Use posix_memalign for network buffer allocation on host memory (#221)

* Use posix_memalign for network buffer allocation on host memory

* ib-test: add ability to specify run iterations

* ib-test: define iterations as multiple of default cycles

* Add checking to posix_memalign return value

src/include/devcomm.h

@@ -92,6 +92,9 @@ static_assert(NCCL_LL_CLEAN_MASK % NCCL_STEPS == 0, "Invalid NCCL_LL_CLEAN_MASK
 struct ncclConnInfo {
   // Regular comm mechanism
   char *buffs[NCCL_NUM_PROTOCOLS]; // Local for recv, remote for send
+#ifdef USE_MEMALIGN
+  char *hostBuffs[NCCL_NUM_PROTOCOLS]; // Local for recv, remote for send
+#endif
   uint64_t *tail;     // Local for recv, remote for send
   uint64_t *head;     // Local for send, remote for recv
   uint64_t *opCountLoc; // opCount of local rank