Instead of reusing instructions from the generic entity in the eval
block, rebuild constants in the same way we rebuild types. The previous
attempt to not rebuild these constants assumed that every constant used
in a generic would be built in that generic, and not referenced directly
or referenced from some enclosing scope, which isn't true in practice
and is a fragile assumption in any case.
We could add back some reuse of instructions from the generic -- if we
happen to see the right instruction to build a constant, we could
opportunistically reuse it -- but given the complexity added by doing
so, I'm not pursuing that here.
Now that the eval block for a generic consists of instructions uniquely
owned by that generic, rather than often being shared with another
entity, include the generic in the formatted SemIR output. I'm using the
same scope name for the generic object itself as for the parameterized
class / function / interface, because there are very frequently
references between them and this keeps the IR simpler and more readable,
and avoids needing to invent a second name for the scope.
Also adds import_ir_scope to namespace formatting. I'd done this as an
aid for #4153, and am splitting it out.
---------
Co-authored-by: Geoff Romer <gromer@google.com>
Changes crash messages to start printing verbose forms of instructions,
rather than just the ID. Fixes some indentation issues with stacks. Also
switches unexpected inst formatting, because now there are lots, and
it'd be helpful to know where they are.
This uses a pimpl pattern for Formatter due to the number of member
functions on Formatter. Maybe we should refactor that, but this didn't
feel like a good place to do so.
Note, I have two concerns about this change... to note them here, to
make sure others are considering them when evaluating the
implementation:
1. Some instructions are very verbose to print, as evidenced by the
fn_decl printing (which includes function params) or scope printing
(which includes scope members).
- I'm not sure whether there's a way to simply reduce this, as it seems
essential to the requested printing of instructions.
- Long-term, we may at least want to limit the number of lines printed
here. However, I've already spent a fair amount of time here and I think
it's in a good state to evaluate.
2. Increased complexity in the crash handler may result in crash
messages failing to generate.
- For example, a crash in Formatter (and its deps, such as InstNamer or
location handling) prevents a stack from being printed. I'm pretty sure
I've written crashes in Formatter before.
Here's an example crash snippet (generated by adding a crash inside
`return` handling) before:
```
2. NodeStack:
0. FunctionDefinitionStart -> function2
1. ReturnStatementStart -> no value
2. IntLiteral -> inst+26
inst_block_stack_:
0. block<invalid> {inst+0, inst+1, inst+2, inst+23}
1. block9 {inst+26}
param_and_arg_refs_stack:
args_type_info_stack_:
```
And after:
```
2. Check::Context
NodeStack:
0. FunctionDefinitionStart: function2
1. ReturnStatementStart: no value
2. IntLiteral:
unexpected.inst+26.loc12_10: i32 = int_literal 0 [template = constants.%.2]
inst_block_stack_:
0. block<invalid> {
package: <namespace> = namespace [template] {
.Core = unexpected.inst+2
.F = unexpected.inst+23.loc11_22
}
unexpected.inst+1 = import Core
unexpected.inst+2: <namespace> = namespace unexpected.inst+1, [template] {}
unexpected.inst+23.loc11_22: %F.type = fn_decl @F [template = constants.%F] {
unexpected.inst+9.loc11_9: init type = call constants.%Bool() [template = bool]
unexpected.inst+10.loc11_9: type = value_of_initializer unexpected.inst+9.loc11_9 [template = bool]
unexpected.inst+11.loc11_9: type = converted unexpected.inst+9.loc11_9, unexpected.inst+10.loc11_9 [template = bool]
unexpected.inst+12.loc11_6: bool = param b
@F.%b: bool = bind_name b, unexpected.inst+12.loc11_6
unexpected.inst+19.loc11_18: init type = call constants.%Int32() [template = i32]
unexpected.inst+20.loc11_18: type = value_of_initializer unexpected.inst+19.loc11_18 [template = i32]
unexpected.inst+21.loc11_18: type = converted unexpected.inst+19.loc11_18, unexpected.inst+20.loc11_18 [template = i32]
@F.%return: ref i32 = var <return slot>
}
}
1. block9 {
unexpected.inst+26.loc12_10: i32 = int_literal 0 [template = constants.%.2]
}
param_and_arg_refs_stack:
args_type_info_stack_:
```
This executes on a TODO in AddImportRef to add instructions to their own
block instead of the File block. This has an important consequence of
removing a pattern from InstBlockStack that added to blocks not
currently at the top, cleaning up an issue for ArrayStack. The delta
here is then mostly in different formatting of the import refs, a
consequence of the separation.
