8258393: Shenandoah: "graph should be schedulable" assert failure #41

rwestrel · 2020-12-17T08:20:20Z

This is a shenandoah bug but the fix is in shared code.

At barrier expansion time, we need the raw memory state for the
control at which the barrier is expanded. Before expansion, the memory
graph is traversed and memory state for each control is recorded. In
that process, if opportunities to improve the memory graph are found,
the graph is modified. In the case of the failure, a raw memory load
was assigned a call projection as control but for that control, no
memory state was recorded and the fall back is to use the call's input
memory state. As a consequence the load's memory input is updated to a
wrong memory state. This causes the assert failure.

The call has 2 memory projections: one for the fallthrough and one for
the exception path. One projection is the memory state for the
CatchProj for the fallthrough, the other one for the CatchProj of the
exception path. For the control projection out of the call, there's no
memory state that makes sense and assigning the control projection of
the call as control for the load is the root cause of the
problem. This happens because anti-dependence analysis is too
conservative: a memory Phi that merges states from the exception and
fallthrough path is considered a dependency and that pushed the load
right after the call.

I propose to be less conservative in anti-dependence analysis for
Phis. For a Phi, when computing the LCA, I think it's sufficient to
only consider region's inputs that we actually reach by following the
memory edges and that's what I propose here.

Progress

Change must not contain extraneous whitespace
Commit message must refer to an issue
Change must be properly reviewed

Issue

JDK-8258393: Shenandoah: "graph should be schedulable" assert failure

Reviewers

Roman Kennke (@rkennke - Reviewer) ⚠️ Review applies to 31f3bd1
Tobias Hartmann (@TobiHartmann - Reviewer)

Download

$ git fetch https://git.openjdk.java.net/jdk16 pull/41/head:pull/41
$ git checkout pull/41

bridgekeeper · 2020-12-17T08:20:38Z

👋 Welcome back roland! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2020-12-17T08:22:05Z

@rwestrel The following labels will be automatically applied to this pull request:

hotspot
shenandoah

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing lists. If you would like to change these labels, use the /label pull request command.

mlbridge · 2020-12-17T08:25:04Z

Webrevs

01: Full (b0f679d)
00: Full (31f3bd1)

rwestrel · 2020-12-17T08:27:33Z

/label add hotspot-compiler

openjdk · 2020-12-17T08:27:42Z

@rwestrel
The hotspot-compiler label was successfully added.

rwestrel · 2020-12-17T08:27:46Z

/label remove hotspot

openjdk · 2020-12-17T08:28:13Z

@rwestrel
The hotspot label was successfully removed.

rkennke

As far as I can tell, the change looks good. I'm not very familiar with the code, it would be good to have somebody else check it too.
I verified that the tests that failed before are passing now and also have run some additional tests.

openjdk · 2020-12-21T15:32:11Z

@rwestrel This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8258393: Shenandoah: "graph should be schedulable" assert failure

Reviewed-by: rkennke, thartmann

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been no new commits pushed to the master branch. If another commit should be pushed before you perform the /integrate command, your PR will be automatically rebased. If you prefer to avoid any potential automatic rebasing, please check the documentation for the /integrate command for further details.

➡️ To integrate this PR with the above commit message to the master branch, type /integrate in a new comment.

TobiHartmann

Otherwise looks reasonable to me.

TobiHartmann · 2021-01-05T11:38:20Z

test/hotspot/jtreg/gc/shenandoah/compiler/TestBadRawMemoryAfterCall.java

+
+/**
+ * @test
+ *


Missing @requires vm.gc.Shenandoah

TobiHartmann · 2021-01-05T11:38:38Z

test/hotspot/jtreg/gc/shenandoah/compiler/TestBadRawMemoryAfterCall.java

@@ -0,0 +1,70 @@
+/*
+ * Copyright (c) 2020, Red Hat, Inc. All rights reserved.


Copyright should be updated to 2021.

TobiHartmann · 2021-01-05T11:39:14Z

src/hotspot/share/gc/shenandoah/c2/shenandoahSupport.cpp

+  assert(!ctrl->is_Call() || ctrl == n, "projection expected");
+#ifdef ASSERT
+  if ((ctrl->is_Proj() && ctrl->in(0)->is_Call()) ||
+          (ctrl->is_Catch() && ctrl->in(0)->in(0)->is_Call())){


Indentation is wrong and whitespace is missing here ))){

TobiHartmann · 2021-01-05T11:54:07Z

src/hotspot/share/opto/loopnode.cpp

+            for (uint j = 1; j < s->req(); j++) {
+              Node* in = s->in(j);
+              Node* r_in = r->in(j);
+              if (worklist.member(in) && is_dominator(early, r_in)) {


Above we check for !sctrl->is_top(), do we need the same check for r_in?

