ARMC6: Add a build profile extension with the link-time optimizer enabled #11874

fkjagodzinski · 2019-11-15T14:16:52Z

Description (required)

Summary of change (What the change is for and why)

Added an lto build profile extension for ARMC6 toolchain.

added -flto to common flags,
added --lto & --lto_level=Oz to ld flags.

Results

Product: ARM Compiler 6.11 Professional Component: ARM Compiler 6.11 Tool: armclang [5d3b4200]

[email protected]

Mbed OS branch	commit SHA	RAM (data + bss)	Flash (text + data)	build time	details
mbed-os-5.14.1	`679d248`	206047	39168	739.53	blinky-679d248.txt
armc6_build-enable_lto_for_release	`c6a655f`	205848	36601	610.81	blinky-c6a655f.txt

[email protected]

Mbed OS branch	commit SHA	RAM (data + bss)	Flash (text + data)	build time	details
mbed-os-5.14.1	`679d248`	252457	361038	828.68	ccexample-679d248.txt
armc6_build-enable_lto_for_release	`c6a655f`	251580	329505	835.21	ccexample-c6a655f.txt

Build commands used to produce the above results

mbed-os-example-blinky

cd $(mktemp -d)
git clone [email protected]:ARMmbed/mbed-os-example-blinky.git
cd mbed-os-example-blinky/
git checkout mbed-os-5.14.1
git clean -dxff && git reset --hard

# Define Mbed OS rev
# 1. mbed-os-5.14.1
echo 'https://github.com/ARMmbed/mbed-os/#679d24833acf0a0b5b0d528576bb37c70863bc4e' > mbed-os.lib
# 2. LTO branch on top of mbed-os-5.14.1
echo 'https://github.com/fkjagodzinski/mbed-os/#d7df3e673144232d3e162de19ec13f0472d725c6' > mbed-os.lib

mbed deploy
cd mbed-os/ && git log --oneline -1 && cd ..
time -p mbed compile -t ARM -m K64F --profile mbed-os/tools/profiles/release.json

# More details
mbed compile -t ARM -m K64F --profile mbed-os/tools/profiles/release.json --stats-depth=100

mbed-cloud-client-example

cd $(mktemp -d)
git clone [email protected]:ARMmbed/mbed-cloud-client-example.git
cd mbed-cloud-client-example/
git checkout 4.0.0 
git clean -dxff && git reset --hard
cp ../mbed_cloud_dev_credentials.c .

# Define Mbed OS rev
# 1. mbed-os-5.14.1
echo 'https://github.com/ARMmbed/mbed-os/#679d24833acf0a0b5b0d528576bb37c70863bc4e' > mbed-os.lib
# 2. LTO branch on top of mbed-os-5.14.1
echo 'https://github.com/fkjagodzinski/mbed-os/#d7df3e673144232d3e162de19ec13f0472d725c6' > mbed-os.lib

mbed deploy
cd mbed-os/ && git log --oneline -1 && cd ..
time -p mbed compile -t ARM -m K64F --profile mbed-os/tools/profiles/release.json

# More details
mbed compile -t ARM -m K64F --profile mbed-os/tools/profiles/release.json --stats-depth=100

Documentation (Details of any document updates required)

Pull request type (required)

[] Patch update (Bug fix / Target update / Docs update / Test update / Refactor)
[x] Feature update (New feature / Functionality change / New API)
[] Major update (Breaking change E.g. Return code change / API behaviour change)

Test results (required)

[] No Tests required for this change (E.g docs only update)
[x] Covered by existing mbed-os tests (Greentea or Unittest)
[] Tests / results supplied as part of this PR

Reviewers (optional)

@jamesbeyond @kjbracey-arm @maciejbocianski

Release Notes (required for feature/major PRs)

Summary of changes

Add a build profile extension with the link-time optimizer enabled for ARMC6 toolchain.

Impact of changes

Migration actions required

ciarmcom · 2019-11-15T16:00:27Z

@fkjagodzinski, thank you for your changes.
@maciejbocianski @jamesbeyond @kjbracey-arm @ARMmbed/mbed-os-tools @ARMmbed/mbed-os-maintainers please review.

bulislaw

Should we also do it for develop?

jamesbeyond · 2019-11-22T11:47:53Z

based on the original plan this is for 5.15. So now is this 5.15 or 6.0 @bulislaw @0xc0170

fkjagodzinski · 2019-11-22T13:09:54Z

[email protected], ARMC6, develop profile, Windows host

Mbed OS branch	commit SHA	RAM (data + bss)	Flash (text + data)	build time	rebuild time
mbed-os-5.14.1	`679d248`	252165	406682	875.66	26.50
armc6_build-enable_lto	`d7df3e6`	251709	390132	917.86	37.90

0xc0170 · 2019-11-22T13:42:00Z

based on the original plan this is for 5.15. So now is this 5.15 or 6.0 @bulislaw @0xc0170

I don't recall seeing a move to 6.0 for this one. @bulislaw ?

