* Sema: implement peer type resolution for optionals and null.
* Rename `Module.optionalType` to `Type.optional`.
* LLVM backend: re-use anonymous values. This is especially useful when
isByRef()=true because it means re-using the same generated LLVM globals.
* LLVM backend: rework the implementation of is_null and is_non_null
AIR instructions. Generate slightly better LLVM code, and also fix
the behavior for optionals whose payload type is 0-bit.
* LLVM backend: improve `cmp` AIR instruction lowering to support
pointer-like optionals.
* `Value`: implement support for equality-checking optionals.
* LLVM backend: The `alloc` AIR instruction as well as pointer
constants which point to a 0-bit element type now call a common
codepath to produce a `*const llvm.Value` which is a non-zero pointer
with a bogus-but-properly-aligned address.
* LLVM backend: improve the lowering of optional types.
* Type: `hasCodeGenBits()` now returns `true` for pointers even when
it returns `false` for their element types.
Effectively, #6706 is now implemented in stage2 but not stage1.
* std.os: take advantage of `@minimum`. It's probably time to
deprecate `std.min` and `std.max`.
* New AIR instructions: min and max
* Introduce SIMD vector support to stage2
* Add `@Type` support for vectors
* Sema: add `checkSimdBinOp` which can be re-used for other arithmatic
operators that want to support vectors.
* Implement coercion from vectors to arrays.
- In backends this is handled with bitcast for vector to array,
however maybe we want to reduce the amount of branching by
introducing an explicit AIR instruction for it in the future.
* LLVM backend: implement lowering vector types
* Sema: Implement `slice.ptr` at comptime
* Value: improve `numberMin` and `numberMax` to support floats in
addition to integers, and make them behave properly in the presence
of NaN.
* `Module.Union.getFullyQualifiedName` returns a sentinel-terminated
slice so that backends that need null-termination do not need an
additional copy.
* Module.Union: implement a `getLayout` function which returns
information about ABI size and alignment so that the LLVM backend can
properly lower union types into llvm types.
* Sema: `resolveType` now returns `error.GenericPoison` rather than a
Type with tag `generic_poison`. Callsites that want to allow that
need to bypass this higher-level function.
* Sema: implement coercion of enums and enum literals to unions.
* Sema: fix comptime mutation of pointers to unions
* LLVM backend: fully implement proper lowering of union types and
values according to the union layout, and update the handling of AIR
instructions that deal with unions to support union layouts.
* LLVM backend: handle `decl_ref_mut`
- Maybe this should be unreachable since comptime vars should be
changed to be non-mutable when they go out of scope, but it's
harmless for the LLVM backend to support lowering the value.
* Type: fix `requiresComptime` for optionals, pointers, and some other
types. This function is still wrong for structs, unions, and enums.
* Make `alloc` AIR instructions call `resolveTypeLayout`.
* `Sema.zirResolveInferredAlloc` now calls `requireRuntimeBlock` in the
case that it operates on a non-comptime instruction.
* `Type.abiSize` and `Type.abiAlignment` now return 0 for `void`
* Sema: implement `resolveTypeFields` for unions.
* LLVM Backend: support `ptr_elem_ptr` when the element type is 0-bit.
* Type: improve `abiAlignment` implementation for structs to properly
handle fields with non-default alignment.
* Value: implement hashing array, vector, and structs.
The logic for `buildAlloca` had a null deref when the latest alloca was
the last instruction in the entry block. Now the logic is simplified to
always insert alloca instructions first before all other instructions.
There is no longer a need to track `entry_block` or `latest_alloca_inst`;
these fields are deleted frem `FuncGen`.
* ZIR: the `array_type_sentinel` now has a source node attached to it
for proper error reporting.
* Refactor: move `Module.arrayType` to `Type.array`
* Value: the `bytes` and `array` tags now include the sentinel, if the
type has one. This simplifies comptime evaluation logic.
* Sema: fix `zirStructInitEmpty` to properly handle when the type is
void or a sentinel-terminated array. This handles the syntax `void{}`
and `[0:X]T{}`.
* Sema: fix the logic for reporting "cannot store runtime value in
compile time variable" as well as for emitting a runtime store when a
pointer value is comptime known but it is a global variable.
* Sema: implement elemVal for double pointer to array. This can happen
with this code for example: `var a: *[1]u8 = undefined; _ = a[0];`
* Sema: Rework the `storePtrVal` function to properly handle nested
structs and arrays.
