[interp] always widen i4 constants to i8 when pushing to stack #17652
Conversation
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
Fixes:
```console
$ ./mono/mini/mono-sgen --interp=-all ./mono/mini/iltests.exe
```
This is the problematic trace:
```
(0x103e90130) Tests:i8_ptr_arithm_x64 -> IR_004e: ldc.i4.5 0:
(0x103e90130) Tests:i8_ptr_arithm_x64 -> IR_004f: ldc.i8 1229801703532086340 0:[0x5 (5)]
(0x103e90130) Tests:i8_ptr_arithm_x64 -> IR_0054: ldc.i4 -65536 0:[0x5 (5)] [0x1111222233334444 (1229801703532086340)]
(0x103e90130) Tests:i8_ptr_arithm_x64 -> IR_0057: call 2 0:[0x5 (5)] [0x1111222233334444 (1229801703532086340)] [0xffff0000 (4294901760)]
(0x103e90130) Entering Tests:and_ptr_i8 ([0x1111222233334444] [{long} 4294901760/0xffff0000] )
(0x103e90130) Tests:and_ptr_i8 -> IR_0000: safepoint 0:
(0x103e90130) Tests:and_ptr_i8 -> IR_0001: ldarg.p0 0:
(0x103e90130) Tests:and_ptr_i8 -> IR_0002: ldarg.i8 1 0:[0x1111222233334444 (1229801703532086340)]
(0x103e90130) Tests:and_ptr_i8 -> IR_0004: and.i8 0:[0x1111222233334444 (1229801703532086340)] [0xffff0000 (4294901760)]
(0x103e90130) Tests:and_ptr_i8 -> IR_0005: ret 0:[0x33330000 (858980352)]
(0x103e90130) Leaving Tests:and_ptr_i8 => [{long} 858980352/0x33330000]
(0x103e90130) Tests:i8_ptr_arithm_x64 -> IR_0059: ldc.i8 1229801703532068864 0:[0x5 (5)] [0x33330000 (858980352)]
```
`and_ptr_i8` expects a ptr and i8 on the stack. In the default configuration the interpreter inlines `and_ptr_i8` and thus stores the arguments into some locals:
https://github.com/mono/mono/blob/b484979aa9741673c2f820bdac7f5c925128ad9d/mono/mini/interp/transform.c#L3360-L3364
where `store_local` will take care of a proper i4->i8 conversion:
https://github.com/mono/mono/blob/b484979aa9741673c2f820bdac7f5c925128ad9d/mono/mini/interp/transform.c#L820-L825
I'm not sure how to do that (effeciently) for regular calls. The suggested change here is a bit of a hack. Properly harmless on 64 bit systems, but potentially a performance penalty on 32 bit.
Maybe also addresses mono#17440
Member
|
@lewurm I'm not convinced the il from the test is valid. Checking the ecma spec |
Contributor
|
Obvious questions, inspired by Vlad:
|
lewurm
added a commit
to lewurm/mono
that referenced
this pull request
Jan 7, 2020
Contributor
Author
|
Closing it for now. My attempt to fix it is obviously flawed. Not sure about the test, this needs further investigation as suggested. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
Fixes:
$ ./mono/mini/mono-sgen --interp=-all ./mono/mini/iltests.exeThis is the problematic trace:
and_ptr_i8expects a ptr and i8 on the stack. In the default configuration the interpreter inlinesand_ptr_i8and thus stores the arguments into some locals:mono/mono/mini/interp/transform.c
Lines 3360 to 3364 in b484979
where
store_localwill take care of a proper i4->i8 conversion:mono/mono/mini/interp/transform.c
Lines 820 to 825 in b484979
I'm not sure how to do that (effeciently) for regular calls. The suggested change here is a bit of a hack. Properly harmless on 64 bit systems, but potentially a performance penalty on 32 bit.
Maybe also addresses #17440it's a different problem@BrzVlad do you have a better suggestion how to fix this?