Arm64 support as documented here: … #5360

ArielleA · 2018-06-10T20:31:41Z

http://zhiyisun.github.io/2017/02/15/Running-Google-Machine-Learning-Library-Tensorflow-On-ARM-64-bit-Platform.html

These instructions are a little out of date so I have re-biased the patch.
Not included in this patch are changes to the iOS and Android rulebases.

…5/Running-Google-Machine-Learning-Library-Tensorflow-On-ARM-64-bit-Platform.html

googlebot · 2018-06-10T20:31:45Z

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project (if not, look below for help). Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed (or fixed any issues), please reply here (e.g. I signed it!) and we'll verify it.

What to do if you already signed the CLA

Individual signers

It's possible we don't have your GitHub username or you're using a different email address on your commit. Check your existing CLA data and verify that your email is set on your git commits.

Corporate signers

Your company has a Point of Contact who decides which employees are authorized to participate. Ask your POC to be added to the group of authorized contributors. If you don't know who your Point of Contact is, direct the Google project maintainer to go/cla#troubleshoot (Public version).
The email used to register you as an authorized contributor must be the email used for the Git commit. Check your existing CLA data and verify that your email is set on your git commits.
The email used to register you as an authorized contributor must also be attached to your GitHub account.

ArielleA · 2018-06-10T20:34:44Z

I signed it

googlebot · 2018-06-10T20:34:46Z

CLAs look good, thanks!

laszlocsomor · 2018-06-11T11:25:49Z

What's the motivation of this PR? Is there a corresponding GitHub issue?

ArielleA · 2018-06-11T12:04:55Z

My main workstation is a Samsung Chromebook Plus running Debian Stretch, which is arm64 architecture. The patch is allow me to build my project locally using Bazel.
Also as Tensorflow is quite widely used, adding support for Arm64 would be useful especially as SOCs with accelerators are becoming available.
There is a related issue is bazel-contrib/rules_go#1506

laszlocsomor · 2018-06-13T06:01:04Z

@ulfjack: ping

futurejones · 2018-06-13T08:43:07Z

I have been testing Bazel on aarch64 systems with these changes.
https://github.com/futurejones/bazel/tree/futurejones-aarch64-patch
(Obtained from the same source as above.)
http://zhiyisun.github.io/2017/02/15/Running-Google-Machine-Learning-Library-Tensorflow-On-ARM-64-bit-Platform.html
I can confirm successful builds and use of Bazel on a variety of arm64/aarch64 machines.

ulfjack · 2018-06-15T11:17:26Z

tools/platforms/platforms.BUILD

@@ -4,6 +4,18 @@ package(
    default_visibility = ["//visibility:public"],
 )

+filegroup(


What is this needed for?

ArielleA · 2018-06-15T13:03:20Z

Sorry, that was my error, those lines should not have been there.

ulfjack · 2018-06-15T14:38:59Z

@dslomov

steeve · 2018-06-15T22:38:21Z

tools/platforms/BUILD

@@ -107,6 +112,7 @@ platform(
        ":x86_64",
        ":ppc",
        ":arm",
+	":aarch64",


indentation is slightly wrong

steeve · 2018-06-15T22:38:27Z

tools/platforms/BUILD

@@ -128,6 +134,7 @@ platform(
        ":x86_64",
        ":ppc",
        ":arm",
+	":aarch64",


indentation is slightly wrong

steeve · 2018-06-15T22:38:37Z

tools/platforms/platforms.BUILD

@@ -95,6 +100,7 @@ platform(
        ":x86_64",
        ":ppc",
        ":arm",
+	":aarch64",


indentation is slightly wrong

steeve · 2018-06-15T22:38:41Z

tools/platforms/platforms.BUILD

@@ -116,6 +122,7 @@ platform(
        ":x86_64",
        ":ppc",
        ":arm",
+	":aarch64",


indentation is slightly wrong

ArielleA · 2018-06-18T05:47:33Z

Indentation fixed thankyou :)

Fixes bazelbuild#5272 Change-Id: Ia49c084f1fe036ff0e5b18fea096f31642cf8b5b Closes bazelbuild#5280. Change-Id: Ia49c084f1fe036ff0e5b18fea096f31642cf8b5b PiperOrigin-RevId: 200005785

- use an internal event wrapper to unify handling - rewrite the test to be more explicit about event names and ordering This a part split out of unknown commit. PiperOrigin-RevId: 200015904