- Also it now handles comptime stores through a bitcasted pointer.
When the pointer element type and the type according to the Decl
don't match, the element value is bitcasted before storage.
* Add AIR instructions: ret_ptr, ret_load
- This allows Sema to be blissfully unaware of the backend's decision
to implement by-val/by-ref semantics for struct/union/array types.
Backends can lower these simply as alloc, load, ret instructions,
or they can take advantage of them to use a result pointer.
* Add AIR instruction: array_elem_val
- Allows for better codegen for `Sema.elemVal`.
* Implement calculation of ABI alignment and ABI size for unions.
* Before appending the following AIR instructions to a block,
resolveTypeLayout is called on the type:
- call - return type
- ret - return type
- store_ptr - elem type
* Sema: fix memory leak in `zirArrayInit` and other cleanups to this
function.
* x86_64: implement the full x86_64 C ABI according to the spec
* Type: implement `intInfo` for error sets.
* Type: implement `intTagType` for tagged unions.
The Zig type tag `Fn` is now used exclusively for function bodies.
Function pointers are modeled as `*const T` where `T` is a `Fn` type.
* The `call` AIR instruction now allows a function pointer operand as
well as a function operand.
* Sema now has a coercion from function body to function pointer.
* Function type syntax, e.g. `fn()void`, now returns zig tag type of
Pointer with child Fn, rather than Fn directly.
- I think this should probably be reverted. Will discuss the lang
specs before doing this. Idea being that function pointers would
need to be specified as `*const fn()void` rather than `fn() void`.
LLVM backend:
* Enable calling the panic handler (previously this just
emitted `@breakpoint()` since the backend could not handle the panic
function).
* Implement sret
* Introduce `isByRef` and implement it for structs and arrays. Types
that are `isByRef` are now passed as pointers to functions, and e.g.
`elem_val` will return a pointer instead of doing a load.
* Move the function type creating code from `resolveLlvmFunction` to
`llvmType` where it belongs; now there is only 1 instance of this
logic instead of two.
* Add the `nonnull` attribute to non-optional pointer parameters.
* Fix `resolveGlobalDecl` not using fully-qualified names and not using
the `decl_map`.
* Implement `genTypedValue` for pointer-like optionals.
* Fix memory leak when lowering `block` instruction and OOM occurs.
* Implement volatile checks where relevant.
* Sema: fix a missing copy on enum tag values
* LLVM backend: fix lowering of enum constant values for enums with
specified tag values.
* Value: fix enumToInt for `enum_numbered` cases.
The float widening behavior tests which rely on compiler-rt symbols are
now passing.
* Update AIR instruction `intcast` to allow the dest type to be an
enum.
* LLVM backend: update `intcast` to support when the bit counts of
operand and dest type are the same. This was already a requirement of
the instruction previously.
* Type: `intInfo` supports the case when the type is an enum, and
retrieves the info for the integer tag type. This makes it pretty
easy for backends to implement `intcast` without having to care
explicitly that the new type is an enum. As a bonus, simple enums
never have to go through the type system; their signedness and bit
count are computed directly.
The "int to enum" behavior test case is now passing for stage2 in the
LLVM backend.
* work around a stage1 miscompilation leading to the wrong integer
comparison predicate being emitted.
* fix the bug of not annotating callsites with the calling convention
of the callee, leading to undefined behavior.
* add the `nobuiltin` attribute when building freestanding libc or
compiler_rt libraries to prevent e.g. memcpy from being "optimized"
into a call to itself.
* compiler-rt: change a call to be comptime to make the generated LLVM
IR simpler and easier to study.
I still can't enable the widening tests due to the compiler-rt compare
function being miscompiled in some not-yet-diagnosed way.
Before, Sema for comptime `@bitCast` would return the same Value but
change the Type. This gave invalid results because, for example, an
integer Value when the Type is a float would be interpreted numerically,
but `@bitCast` needs it to reinterpret how they would be stored in
memory.
This requires a mechanism to serialize a Value to a byte buffer and
deserialize a Value from a byte buffer.
Not done yet, but needs to happen: comptime dereferencing a pointer
to a Decl needs to perform a comptime bitcast on the loaded value.
Currently the value is silently wrong in the same way that `@bitCast`
was silently wrong before this commit.
The logic in Value for handling readFromMemory for large integers is
only correct for small integers. It needs to be fleshed out for proper
big integers.
As part of this change:
* std.math.big.Int: initial implementations of readTwosComplement and
writeTwosComplement. They only support bit_count <= 128 so far and
panic otherwise.
