Add local user buffer registration for NVLink SHARP.
Add tuning plugin support.
Increase net API to v7 to allow for device-side packet reordering;
remove support for v4 plugins.
Add support for RoCE ECE.
Add support for C2C links.
Better detect SHM allocation failures to avoid crash with Bus Error.
Fix missing thread unlocks in bootstrap (Fixes #936).
Disable network flush by default on H100.
Move device code from src/collectives/device to src/device.


[ROCm/rccl commit: f9c3dc251e]
This commit is contained in:
Sylvain Jeaugey
2023-09-26 05:47:28 -07:00
parent 26b91b9dbb
commit 506d6c332c
108 changed files with 4749 additions and 2059 deletions
+1 -1
View File
@@ -101,7 +101,7 @@ static inline ncclResult_t ncclCuMemAlloc(void **ptr, CUmemGenericAllocationHand
/* Allocate the physical memory on the device */
CUCHECK(cuMemCreate(&handle, size, &prop, 0));
/* Reserve a virtual address range */
CUCHECK(cuMemAddressReserve((CUdeviceptr *)ptr, size, 0, 0, 0));
CUCHECK(cuMemAddressReserve((CUdeviceptr *)ptr, size, granularity, 0, 0));
/* Map the virtual address range to the physical allocation */
CUCHECK(cuMemMap((CUdeviceptr)*ptr, size, 0, handle, 0));
/* Now allow RW access to the newly mapped memory */