Put system to sleep when going idle #3566

betzw · 2017-01-11T10:48:46Z

Description

While waiting for tickless mode being implemented & stable, put system to sleep when going idle.

Status

READY

Migrations

NO

Related PRs

#3607
#3602
#3693

bulislaw · 2017-01-11T11:40:09Z

I don't think that will give us much power savings, due to systick set to 1ms by default for RTX. On the other hand it shouldn't cause us any issues so might be worth doing anyway (on top of which we can build a tickless solution). We should actually measure that so we have some hard points rather than 'I think'.

bulislaw · 2017-01-11T11:40:19Z

/morph test

betzw · 2017-01-11T12:32:36Z

Any idea why continuous-integration/jenkins/pr-head fails. On my side the link fails to open with a timeout.

mbed-bot · 2017-01-11T12:34:09Z

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph test

Output

mbed Build Number: 1380

All builds and test passed!

0xc0170 · 2017-01-11T13:35:43Z

Any idea why continuous-integration/jenkins/pr-head fails. On my side the link fails to open with a timeout.

Just restarted. Reason was not able to allocate resources.

betzw · 2017-01-11T13:36:19Z

We have performed a little test on a NUCLEO_F401RE with the following application:

#include "mbed.h"

int main() {
    printf("Start waiting forever ...\r\n");

    while(1) { // wait forever
	wait(24*60); // wait a day
    }
}

Compiling this program with mbed compile the MCU consumes 18.71mA.
Compiling this program with mbed compile -DNDEBUG the MCU consumes 5.84mA.

bulislaw · 2017-01-11T13:48:53Z

I just thought that the test run didn't test much as we don't run it with NDEBUG, we use the default profile. @bridadan is there a way of triggering a NDEBUG run?

betzw · 2017-01-11T14:31:31Z

Well, I think compiling with mbed compile -DNDEBUG will define the macro NDEBUG in whatever profile you use, isn't it?

betzw · 2017-01-11T14:34:06Z

And using option -vv seems to confirm my hypothesis.

bulislaw · 2017-01-11T14:37:00Z

Sure, but I want to run the tests on CI with that option so we can see how other platforms behave.

betzw · 2017-01-11T14:40:25Z

What about the small.json profile?

bridadan · 2017-01-11T16:31:23Z

@bulislaw There currently isn't a way to change the profile for CI tests.

PR #3019 tried to add this as well but it was closed. @sg- or @0xc0170 do you have concerns over this?

betzw · 2017-01-12T05:55:18Z

Just in order to underline that in contrast to #3019 I added both #ifdef NDEBUG and #if DEVICE_SLEEP around the call to sleep()!
Pls. consider also that this would help a bit in contradicting the rumour I heard from customers that mbed-os wouldn't be power efficient ...

bulislaw · 2017-01-12T09:28:26Z

Could you submit temporary change, like that:

diff --git a/tools/profiles/default.json b/tools/profiles/default.json
index d4755e9..1854374 100644
--- a/tools/profiles/default.json
+++ b/tools/profiles/default.json
@@ -5,7 +5,7 @@
                    "-fmessage-length=0", "-fno-exceptions", "-fno-builtin",
                    "-ffunction-sections", "-fdata-sections", "-funsigned-char",
                    "-MMD", "-fno-delete-null-pointer-checks",
-                   "-fomit-frame-pointer", "-Os"],
+                   "-fomit-frame-pointer", "-Os", "-DNDEBUG"],
         "asm": ["-x", "assembler-with-cpp"],
         "c": ["-std=gnu99"],
         "cxx": ["-std=gnu++98", "-fno-rtti", "-Wvla"],
@@ -14,19 +14,19 @@
                "-Wl,--wrap,_calloc_r", "-Wl,--wrap,exit", "-Wl,--wrap,atexit"]
     },
     "ARM": {
-        "common": ["-c", "--gnu", "-Otime", "--split_sections",
+        "common": ["-c", "--gnu", "-Ospace", "--split_sections",
                    "--apcs=interwork", "--brief_diagnostics", "--restrict",
-                   "--multibyte_chars", "-O3"],
+                   "--multibyte_chars", "-O3", "-DNDEBUG"],
         "asm": [],
         "c": ["--md", "--no_depend_system_headers", "--c99", "-D__ASSERT_MSG"],
         "cxx": ["--cpp", "--no_rtti", "--no_vla"],
         "ld": []
     },
     "uARM": {
-        "common": ["-c", "--gnu", "-Otime", "--split_sections",
+        "common": ["-c", "--gnu", "-Ospace", "--split_sections",
                    "--apcs=interwork", "--brief_diagnostics", "--restrict",
                    "--multibyte_chars", "-O3", "-D__MICROLIB",
-                   "--library_type=microlib", "-DMBED_RTOS_SINGLE_THREAD"],
+                   "--library_type=microlib", "-DMBED_RTOS_SINGLE_THREAD", "-DNDEBUG"],
         "asm": [],
         "c": ["--md", "--no_depend_system_headers", "--c99", "-D__ASSERT_MSG"],
         "cxx": ["--cpp", "--no_rtti", "--no_vla"],
@@ -35,7 +35,7 @@
     "IAR": {
         "common": [
             "--no_wrap_diagnostics", "-e",
-            "--diag_suppress=Pa050,Pa084,Pa093,Pa082", "-Oh"],
+            "--diag_suppress=Pa050,Pa084,Pa093,Pa082", "-Ohz", "-DNDEBUG"],
         "asm": [],
         "c": ["--vla"],
         "cxx": ["--guard_calls", "--no_static_destruction"],

