edf_schedule: reimplement the whole EDF scheduler by tlauda · Pull Request #1710 · thesofproject/sof

tlauda · 2019-08-07T16:31:01Z

This patch reimplements the whole EDF scheduler.
This is the first basic version, which can be easily extended
in the future. The new implementation uses only one
software interrupt and allows for dynamic context switching
between different tasks based on the priority and deadline.
Task chosen to be run is executed on passive level, so automatically
can be preempted by every interrupt. Every task has its own private
stack on which current context is saved and restored if needed.
Note that main firmware loop is task scheduled with idle priority,
so it will be executed only after every other task is completed.

Signed-off-by: Tomasz Lauda tomasz.lauda@linux.intel.com

lgirdwood · 2019-08-07T19:21:22Z

@mengdonglin @keqiaozhang this is P1 for testing on legacy.

lgirdwood · 2019-08-07T19:22:38Z

@tlauda this should be our last big ticket item prior to v1.4-rc1

tlauda · 2019-08-07T20:59:09Z

@lgirdwood @keqiaozhang @mengdonglin There seems to be some problem with non SMP configurations, so let me fix it first thing tomorrow and then you can test it on legacy.

jajanusz · 2019-08-07T22:36:17Z

@lgirdwood @keqiaozhang @mengdonglin There seems to be some problem with non SMP configurations, so let me fix it first thing tomorrow and then you can test it on legacy.

CI builds also APL for 1 core and it works.

tlauda · 2019-08-08T06:03:14Z

@jajanusz Yes, I fixed it after my last comment and it worked for APL. Still doesn't work for legacy platforms.

dbaluta · 2019-08-08T07:58:55Z

@tlauda can you please mention in the commit message what are the limitation of the old scheduler and why do we need to change it?

tlauda · 2019-08-08T14:29:03Z

@mengdonglin @keqiaozhang You can now test it on legacy platforms. There is still minor issue with APL, because we don't have enough buffer space for the topology.

tlauda · 2019-08-09T08:58:46Z

@xiulipan Is QEMU somehow dependent on memory window address from DSP side? I needed to increase HP buffer heap for APL based platforms, so automatically the mailbox addresses have changed.

Edit: Figured it out thanks to @jajanusz.

jajanusz · 2019-08-12T13:49:59Z

SOFCI TEST

tlauda · 2019-08-20T07:51:32Z

SOFCI TEST

tlauda · 2019-08-20T13:28:16Z

@lgirdwood @mmaka1 Any comments?

tlauda · 2019-08-26T21:10:17Z

SOFCI TEST

tlauda · 2019-08-30T06:45:19Z

@dbaluta @lgirdwood @mmaka1 There is no regression in the CI. Anyone dare to review?

lgirdwood

I would split this into some smaller incremental PRs too. The HAL changes could be seperated out and merged first if they dont break any existing code. Btw, does this align the HAL with latest ??

lgirdwood · 2019-08-31T17:48:13Z

We need to comment what we are changing in xtensa HAL and why. Does this also align with latest CAVS version ??

Every single operation in the assembly code is commented.
What do you mean by "latest CAVS version"?

tlauda · 2019-09-02T09:44:28Z

@dbaluta @mmaka1 Solved all your suggestions. I hope now it is more clear.
@lgirdwood I cannot divide it into smaller PRs, because it simply won't work. I've tried to do it, but it wasn't worth the effort.

tlauda · 2019-09-05T08:15:57Z

SOFCI TEST

Defines and adds dedicated stacks for interrupt handlers to core data. They are required to implement the execution context switching, which is useful for any preemptive scheduler. EDF is the first scheduler for which they will be used. Signed-off-by: Tomasz Lauda <tomasz.lauda@linux.intel.com>

Passes sof context structure to the schedulers during scheduler initialization. It will be required by the new EDF scheduler. Signed-off-by: Tomasz Lauda <tomasz.lauda@linux.intel.com>

Removes two tests which don't really test anything and are based on magic numbers. Changing memory map can easily break them. Signed-off-by: Tomasz Lauda <tomasz.lauda@linux.intel.com>

Increases HP Buffer Heap and Runtime System Heap sizes in order to fit all the pipeline buffers. Signed-off-by: Tomasz Lauda <tomasz.lauda@linux.intel.com>

This patch implements the new version of EDF scheduler. This is the first basic implementation, which can be easily extended in the future. Previous implementation had two major flaws: - Tasks could only be preempted based on the interrupt level on which they have been scheduled, so it was limited based on the hardware. - This led to the system not fully utilized, so we could never achieve full EDF functionality. The new implementation uses only one software interrupt and allows for dynamic context switching between different tasks based on the deadline. Also for some cases priority is taken into consideration e.g. for idle and almost idle tasks, which don't have deadline at all and should be executed only if the system is free. All other normal tasks should have the same priority. Task chosen to be run is executed on passive level, so automatically can be preempted by every interrupt. Every task has its own private stack on which current context is saved and restored if needed. Note that main firmware loop is task scheduled with idle priority, so it will be executed only after every other task is completed. Signed-off-by: Tomasz Lauda <tomasz.lauda@linux.intel.com>

tlauda · 2019-09-05T10:04:03Z

SOFCI TEST

Changes applied.

