpleGuessLocalWorkSize(size_t *ThreadsPerBlock,
                               const size_t *GlobalWorkSize,
-                              const size_t MaxThreadsPerBlock[3],
-                              ur_kernel_handle_t Kernel) {
+                              const size_t MaxThreadsPerBlock[3]) {
   assert(ThreadsPerBlock != nullptr);
   assert(GlobalWorkSize != nullptr);
-  assert(Kernel != nullptr);
-
-  std::ignore = Kernel;
 
   ThreadsPerBlock[0] = std::min(MaxThreadsPerBlock[0], GlobalWorkSize[0]);
 
@@ -345,7 +341,7 @@ UR_APIEXPORT ur_result_t UR_APICALL urEnqueueKernelLaunch(
       }
     } else {
       simpleGuessLocalWorkSize(ThreadsPerBlock, pGlobalWorkSize,
-                               MaxThreadsPerBlock, hKernel);
+                               MaxThreadsPerBlock);
     }
   }