The main purpose of this commit is to prepare to implement support for
callconv(), align(), linksection(), and addrspace() annotations on
generic functions where the provided expression depends on comptime
parameters (making the function generic).
It's a rather involved change, so this commit only makes the necessary
changes to AstGen without regressing any behavior, and a follow-up
commit can finish the task by making the enhancements to Sema.
By my quick estimation, the new encoding for functions is a negligible
improvement - along the lines of 0.005% fewer total ZIR bytes on
average. Still, it's nice that this commit, while adding more
data into ZIR, actually ends up reducing the storage size thanks to a
slightly more sophisticated encoding.
Zir.Inst.ExtendedFunc is renamed to Zir.Inst.FuncFancy to eliminate
confusion about it being an extended instruction (it used to be but is
no longer). The encoding for this instruction is completely reworked.
The encoding for Zir.Inst.Func is also changed slightly - when the
return type body length is 1, then only a Zir.Inst.Ref is provided; not
a full body.
linksection() and addrspace() are now communicated via func_fancy ZIR
instruction rather than as part of the corresponding decl. This allows
their expressions to observe comptime parameters.
`@call` allows specifying the modifier explicitly, however it can still
appear in a context that overrides the modifier. This commit adds flags
to the BuiltinCall ZIR encoding. Since we have unused bits I also threw
in the ensure_result_used mechanism.
I also deleted a behavior test that was checking for bound function
behavior where I think stage2 behavior is correct and stage1 behavior
is incorrect.
ZIR instructions updated: atomic_load, atomic_rmw, atomic_store, cmpxchg
These no longer construct a pointer type as the result location. This
solves a TODO that was preventing the pointer from possibly being
volatile, as well as properly handling allowzero and addrspace.
It also allows the pointer to be over-aligned, which may be needed
depending on the target. As a consequence, the element type needs to be
communicated in the ZIR. This is done by strategically making one of the
operands be ResultLoc.ty instead of ResultLoc.coerced_ty if possible, or
otherwise explicitly adding elem_type into the ZIR encoding, such as in
the case of atomic_load.
The pointer type of atomic operations is now checked in Sema by coercing
it to an expected pointer type, that maybe over-aligned according to
target requirements.
Together with the previous commit, Zig now has smaller alignment for
large integers, depending on the target, and yet still has type safety
for atomic operations that specially require higher alignment.
The reason for having `@tan` is that we already have `@sin` and `@cos`
because some targets have machine code instructions for them, but in the
case that the implementation needs to go into compiler-rt, sin, cos, and
tan all share a common dependency which includes a table of data. To
avoid duplicating this table of data, we promote tan to become a builtin
alongside sin and cos.
ZIR: The tag enum is at capacity so this commit moves
`field_call_bind_named` to be `extended`. I measured this as one of
the least used tags in the zig codebase.
Fix libc math suffix for `f32` being wrong in both stage1 and stage2.
stage1: add missing libc prefix for float functions.
Sema avoids adding map entries for certain instructions such as
`set_eval_branch_quota` and `atomic_store`. This means that result
location semantics in AstGen must not emit any instructions that attempt
to use the result of any of these instructions.
This commit makes AstGen replace such instructions with
`Zir.Inst.Ref.void_value` if their result value ends up being
referenced.
This fixes a compiler crash when running std lib atomic tests.
* In semaStructFields and semaUnionFields we return error.GenericPoison
if one of the field types ends up being generic poison.
- This requires handling function calls and function types taking
this into account when calling `typeRequiresComptime` on the return
type.
* Unrelated: I noticed using Valgrind that struct reification did not
populate the `known_opv` field. After fixing it, the behavior tests
run Valgrind-clean.
* ZIR: use `@ptrCast` to cast between slices instead of exploiting
the fact that stage1 incorrectly allows `@bitCast` between slices.
- A future enhancement will make Zig support `@ptrCast` to directly
cast between slices.