lgirdwood · 2019-09-06T14:50:50Z

The CI failures look like they are not related to this PR but to this bug thesofproject/linux#1183

tlauda requested review from RanderWang, jajanusz, lbetlej, lgirdwood, mmaka1, mrajwa, plbossart and xiulipan as code owners August 7, 2019 16:31

tlauda force-pushed the topic/new-edf branch from 3f83671 to d31f2f2Compare August 7, 2019 16:45

lgirdwood added this to the v1.4 milestone Aug 7, 2019

lgirdwood added the P1 Blocker bugs or important features label Aug 7, 2019

tlauda force-pushed the topic/new-edf branch from d31f2f2 to 4fc3d2cCompare August 7, 2019 21:16

tlauda force-pushed the topic/new-edf branch from 4fc3d2c to 71a8864Compare August 8, 2019 06:45

tlauda force-pushed the topic/new-edf branch 4 times, most recently from 090d2b1 to 7d78e15Compare August 8, 2019 10:39

jajanusz approved these changes Aug 8, 2019
View reviewed changes

tlauda force-pushed the topic/new-edf branch from 7d78e15 to 194bca5Compare August 8, 2019 11:50

tlauda force-pushed the topic/new-edf branch from 194bca5 to 01cc726Compare August 9, 2019 08:19

tlauda force-pushed the topic/new-edf branch 2 times, most recently from 02b26cf to 4146ec6Compare August 9, 2019 13:15

tlauda requested a review from dbaluta August 30, 2019 06:45

lgirdwood modified the milestones: v1.4, v1.5 Aug 30, 2019

mmaka1 reviewed Aug 30, 2019
View reviewed changes

Comment threadsrc/arch/xtensa/xtos/xtos-structs.h Outdated
Comment threadsrc/arch/xtensa/xtos/xtos-internal.h Outdated
Comment threadsrc/arch/xtensa/include/xtensa/xtruntime-frames.h Outdated
Comment threadsrc/platform/apollolake/include/platform/lib/memory.h Outdated
Comment threadKconfig Outdated

dbaluta reviewed Aug 31, 2019
View reviewed changes

Comment threadsrc/arch/xtensa/include/arch/schedule/task.h Outdated

dbaluta reviewed Aug 31, 2019
View reviewed changes

Comment threadKconfig Outdated

lgirdwood previously requested changes Aug 31, 2019
View reviewed changes

tlauda force-pushed the topic/new-edf branch from 4146ec6 to 6a4a284Compare September 2, 2019 09:34

tlauda added 5 commits September 5, 2019 11:34

schedule: pass sof context during scheduler init
d3b4d4a
Passes sof context structure to the schedulers during scheduler initialization. It will be required by the new EDF scheduler. Signed-off-by: Tomasz Lauda <tomasz.lauda@linux.intel.com>

test: alloc: remove tests which don't test anything
d2fc3ed
Removes two tests which don't really test anything and are based on magic numbers. Changing memory map can easily break them. Signed-off-by: Tomasz Lauda <tomasz.lauda@linux.intel.com>

apl: increase HP Buffer Heap and Runtime Sys Heap sizes
6da32e0
Increases HP Buffer Heap and Runtime System Heap sizes in order to fit all the pipeline buffers. Signed-off-by: Tomasz Lauda <tomasz.lauda@linux.intel.com>

tlauda force-pushed the topic/new-edf branch from 6a4a284 to fa4e495Compare September 5, 2019 09:35

tlauda requested a review from mmaka1 September 6, 2019 07:59

lbetlej approved these changes Sep 6, 2019
View reviewed changes

lgirdwood merged commit 20acc16 into thesofproject:masterSep 6, 2019

Jiangxinx mentioned this pull request Sep 10, 2019
[BUG][BYT]Capture failed (arecord: set_params:1363: Unable to install hw params) on byt nocodec platform. #1811
Closed

tlauda deleted the topic/new-edf branch November 22, 2019 11:24

marc-hb mentioned this pull request Jun 15, 2021
ipc_send_queued_msg() has a hard dependency on the EDF scheduler #4356
Closed

Uh oh!

Conversation

tlauda commented Aug 7, 2019

Uh oh!

lgirdwood commented Aug 7, 2019

Uh oh!

lgirdwood commented Aug 7, 2019

Uh oh!

tlauda commented Aug 7, 2019

Uh oh!

jajanusz commented Aug 7, 2019

Uh oh!

tlauda commented Aug 8, 2019

Uh oh!

dbaluta commented Aug 8, 2019

Uh oh!

tlauda commented Aug 8, 2019

Uh oh!

tlauda commented Aug 9, 2019• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jajanusz commented Aug 12, 2019

Uh oh!

tlauda commented Aug 20, 2019

Uh oh!

tlauda commented Aug 20, 2019

Uh oh!

tlauda commented Aug 26, 2019

Uh oh!

tlauda commented Aug 30, 2019

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

lgirdwood left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

lgirdwoodAug 31, 2019

Choose a reason for hiding this comment

Uh oh!

tlaudaSep 2, 2019

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

tlauda commented Sep 2, 2019

Uh oh!

tlauda commented Sep 5, 2019

Uh oh!

tlauda commented Sep 5, 2019

Uh oh!

lgirdwood commented Sep 6, 2019

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

tlauda commented Aug 9, 2019•
edited
Loading