By adding an `ImportDecl` instruction, this creates something that can
be referenced through `ImportIRInst`.
packages/no_prelude/implicit_imports_entities.carbon is getting a test
of this (import_conflict and import_conflict_reverse).
Also re-packs ImportIR from 24 bytes to 16 on 64-bit, since I'm touching
everywhere that makes one anyways.
Name scopes store the names in their scope in a `DenseMap`. Several
places reasonably avoid depending on the iteration order by sorting the
names -- they're in the formatting code path where that's a solid
approach.
Unfortunately, when we're importing one scope into another, we also need
to walk the entire scope and do something for each name. =[ This doesn't
seem like a great place to sort things to stabilize them.
I've switched to a fairly simplistic solution of having a vector of name
entries that can be iterated stably, and a separate map for lookups. I
didn't use the set-of-indices trick here because it's not clear that's
the right trade-off for a scope: likely a lot of small scopes here with
relatively hot name lookups. And the key here isn't a large or
dynamically sized thing that we're canonicalizing, it's a `NameId`. That
made me lean towards duplicating the name in the hashtable for lookup
and the vector for iteration.
I thought about a fancy approach of sorting the hashtable keys by their
values (the indices), but that would still require a bit of copying and
more code.
I also thought a bit about other optimizations, but decided to leave a
comment for now -- it's not obvious to me exactly how hot this is and
whether it's better served by faster lookups, being more memory dense,
etc. And that might involve more of an SOA layout change or some other
approach. Rather than do that here, and especially before switching
hashtables, I stuck with a simple approach to address the ordering.
---------
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
Merges handle_let.cpp and handle_variable.cpp into a single file in
order to take better advantage of commonalities.
Both still have a bunch of kludges, and there's still divergent handling
in handle_binding_pattern that will yield different results. However,
this fixes some things about `let` like starting to make use of
global_init (imperfectly, due to how VarStorage works, I grant), and in
particular adding `let` names to a name scope, not just lexical lookup.
Don't use the pretty-printed type name, because that's intended for
diagnostics, not for a theoretically machine-readable format like SemIR.
Types are always constants, so omit the leading `constant.` on the type
instruction name.
Don't format the *type of* the function `F` as `"F"`. Instead use
`"<type of F>"`.
Also improve instruction naming for function and generic class name
values: name the value after the function or generic class, and add a
`.type` suffix to the type.
This has is a nice-to-have for me. Frequently I want to run a specific
test, and end up digging through output to be able to copy-paste the run
line. This uses TIP lines to inject the command into the file when using
AUTOUPDATE.
Note, one of the reasons I want this is because "bazel test
//toolchain/testing:file_test --test_output=all" has been regularly
exceeding bazel's output limit for me (workaround is either opening the
output file or specifying an obscure output limit flag), making it a
little harder for me to get the commands. However, frequently I'm adding
a file and want to iterate on it, so that's really the use case I have
in mind here.
Modifies the core package and literal handling to use factory functions
for standard type literals.
Updates function/builtin/import.carbon to stop depending on the prelude,
since it would list all the impls in the core file. Updates
alias/builtins.carbon to be failing (the type values cannot be aliased).
Note, I assume this doesn't affect all the TODOs (like for i32.carbon),
but does cut some things down (and requires updating some tests that
have prelude name conflicts, but I think they were intended to be
updated this way).
I think the template files are simply an oversight. The fuzz files
should probably be updated because failing on the `package` line isn't
an interesting test of logic. Plus one minor comment edit.
Change syntax for package declaration to put the `impl` keyword at the
start and remove the `api` keyword.
To support this, rearrange processing of package, library, and import
declarations to use the general modifier handling support in declaration
parsing rather than special-case logic.
