2.6.4-1

Add support for network collectives.
Add support for XML topology dump/injection.
Add text values for GDR and P2P Levels, including "NVL".
Add speed detection for PCI, Infiniband and Ethernet cards.
Add CPU detection for ARM and AMD CPUs.
Add support for adaptive routing on Infiniband.
Change NET plugin API to v3 : merge PCI path and GPU pointer
  capability into a single structure and add other properties.

src/include/alloc.h

@@ -9,6 +9,7 @@
 
 #include "nccl.h"
 #include "checks.h"
+#include "align.h"
 #include <sys/mman.h>
 
 static inline ncclResult_t ncclCudaHostAlloc(void** ptr, void** devPtr, size_t size) {
@@ -48,4 +49,18 @@ static ncclResult_t ncclCudaMemcpy(T* dst, T* src, size_t nelem) {
   return ncclSuccess;
 }
 
+// Allocate memory to be potentially ibv_reg_mr'd. This needs to be
+// allocated on separate pages as those pages will be marked DONTFORK
+// and if they are shared, that could cause a crash in a child process
+static ncclResult_t ncclIbMalloc(void** ptr, size_t size) {
+  size_t page_size = sysconf(_SC_PAGESIZE);
+  void* p;
+  int size_aligned = ROUNDUP(size, page_size);
+  int ret = posix_memalign(&p, page_size, size_aligned);
+  if (ret != 0) return ncclSystemError;
+  memset(p, 0, size);
+  *ptr = p;
+  return ncclSuccess;
+}
+
 #endif