I would mark this as Major update as it might have consequences to apps .

0xc0170 · 2019-11-22T15:07:38Z

Set to 5.15 as agreed.

I'll trigger the first CI run

mbed-ci · 2019-11-22T16:10:38Z

Test run: FAILED

Summary: 3 of 4 test jobs failed
Build number : 1
Build artifacts

Failed test jobs:

jenkins-ci/mbed-os-ci_build-IAR
jenkins-ci/mbed-os-ci_build-GCC_ARM
jenkins-ci/mbed-os-ci_build-ARM

jamesbeyond · 2019-11-24T21:46:40Z

Seems this PR only touched ARMC6 options, not sure why IAR and GCC failed. I'll try to restart CI

mbed-ci · 2019-11-24T22:43:58Z

Test run: FAILED

Summary: 3 of 4 test jobs failed
Build number : 2
Build artifacts

Failed test jobs:

jenkins-ci/mbed-os-ci_build-IAR
jenkins-ci/mbed-os-ci_build-GCC_ARM
jenkins-ci/mbed-os-ci_build-ARM

0xc0170 · 2019-11-25T08:00:51Z

The artifacts report green for GCC and IAR. just focus on failures in ARM:

Error: L6915E: Library reports error: Heap was used, but no heap region was defined
Finished: 0 information, 1 warning and 1 error messages.

@fkjagodzinski Please review

maciejbocianski · 2019-11-26T14:05:38Z

The artifacts report green for GCC and IAR. just focus on failures in ARM:
Error: L6915E: Library reports error: Heap was used, but no heap region was defined
Finished: 0 information, 1 warning and 1 error messages.
@fkjagodzinski Please review

This error is due to lack of ARM_LIB_HEAP section in scatter file

adbridge · 2019-11-27T11:21:05Z

@kjbracey-arm if the GCC version of this PR has been marked as a breaking change is this one different ?

adbridge · 2019-11-27T11:37:21Z

@fkjagodzinski similarly to the GCC equivalent changes if the user has to do anything to use these changes (such as adding an additional header file), could you please fill in the impact and migration sections in the release notes accordingly ? Thanks

fkjagodzinski · 2019-11-27T11:39:55Z

Added one commit from @maciejbocianski that fixes RTC_IRQHandler linker error. Thanks Maciej! 👍

0xc0170 · 2019-11-27T11:46:05Z

CI restarted

mbed-ci · 2019-11-27T12:37:39Z

Test run: FAILED

Summary: 3 of 4 test jobs failed
Build number : 3
Build artifacts

Failed test jobs:

jenkins-ci/mbed-os-ci_build-IAR
jenkins-ci/mbed-os-ci_build-GCC_ARM
jenkins-ci/mbed-os-ci_build-ARM

kjbracey · 2019-11-27T12:54:31Z

That RTC_IRQHandler thing smells like a compiler/linker bug to me. Don't see why that should fail.

There is a similar trap for types that can hit LTO - you're not allowed to have duplicate type names in different files in C++. To avoid that you have to stick them in an anonymous namespace, which is the equivalent of making the type static/local.

But this seems legal to me. At least as long as you weren't including a header file declaring the external RTC_IRQHandler, which you're not.

maciejbocianski · 2019-11-27T14:53:15Z

Another problem which I haven't solved yet is:
L6137E: Symbol $Sub$$calloc was not preserved by the LTO codegen but is needed by the image.

Error is visible when -DMBED_HEAP_STATS_ENABLED or -DMBED_ALL_STATS_ENABLED is defined

fkjagodzinski · 2019-11-29T09:49:09Z

I've added two more fixes form https://github.com/maciejbocianski/mbed-os/commits/fix_armc6_lto. @jamesbeyond, @0xc0170 I think this is ready to pass CI now.

0xc0170 · 2019-11-29T09:55:12Z

CI started

Isn't there a better way for d71637b (using dummy value there to keep the function in) ?

Add a "used" attribute to __vector_handlers to fix ARMC6 build with the "-flto" flag. (Error: L6236E: No section matches selector - no section to be FIRST/LAST.) This attribute, attached to a function/variable, means that code must be emitted for the function even if it appears that the function is not referenced.

Add a "used" attribute to __vector_handlers to fix ARMC6 build with the "-flto" flag. (Error: L6236E: No section matches selector - no section to be FIRST/LAST. ) This attribute, attached to a function/variable, means that code must be emitted for the function even if it appears that the function is not referenced.

…LTO builds Add a "used" attribute to SVCHandler_main/tfm_pendsv_do_schedule to fix ARMC6 build with the "-flto" flag. This attribute, attached to a function/variable, means that code must be emitted for the function even if it appears that the function is not referenced.

maciejbocianski · 2020-02-04T12:49:44Z

rebase done

0xc0170 · 2020-02-04T13:21:57Z

CI started

mbed-ci · 2020-02-04T15:52:04Z

Test run: SUCCESS

Summary: 11 of 11 test jobs passed
Build number : 11
Build artifacts

0xc0170

Is there any docs PR for this profile extension?

