frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-21 05:34:18 +08:00
Code Issues Packages Projects Releases Wiki Activity

Updated Multiprocessing Technical Notes (markdown)

Soumith Chintala
2016-08-31 23:15:19 -04:00
parent ccc536e15b
commit 394abfcfe1
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
Multiprocessing-Technical-Notes.md

@ -25,7 +25,7 @@ Let's say process `A` wants to share memory `[M, M+D]`, i.e. starting at `M` wit
3. Then, process `A` copies over memory `[M, M+D]` to `[SHM, SHM+D]`. 3. Then, process `A` copies over memory `[M, M+D]` to `[SHM, SHM+D]`.
4. Then, process `A` swaps all pointer references of `M` to be `SHM`. There is no straight forward way to do this in general. 4. Then, process `A` swaps all pointer references of `M` to be `SHM`. There is no straight forward way to do this in general.
For (4), Torch has a Tensor / Storage abstraction. We can simply swap the data pointer inside the Storage and all Tensors referring to this Storage do not notice anything changing. For (5), Torch has a Tensor / Storage abstraction. We can simply swap the data pointer inside the Storage and all Tensors referring to this Storage do not notice anything changing.
Now, `A` communicates the file path of the shared memory along with the size to `B`, i.e. `("/dev/shm/torch_shmfile_1", D)`. Now, `A` communicates the file path of the shared memory along with the size to `B`, i.e. `("/dev/shm/torch_shmfile_1", D)`.