Enable Direct Use of Arm Compute Library (ACL) in ATen #148584
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/148584
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 5e27a25 with merge base 6c3492b ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@pytorchbot label "module: arm" |
pytorch-bot
bot
added
the
module: arm
|
@pytorchbot label "topic: not user facing" |
|
@pytorchbot label "ciflow/linux-aarch64" |
| @@ -0,0 +1,62 @@ | |||
| # Copied from: https://github.com/oneapi-src/oneDNN/blob/main/cmake/FindACL.cmake | |||
| # ---------- | |||
| # FindACL | |||
There was a problem hiding this comment.
@fadara01 what is the provenance of this file? Was it written for PyTorch or copied from someplace else?
If for PyTorch can it use modern cmake paradigms of targets rather that old style global variables definition
There was a problem hiding this comment.
It was copied from oneDNN as the you can see in the header:
# Copied from: https://github.com/oneapi-src/oneDNN/blob/main/cmake/FindACL.cmake
malfet
left a comment
malfet
left a comment
There was a problem hiding this comment.
Would be nice to have a more modern .cmake file but as it is a verbatim copy then LGTM as is
|
@pytorchbot merge -f "Lint + ARM builds are green" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
ACL is already built with PyTorch as a shared library when USE_MKLDNN_ACL is set. Currently, it is only used indirectly in ATen via oneDNN for AArch64 targets. However there are cases where it makes sense to utilize ACL directly without oneDNN as an intermediary - e.g. quantization. See #145942, #147337, #146620. This patch enables such use cases by exposing ACL to ATen ghstack-source-id: 266c621 Pull Request resolved: pytorch/pytorch#148584
ACL is already built with PyTorch as a shared library when USE_MKLDNN_ACL is set. Currently, it is only used indirectly in ATen via oneDNN for AArch64 targets. However there are cases where it makes sense to utilize ACL directly without oneDNN as an intermediary - e.g. quantization. See #145942, #147337, #146620. This patch enables such use cases by exposing ACL to ATen ghstack-source-id: 1578e06 Pull Request resolved: pytorch/pytorch#148584
Stack from ghstack (oldest at bottom):
ACL is already built with PyTorch as a shared library when USE_MKLDNN_ACL is set.
Currently, it is only used indirectly in ATen via oneDNN for AArch64 targets. However there are cases where it makes sense to utilize ACL directly without oneDNN as an intermediary - e.g. quantization. See #145942, #147337, #146620.
This patch enables such use cases by exposing ACL to ATen
cc @malfet @snadampal @milpuz01