MIR unified ADTs #30890
MIR unified ADTs #30890
Conversation
nagisa
force-pushed
the
mir-tuple-adts
branch
2 times, most recently
from
f39bfd1 to
4e9c88f
Compare
| } | ||
| } else { | ||
| ExprKind::Call { | ||
| ty: &cx.tcx.node_id_to_type(fun.id), |
There was a problem hiding this comment.
This was present before, but I think this & is redundant and will just get deref-coerced out of existence.
There was a problem hiding this comment.
Yeah, probably a remnant of when the code was generic over the HIR.
|
Can you add a few more tests:
Otherwise, looks good to me. |
|
To be clear, r=me with more tests. |
nagisa
force-pushed
the
mir-tuple-adts
branch
from
4e9c88f to
b9d54b6
Compare
|
@bors r=nikomatsakis |
|
📌 Commit b9d54b6 has been approved by |
nagisa
force-pushed
the
mir-tuple-adts
branch
from
b9d54b6 to
949dc72
Compare
|
@bors r=nikomatsakis 949dc72 (had omitted a rustc_mir attribute on test) |
|
We might want a more generic infrastructure to handle things like intrinsics, and maybe even trivial |
|
⌛ Testing commit 949dc72 with merge fef6ad1... |
|
💔 Test failed - auto-mac-64-nopt-t |
Previously we would generate regular calls for these, which is likely to result in worse LLVM code, especially in presence of cleanups – we needn’t unecessarilly generate landing pads to construct an ADT!
nagisa
force-pushed
the
mir-tuple-adts
branch
2 times, most recently
from
bb50c48 to
72eacc1
Compare
|
@bors r=nikomatsakis 72eacc1 (merge conflict with another patch which was not realised by git) |
Previously it would go through def_id and retrieve a type that’s not always correct, monomorphized, etc.
These tests test building and matching some non-trivial ADT configurations such as C-like enums and packed structs.
nagisa
force-pushed
the
mir-tuple-adts
branch
from
72eacc1 to
7d6da8e
Compare
This PR changes translation of tuple-like ADTs from being calls to being proper aggregates. This change is done in hope to make code generation better. Namely, now we can avoid: 1. Call overhead; 2. Generating landingpads in presence of cleanups (we know for sure constructing ADTs can’t panic); 3. And probably much more, gaining better MIR introspectablilty. Along with that a few serious deficiencies with translation of ADTs and switches have been fixed as well (commits 2 and 3). r? @nikomatsakis cc @tsion
This PR changes translation of tuple-like ADTs from being calls to being proper aggregates. This change is done in hope to make code generation better. Namely, now we can avoid:
Along with that a few serious deficiencies with translation of ADTs and switches have been fixed as well (commits 2 and 3).
r? @nikomatsakis
cc @tsion