Revert "[SRC] Enable unroll=1 for gfx950 (#1602)" (#1667)

* Revert "[SRC] Enable unroll=1 for gfx950 (#1602)"
This reverts commit 307bc10781.

* Update Changelog

---------

Signed-off-by: nileshnegi <Nilesh.Negi@amd.com>

src/device/common_kernel.h

@@ -263,29 +263,9 @@ __device__ __forceinline__ void reduceCopy(
     }
   }
 
-/*
- * For gfx90a,
-* Before we had `Unroll/2*(16/sizeof(T))/2`, which does not work with unroll=1
-* as unroll=1; `Unroll/2` = 0, which results in the above expression to 0, and is not supported
-* This was reformulated to `(Unroll*4 + sizeof(T) - 1)/sizeof(T)`
-*
-* Before: `Unroll/2*(16/sizeof(T))/2`
-*         sizeof(T)
-* unroll  1   2   4   8
-*   4     16  8   4   2
-*   2     8   4   2   1
-*   1     0   0   0   0
-*
-* After: `(Unroll*4 + sizeof(T) - 1)/sizeof(T)`
-*         sizeof(T)
-* unroll  1   2   4   8
-*   4     16  8   4   2
-*   2     8   4   2   1
-*   1     4   2   1   1
-*/
 #if defined(__gfx90a__)
   if (MinSrcs > 1) {
-    reduceCopyPacks<RedFn, T, (Unroll*4 + sizeof(T) - 1)/sizeof(T), sizeof(T),
+    reduceCopyPacks<RedFn, T, Unroll/2*(16/sizeof(T))/2, sizeof(T),
     MultimemSrcs, MinSrcs, MaxSrcs, MultimemDsts, MinDsts, MaxDsts, PreOpSrcs>
     (nThreads, thread, redArg, preOpArgs, postOp,
      nSrcs, srcPtrFn, nDsts, dstPtrFn, nBytesBehind, nBytesAhead);
@@ -327,4 +307,4 @@ __device__ __forceinline__ void reduceCopy(
      nDsts, [=]__device__(int i) { return dstPtrs[i]; }, nElts);
 }
 
-#endif // COMMON_KERNEL_H_
+#endif // COMMON_KERNEL_H_