Instead, add some simple APIs to ExecutorBuilder and inline all the previous subclasses into their corresponding modules. This removes a bunch of boilerplate. PiperOrigin-RevId: 200017162

This adds a flag to the installer to skip uncompressing the base image during installation. Uncompressing the base image during installation can be problematic: 1. If you configure the startup option `output_user_root` you'll wind up extracting the image twice. Once during installation and a second time during the first bazel invocation in your project. 2. If you're bootstrapping Bazel in a controlled environment (in my case via Nix) then you can get permission errors if Bazel defaults to extracting to the home directory. There's no way to skip this extraction and for the time being I have resorted to seting `HOME` to a throw away directory during install. Closes bazelbuild#5303. PiperOrigin-RevId: 201004235

This functionality will be useful for node restarting. More than one parent node may request the restart of a shared child node, and this should not fail. Instead, the returned MarkedDirtyResult indicates whether the dirtying was redundant, and the calling code can assert on that. RELNOTES: None. PiperOrigin-RevId: 201005663

Closes bazelbuild#5403. PiperOrigin-RevId: 201007405

Change-Id: I3655868d34dd98f5cde084a61995dfae8e4b94c0 Closes bazelbuild#5374. Change-Id: I3655868d34dd98f5cde084a61995dfae8e4b94c0 PiperOrigin-RevId: 201032126

…ache. ConfiguredRuleClassProvider can specify a predicate which accepts an OptionsDiff and the new BuildOptions in order to make a decision on whether the given diff requires the analysis cache to be dropped. This predicate is only called if all of the following hold: * there was an old configuration collection (if not, the cache is not dropped because there wasn't one yet) * the old configuration collection is not exactly equal to the new configuration collection (if it is, the cache is not dropped because it definitely hasn't changed) * the old configuration collection has the same number of configurations as the new collection (if not, the cache is always dropped because experimental_multi_cpu has changed, definitely not a flag which should cause old analysis cache to stick around!) If all of these hold, the old target configurations are paired up with the new target configurations by index in the configuration collection, and each pair is diffed. The predicate is called with each diff and the corresponding new configuration's build options. If any of these invocations returns true, the cache is dropped. Otherwise, the cache is kept for the next build. No implementation of this predicate is actually supplied for this change, so the old behavior (always drop the cache if the configuration changes at all) holds. RELNOTES: None. PiperOrigin-RevId: 201050049

*** Reason for rollback *** See linked bug. *** Original change description *** C++: Refactors PyWrapCc to make it easier to migrate to Skylark Rolling forward bazelbuild@6c87715. When I first tried to roll forward this CL I missed one place where PyCcLinkParamsProvider.TO_LINK_PARAMS should have been called. The target //production/midas/config:client_config_pb builds fine now. RELNOTES:none PiperOrigin-RevId: 201070354

The memory savings from this flag are not worth the complexity, and it interferes with action restarting. RELNOTES: Remove support for --discard_actions_after_execution. PiperOrigin-RevId: 201077905

…xception. We can construct the exception message eagerly, since it is going to be thrown with very high likelihood in any case. The complex objects inside it made it hard to serialize. PiperOrigin-RevId: 201130522

…ation as this isn't supported by `((... == 0))` below. See bazelbuild#5413 PiperOrigin-RevId: 201135798

PiperOrigin-RevId: 201144030

…ionConfig that are needed for FeatureConfiguration creation In the future CcToolchainFeatures will not be created from a CToolchain but from a CrosstoolInfo provider that will encapsule all relevant information from the CROSSTOOL. Feature and ActionConfig classes need to carry all the information that is currently obtained from CToolchain. Work towards issue bazelbuild#5380 RELNOTES: None. PiperOrigin-RevId: 201147336

Rolls forward bazelbuild@c720152. This CL includes testing to make sure that linker parameters coming from every possible provider in Python rules is propagated. RELNOTES:none PiperOrigin-RevId: 201160720

For paths passed to Bazel on the command line, the shell expands these variables, but for hardcoded defaults, we must make the library call ourselves. We do not add similar support in Posix systems, where it is less common to rely on standard path-related variables. Prerequisit for issue bazelbuild#4502. RELNOTES: None. PiperOrigin-RevId: 201183214

…'t expect equality, not because there's no implementation. PiperOrigin-RevId: 201191262

