Change how we store and sign/zero extend integers. #1557
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
|
We are talking only about constant integers here, right? I think we store local variables in as many bytes as is required (not always a u64). |
|
Yes, more-or-less trace-level |
ptersilie
reviewed
Jan 23, 2025
vext01
reviewed
Jan 24, 2025
|
LGTM. This API is basically what LLVM does with |
|
Please squash. |
Previously we stored raw `u64`s and expected the user to remember that they needed to zero/sign extend the underlying integer whenever they wanted to do anything with that. We did not always do this, and we did it incorrectly in a couple of places! This commit introduces a new struct `ArbBitInt` which is basically a pair `(bit_width: u32, value: u64)` which hides the underlying value. To get a raw Rust-level integer you have to call a method, and those methods have names with `sign_extend` and `zero_extend` in them. While one can, of course, call the wrong one, it is now impossible not to sign/zero extend. This struct is currently rather simple, but the API is flexible enough to extend to beyond-64-bit ints transparently. The (fairly extensive) test suite is a bit overkill right now, but is intended to help give us confidence if/when we support more than 64 bit ints in the future. This commit also necessarily requires a full audit of everything to do with ints-in-traces. That means a lot of code churn, but it's absolutely necessary, and (a) makes much clearer where we should sign/zero extend (b) catches some places where we didn't do this but should. This commit isn't perfect. In particular, I'm not very happy that `Const::Int` has both a `TyIdx` that contains a bit width *and* an `ArbBitInt` that separately records a bit width. That feels icky, but doing something neater will require at least some ickiness elsewhere. I'll worry about that another day.
|
Squashed. |
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.
Previously we stored raw
u64s and expected the user to remember that they needed to zero/sign extend the underlying integer whenever they wanted to do anything with that. We did not always do this, and we did it incorrectly in a couple of places!This commit introduces a new struct
ArbBitIntwhich is basically a pair(bit_width: u32, value: u64)which hides the underlying value. To get a raw Rust-level integer you have to call a method, and those methods have names withsign_extendandzero_extendin them. While one can, of course, call the wrong one, it is now impossible not to sign/zero extend. This struct is currently rather simple, but the API is flexible enough to extend to beyond-64-bit ints transparently. The (fairly extensive) test suite is a bit overkill right now, but is intended to help give us confidence if/when we support more than 64 bit ints in the future.This commit also necessarily requires a full audit of everything to do with ints-in-traces. That means a lot of code churn, but it's absolutely necessary, and (a) makes much clearer where we should sign/zero extend (b) catches some places where we didn't do this but should.
This commit isn't perfect. In particular, I'm not very happy that
Const::Inthas both aTyIdxthat contains a bit width and anArbBitIntthat separately records a bit width. That feels icky, but doing something neater will require at least some ickiness elsewhere. I'll worry about that another day.