Which works around our CIs issues and would use the small profile instead of default one. We could kick off the run and revert the change after. That should give us some more information on how it works across multiple targets.

betzw · 2017-01-12T09:38:30Z

Done!

bulislaw · 2017-01-12T09:40:27Z

Thanks!

/morph test-nightly

mbed-bot · 2017-01-12T13:01:30Z

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph test-nightly

Output

mbed Build Number: 1385

All builds and test passed!

bulislaw · 2017-01-12T13:29:43Z

@betzw thanks, we can get rid of this test commit now.

@sg- what do you think about that?

betzw · 2017-01-12T14:15:10Z

Will do it tomorrow morning!

adbridge · 2017-01-12T15:39:48Z

retest uvisor

bridadan · 2017-01-12T16:11:24Z

Maintainers: I'd like @sg- and/or @0xc0170 to review this first before this comes in.

Also, as a developer, I would be surprised to see my device going to sleep just by enabling NDEBUG. It may be a welcome surprise in some cases, but I would be worried to see my device's power consumption dropping significantly with me only enabling NDEBUG.

If we are set on enabling this with NDEBUG, can this please be documented somewhere?

I think it might be more fitting to make this a config setting.

bulislaw · 2017-01-12T16:14:44Z

Sure documenting would be great, any ideas what would be the right place to do it?

Also the uvisor tests need to be rerun, when we revert the change to build with 'small' profile. I'll put a 'do not merge' label so that happens before merging.

bridadan · 2017-01-12T16:26:32Z

I'd suggest placing it in the build profiles documentation: https://github.com/ARMmbed/mbed-os/blob/master/docs/build_profiles.md#small-profile

betzw · 2017-01-13T09:03:23Z

Reverted PR back, i.e. CIs use default profile again.

bulislaw · 2017-01-13T11:33:28Z

@0xc0170 what do you think?
@betzw could you add to the docs, mentioned above, that the platform will go to sleep on idle (between systicks) please.

betzw · 2017-01-13T12:12:56Z

You can find a proposal for a change in documentation here.

betzw · 2017-01-13T12:36:36Z

@bulislaw Have you seen this? If you want I can merge it in!

sg- · 2017-01-18T10:42:28Z

resolves #2769

betzw · 2017-01-19T11:04:51Z

Requires PR #3607!!!

bulislaw · 2017-01-26T07:46:50Z

retest uvisor

bulislaw · 2017-01-26T07:54:05Z

@mazimkhan it seems I don't have the runes to do the retest uvisor could you kick off the run and add me to the acl please.

mazimkhan · 2017-01-26T08:38:34Z

retest uvisor

betzw · 2017-01-27T05:36:22Z

Ciao Martin (@0xc0170), could you pls. proceed on & merge this PR too, or are there still some blocking points I am not aware of?

bulislaw · 2017-01-27T09:25:23Z

I removed the labels: 'do not merge' as i added it before and it does not apply anymore and 'needs review' as it was reviewed.

sg- · 2017-02-02T16:42:21Z

/morph test-nightly

mbed-bot · 2017-02-02T18:46:50Z

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph test-nightly

Output

mbed Build Number: 1493

All builds and test passed!

