Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

JIT: Add ordering side effects in fgMorphExpandInstanceField #92998

Merged
merged 2 commits into from
Oct 5, 2023

Conversation

jakobbotsch
Copy link
Member

In #92710 I removed an ordering side effect set on a LCL_VAR node in fgMorphExpandInstanceField, since the side effect is meaningless on a local. However, I did not realize that setting the ordering effect was important because propagating it to parent nodes was actually necessary.

This PR explicitly sets the ordering effect on the ADD nodes created, since forming these and reporting them to GC is illegal if the base is invalid.

Additionally, it sets the ordering effect on the consuming indirection; if that indirection made use of the fact that the FIELD_ADDR is non-null to become non-faulting, then it has an ordering dependency once we expand the FIELD_ADDR to a control-flow based null-check. This is not actually necessary since we always set end up with an ADD that propagates it, but that seems more like a coincidence than anything else.

Fix #92990

In dotnet#92710 I removed an ordering side effect set on a LCL_VAR node in
fgMorphExpandInstanceField, since the side effect is meaningless on a
local. However, I did not realize that setting the ordering effect was
important because propagating it to parent nodes was actually necessary.

This PR explicitly sets the ordering effect on the ADD<byref> nodes
created, since forming these and reporting them to GC is illegal if the
base is invalid.

Additionally, it sets the ordering effect on the consuming indirection;
if that indirection made use of the fact that the FIELD_ADDR is non-null
to become non-faulting, then it has an ordering dependency once we
expand the FIELD_ADDR to a control-flow based null-check. This is not
actually necessary since we always set end up with an ADD<byref> that
propagates it, but that seems more like a coincidence than anything
else.

Fix dotnet#92990
@dotnet-issue-labeler dotnet-issue-labeler bot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Oct 4, 2023
@ghost ghost assigned jakobbotsch Oct 4, 2023
@ghost
Copy link

ghost commented Oct 4, 2023

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

Issue Details

In #92710 I removed an ordering side effect set on a LCL_VAR node in fgMorphExpandInstanceField, since the side effect is meaningless on a local. However, I did not realize that setting the ordering effect was important because propagating it to parent nodes was actually necessary.

This PR explicitly sets the ordering effect on the ADD nodes created, since forming these and reporting them to GC is illegal if the base is invalid.

Additionally, it sets the ordering effect on the consuming indirection; if that indirection made use of the fact that the FIELD_ADDR is non-null to become non-faulting, then it has an ordering dependency once we expand the FIELD_ADDR to a control-flow based null-check. This is not actually necessary since we always set end up with an ADD that propagates it, but that seems more like a coincidence than anything else.

Fix #92990

Author: jakobbotsch
Assignees: -
Labels:

area-CodeGen-coreclr

Milestone: -

@jakobbotsch
Copy link
Member Author

/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress

@azure-pipelines
Copy link

Azure Pipelines successfully started running 2 pipeline(s).

@jakobbotsch jakobbotsch marked this pull request as ready for review October 4, 2023 19:18
@jakobbotsch
Copy link
Member Author

cc @dotnet/jit-contrib PTAL @BruceForstall

Diffs

Copy link
Member

@BruceForstall BruceForstall left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM with one nit

Comment on lines 5920 to 5921
GenTreeIndir* Consumer = nullptr; // Indirection using this address.
size_t TotalOffset = 0; // Sum of offsets between the top-level indirection and here (current context).
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: why change the naming style? We aren't very consistent with https://github.com/dotnet/runtime/blob/main/docs/coding-guidelines/clr-jit-coding-conventions.md#8.7, but we are pretty consistent to not capitalize local/field names.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I reverted this... mainly I think the m_ prefix on fields meant to be used publicly looks off (I like the convention, but only for private fields, and I thought that's only where it applied).

@jakobbotsch jakobbotsch merged commit ade2459 into dotnet:main Oct 5, 2023
126 of 129 checks passed
@jakobbotsch jakobbotsch deleted the fix-92990 branch October 5, 2023 14:38
@ghost ghost locked as resolved and limited conversation to collaborators Nov 4, 2023
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI
Projects
None yet
Development

Successfully merging this pull request may close these issues.

jitstress failing widely with NRE
2 participants