… tag lambdas as Serializable. PiperOrigin-RevId: 201191461

…t be checked for restricted_to and compatible_with constraints, because they are part of the execution, not providing new dependencies. PiperOrigin-RevId: 201196891

…mbdas. Get rid of a useless tag, because the Function being tagged is a concrete class, so won't be serializable. Will deal with it in a follow-up. Implement equality for BazelInfo. PiperOrigin-RevId: 201199255

PiperOrigin-RevId: 201203706

PiperOrigin-RevId: 201205388

…ion, and have DynamicCodec add to it. It's very useful to have this trail available when debugging or whitelisting. PiperOrigin-RevId: 201205884

Turns out the event handler (BlazeCommandEventHandler) prints almost all event types, and I don't believe there's a way to tune it. We certainly don't want these messages printed to the console unless we're debugging the debugger, so turn them off by default. PiperOrigin-RevId: 201211355

… by default. PiperOrigin-RevId: 201218341

…ER_TARGET but the "exec_compatible_with" attribute is present. PiperOrigin-RevId: 201219412

When set, any action parameter files are written locally upon action execution, even when the action is executed remotely. This is mainly useful for debugging. This option is effectively implied by --subcommands and --verbose_failures, as it is likely that the user is debugging actions when using these flags. RELNOTES: Add --materialize_param_files flag to write parameter files even when actions are executed remotely. PiperOrigin-RevId: 201225566

googlebot · 2018-06-19T20:06:31Z

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 have the pull request author add another comment and the bot will run again. If the bot doesn't comment, it means it doesn't think anything has changed.

Based on work of Sun Zhiyi <[email protected]>. Closes bazelbuild#5360. PiperOrigin-RevId: 201339882

Arm64 support as documented here: http://zhiyisun.github.io/2017/02/1…

cc46dac

…5/Running-Google-Machine-Learning-Library-Tensorflow-On-ARM-64-bit-Platform.html

googlebot added the cla: no label Jun 10, 2018

googlebot added cla: yes and removed cla: no labels Jun 10, 2018

laszlocsomor self-assigned this Jun 11, 2018

laszlocsomor requested a review from ulfjack June 11, 2018 12:33

laszlocsomor assigned ulfjack and unassigned laszlocsomor Jun 11, 2018

Add in platform export definitions

1d4a366

ArielleA mentioned this pull request Jun 14, 2018

Arm64 support on Linux bazel-contrib/rules_go#1550

Merged

ulfjack reviewed Jun 15, 2018

View reviewed changes

Fix error pointed out by ulfjack. Thankyou :)

710aa55

steeve reviewed Jun 15, 2018

View reviewed changes

Fix indentation

878bcf8

laszlocsomor and others added 3 commits June 19, 2018 22:05

docs,windows: emphasize that BAZEL_VS != BAZEL_VC

469530d

Fixes bazelbuild#5272 Change-Id: Ia49c084f1fe036ff0e5b18fea096f31642cf8b5b Closes bazelbuild#5280. Change-Id: Ia49c084f1fe036ff0e5b18fea096f31642cf8b5b PiperOrigin-RevId: 200005785

Refactor the build event service transport

4c448de

- use an internal event wrapper to unify handling - rewrite the test to be more explicit about event names and ordering This a part split out of unknown commit. PiperOrigin-RevId: 200015904

Remove ActionContextConsumer

f11165f

Instead, add some simple APIs to ExecutorBuilder and inline all the previous subclasses into their corresponding modules. This removes a bunch of boilerplate. PiperOrigin-RevId: 200017162

andyscott and others added 24 commits June 19, 2018 22:05

Make the embedded JDK targets the default

1311886

Closes bazelbuild#5403. PiperOrigin-RevId: 201007405

Clean up uses of SkylarkList<String> and ast in toolchains parameters.

325ab39

Change-Id: I3655868d34dd98f5cde084a61995dfae8e4b94c0 Closes bazelbuild#5374. Change-Id: I3655868d34dd98f5cde084a61995dfae8e4b94c0 PiperOrigin-RevId: 201032126

Remove support for --discard_actions_after_execution.