* AstGen: restore the param_type ZIR instruction and pass it to the
expression for function call arguments. This does not solve the
problem for generic function parameters, but it catches stage2 up to
stage1 which also does not solve the problem for generic function
parameters.
- Most of the enhancements in this commit will still be needed for a
more sophisticated further improvement to handle generic function
types.
- In Sema, handling of `as` coercion recognizes the `var_args_param`
Type Tag and passes the operand through doing no coercion.
- That was the last ZIR tag and we are now using all 256 ZIR tags.
* AstGen: array init and struct init expressions use the anon form even
when the result location has a type. Prevents the type system
incorrectly believing, for example, that a tuple is actually an array
when the result location is a param_type of a function with `anytype`
parameter.
* Sema: add missing coercion in `unionInit` to coerce the init to the
corresponding union field type.
* `Value.fieldValue` now takes a type and does not take an allocator.
closes#11293
After this commit, stage2 passes all the parser tests.
Adds 2 new AIR instructions:
* dbg_var_ptr
* dbg_var_val
Sema no longer emits dbg_stmt AIR instructions when strip=true.
LLVM backend: fixed lowerPtrToVoid when calling ptrAlignment on
the element type is problematic.
LLVM backend: fixed alloca instructions improperly getting debug
location annotated, causing chaotic debug info behavior.
zig_llvm.cpp: fixed incorrect bindings for a function that should use
unsigned integers for line and column.
A bunch of C test cases regressed because the new dbg_var AIR
instructions caused their operands to be alive, exposing latent bugs.
Mostly it's just a problem that the C backend lowers mutable
and const slices to the same C type, so we need to represent that in the
C backend instead of printing two duplicate typedefs.
This uses a new ZIR inst `array_init_sent` (and a ref equivalent) to
represent array init expressions that terminate in a a sentinel value.
The sentienl value is the last value in the `MultiOp` payload. This
makes it a bit more awkward to deal with (lots of "len - 1") but makes
it so that the payload matches the fact that sentinels appear at the end
of arrays. However, this is not a hill I want to die on so if we want to
change it to index 0, I'm happy to do so.
This makes the following work properly:
try expect(@TypeOf([_:0]u8{}) == [0:0]u8);
* mul_add AIR instruction: use `pl_op` instead of `ty_pl`. The type is
always the same as the operand; no need to waste bytes redundantly
storing the type.
* AstGen: use coerced_ty for all the operands except for one which we
use to communicate the type.
* Sema: use the correct source location for requireRuntimeBlock in
handling of `@mulAdd`.
* native backends: handle liveness even for the functions that are
TODO.
* C backend: implement `@mulAdd`. It lowers to libc calls.
* LLVM backend: make `@mulAdd` handle all float types.
- improved fptrunc and fpext to handle f80 with compiler-rt calls.
* Value.mulAdd: handle all float types and use the `@mulAdd` builtin.
* behavior tests: revert the changes to testing `@mulAdd`. These
changes broke the test coverage, making it only tested at
compile-time.
Improved f80 support:
* std.math.fma handles f80
* move fma functions from freestanding libc to compiler-rt
- add __fmax and fmal
- make __fmax and fmaq only exported when they don't alias fmal.
- make their linkage weak just like the rest of compiler-rt symbols.
* removed `longDoubleIsF128` and replaced it with `longDoubleIs` which
takes a type as a parameter. The implementation is now more accurate
and handles more targets. Similarly, in stage2 the function
CTypes.sizeInBits is more accurate for long double for more targets.
The ZIR instruction `union_init_ptr` is renamed to `union_init`.
I made it always use by-value semantics for now, not taking the time to
invest in result location semantics, in case we decide to change the
rules for unions. This way is much simpler.
There is a new AIR instruction: union_init. This is for a comptime known
tag, runtime-known field value.
vector_init is renamed to aggregate_init, which solves a TODO comment.
* AstGen: remove the setBlockBodyEliding function. This is no longer
needed after 63788b2a511eb87974065a052e2436b0c6202544.
* Sema: store_to_block_ptr instruction is handled as
store_to_inferred_ptr or store, as necessary.