0xc0170 added the needs: review label Jan 11, 2017

bulislaw added the do not merge label Jan 12, 2017

betzw force-pushed the betzw_wb_sleep branch from 3d013d5 to 39ad07f Compare January 13, 2017 08:58

bulislaw mentioned this pull request Jan 18, 2017

Platform: Add sleep/deepsleep user facing functions #3607

Merged

betzw force-pushed the betzw_wb_sleep branch from 968fa2a to 0551195 Compare January 19, 2017 11:00

betzw changed the title ~~Put system to sleep (when available)~~ Put system to sleep when going idle Jan 19, 2017

bulislaw approved these changes Jan 26, 2017

View reviewed changes

betzw added 2 commits January 26, 2017 12:59

Put system to sleep when going idle

c6897c5

Improve build profiles documentation

109faa1

betzw force-pushed the betzw_wb_sleep branch from 0551195 to 109faa1 Compare January 26, 2017 12:00

Include new "platform/sleep.h" file

2a585ea

bulislaw removed do not merge needs: review labels Jan 27, 2017

0xc0170 approved these changes Jan 27, 2017

View reviewed changes

0xc0170 added the ready for merge label Jan 27, 2017

sg- added release-version: 5.3.5 and removed ready for merge labels Feb 2, 2017

sg- added ready for merge release-version: 5.4.0 and removed release-version: 5.3.5 labels Feb 2, 2017

bulislaw mentioned this pull request Feb 3, 2017

[Proposal] Simplify build profiles: debug and release only #3693

Merged

sg- merged commit e852d7f into ARMmbed:master Feb 7, 2017

sg- removed the ready for merge label Feb 7, 2017

bridadan mentioned this pull request Feb 16, 2017

Renamed files in platform to match source names #3654

Merged

Put system to sleep when going idle #3566

Put system to sleep when going idle #3566

Uh oh!

Conversation

betzw commented Jan 11, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Status

Migrations

Related PRs

Uh oh!

bulislaw commented Jan 11, 2017

Uh oh!

bulislaw commented Jan 11, 2017

Uh oh!

betzw commented Jan 11, 2017

Uh oh!

mbed-bot commented Jan 11, 2017

Result: SUCCESS

Output

Uh oh!

0xc0170 commented Jan 11, 2017

Uh oh!

betzw commented Jan 11, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bulislaw commented Jan 11, 2017

Uh oh!

betzw commented Jan 11, 2017

Uh oh!

betzw commented Jan 11, 2017

Uh oh!

bulislaw commented Jan 11, 2017

Uh oh!

betzw commented Jan 11, 2017

Uh oh!

bridadan commented Jan 11, 2017

Uh oh!

betzw commented Jan 12, 2017

Uh oh!

bulislaw commented Jan 12, 2017 • edited by bridadan Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

betzw commented Jan 12, 2017

Uh oh!

bulislaw commented Jan 12, 2017

Uh oh!

mbed-bot commented Jan 12, 2017

Result: SUCCESS

Output

Uh oh!

bulislaw commented Jan 12, 2017

Uh oh!

betzw commented Jan 12, 2017

Uh oh!

adbridge commented Jan 12, 2017

Uh oh!

bridadan commented Jan 12, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bulislaw commented Jan 12, 2017

Uh oh!

bridadan commented Jan 12, 2017

Uh oh!

betzw commented Jan 13, 2017

Uh oh!

bulislaw commented Jan 13, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

betzw commented Jan 13, 2017

Uh oh!

betzw commented Jan 13, 2017

Uh oh!

sg- commented Jan 18, 2017

Uh oh!

betzw commented Jan 19, 2017

Uh oh!

bulislaw commented Jan 26, 2017

Uh oh!

bulislaw commented Jan 26, 2017

betzw commented Jan 11, 2017 •

edited

Loading

betzw commented Jan 11, 2017 •

edited

Loading

bulislaw commented Jan 12, 2017 •

edited by bridadan

Loading

bridadan commented Jan 12, 2017 •

edited

Loading

bulislaw commented Jan 13, 2017 •

edited

Loading

betzw commented Jan 27, 2017 •

edited

Loading