[BUILD] Move code generation to python from CMake (#1360)

* Use generate.py for func generation

* Convert AddUnroll.cmake to bash

src/include/device.h

@@ -552,7 +552,7 @@ inline bool ncclNvlsSupported(int devRedOp, int type) {
 // Map the rowIdx to funcIdx
 extern int const ncclDevFuncRowToId[];
 
-// `ncclFuncIndex()` needs to be in sync with 'ALL_COLLS' in Generate.cmake
+// `ncclDevFuncId()` needs to be in sync with 'ALL_COLLS' in generate.py
 inline int ncclDevFuncId(int coll, int devRedOp, int type, int algo, int proto) {
   int row = 0;
   do {
@@ -568,7 +568,7 @@ inline int ncclDevFuncId(int coll, int devRedOp, int type, int algo, int proto)
       row += (((algo * NCCL_NUM_PROTOCOLS + proto) * ncclNumDevRedOps + devRedOp) * ncclNumTypes + type) - NCCL_NUM_FLOATS * (algo * NCCL_NUM_PROTOCOLS + proto);
       break;
     }
-    row += (NCCL_NUM_ALGORITHMS - 2) * NCCL_NUM_PROTOCOLS * (ncclNumDevRedOps * ncclNumTypes - NCCL_NUM_FLOATS);
+    row += (NCCL_NUM_ALGORITHMS - 4) * NCCL_NUM_PROTOCOLS * (ncclNumDevRedOps * ncclNumTypes - NCCL_NUM_FLOATS);
 
     // RING / SIMPLE / Sum / int8_t
     if (coll == ncclFuncAllToAllPivot) break;

src/include/nccl_common.h

@@ -13,7 +13,7 @@ typedef enum {NCCL_INIT=1, NCCL_COLL=2, NCCL_P2P=4, NCCL_SHM=8, NCCL_NET=16, NCC
 typedef void (*ncclDebugLogger_t)(ncclDebugLogLevel level, unsigned long flags, const char *file, int line, const char *fmt, ...);
 
 #define NCCL_NUM_ONERANK 12
-#define FUNC_INDEX_TOTAL 980 + NCCL_NUM_ONERANK
+#define FUNC_INDEX_TOTAL 656 + NCCL_NUM_ONERANK
 
 #define NCCL_NUM_FUNCTIONS 5 // Send/Recv not included for now
 typedef enum {