Refactor singlepass init stack assembly #2012
Merged
Conversation
This file contains 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
slave5vw
requested review from
losfair,
MarkMcCaskey and
nlewycky
as code owners
slave5vw
changed the title
Since the direction of rep stosq is in reverse order (DF=0), The trap handler is judged as a heap oob exception based on the start address.
brew0722
force-pushed
the
feature/refactor-init-stack-assembly
branch
from
99ad186
to
71f4e15
Compare
|
Thanks for the pull request! I added stack probe for large allocations before |
…it-stack-assembly
|
bors try |
1 similar comment
|
bors try |
tryAlready running a review |
|
bors try- |
|
bors try |
Co-authored-by: Ivan Enderlin <ivan@mnt.io>
Co-authored-by: Ivan Enderlin <ivan@mnt.io>
|
@losfair Can you approve this PR if it's OK please? :-) |
losfair
approved these changes
Jan 14, 2021
|
bors r+ |
bors bot
added a commit
that referenced
this pull request
Jan 14, 2021
2012: Refactor singlepass init stack assembly r=syrusakbary a=slave5vw <!-- Prior to submitting a PR, review the CONTRIBUTING.md document for recommendations on how to test: https://github.com/wasmerio/wasmer/blob/master/CONTRIBUTING.md#pull-requests --> # Description <!-- Provide details regarding the change including motivation, links to related issues, and the context of the PR. --> <!-- Prior to submitting a PR, review the CONTRIBUTING.md document for recommendations on how to test: https://github.com/wasmerio/wasmer/blob/master/CONTRIBUTING.md#pull-requests --> The old method generated 1 mov assembly instruction per stack slot. This takes up 15 bytes * slots count size. So more local stack variables use, the larger the code size. The suggested method is the same as c memset. Initialize with a fixed instructions (24 bytes) using the rep stosq assembly instruction. It works on a range basis. https://www.felixcloutier.com/x86/stos:stosb:stosw:stosd:stosq However, in this way, the skip_stack_guard_page test fails. https://github.com/wasmerio/wasmer/blob/1b49fe8475e335c6cdbb702a7100ba044201afd0/lib/vm/src/trap/traphandlers.rs#L124 The reason is that DF Flag is 0, so it is accessed in reverse order and initialized. However, The vm trap handler checks the last address and determines it as a heap oob without range. the stack oob is correct, and the test condition fails because it is a stack oob occurrence. I think there are two ways. 1. improve stack oob check in VM trap handler -> Abstract. There is no information on the range of access, so this may not be possible. 2. pushf/popf/std(set DF=1), and access to sequential order -> Not recommended. Unnecessarily increases assembly size. So for now, I added it to ignores.txt like cranelift/llvm. # Review - [ ] Add a short description of the the change to the CHANGELOG.md file Co-authored-by: Jiyong Ha <slave5vw@gmail.com> Co-authored-by: losfair <zhy20000919@hotmail.com> Co-authored-by: Jiyong Ha <70933233+brew0722@users.noreply.github.com> Co-authored-by: Syrus Akbary <me@syrusakbary.com>
|
Build failed: |
|
bors r+ |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Description
The old method generated 1 mov assembly instruction per stack slot.
This takes up 15 bytes * slots count size. So more local stack variables use, the larger the code size.
The suggested method is the same as c memset.
Initialize with a fixed instructions (24 bytes) using the rep stosq assembly instruction. It works on a range basis.
https://www.felixcloutier.com/x86/stos:stosb:stosw:stosd:stosq
However, in this way, the skip_stack_guard_page test fails.
wasmer/lib/vm/src/trap/traphandlers.rs
Line 124 in 1b49fe8
The reason is that DF Flag is 0, so it is accessed in reverse order and initialized.
However, The vm trap handler checks the last address and determines it as a heap oob without range.
the stack oob is correct, and the test condition fails because it is a stack oob occurrence.
I think there are two ways.
-> Abstract. There is no information on the range of access, so this may not be possible.
-> Not recommended. Unnecessarily increases assembly size.
So for now, I added it to ignores.txt like cranelift/llvm.
Review