Sorry for the mega-commit, this diff got a little out of control.
The main thing here is a complete rework of how Elf2 handles the symbol
table. I messed around with the design for a while and landed on
something which is fairly memory-efficient (in particular the overhead
for STB_LOCAL symbols is as low as possible) and fulfils some of the
more awkward constraints of the ELF format. The main such constraint is
that all STB_LOCAL symbols in a symbol table are required to appear
before any STB_GLOBAL/STB_WEAK symbols. This is further complicated by
the fact that when producing a DSO, symbols with STV_HIDDEN or
STV_INTERNAL visibility are required to have STB_LOCAL binding in the
symbol table, even though they are global symbols from the perspective
of the link editor. Plus, when combining multiple symbols with the same
name, the resulting visibility is the strictest of all of the inputs, so
it is possible at any point in compilation to discover an extern/export
symbol which forces an existing STB_GLOBAL symbol to become STB_LOCAL
and therefore requires it to move to an earlier symtab index. Dealing
with all of this was quite awkward.
But I got there! I also implemented a lot of features in the process. I
don't remember everything perfectly, but here's a vague list:
* Multiple definitions of and/or unresolved references to symbols are
now combined correctly in all cases
* `.bss` sections from inputs are correctly lowered (we don't actually
emit a `.bss` section of our own yet, but I was able to put that data
into the `.data` section so that the functionality is correct)
* Relocations in link inputs are now always processed (previously they
would be silently ignored in most cases)
* Linker errors are triggered if a supported input section has a
relocation which targets an unsupported input section (previously
the unsupported section's symbol was dropped and associated
relocations would be silently ignored)
* When linking a static executable, an error is emitted if a required
symbol (i.e. an undefined reference with strong linkage) was never
defined
* Duplicate symbol errors now work correctly
* When emitting a relocatable, the offsets of relocation entries are now
correct (previously the offsets written were relative to a symbol
rather than a section, meaning that e.g. almost all text relocations
were just in a single function)
The changes in all of the other linkers and codegen backends are some
added type-safety at the codegen-linker API boundary. There are now
distinct `u32`-backed types for identifying an "atom" (the thing we're
codegenning) and a "symbol" (the thing which a relocation targets).
Linker implementations can use a couple of private helper functions to
convert between this implementation-agnostic type and their specific
type; for instance, `Elf2` can convert between a `Symbol.Id` and a
`link.File.SymbolId` with `Symbol.Id.fromTypeErased` and
`Symbol.Id.toTypeErased`. I didn't implement this nicely for any other
linker, so right now there's a lot of `@enumFromInt`/`@intFromEnum`
sprinkled all over the place, particularly with the legacy ELF and
Mach-O linkers.
I tested that I could still perform incremental updates to the Zig
compiler using this commit. In terms of the new behaviors, the most
interesting stuff is symbol and relocation resolution, so I ran a few
tests involving building a "Hello World" binary in various different
ways:
* `build-exe` correctly succeeds
* `build-exe -fno-compiler-rt` correctly reports undefined symbols
* `build-obj` linked with `build-exe` correctly succeeds
* `build-obj` linked with `build-exe -fno-compiler-rt` correctly reports
undefined symbols
* `build-obj -fcompiler-rt` linked with `build-exe -fno-compiler-rt`
correctly succeeds
* `build-obj -fcompiler-rt` linked with `build-exe` correctly succeeds
(the compiler-rt symbols are weak so the global symbols are
arbitrarily resolved to one of the two implementations)
I also manually verified with `readelf` that symbol tables were always
ordered correctly (before this PR, `readelf -s` would usually emit
warnings about incorrectly-ordered symtabs!), and verified that various
visibility attributes worked as expected.
No actual test coverage is added due to the current lack of a useful
linker test harness. Once a good test harness is available I will be
willing to write some tests.
I've realised that the cause of at least some of our weird CI flakiness
was a bug in how `Nav` values were resolved. Consider this scenario: the
frontend resolves the type of a `Nav`, and then sends a function to the
backend, which requires the backend to lower a pointer to that `Nav`.
The backend calls `InternPool.getNav` to determine the `Nav`'s type.
However, this races with the frontend resolving the *value* of that
`Nav`. This involves writing separately to two fields, `bits` and
`type_or_value`. If only one of these changes is observed, then the
backend will incorrectly interpret the type as the value or vice versa,
leading to a crash or even a miscompilation. (Of course, there's also
the straightforward issue that the racing loads were non-atomic, making
them illegal).