* compiler-rt: move the compareXf2 exports over to the stage2 section.
Even with the improvements in this commit, I'm still seeing test
failures in the widening behavior tests; more investigation is
needed.
Decl names are now rendered fully qualified, so that e.g. "main" in two
different files does not conflict with each other.
Additionally, they are now mangled so as to not generate invalid C
identifier names.
Companion commit to 61a53a587558ff1fe1b0ec98bb424022885edccf.
This commit also moves over a bunch of behavior test cases to the
passing-for-stage2 section.
Also improve the LLVM backend to support lowering bigints to LLVM
values.
Moves over a bunch of math.zig test cases to the "passing for stage2"
section.
* Remove the builtins `@addWithSaturation`, `@subWithSaturation`,
`@mulWithSaturation`, and `@shlWithSaturation` now that we have
first-class syntax for saturating arithmetic.
* langref: Clarify the behavior of `@shlExact`.
* Ast: rename `bit_shift_left` to `shl` and `bit_shift_right` to `shr`
for consistency.
* Air: rename to include underscore separator with consistency with
the rest of the ops.
* Air: add shl_exact instruction
* Use non-extended tags for saturating arithmetic, to keep it
simple so that all the arithmetic operations can be done the same
way.
- Sema: unify analyzeArithmetic with analyzeSatArithmetic
- implement comptime `+|`, `-|`, and `*|`
- allow float operands to saturating arithmetic
* `<<|` allows any integer type for the RHS.
* C backend: fix rebase conflicts
* LLVM backend: reduce the amount of branching for arithmetic ops
* zig.h: fix magic number not matching actual size of C integer types
- modify AstGen binOpExt()/assignBinOpExt() to accept generic extended payload T
- rework Sema zirSatArithmetic() to use existing sema.analyzeArithmetic() by adding an `opt_extended` parameter.
- add airSatOp() to codegen/c.zig
- add saturating functions to src/link/C/zig.h
- adds initial support for the operators +|, -|, *|, <<|, +|=, -|=, *|=, <<|=
- uses operators in addition to builtins in behavior test
- adds binOpExt() and assignBinOpExt() to AstGen.zig. these need to be audited
* AIR: add `mod` instruction for modulus division
- Implement for LLVM backend
* Sema: implement `@mod`, `@rem`, and `%`.
* Sema: fix comptime switch evaluation
* Sema: implement comptime shift left
* Sema: fix the logic inside analyzeArithmetic to handle all the
nuances between the different mathematical operations.
- Implement comptime wrapping operations
* AIR: add `get_union_tag` instruction
- implement in LLVM backend
* Sema: implement == and != for union and enum literal
- Also implement coercion from union to its own tag type
* Value: implement hashing for union values
The motivating example is this snippet:
comptime assert(@typeInfo(T) == .Float);
This was the next blocker for stage2 building compiler-rt.
Now it is switch at compile-time on an integer.
* AIR instructions struct_field_ptr and related functions now are also
emitted by the frontend for unions. Backends must inspect the type
of the pointer operand to lower the instructions correctly.
- These will be renamed to `agg_field_ptr` (short for "aggregate") in
the future.
* Introduce the new `set_union_tag` AIR instruction.
* Introduce `Module.EnumNumbered` and associated `Type` methods. This
is for enums which have no decls, but do have the possibility of
overriding the integer tag type and tag values.
* Sema: Implement support for union tag types in both the
auto-generated and explicitly-provided cases, as well as explicitly
provided enum tag values in union declarations.
* LLVM backend: implement lowering union types, union field pointer
instructions, and the new `set_union_tag` instruction.
* LLVM backend: respect `sub_path` just like the other stage2 backends
do.
* Compilation has some new logic to only emit work queue jobs for
building stuff when it believes itself to be capable. The linker
backends no longer have duplicate logic; instead they respect the
optional bit on the respective asset.
Introduce an explicit decl_map for *Decl to LLVMValueRef. Doc comment
reproduced here:
Ideally we would use `llvm_module.getNamedFunction` to go from *Decl to
LLVM function, but that has some downsides:
* we have to compute the fully qualified name every time we want to do the lookup
* for externally linked functions, the name is not fully qualified, but when
a Decl goes from exported to not exported and vice-versa, we would use the wrong
version of the name and incorrectly get function not found in the llvm module.
* it works for functions not all globals.
Therefore, this table keeps track of the mapping.
Non-exported functions now use fully-qualified symbol names.
