diff --git a/src/device/prims_simple.h b/src/device/prims_simple.h
index 1887ea154b..11251684ee 100644
--- a/src/device/prims_simple.h
+++ b/src/device/prims_simple.h
@@ -369,7 +369,7 @@ private:
     // worker perf is the limiter, perf-wise this loop is effectively unentered,
     // hence just a single branch insn.
     #pragma unroll 1
-    while (slice < SlicePerChunk) {
+    while (false) {
       sliceSize = sliceSize < nelem-offset ? sliceSize : nelem-offset;
       { // Only workers could have Wait roles so we know the slice must be empty
         // since we've exited the loop above.
@@ -942,4 +942,4 @@ private:
   __device__ __forceinline__ void localCopy(T* srcs, T* dsts, int eltN) {
     return mscclGenericOp<0,1,0,0>(&srcs, 1, &dsts, 1, eltN);
   }
-};
\ No newline at end of file
+};