I don't think a Region can have a top input at this point (previous IGVN should have taken care of it).

rwestrel · 2021-01-05T12:59:51Z

Otherwise looks reasonable to me.

Thanks for the review. I took care of your comments (except for the is_top() check that I don't think is needed).

TobiHartmann

Thanks for updating, looks good to me.

rwestrel · 2021-01-05T13:06:08Z

As far as I can tell, the change looks good. I'm not very familiar with the code, it would be good to have somebody else check it too.
I verified that the tests that failed before are passing now and also have run some additional tests.

thanks for the review and running tests.

rwestrel · 2021-01-05T13:06:16Z

/integrate

openjdk · 2021-01-05T13:07:10Z

@rwestrel Pushed as commit 6775113.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

openjdk bot added the rfr Pull request is ready for review label Dec 17, 2020

openjdk bot added hotspot shenandoah shenandoah-dev@openjdk.java.net labels Dec 17, 2020

openjdk bot added the hotspot-compiler hotspot-compiler-dev@openjdk.java.net label Dec 17, 2020

openjdk bot removed the hotspot label Dec 17, 2020

rkennke approved these changes Dec 21, 2020

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Dec 21, 2020

TobiHartmann approved these changes Jan 5, 2021

View reviewed changes

rwestrel added 2 commits January 5, 2021 13:35

test & fix

849aa11

review

b0f679d

rwestrel force-pushed the JDK-8258393 branch from 31f3bd1 to b0f679d Compare January 5, 2021 12:56

TobiHartmann approved these changes Jan 5, 2021

View reviewed changes

openjdk bot closed this Jan 5, 2021

openjdk bot added integrated Pull request has been integrated and removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels Jan 5, 2021

8258393: Shenandoah: "graph should be schedulable" assert failure #41

8258393: Shenandoah: "graph should be schedulable" assert failure #41

rwestrel commented Dec 17, 2020 •

edited by openjdk bot

bridgekeeper bot commented Dec 17, 2020

openjdk bot commented Dec 17, 2020

mlbridge bot commented Dec 17, 2020 •

edited

rwestrel commented Dec 17, 2020

openjdk bot commented Dec 17, 2020

rwestrel commented Dec 17, 2020

openjdk bot commented Dec 17, 2020

rkennke left a comment

openjdk bot commented Dec 21, 2020 •

edited

TobiHartmann left a comment

TobiHartmann Jan 5, 2021

TobiHartmann Jan 5, 2021

TobiHartmann Jan 5, 2021

TobiHartmann Jan 5, 2021

rwestrel Jan 5, 2021

rwestrel commented Jan 5, 2021

TobiHartmann left a comment

rwestrel commented Jan 5, 2021

rwestrel commented Jan 5, 2021

openjdk bot commented Jan 5, 2021

		@@ -0,0 +1,70 @@
		/*
		* Copyright (c) 2020, Red Hat, Inc. All rights reserved.

8258393: Shenandoah: "graph should be schedulable" assert failure #41

8258393: Shenandoah: "graph should be schedulable" assert failure #41

Conversation

rwestrel commented Dec 17, 2020 • edited by openjdk bot

Progress

Issue

Reviewers

Download

bridgekeeper bot commented Dec 17, 2020

openjdk bot commented Dec 17, 2020

mlbridge bot commented Dec 17, 2020 • edited

Webrevs

rwestrel commented Dec 17, 2020

openjdk bot commented Dec 17, 2020

rwestrel commented Dec 17, 2020

openjdk bot commented Dec 17, 2020

rkennke left a comment

Choose a reason for hiding this comment

openjdk bot commented Dec 21, 2020 • edited

TobiHartmann left a comment

Choose a reason for hiding this comment

TobiHartmann Jan 5, 2021

Choose a reason for hiding this comment

TobiHartmann Jan 5, 2021

Choose a reason for hiding this comment

TobiHartmann Jan 5, 2021

Choose a reason for hiding this comment

TobiHartmann Jan 5, 2021

Choose a reason for hiding this comment

rwestrel Jan 5, 2021

Choose a reason for hiding this comment

rwestrel commented Jan 5, 2021

TobiHartmann left a comment

Choose a reason for hiding this comment

rwestrel commented Jan 5, 2021

rwestrel commented Jan 5, 2021

openjdk bot commented Jan 5, 2021

rwestrel commented Dec 17, 2020 •

edited by openjdk bot

mlbridge bot commented Dec 17, 2020 •

edited

openjdk bot commented Dec 21, 2020 •

edited