There is an ambiguity in `impl package.Foo as Bar`, which we resolve by
treating `package` as an introducer after a modifier only if it's not
followed by `.`.
This was to track use of a declaration after import, prior to a
redeclaration. Per [discussion on
Discord](https://discord.com/channels/655572317891461132/1217182321933815820/1236016521059237962),
we likely don't need this check due to the change in behavior of
`extern`.
Rather than potentially getting one of many `extern` decls and depending
on it by accident, it is now planned to be _required_ to be imported,
and the library doing a non-`extern` decl must _know_ it's importing the
`extern` decl. The stricter requirement on the library means it now
seems more reasonable to use the `extern` decl.
So kind of rolling back #3831, though keeping `ImportIRInstId` (at least
for now) and keeping `Loaded`/`Unloaded` terminology (seems a nicer
fit).
This removes the builtin FunctionType, replacing it with a FunctionType
instruction. The constant for a FunctionDecl is now a StructValue with
type of FunctionType.
Note this means a function declaration produces _both_ a type, and a
value of the type. This has some consequences in terms of circularity,
and makes the importing of function declarations a little more complex.
It'll get particularly peculiar for imports because of the behavior of
the reference, but that's a known issue due to other things such as
`alias`. The impact will hopefully be contained to
ResolvePrevInstForMerge (and ImportRefs).
To note a small formatting change in diagnostics:
```
- // CHECK:STDERR: fail_member_lookup.carbon:[[@LINE+4]]:3: ERROR: Value of type `<associated <function> in Interface>` is not callable.
+ // CHECK:STDERR: fail_member_lookup.carbon:[[@LINE+4]]:3: ERROR: Value of type `<associated F in Interface>` is not callable.
- // CHECK:STDERR: fail_todo_facet_lookup.carbon:[[@LINE+4]]:3: ERROR: Value of type `<associated <function> in Interface>` is not callable.
+ // CHECK:STDERR: fail_todo_facet_lookup.carbon:[[@LINE+4]]:3: ERROR: Value of type `<associated F in Interface>` is not callable.
```
---------
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
Change the names for emitted globals for constants with storage to
include both the name of the constant and the name of the use.
This causes the instructions to also be named in SemIR and in LLVM IR
constants.
We don't need it any more, and removing it simplifies a few things:
- One fewer predefined `File` and reserved ID.
- We now have simply `Builtin` instructions for builtins, instead of
having an `ImportRef` that indirectly references a `Builtin`.
- `ConstantId`s now always refer directly to a local constant, instead
of sometimes referring to an `ImportRef` for a constant in the builtins
IR.
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
First steps towards using constant values in lowering.
For now, we reuse the regular instruction lowering to lower constants.
This mostly works, because we don't actually need an `llvm::Function` or
a current basic block when lowering a constant most of the time.
However, a special case is needed for lowering aggregate value constants
because they would otherwise create a stack alloca to store the
constant. Separate constant lowering code will be added in a future
change to clean this up.
When lowering a constant initializing expression, the result is a value
of the destination type, rather than code to initialize the destination,
so a separate copy step is required when finishing initialization from a
constant for a type that uses in-place initialization. Handling this
required extending `ReturnExpr` to track its destination location.
We currently often create non-constant `*_access` SemIR instructions
that are only used by constant `*_init` instructions. These cause
lowering to leave behind `getelementptr` instructions in the lowered IR
that are now unused. It should be possible to detect this case and avoid
producing these instructions, or to produce them lazily, but for now
we're just leaving them around for LLVM to clean up.
Use a level comparison during substitution to determine whether we're
substituting a particular binding. Evaluate symbolic bindings with the
same name and the same level to the same symbolic constant, for example
across redeclarations of a generic function.
In handle_class and handle_interface we assign TypeType, so is more
consistent. I think this had been missed because we haven't really been
using these declarations (historically, declarations didn't have a
type). It seems not to significantly affect output at present, although
I found this while trying to change merge behavior.
Adds support for unary `-` and binary `+`, `-`, `*`, `/` for floating
point types.
Real literals are now transformed to `llvm::APFloat`s during the check
phase into the `FloatLiteral` instruction.
