add 256 channels support

[ROCm/rccl commit: 506f16c506]

projects/rccl/src/include/device.h

@@ -61,7 +61,8 @@ union ncclLLFifoLine {
 };
 
 #define WARP_SIZE warpSize
-#define MAXCHANNELS 64
+#define MAXCHANNELS 256
+
 #define NCCL_MAX_NTHREADS 256
 #define NCCL_SIMPLE_MAX_NTHREADS NCCL_MAX_NTHREADS
 #define NCCL_LL_MAX_NTHREADS NCCL_MAX_NTHREADS
@@ -586,4 +587,4 @@ inline int ncclDevFuncId(int coll, int devRedOp, int type, int algo, int proto)
 
 inline int ncclDevFuncId_P2p() { return ncclDevFuncRowToId[FUNC_INDEX_TOTAL - NCCL_NUM_ONERANK - 1]; }
 
-#endif
+#endif