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

Revert "Do not generate TYP_STRUCT LCL_FLD nodes (#66251)" #67360

Merged
merged 1 commit into from
Mar 30, 2022

Conversation

SingleAccretion
Copy link
Contributor

In #66251 I've enhanced IsLocalAddrExpr to look for zero-offset sequences on ADDRs.

This was a correctness fix.

Unfortunately, it also broke the structreturn test because we have the following tree there:

***** BB01, STMT00006(before)
N005 ( 10,  8) [000064] -A------R---              *  ASG       long  
N004 (  3,  2) [000062] D------N----              +--*  LCL_VAR   long   V04 tmp2         d:2
N003 (  6,  5) [000063] n-----------              \--*  IND       long  
N002 (  3,  3) [000043] ------------                 \--*  ADDR      byref  Zero Fseq[a]
N001 (  3,  2) [000044] -------N----                    \--*  LCL_VAR   struct<TestUnsafeCasts+largeStruct, 32> V00 loc0         u:6 (last use)

Where the field a does not belong to TestUnsafeCasts+largeStruct. This is the well-known VN problem with invalid field sequences, there is no expedient fix for it.

So I am reverting #66251, it seems TYP_STRUCT LCL_FLD is fully gated on fixing this problem (I knew it was gated at least partially, I did not anticipate it to break so quickly though).

Fixes #67346, #67355.

@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 Mar 30, 2022
@ghost ghost added the community-contribution Indicates that the PR has been added by a community member label Mar 30, 2022
@ghost
Copy link

ghost commented Mar 30, 2022

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

Issue Details

In #66251 I've enhanced IsLocalAddrExpr to look for zero-offset sequences on ADDRs.

This was a correctness fix.

Unfortunately, it also broke the structreturn test because we have the following tree there:

***** BB01, STMT00006(before)
N005 ( 10,  8) [000064] -A------R---              *  ASG       long  
N004 (  3,  2) [000062] D------N----              +--*  LCL_VAR   long   V04 tmp2         d:2
N003 (  6,  5) [000063] n-----------              \--*  IND       long  
N002 (  3,  3) [000043] ------------                 \--*  ADDR      byref  Zero Fseq[a]
N001 (  3,  2) [000044] -------N----                    \--*  LCL_VAR   struct<TestUnsafeCasts+largeStruct, 32> V00 loc0         u:6 (last use)

Where the field a does not belong to TestUnsafeCasts+largeStruct. This is the well-known VN problem with invalid field sequences, there is no expedient fix for it.

So I am reverting #66251, it seems TYP_STRUCT LCL_FLD is fully gated on fixing this problem (I knew it was gated at least partially, I did not anticipate it to break so quickly though).

Fixes #67346, #67355.

Author: SingleAccretion
Assignees: -
Labels:

area-CodeGen-coreclr

Milestone: -

@SingleAccretion
Copy link
Contributor Author

@dotnet/jit-contrib

@BruceForstall BruceForstall merged commit 59bb459 into dotnet:main Mar 30, 2022
@SingleAccretion SingleAccretion deleted the Revert-NoStrucLclFld branch March 30, 2022 23:14
@ghost ghost locked as resolved and limited conversation to collaborators Apr 30, 2022
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 community-contribution Indicates that the PR has been added by a community member
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Assertion failed 'lclNode->GetSsaNum() == SsaConfig::FIRST_SSA_NUM' during 'VN based copy prop'
3 participants