This PR likely collides a bit with #3892 and might need to be updated
when that one is merged.
Factor out `SemIR::InstNamer` and also use it when lowering to LLVM IR.
Automatically name all instructions created with our `IRBuilder` based
on the name computed by the `InstNamer`, and likewise name basic blocks
using the label generated by the `InstNamer`.
Move some of the existing naming logic out from lower into `InstNamer`
so that it's also used in SemIR. In particular, we now name call
instructions after their callee, or after the builtin name for calls to
builtins.
Computing and adding these names isn't completely free. This instruction
naming is designed to be optional, so that we can turn it off for builds
where the LLVM IR will only be converted to assembly and won't be seen
by a human, but so far it's enabled unconditionally. We can tune that
later as needed.
Use the complete location of the `as` conversion rather than the
location of the first operand, so diagnostics referring to the result
point at the whole thing.
Move completeness check to the point where the function is defined or
first called. This means we also defer deciding whether the function has
a return slot until that point. Instead of storing a return slot per
function, store the location of the return storage, which may or may not
be used, and compute and store a separate flag saying whether to use it
at the point of first use or definition.
This is the final piece in supporting simple `Make` functions in classes
as a replacement for constructors.
- Adds an empty prelude.carbon file
- Imports that file in any non-Core package file
- Adds --disable-prelude-import to avoid that
- Adds --exclude-dump-file-prefix to be able to hide files from dumping
- Used to hide core files (we can't do this by package name due to lex
dumps, for example)
- Restructures some tests to not rely on `i32`, particularly `alias`
tests (which rely on a name ref) and tests with no prelude.
I'm adding the framework for switching i32 to calling Int32 in the
prelude, but I'm running into a separate error actually switching over.
So that *mostly* works, but isn't quite ready for prime time. However,
maybe the current state of this PR is still useful to review since it
does a lot of the infrastructure work and adds the %Core everywhere?
Adds ImportIRId::ApiForImpl to reserve a specific slot for the `api`
import, so that the code can trivially determine whether an import is
from the same library. This is then used for merging function
declarations, because the rules for redeclarations in the same library
slightly differ as compared to other imports (note they're also not
identical to same-file rules).
The main thing this leaves from the recent #3762 is verifying that
entities forward declared in the `impl` file are also defined, but
that's not in-scope for merging; it's moreso post-checking validation.
Note, a lot of our `invalid <entity> ID` comments in ids.h were
incorrectly copy-pasted, so I've cut `<entity>`.
This doesn't significantly change logic, although I'm trying to add the
location to used state.
The issue I'm trying to address is how to identify a declaration as
"allowed to be redeclared". Consider:
```
library "a" api;
extern fn F();
```
```
library "b" api;
extern fn F();
```
```
library "c" api;
import library "a";
import library "b";
var x: auto = F();
fn F();
```
What currently happens is:
1. On import of "a", `F` becomes ImportRefUnused
2. On import of "b", `F` becomes ImportRefUsed in order to merge.
3. In "c", the call `F()` doesn't change the state.
4. In "c", the declaration `fn F();` needs some breadcrumb to understand
whether "F" has been referenced, as in step (3) here.
What I want to happen is:
1. On import of "a", `F` becomes ImportRefUnloaded
2. On import of "b", `F` becomes ImportRefLoaded in order to merge.
3. In "c", the call `F()` causes `F` to become ImportRefUsed
4. In "c", the declaration `fn F();` detects that `F` is already
ImportRefUsed, and can use the associated `used_id` for a diagnostic
about why redeclaring is invalid.
Note this PR isn't implementing (4). I'm focused on the refactoring to
add a new ImportRef state here.
These can't be supported by `TryEvalInst`, because we don't track
sufficient information about predecessors and branch conditions in SemIR
to efficiently compute the constant result. Even if we could do so, we
may not want to treat all `BlockArg`s for which we can infer a constant
value as being constants. Instead, set the constant value explicitly
after creating the corresponding `BlockArg` instruction.
In parse, form a list of methods that are defined inline, tracking where
they start, where they end, and which other inline methods are nested
within them.
In check, when we reach an inline method body, skip it and add it to a
worklist to be processed later. We also track when we reach the start
and end of a context in which inline method bodies are deferred, so that
we know when to replay the bodies.
