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>
This commit is contained in:
Wenkai Du
2024-07-19 08:15:59 -07:00
committed by GitHub
parent a1ef217b32
commit 89349f2ce4
14 changed files with 149 additions and 55 deletions
+8 -2
View File
@@ -18,11 +18,17 @@ struct RunWorkNop {
};
__launch_bounds__(NCCL_MAX_NTHREADS, 1) __global__ void ncclDevKernel_Generic(struct ncclDevComm* comm, struct channelMasks channelMask, struct ncclWork* workHead) {
ncclKernelMain<-1, RunWorkNop, false>(comm, channelMask, workHead);
ncclKernelMain<-1, RunWorkNop, false, 2>(comm, channelMask, workHead);
}
__launch_bounds__(NCCL_MAX_NTHREADS, 1) __global__ void ncclDevKernel_Generic_4(struct ncclDevComm* comm, struct channelMasks channelMask, struct ncclWork* workHead) {
ncclKernelMain<-1, RunWorkNop, false, 4>(comm, channelMask, workHead);
}
#ifdef ENABLE_COLLTRACE
__launch_bounds__(NCCL_MAX_NTHREADS, 1) __global__ void ncclDevKernelDebug_Generic(struct ncclDevComm* comm, struct channelMasks channelMask, struct ncclWork* workHead) {
ncclKernelMain<-1, RunWorkNop, true>(comm, channelMask, workHead);
ncclKernelMain<-1, RunWorkNop, true, 2>(comm, channelMask, workHead);
}
__launch_bounds__(NCCL_MAX_NTHREADS, 1) __global__ void ncclDevKernelDebug_Generic_4(struct ncclDevComm* comm, struct channelMasks channelMask, struct ncclWork* workHead) {
ncclKernelMain<-1, RunWorkNop, true, 4>(comm, channelMask, workHead);
}
#endif