0xc0170 · 2020-02-05T10:28:35Z

@bulislaw @kjbracey-arm Please review the final version, it is be ready to go in.

bulislaw

Looks good, do we have docs following this PR?

maciejbocianski · 2020-02-05T11:09:10Z

Looks good, do we have docs following this PR?

There is no docs yet @jamesbeyond @bulislaw

jamesbeyond · 2020-02-05T11:58:59Z

There is no docs yet @jamesbeyond @bulislaw

We are prepare docs, and will be put into:
https://os.mbed.com/docs/mbed-os/v5.15/tutorials/optimizing.html

ciarmcom requested review from jamesbeyond, kjbracey, maciejbocianski and a team November 15, 2019 16:00

ciarmcom added the needs: review label Nov 15, 2019

bulislaw reviewed Nov 21, 2019

View reviewed changes

fkjagodzinski changed the title ~~ARMC6: Enable link-time optimizer for release profile~~ ARMC6: Enable link-time optimizer for release & develop profiles Nov 22, 2019

0xc0170 added the release-version: 5.15.0-rc1 label Nov 22, 2019

0xc0170 added needs: work and removed needs: review labels Nov 25, 2019

maciejbocianski added 5 commits February 4, 2020 12:30

armc6: make lto an optional profile

083e3e5

maciejbocianski force-pushed the armc6_build-enable_lto_for_release branch from cd42f67 to 083e3e5 Compare February 4, 2020 11:40

0xc0170 added needs: CI and removed needs: work labels Feb 4, 2020

jamesbeyond approved these changes Feb 4, 2020

View reviewed changes

mergify bot added ready for merge and removed needs: CI labels Feb 4, 2020

0xc0170 requested review from bulislaw and kjbracey February 5, 2020 10:20

0xc0170 approved these changes Feb 5, 2020

View reviewed changes

0xc0170 added needs: review and removed ready for merge labels Feb 5, 2020

bulislaw approved these changes Feb 5, 2020

View reviewed changes

maciejbocianski mentioned this pull request Feb 5, 2020

LTO build profile extension documentation ARMmbed/mbed-os-5-docs#1219

Merged

mark-edgeworth approved these changes Feb 5, 2020

View reviewed changes

0xc0170 added ready for merge and removed needs: review labels Feb 5, 2020

0xc0170 merged commit 32675cc into ARMmbed:master Feb 5, 2020

mergify bot removed the ready for merge label Feb 5, 2020

fkjagodzinski deleted the armc6_build-enable_lto_for_release branch February 10, 2020 09:39

ARMC6: Add a build profile extension with the link-time optimizer enabled #11874

ARMC6: Add a build profile extension with the link-time optimizer enabled #11874

Uh oh!

Conversation

fkjagodzinski commented Nov 15, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description (required)

Summary of change (What the change is for and why)

Results

[email protected]

[email protected]

Build commands used to produce the above results

Documentation (Details of any document updates required)

Pull request type (required)

Test results (required)

Reviewers (optional)

Release Notes (required for feature/major PRs)

Summary of changes

Impact of changes

Migration actions required

Uh oh!

ciarmcom commented Nov 15, 2019

Uh oh!

bulislaw left a comment

Choose a reason for hiding this comment

Uh oh!

jamesbeyond commented Nov 22, 2019

Uh oh!

fkjagodzinski commented Nov 22, 2019

[email protected], ARMC6, develop profile, Windows host

Uh oh!

0xc0170 commented Nov 22, 2019

Uh oh!

0xc0170 commented Nov 22, 2019

Uh oh!

mbed-ci commented Nov 22, 2019

Test run: FAILED

Uh oh!

jamesbeyond commented Nov 24, 2019

Uh oh!

mbed-ci commented Nov 24, 2019

Test run: FAILED

Uh oh!

0xc0170 commented Nov 25, 2019

Uh oh!

maciejbocianski commented Nov 26, 2019

Uh oh!

adbridge commented Nov 27, 2019

Uh oh!

adbridge commented Nov 27, 2019

Uh oh!

fkjagodzinski commented Nov 27, 2019

Uh oh!

0xc0170 commented Nov 27, 2019

Uh oh!

mbed-ci commented Nov 27, 2019

Test run: FAILED

Uh oh!

kjbracey commented Nov 27, 2019

Uh oh!

maciejbocianski commented Nov 27, 2019

Uh oh!

fkjagodzinski commented Nov 29, 2019

Uh oh!

0xc0170 commented Nov 29, 2019

Uh oh!

maciejbocianski commented Feb 4, 2020

Uh oh!

0xc0170 commented Feb 4, 2020

Uh oh!

mbed-ci commented Feb 4, 2020

Test run: SUCCESS

Uh oh!

0xc0170 left a comment

Choose a reason for hiding this comment

Uh oh!

0xc0170 commented Feb 5, 2020

Uh oh!

bulislaw left a comment

Choose a reason for hiding this comment

fkjagodzinski commented Nov 15, 2019 •

edited

Loading