When suspending a function definition to be processed later, the
`DeclNameStack` entry is moved to separate storage, including popping
the corresponding scopes from the scope stack and removing the
corresponding lexical names from lexical lookup. Later, when we return
to the function and parse its definition, the `DeclNameStack` entry is
restored. The same is done when we reach the end of a nested context
that can have inline methods, so that we can reenter the nested scope
before processing its members.
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
The purpose of the newline is to make it clearer where a given
diagnostic begins and ends, particularly as the first message of a
diagnostic may not be the error.
This is a trivial code change, but ripples edits through test files.
For now, a builtin function is defined by specifying a string literal
initializer in a function declaration:
```carbon
fn MyBuiltin(a: i32) -> i32 = "builtin.name";
```
End-to-end support is included for a sample `"int.add"` builtin
performing integer addition, covering constant evaluation and code
generation.
The implementation here needs substantial refactoring before we'll be
ready to start adding more builtins. That refactoring work will be
coming next. This change is aiming to checkpoint some incremental
progress.
Fix a collection of issues that were preventing lowering for overloaded
operators from working.
Instead of creating `import_ref` instructions during name lookup in the
current block, whatever that might be, we now create them in the `file`
block always. This avoids inserting them into blocks that might not be
intended to contain them, such as functions, and avoids the IR generated
for a function depending on which names we happen to have looked up
first.
When importing a class, function, or interface, import its enclosing
scope ID. This is necessary to allow us to distinguish between functions
at interface scope, which shouldn't be lowered, and other functions, and
will also be used in future to provide qualified names for declarations
when printing types. In order to support this:
- Track the constant values of namespaces created during importing so
that we can find them when resolving an import ref. Use those constant
values to convert an enclosing scope ID from the imported IR into a
corresponding ID in the current IR.
- Change how we do two-pass import of classes and namespaces so that we
can do two-pass import even for non-defining declarations, so that we
can import the enclosing scope.
While working on the final point above, I reworked `TryResolveInst` to
return a flag indicating whether another pass is necessary instead of
implicitly encoding this in the `ConstantId`. This permits the handling
of classes to be simplified; now `import_ir_constant_values` is only
accessed in a single place.
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
Support is added for all overloaded operator interfaces in the current
design apart from `Assign`, which is going to require some more work to
properly handle, given that primitive assignment currently has a special
implementation for quite a few builtin types.
As we don't have support for generics yet -- in particular, generic
interfaces -- there is no support for `*With` interfaces, but homogenous
interfaces such as `Add` are supported instead.
Factor out building of call expressions so that overloaded operators can
generate calls.
Switch a few places from using specific kinds of NodeId to a general
NodeId. Because overloaded operators and other things like implicit
conversions can result in member access and function calls, those
operations can't require a specific kind of NodeId.
Add import support for associated entities, and fix import support for
interfaces and symbolic bindings. We now import interfaces in two steps,
first importing a forward declaration then a definition, just like we do
for classes. For symbolic bindings, we ensure that each BindSymbolicName
is imported only once, because its ID is used as its symbolic identity.
This is necessary because we (only) support operator interfaces that are
defined in an imported Carbon package for now.
The entire contents of `check/operator.cpp` should probably be
rethought. In particular, doing a lot of name lookups on each operator
is likely to be bad for performance. But this gets us to the point where
overloaded operators are basically working, which seems like a good
place to iterate from.
For now, the tests that the individual operators map to the right
interfaces are mostly generated by a script, but that's just because I'm
expecting a fair bit of churn in how we define the prelude and the
`impl`s -- in particular, when we add support for `AddWith`, we'll need
to update all the tests. The plan is to remove the script once things
settle down.
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
Co-authored-by: Carbon Infra Bot <carbon-external-infra@google.com>
On the parsing side, we treat `a.(b)` as a member access whose second
operand is a `ParenExpr` rather than a `MemberName`. A new node category
is added for the union of `MemberName` and `ParenExpr` to support this.
Checking is mostly reusing the same pieces we already have for simple
member access. Compound member access is in most ways a simplified form
of simple member access because it doesn't need to do any lookup.