The only good solution to this was to make `Nav` 4 bytes bigger, giving
it separate `type` and `value` fields. In theory that's a quite small
change, but it ended up having a bunch of nice consequences which led to
this diff being a bit bulkier than expected:
* `Nav.Repr.Bits` was simplified, because it no longer has to track
"resolution status": we can use `.none` for that. This frees up some
bits to make things more consistent between the "type resolved" and
"fully resolved" states.
* This consistency allowed the `Nav.status` union to be replaced with a
simpler field `Nav.resolved`, which is a bit nicer to work with.
* Most of the "getter" functions were able to be removed from `Nav`
because the state they were fetching had been moved to simple fields
on `Nav.resolved`.
* There were still a handful of free bits in `Nav.Repr.Bits`, which
could be used to represent the "const" and "threadlocal" flags rather
than these being stored on `Key.Extern` and `Key.Variable`. This is a
bit more convenient for linkers.
* With those bits gone, `Key.Variable` is a trivial wrapper around a
type and an initial value, and the fact that a declaration is mutable
can be represented solely through the "const" flag. Therefore,
`Key.Variable` no longer served a purpose, and could be eliminated
entirely in favour of storing the variable's initial value directly in
the "value" field of the `Nav`.
So, I'm quite pleased with this refactor! But anyway, regarding the bug
fix which actually motivated this: if I've done my job correctly, this
should solve some crashes, such as these (which were what tipped me off
to this bug in the first place):
https://codeberg.org/ziglang/zig/actions/runs/2306/jobs/7/attempt/1https://codeberg.org/ziglang/zig/actions/runs/2173/jobs/6/attempt/1
...and, who knows, perhaps even the random SIGSEGVs we've seen on some
targets! Probably not, but one can hope.
Introduces a small abstraction, `link.DebugConstPool`, to deal with
lowering type/value information into debug info when it may not be known
until type resolution (which in some cases will *never* happen). It is
currently only used by self-hosted DWARF logic, but it will also be of
use to the LLVM backend (which is my next focus).
Now that https://github.com/ziglang/zig/issues/24657 has been
implemented, the compiler can simplify its internal representation of
comptime-known `packed struct` and `packed union` values. Instead of
storing them field-wise, we can simply store their backing integer
value. This simplifies many operations and improves efficiency in some
cases.
This commit expands on the foundations laid by https://github.com/ziglang/zig/pull/23177
and moves even more `Sema`-only functionality from `Value`
to `Sema.arith`. Specifically all shift and bitwise operations,
`@truncate`, `@bitReverse` and `@byteSwap` have been moved and
adapted to the new rules around `undefined`.
Especially the comptime shift operations have been basically
rewritten, fixing many open issues in the process.
New rules applied to operators:
* `<<`, `@shlExact`, `@shlWithOverflow`, `>>`, `@shrExact`: compile error if any operand is undef
* `<<|`, `~`, `^`, `@truncate`, `@bitReverse`, `@byteSwap`: return undef if any operand is undef
* `&`, `|`: Return undef if both operands are undef, turn undef into actual `0xAA` bytes otherwise
Additionally this commit canonicalizes the representation of
aggregates with all-undefined members in the `InternPool` by
disallowing them and enforcing the usage of a single typed
`undef` value instead. This reduces the amount of edge cases
and fixes a bunch of bugs related to partially undefined vecs.
List of operations directly affected by this patch:
* `<<`, `<<|`, `@shlExact`, `@shlWithOverflow`
* `>>`, `@shrExact`
* `&`, `|`, `~`, `^` and their atomic rmw + reduce pendants
* `@truncate`, `@bitReverse`, `@byteSwap`
This experimental target was never fully completed. The operating system
is not that interesting or popular anyway, and the maintainer is no
longer around.
Not worth the maintenance burden. This code can be resurrected later if
it is worth it. In such case it will be subject to greater scrutiny.
My original goal here was just to get the self-hosted Wasm backend
compiling again after the pipeline change, but it turned out that from
there it was pretty simple to entirely eliminate the shared state
between `codegen.wasm` and `link.Wasm`. As such, this commit not only
fixes the backend, but makes it the second backend (after CBE) to
support the new 1:N:1 threading model.
As of this commit, every backend other than self-hosted Wasm and
self-hosted SPIR-V compiles and (at least somewhat) functions again.
Those two backends are currently disabled with panics.
