Template unroll for RCCL kernels (#1250)

* Template unroll for RCCL kernels

* Adding unroll template arg during CMake hipification

* Reduce linking parallel jobs to avoid OOM in CI

* Workaround issues with UT tests

SWDEV-469533: register spill fix is needed for mainline build
LWPCOMMLIBS-369: cannot enable 112 channels with 80 CUs
Use -parallel-jobs=8 for linking

* CI: do not use -j 16 when building

* CI: use -j 8 when building

* Only reduce parallel linking job for CI extended

* Restore original jenkins command. Change parallel linking jobs in cmake

* Disable MSCCLPP

---------

Co-authored-by: gilbertlee-amd <gilbert.lee@amd.com>

[ROCm/rccl commit: 89349f2ce4]

projects/rccl/src/enqueue.cc

@@ -30,13 +30,16 @@ struct ncclKernelMatch {
 };
 
 #ifdef ENABLE_COLLTRACE
-static ncclKernelMatch const ncclKerns[2] = {
+static ncclKernelMatch const ncclKerns[4] = {
   {(void *)ncclDevKernel_Generic, true},
+  {(void *)ncclDevKernel_Generic_4, true},
   {(void *)ncclDevKernelDebug_Generic, true},
+  {(void *)ncclDevKernelDebug_Generic_4, true},
 };
 #else
-static ncclKernelMatch const ncclKerns[1] = {
-  {(void*)ncclDevKernel_Generic, true}
+static ncclKernelMatch const ncclKerns[2] = {
+  {(void*)ncclDevKernel_Generic, true},
+  {(void*)ncclDevKernel_Generic_4, true},
 };
 #endif
 
@@ -57,6 +60,21 @@ static ncclResult_t getPatternInfo(struct ncclInfo* collInfo);
 static ncclResult_t getLoopInfo(struct ncclInfo* collInfo);
 static ncclResult_t getCollNetSupport(struct ncclInfo* info, int* collNetSupport);
 
+int ncclGetKernelIndex(struct ncclComm* comm) {
+#if ENABLE_COLLTRACE
+  int start_idx = comm->collTraceThread ? 2 : 0;
+#else
+  int start_idx = 0;
+#endif
+  hipDeviceProp_t devProp;
+  CUDACHECK(hipGetDeviceProperties(&devProp, comm->cudaDev));
+  if(IsArchMatch(devProp.gcnArchName, "gfx908") || (IsArchMatch(devProp.gcnArchName, "gfx94")
+    && devProp.multiProcessorCount > 80))
+    return start_idx;
+  else
+    return start_idx + 1;
+}
+
 // Returns maximum kernel stack size of all CUDA kernels
 ncclResult_t ncclInitKernelsForDevice(int cudaArch, size_t* maxStackSize) {
   constexpr int KernelCount = sizeof(ncclKerns)/sizeof(ncclKerns[0]);