`const` declarations inside comptime blocks were not getting properly
evaluated at compile-time. To accomplish this there is a new ZIR
instruction, `alloc_inferred_comptime`. Actually we already had one
named that, but it got renamed to `alloc_inferred_comptime_mut` to match
the naming convention with the other similar instructions.
Clarify that `astgen.advanceSourceCursor` already increments absolute
values of the line and columns numbers; i.e., `GenZir.calcLine` is thus
not only obsolete but wrong by design.
Incidentally, this clean up allows for specifying the `FnDecl` line
numbers for DWARF use correctly as relative values with respect to
the start of the parent `Decl`. This `Decl` in turn has its line number
information specified relatively to its parent `Decl`, and so on, until
we reach the global scope.
AstGen:
* rename the known_has_bits flag to known_non_opv to make it better
reflect what it actually means.
* add a known_comptime_only flag.
* make the flags take advantage of identifiers of primitives and the
fact that zig has no shadowing.
* correct the known_non_opv flag for function bodies.
Sema:
* Rename `hasCodeGenBits` to `hasRuntimeBits` to better reflect what it
does.
- This function got a bit more complicated in this commit because of
the duality of function bodies: on one hand they have runtime bits,
but on the other hand they require being comptime known.
* WipAnonDecl now takes a LazySrcDecl parameter and performs the type
resolutions that it needs during finish().
* Implement comptime `@ptrToInt`.
Codegen:
* Improved handling of lowering decl_ref; make it work for
comptime-known ptr-to-int values.
- This same change had to be made many different times; perhaps we
should look into merging the implementations of `genTypedValue`
across x86, arm, aarch64, and riscv.
* AstGen: use Ast.zig helper methods to avoid copy pasting token counting logic
- take advantage of the `first_doc_comment` field we already have for
param AST nodes
* Add missing ZIR docs
The ZIR instructions `switch_capture_else` and `switch_capture_ref` are
removed because they are not needed. Instead, the prong index is set to
max int for the special prong.
Else prong with error sets is not handled yet.
Adds a new behavior test because there was not a prior on to cover only
the capture value of else on a switch.
Introduce `validate_array_init_comptime`, similar to
`validate_struct_init_comptime` introduced in
713d2a9b3883942491b40738245232680877cc66.
`zirValidateArrayInit` is improved to detect comptime array literals and
emit AIR accordingly. This code is very similar to the changes
introduced in that same commit for `zirValidateStructInit`.
The C backend needed some improvements to continue passing the same set
of tests:
* `resolveInst` for arrays now will add a local `static const` with the
array value and so then `elem_val` instructions reference that local.
It memoizes accesses using `value_map`, which is changed to use
`Air.Inst.Ref` as the key rather than `Air.Inst.Index`.
* This required a mechanism for writing to a "header" which is lines
that appear at the beginning of a function body, before everything
else.
* dbg_stmt output comments rather than `#line` directives.
TODO comment reproduced here:
We need to re-evaluate whether to emit these or not. If we naively emit
these directives, the output file will report bogus line numbers because
every newline after the #line directive adds one to the line.
We also don't print the filename yet, so the output is strictly unhelpful.
If we wanted to go this route, we would need to go all the way and not output
newlines until the next dbg_stmt occurs.
Perhaps an additional compilation option is in order?
`Value.elemValue` is improved to support `elem_ptr` values.
Add a variant of the `validate_struct_init` ZIR instruction:
`validate_struct_init_comptime` which is the same thing except it
indicates a comptime scope.
Sema code for this instruction now handles default struct field
values and detects when the struct initialization resulted in a
comptime value, replacing the already-emitted AIR instructions
to store each individual field with a single `store` instruction
with a comptime struct value as the operand.
In the case of a comptime scope, there is a simpler path that only
evals the implicit store instructions for default field values, avoiding
the mechanism for detecting comptime values.
This regressed one test case for the wasm backend, but it's just hitting
a different prong of `emitConstant` which currently has "TODO" in there,
so I think it's fine.
