Update of --config=mkl_aarch64 with Compute Library support

[alenik01]

This PR introduces the support of Compute Library for the Arm Architecture (ACL) as part of Bazel build for mkl_aarch64 build config. The code was developed together with @joeramsay
The PR is related to

• previously raised issue #47415
• PRs [INTEL MKL] Enabling TF + OneDNN in stock TF build #47745 and [Intel MKL] Make Vanilla TF and MKL-based TF use common oneDNN build #47679

[alenik01]

@agramesh1 @gzmkl please have a look.

[gbaned]

@alenik01 Can you please resolve conflicts? Thanks!

[google-cla]

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

[alenik01]

@gbaned, GitHub GUI for the conflict resolution fails to pass google-cla bot for some reason. Would force push be the best way to proceed here?

[penpornk]

@alenik01 Please don't force push.

@google-cla I have verified that the commits are from the same user (@alenik01). Manually setting CLA to yes.

[agramesh1]

Thanks @alenik01 for the PR.
Also some of the files have missing newlines, can you fix those.

[gbaned]

@alenik01 Can you please check @agramesh1's comments and keep us posted ? Thanks!

[penpornk]

Thank you for the PR! I have minor comments.

[penpornk]

Nit: Please add a new line at the end of file. Same for other files.

[penpornk]

Since this PR doesn't seem to need any more significant changes, I'm going to approve this now and make the remaining minor fixes myself to save time. Thank you again for the PR!

[alenik01]

@penpornk, thanks. My main concern is how this PR stacks with recently merged PR #47745. If the functionality will be broken, then what way would be the best to proceed, open another PR for the fix?

[alenik01]

@penpornk, arm_any leads to the following conflict during the build:

ERROR: Analysis of target '//tensorflow/tools/pip_package:build_pip_package' failed; build aborted: /tensorflow/tools/pip_package/BUILD:179:10: Illegal ambiguous match on configurable attribute "data" in //tensorflow/tools/pip_package:licenses:
//third_party/mkl:build_with_mkl_aarch64
//tensorflow:arm_any

thus I removed it. May you please advise on the possible fix of "Illegal ambiguous match on configurable attribute"? The only new LICENSE file which this PR adds is in ./third_party/compute_library/, and I don't see from where this "ambiguous match" arises.

Have you tested with all targets that //tensorflow:arm_any covers?

The tests were performed with linux_aarch64 and arm64-v8a.

[alenik01]

@penpornk, the new fix 37c1f29 actually excludes arm_any from if_mkl and mkl_deps.

Could you please help resolve merge conflicts again?

I am now building TF to test.

[google-cla]

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

[alenik01]

Could you please help resolve merge conflicts again?

Done.

[penpornk]

Thank you very much!

[agramesh1]

Hi @penpornk sorry for the confusion. This commit 37c1f29 broke the oneDNN builds due to ambiguous definitions of
//third_party/mkl:build_with_mkl
//tensorflow:linux_x86_64

We verified that this change fixes it. Let me know if you can make the change or should we submit a PR. Thanks.

diff --git a/third_party/mkl/build_defs.bzl b/third_party/mkl/build_defs.bzl
index d913f22a110..7a4f08ed6a3 100644
--- a/third_party/mkl/build_defs.bzl
+++ b/third_party/mkl/build_defs.bzl
@@ -32,7 +32,6 @@ def if_mkl(if_true, if_false = []):
       may need it. It may be deleted in future with refactoring.
     """
     return select({
-        "@org_tensorflow//third_party/mkl:build_with_mkl": if_true,
         "//tensorflow:linux_x86_64": if_true,
         "//tensorflow:windows": if_true,
         "//conditions:default": if_false,
@@ -101,7 +100,6 @@ def mkl_deps():
       inclusion in the deps attribute of rules.
     """
     return select({
-        "@org_tensorflow//third_party/mkl:build_with_mkl": ["@mkl_dnn_v1//:mkl_dnn"],
         "@org_tensorflow//third_party/mkl:build_with_mkl_aarch64": ["@mkl_dnn_v1//:mkl_dnn_aarch64"],
         "//tensorflow:linux_x86_64": ["@mkl_dnn_v1//:mkl_dnn"],
         "//tensorflow:windows": ["@mkl_dnn_v1//:mkl_dnn"],

[penpornk]

@agramesh1 Thank you for the patch! I'll make the change.

@googlebot I've verified that all the commits are from the same github user. Manually setting CLA to yes.

[google-cla]

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

[penpornk]

The fix is in. 5eabb16

No need to resolve merge conflicts again. I've already pulled this PR in and running tests internally.

[cfRod]

Thanks everyone for sorting this out!