3659e90

The memory savings from this flag are not worth the complexity, and it interferes with action restarting. RELNOTES: Remove support for --discard_actions_after_execution. PiperOrigin-RevId: 201077905

Ensure seq never formats output with decimals or using scientific not…

78b6563

…ation as this isn't supported by `((... == 0))` below. See bazelbuild#5413 PiperOrigin-RevId: 201135798

do not start server if shutdown requested (+ not in batch mode)

2ae255e

PiperOrigin-RevId: 201144030

C++: Refactors PyWrapCc to make it easier to migrate to Skylark

f0fa03d

Rolls forward bazelbuild@c720152. This CL includes testing to make sure that linker parameters coming from every possible provider in Python rules is propagated. RELNOTES:none PiperOrigin-RevId: 201160720

Clarify that ErrorInfoValue is a NotComparableSkyValue because we don…

39b85a6

…'t expect equality, not because there's no implementation. PiperOrigin-RevId: 201191262

Make some Apple objects more serializable: tag constants as constant,…

d62d82a

… tag lambdas as Serializable. PiperOrigin-RevId: 201191461

Make variable providers coming from the toolchains attribute shouldn'…

9c0fc84

…t be checked for restricted_to and compatible_with constraints, because they are part of the execution, not providing new dependencies. PiperOrigin-RevId: 201196891

Make things easier to serialize: make constants constant, tag some la…

92135e2

…mbdas. Get rid of a useless tag, because the Function being tagged is a concrete class, so won't be serializable. Will deal with it in a follow-up. Implement equality for BazelInfo. PiperOrigin-RevId: 201199255

Let MessageLiteCodec handle UnknownFieldSet.

57ca973

PiperOrigin-RevId: 201203706

Implement #equals and #hashCode for TargetPatternPhaseValue.

0a5cc43

PiperOrigin-RevId: 201205388

Put a trail into every SerializationException, not just NoCodecExcept…

6abcf66

…ion, and have DynamicCodec add to it. It's very useful to have this trail available when debugging or whitelisting. PiperOrigin-RevId: 201205884

Delete switch for nested set serialization. It's fast enough to be on…

1367719

… by default. PiperOrigin-RevId: 201218341

Add more logging when executionPlatformConstraintsAllowed is set to P…

74d794c

…ER_TARGET but the "exec_compatible_with" attribute is present. PiperOrigin-RevId: 201219412

googlebot added cla: no and removed cla: yes labels Jun 19, 2018

bazel-io closed this in 886d01c Jun 20, 2018

werkt pushed a commit to werkt/bazel that referenced this pull request Aug 2, 2018

Add Arm64 support

3ce3cbf

Based on work of Sun Zhiyi <[email protected]>. Closes bazelbuild#5360. PiperOrigin-RevId: 201339882

vielmetti mentioned this pull request Dec 1, 2018

ARM64 support? cert-manager/cert-manager#608

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Arm64 support as documented here: … #5360

Arm64 support as documented here: … #5360

ArielleA commented Jun 10, 2018

googlebot commented Jun 10, 2018

ArielleA commented Jun 10, 2018

googlebot commented Jun 10, 2018

laszlocsomor commented Jun 11, 2018

ArielleA commented Jun 11, 2018

laszlocsomor commented Jun 13, 2018

futurejones commented Jun 13, 2018

ulfjack Jun 15, 2018

ArielleA commented Jun 15, 2018

ulfjack commented Jun 15, 2018

steeve Jun 15, 2018

steeve Jun 15, 2018

steeve Jun 15, 2018

steeve Jun 15, 2018

ArielleA commented Jun 18, 2018

googlebot commented Jun 19, 2018

Arm64 support as documented here: … #5360

Arm64 support as documented here: … #5360

Conversation

ArielleA commented Jun 10, 2018

googlebot commented Jun 10, 2018

What to do if you already signed the CLA

Individual signers

Corporate signers

ArielleA commented Jun 10, 2018

googlebot commented Jun 10, 2018

laszlocsomor commented Jun 11, 2018

ArielleA commented Jun 11, 2018

laszlocsomor commented Jun 13, 2018

futurejones commented Jun 13, 2018

ulfjack Jun 15, 2018

Choose a reason for hiding this comment

ArielleA commented Jun 15, 2018

ulfjack commented Jun 15, 2018

steeve Jun 15, 2018

Choose a reason for hiding this comment

steeve Jun 15, 2018

Choose a reason for hiding this comment

steeve Jun 15, 2018

Choose a reason for hiding this comment

steeve Jun 15, 2018

Choose a reason for hiding this comment

ArielleA commented Jun 18, 2018

googlebot commented Jun 19, 2018