Previously we used the default inst block formatting, which writes out a
parenthesized list of references, which would always be 'unexpected
instref's because nothing else prints the instructions in the decl
block.
In addition, track the decl block for function declarations like we do
for other kinds of declaration. This means that the parameter
declarations for a function are now properly rendered into the formatted
IR. Note that this adds a lot of verbosity to `function_decl`, but it
does accurately reflect the IR, and we'll probably want this information
to be printed once we start supporting more complex generic function
declarations.
This removes almost all the 'unexpected instref's in our formatted
output. There are remaining cases when a declarative scope contains
multiple blocks, where we only track one of those blocks. That happens
when there is control flow within declarative scopes, and for error
recovery when a class or interface or similar is defined more than once.
As requested in #3730.
`GlobalInit` block is now static block within a `SemIR` which will be
used to emit initialization instructions for variables in the `Package`
scope.
inst_block_stack now has additional methods to handle `GlobalInit` block
separately, this block can be popped without being finalized allowing to
accumulate between all instances of variables.
At the end of the `check` phase, if this block is not empty , the
function `__global_init` will be added with this block being inserted
into it.
This block is pushed to `inst_block_scope` at the end `BindName`,
allowing instruction to be emitted into it, then popped at the semicolon
(VariableDecl).
This significantly changes the `SemIR` output, that's why this commit
updates a lot of the test cases.
Building on #3636 which handles the general import case, add special
casing for namespaces. Namespaces can be combined cross-IR, so it's a
little more complex.
The implementation adds import_id to the Namespace instruction as a
reference to find the original using the normal structure. This is
achieved by moving the name_id to NameScope to free up space.
---
I considered a few alternatives...
I considered adding import_id to NameScope, but:
1. It's more consistent with things such as Function or Class that
provide name_id on the info object rather than the instruction.
2. I thought it more likely that there would be more NameScope cases
that might want a name_id rather than the import_id, since ImportRef
will typically be used.
I considered putting the import source (cross-ref IR id + inst id) on
the NameScope versus a separate ImportRef, which seems like the
strongest argument towards the NameScope approach because it removes an
instruction. That just felt inconsistent though, and the overhead of
instruction-per-imported-namespace should be low (theoretically few
namespaces should be used). Plus I feel a bit odd adding two
generally-unused ids to NameScope.
A specialized Namespace structure could also have been created to store
the import_id, but that would add an indirection to the NameScope.
Do not create runtime name bindings for `FieldDecl`s even though they're
declared with `:`, so that we can still constant-evaluate references to
fields.
The constant value we associate with an initializing representation is
the object representation that the initializing expression will store to
its destination.
Also include the type in the profile of an instruction. This is now
necessary for array values, which are represented as tuple_value
instructions with array type, to avoid instructions with different types
being merged by constant canonicalization.
Remove the type canonicalization mechanism and instead rely on constant
canonicalization to deduplicate types.
Rename the `Canonicalize*Type` functions to reflect that they're no
longer performing canonicalization. Switch code that creates types due
to semantic checking, rather than due to source syntax, to directly
create type constants through evaluation rather than creating an
instruction and evaluating it to produce a separate constant
representation.
The mapping from `const (const T)` that was previously performed by type
canonicalization is now implemented in expression evaluation instead.
The value `<error>` is now treated as a constant value, with a special
property that an instruction involving `<error>` that could possibly be
constant evaluates to `<error>`. This helps avoid producing follow-on
errors when an error occurs as a subexpression of an expression, such as
a type, that is intended to be constant.
Rather than producing multiple constants with the same value, fold all
instances of a given constant to the same constant instruction.
A future PR will use this to replace the current type canonicalization
system.
Instructions created by splices during conversion are now evaluated, as
are instructions created in cases where we first create a placeholder
instruction and later replace it by a different instruction.
This also removes the ability to set a parse node and instruction
independently after creating an `InstId`, which could lead to them
accidentally not matching.
Richard Smith
Jon Ross-PerkinsandGeoff Romer
Chandler CarruthandRichard Smith
Pablo Paglilla
Gıyaseddin Tanrıkulu