This repository was archived by the owner on Jan 23, 2023. It is now read-only.
Spill side-effects in impAssignMultiRegTypeToVar.#11077
Merged
pgavlin merged 2 commits intodotnet:masterfrom Apr 19, 2017
Merged
Spill side-effects in impAssignMultiRegTypeToVar.#11077pgavlin merged 2 commits intodotnet:masterfrom
pgavlin merged 2 commits intodotnet:masterfrom
Conversation
This importer function converts IR from the form
tree
to
/- tree
STMT - =
\- lclVar
in order to conform to the JIT's representation of multi-register values.
Today, it does not check whether or not the tree being spilled has any
side-effects that may interfere with the side-effects of trees that are
already on the evaluation stack, which can lead to bad codegen. This
change fixes this function to spill any side-effects on the stack
before generating the temp assign.
Fixes #10940.
Author
|
@briansull @CarolEidt @dotnet/jit-contrib PTAL |
Member
|
Would be a good idea to check that we're not breaking trees in such a way as to reintroduce cases of #9562. You should be able to spot the broken opts, if any, from jit-diffs on corelib. Also, please add a test case. |
Author
Yes, I'm working on one. |
|
LGTM |
Author
|
@AndyAyersMS: there is one diff in S.P.CoreLib on Linux and OS X. It is in EventSource:CreateManifestAndDescriptors: G_M43523_IG10:
488BFB mov rdi, rbx
- E800000000 call EventSource:GetGuid(ref):struct
- 48898550FFFFFF mov qword ptr [rbp-B0H], rax
- 48899558FFFFFF mov qword ptr [rbp-A8H], rdx
- 488BFB mov rdi, rbx
8B75D4 mov esi, dword ptr [rbp-2CH]
E800000000 call EventSource:GetName(ref,int):ref
488985B0FEFFFF mov gword ptr [rbp-150H], rax
+ 488BFB mov rdi, rbx
+ E800000000 call EventSource:GetGuid(ref):struct
+ 48898550FFFFFF mov qword ptr [rbp-B0H], rax
+ 48899558FFFFFF mov qword ptr [rbp-A8H], rdx
488D3D00000000 lea rdi, [(reloc)]
E800000000 call CORINFO_HELP_NEWSFAST
488B95B0FEFFFF mov rdx, gword ptr [rbp-150H]
F30F6F8550FFFFFF movdqu xmm0, qword ptr [rbp-B0H]
F30F7F8530FFFFFF movdqu qword ptr [rbp-D0H], xmm0
8B7DD4 mov edi, dword ptr [rbp-2CH]
893C24 mov dword ptr [rsp], edi
488985A8FEFFFF mov gword ptr [rbp-158H], rax
488BF8 mov rdi, rax
488BF2 mov rsi, rdx
488B9530FFFFFF mov rdx, qword ptr [rbp-D0H]
488B8D38FFFFFF mov rcx, qword ptr [rbp-C8H]
4D8BC7 mov r8, r15
4C8B8D08FFFFFF mov r9, gword ptr [rbp-F8H]
E800000000 call ManifestBuilder:.ctor(ref,struct,ref,ref,int):thisIt corresponds to this source code: So this diff is expected: the baseline JIT was incorrectly reordering the calls to |
Member
|
Thanks for checking. Changes LGTM too. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This importer function converts IR from the form
to
in order to conform to the JIT's representation of multi-register values.
Today, it does not check whether or not the tree being spilled has any
side-effects that may interfere with the side-effects of trees that are
already on the evaluation stack, which can lead to bad codegen. This
change fixes this function to spill any side-effects on the stack
before generating the temp assign.
Fixes #10940.