`Module.Decl.getFullyQualifiedName` now returns a sentinel-terminated
slice which is useful to pass to LLVMAddFunction.
Instead of using aliases for all external symbols, now the LLVM backend
takes advantage of LLVMSetValueName to rename functions that become
exported. Aliases are still used for the second and remaining exports.
freeDecl is now handled properly in the LLVM backend, deleting the
LLVMValueRef corresponding to the Decl being deleted. The linker
backends for ELF, COFF, Mach-O, and Wasm had to be updated to forward
the freeDecl call to the LLVM backend.
Extracts lib/std/special/c_stage1.zig from lib/std/special/c.zig.
When the self-hosted compiler is further along, all the logic from c_stage1.zig will
be migrated back c.zig and then c_stage1.zig will be deleted. Until then we have a
simpler implementation of c.zig that only uses features already implemented in self-hosted.
So far it only contains memcpy and memset, with slightly different
(arguably more correct!) implementations that are compatible with
self-hosted.
Additionally, this commit improves the LLVM backend:
* use the more efficient and convenient fnInfo() when lowering function
type info.
* fix incremental compilation not deleting all basic blocks of a
function.
* hook up calling conventions
* hook up the following function attributes:
- noredzone, nounwind, uwtable, minsize, optsize, sanitize_thread
* AstGen: fix not emitting `struct_init_empty` when an explicit type is
present in struct initialization syntax.
* AstGen: these two syntaxes now lower to identical ZIR:
- `var a = A{ .b = c };`
- `var a = @as(A, .{ .b = c });`
* Zir: clarify `auto_enum_tag` in the doc comments.
* LLVM Backend: fix lowering of function return types when the type has
0 bits.
* prepare compiler-rt to support being compiled by stage2
- put in a few minor workarounds that will be removed later, such as
using `builtin.stage2_arch` rather than `builtin.cpu.arch`.
- only try to export a few symbols for now - we'll move more symbols
over to the "working in stage2" section as they become functional
and gain test coverage.
- use `inline fn` at function declarations rather than `@call` with an
always_inline modifier at the callsites, to avoid depending on the
anonymous array literal syntax language feature (for now).
* AIR: replace floatcast instruction with fptrunc and fpext for
shortening and widening floating point values, respectively.
* Introduce a new ZIR instruction, `export_value`, which implements
`@export` for the case when the thing to be exported is a local
comptime value that points to a function.
- AstGen: fix `@export` not properly reporting ambiguous decl
references.
* Sema: handle ExportOptions linkage. The value is now available to all
backends.
- Implement setting global linkage as appropriate in the LLVM
backend. I did not yet inspect the LLVM IR, so this still needs to
be audited. There is already a pending task to make sure the alias
stuff is working as intended, and this is related.
- Sema almost handles section, just a tiny bit more code is needed in
`resolveExportOptions`.
* Sema: implement float widening and shortening for both `@floatCast`
and float coercion.
- Implement the LLVM backend code for this as well.
Previously, linker backends or machine code backends were able to hold
on to references to inside Sema's temporary arena. However there can
be large objects stored there that we want to free after machine code is
generated.
The primary change in this commit is to use a temporary arena for Sema
of function bodies that gets freed after machine code backend finishes
handling `updateFunc` (at the same time that Air and Liveness get freed).
The other changes in this commit are fixing issues that fell out from
the primary change.
* The C linker backend is rewritten to handle updateDecl and updateFunc
separately. Also, all Decl updates get access to typedefs and
fwd_decls, not only functions.
* The C linker backend is updated to the new API that does not depend
on allocateDeclIndexes and does not have to handle garbage collected
decls.
* The C linker backend uses an arena for Type/Value objects that
`typedefs` references. These can be garbage collected every so often
after flush(), however that garbage collection code is not
implemented at this time. It will be pretty simple, just allocate a
new arena, copy all the Type objects to it, update the keys of the
hash map, free the old arena.
* Sema: fix a handful of instances of not copying Type/Value objects
from the temporary arena into the appropriate Decl arena.
* Type: fix some function types not reporting hasCodeGenBits()
correctly.
There were two things to resolve here:
* Snektron's branch edited Zir printing, but in master branch
I moved the printing code from Zir.zig to print_zir.zig. So that
just had to be moved over.
* In master branch I fleshed out coerceInMemory a bit more, which
caused one of Snektron's test cases to fail, so I had to add
addrspace awareness to that. Once I did that the tests passed again.
