Skip to content

Commit 1eb40a8

Browse files
update
1 parent 267e746 commit 1eb40a8

File tree

1 file changed

+5
-3
lines changed

1 file changed

+5
-3
lines changed

docs/software/communication/cray-mpich.md

Lines changed: 5 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -81,7 +81,9 @@ Performance may be negatively affected by this option.
8181

8282
#### `"cxil_map: write error"` when doing inter-node GPU-aware MPI communication
8383

84-
The following environment variable can be set to disable gdrcopy:
84+
This error message is sometimes triggered by applications that use GPU Direct MPI calls when they trigger a bug in gdrcopy (a low-level library used to copy buffers between GPUs).
85+
Setting the following option will completely disable gdrcopy.
86+
Note that this has a performance impact for small message sizes, so it should only be enabled on a case-by-case basis.
8587
```bash
8688
export FI_CXI_SAFE_DEVMEM_COPY_THRESHOLD=0
8789
```
@@ -90,8 +92,8 @@ export FI_CXI_SAFE_DEVMEM_COPY_THRESHOLD=0
9092

9193
#### `"cxil_map: write error"` when doing inter-node GPU-aware MPI communication
9294

93-
??? info "The issue has been resolved on the 7th of October 2024 with a system update"
94-
The issue was caused by a system misconfiguration.
95+
??? info "The issue has been resolved on the 7th of October 2024 with a system
96+
update" The issue was caused by a system misconfiguration.
9597

9698
When doing inter-node GPU-aware communication with Cray MPICH after the update on the 30th of September 2024 on Alps, applications will fail with:
9799
```bash

0 commit comments

Comments
 (0)