2.7.8-1

Fix collective mismatch error when using ncclSend/ncclRecv

src/include/comm.h

@@ -37,7 +37,6 @@ struct ncclSendMem {
       char pad1[CACHE_LINE_SIZE-sizeof(uint64_t)];
       void* ptrExchange;
       char pad2[CACHE_LINE_SIZE-sizeof(void*)];
-      uint64_t opCount;
     };
     char pad3[MEM_ALIGN];
   };
@@ -49,7 +48,6 @@ struct ncclRecvMem {
     struct {
       uint64_t tail;
       char pad1[CACHE_LINE_SIZE-sizeof(uint64_t)];
-      uint64_t opCount;
       char pad2[CACHE_LINE_SIZE-sizeof(uint64_t)];
       int sizesFifo[NCCL_STEPS];
     };
@@ -109,9 +107,6 @@ struct ncclComm {
   // Whether there has been a fatal error in this communicator.
   ncclResult_t fatalError;
 
-  // Error reported by GPU
-  volatile ncclDevError_t* fatalDevError;
-
   // Flag to ask NCCL kernels to abort
   volatile uint32_t *abortFlag;
 

src/include/devcomm.h

@@ -83,8 +83,6 @@ struct ncclConnInfo {
   char *buffs[NCCL_NUM_PROTOCOLS]; // Local for recv, remote for send
   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
-  uint64_t *opCountRem; // opCount of remote rank
 
   int direct;         // Direct communication
   void **ptrExchange; // Pointer exchange for direct communication
@@ -136,7 +134,6 @@ struct ncclDevComm;
 /* Make sure to adjust padding at the end of ncclColl. */
 struct CollectiveArgs {
   struct ncclDevComm* comm;
-  uint64_t opCount;
 
   // local and remote input, output, and buffer
   const void * sendbuff;
@@ -205,12 +202,6 @@ struct ncclChannel {
 };
 static_assert(sizeof(struct ncclChannel) == 0x80*sizeof(int), "ncclChannel must have a pow2 size");
 
-typedef enum {
-  ncclDevSuccess,
-  ncclDevAssertedMismatch,
-  ncclDevSuspectedMismatch
-} ncclDevError_t;
-
 struct ncclDevComm {
   int rank;
   int nRanks;
@@ -218,7 +209,6 @@ struct ncclDevComm {
 
   // Flag to ask NCCL kernels to abort
   volatile uint32_t *abortFlag;
-  volatile ncclDevError_t *fatalDevError;
 
   // Channels, device side
   struct ncclChannel* channels;