Monomorphize in check mode to report (almost) all PMEs#107510
Closed
oli-obk wants to merge 1 commit intorust-lang:masterfrom
Closed
Monomorphize in check mode to report (almost) all PMEs#107510oli-obk wants to merge 1 commit intorust-lang:masterfrom
oli-obk wants to merge 1 commit intorust-lang:masterfrom
Conversation
rustbot
added
S-waiting-on-review
Contributor
Author
|
@bors try @rust-timer queue |
Collaborator
|
Awaiting bors try build completion. @rustbot label: +S-waiting-on-perf |
rustbot
added
the
S-waiting-on-perf
Collaborator
|
⌛ Trying commit b99639861e05bb61a3235429a82a8c232233c751 with merge f792b4425823c469a96cdaff35d25590fe0e5970... |
Member
|
What's the "almost" about? |
Contributor
Author
codegen backends can still bail out if they don't support something (like some huge array size limits that are very target dependent) |
Member
|
We do check for "types can't be larger than 2^48" somewhere on the MIR level already -- so this is a separate check? Anyway, would be good to have a concrete example and an issue tracking that, but doesn't sound high priority to fix. |
Contributor
Author
|
We have tests for
|
Member
|
And const-eval errors, I presume? |
Contributor
Author
those all get emitted in |
Collaborator
|
☔ The latest upstream changes (presumably #100754) made this pull request unmergeable. Please resolve the merge conflicts. |
Member
This comment has been minimized.
This comment has been minimized.
Member
|
I personally think whether optimised mir is used or for collection or not is orthogonal to whether instances are collected eagerly/lazily. Also, wouldn't this approach cause |
This comment has been minimized.
This comment has been minimized.
Contributor
Author
|
cc @tmandry |
Member
|
Thanks for putting this up, I'm very curious to see the perf results! I think the use of PME here is slightly confusing – if I understand correctly we would still have errors that depend on which generic parameters are used, like the example in #104087 (comment). |
oli-obk
changed the title
oli-obk
changed the title
Contributor
|
@oli-obk another implementation strategy that we should consider here is adding a new field to Part of the reason that I bring this up is because I knew the topic of this issue was a possibility, but thought it was a non-issue specifically because I was under the impression that the |
Member
|
How is required_consts not already doing that?
The trouble is that we don't have required_fns so we might not transitively monomorphize functions even though they were mentioned by the user.
|
Contributor
Well, what I didn't realize was that required_consts only contains consts that are unevaluated by the time we do borrowck. What I thought happened is that all consts are put into there, specifically to ensure that things like this don't happen Edit: Oh, I realize what the confusion might have been. I was implicitly assuming that if we put a FnDef-typed constant into the list of constants, that we'd then go and actually mono that function (at least enough to go set off any PMEs) |
Member
Ah -- no, all we do with that list is evaluate constants, we don't monomorphize any further functions. |
Member
|
If we want to make |
Member
That's quite a big new feature that requires an RFC IMO. I don't think there is currently a plan to implement this. |
Contributor
|
Yeah, agreed. That kind of feature would also definitely need explicit Mir support |
Member
|
There's a lang team meeting that will touch on this PR next week (Feb 22). @oli-obk Do you think you could get it fixed up with a perf run by then? 🙏 |
oli-obk
force-pushed
the
no_const_pme
branch
from
b70f67f to
5a2cab4
Compare
Contributor
|
The |
Contributor
Author
I think we can avoid it. We only need to care about generic functions, and those have their MIR encoded no matter what |
Collaborator
|
The job Click to see the possible cause of the failure (guessed by this bot) |
bjorn3
reviewed
Feb 22, 2023
| pub enum MonoItemCollectionMode { | ||
| Eager, | ||
| Eager { | ||
| /// Whether to use optimized mir for collection or just analyis MIR. |
Member
There was a problem hiding this comment.
Suggested change
| /// Whether to use optimized mir for collection or just analyis MIR. | |
| /// Whether to use optimized mir for collection or just analysis MIR. |
Collaborator
|
☔ The latest upstream changes (presumably #108872) made this pull request unmergeable. Please resolve the merge conflicts. |
Dylan-DPC
added
S-experimental
Member
|
Closing this as it was an experiment |
Dylan-DPC
removed
the
S-experimental
Member
|
This never got far enough to get a perf result, did it? Would have been nice to get some numbers, though I think they'd be devastating.^^ |
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
12 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 #99682
see #104087 (comment) for discussion
r? @ghost