Note that `Zcu.Feature.separate_thread` is *not* enabled for the fixed
backends. Avoiding linker references from codegen is a non-trivial task,
and can be done after this branch.
The idea here is that instead of the linker calling into codegen,
instead codegen should run before we touch the linker, and after MIR is
produced, it is sent to the linker. Aside from simplifying the call
graph (by preventing N linkers from each calling into M codegen
backends!), this has the huge benefit that it is possible to
parallellize codegen separately from linking. The threading model can
look like this:
* 1 semantic analysis thread, which generates AIR
* N codegen threads, which process AIR into MIR
* 1 linker thread, which emits MIR to the binary
The codegen threads are also responsible for `Air.Legalize` and
`Air.Liveness`; it's more efficient to do this work here instead of
blocking the main thread for this trivially parallel task.
I have repurposed the `Zcu.Feature.separate_thread` backend feature to
indicate support for this 1:N:1 threading pattern. This commit makes the
C backend support this feature, since it was relatively easy to divorce
from `link.C`: it just required eliminating some shared buffers. Other
backends don't currently support this feature. In fact, they don't even
compile -- the next few commits will fix them back up.
Each target can opt into different sets of legalize features.
By performing these transformations before liveness, instructions
that become unreferenced will have up-to-date liveness information.
Pointers to thread-local variables do not have their addresses known
until runtime, so it is nonsensical for them to be comptime-known. There
was logic in the compiler which was essentially attempting to treat them
as not being comptime-known despite the pointer being an interned value.
This was a bit of a mess, the check was frequent enough to actually show
up in compiler profiles, and it was very awkward for backends to deal
with, because they had to grapple with the fact that a "constant" they
were lowering might actually require runtime operations.
So, instead, do not consider these pointers to be comptime-known in
*any* way. Never intern such a pointer; instead, when the address of a
threadlocal is taken, emit an AIR instruction which computes the pointer
at runtime. This avoids lots of special handling for TLVs across
basically all codegen backends; of all somewhat-functional backends, the
only one which wasn't improved by this change was the LLVM backend,
because LLVM pretends this complexity around threadlocals doesn't exist.
This change simplifies Sema and codegen, avoids a potential source of
bugs, and potentially improves Sema performance very slightly by
avoiding a non-trivial check on a hot path.
Nothing interesting here; literally just the bare minimum so I can work on this
on and off in a branch without worrying about merge conflicts in the non-backend
code.
This commit makes some big changes to how we track state for Zig source
files. In particular, it changes:
* How `File` tracks its path on-disk
* How AstGen discovers files
* How file-level errors are tracked
* How `builtin.zig` files and modules are created
The original motivation here was to address incremental compilation bugs
with the handling of files, such as #22696. To fix this, a few changes
are necessary.
Just like declarations may become unreferenced on an incremental update,
meaning we suppress analysis errors associated with them, it is also
possible for all imports of a file to be removed on an incremental
update, in which case file-level errors for that file should be
suppressed. As such, after AstGen, the compiler must traverse files
(starting from analysis roots) and discover the set of "live files" for
this update.
Additionally, the compiler's previous handling of retryable file errors
was not very good; the source location the error was reported as was
based only on the first discovered import of that file. This source
location also disappeared on future incremental updates. So, as a part
of the file traversal above, we also need to figure out the source
locations of imports which errors should be reported against.
Another observation I made is that the "file exists in multiple modules"
error was not implemented in a particularly good way (I get to say that
because I wrote it!). It was subject to races, where the order in which
different imports of a file were discovered affects both how errors are
printed, and which module the file is arbitrarily assigned, with the
latter in turn affecting which other files are considered for import.
The thing I realised here is that while the AstGen worker pool is
running, we cannot know for sure which module(s) a file is in; we could
always discover an import later which changes the answer.
So, here's how the AstGen workers have changed. We initially ensure that
`zcu.import_table` contains the root files for all modules in this Zcu,
even if we don't know any imports for them yet. Then, the AstGen
workers do not need to be aware of modules. Instead, they simply ignore
module imports, and only spin off more workers when they see a by-path
import.
During AstGen, we can't use module-root-relative paths, since we don't
know which modules files are in; but we don't want to unnecessarily use
absolute files either, because those are non-portable and can make
`error.NameTooLong` more likely. As such, I have introduced a new
abstraction, `Compilation.Path`. This type is a way of representing a
filesystem path which has a *canonical form*. The path is represented
relative to one of a few special directories: the lib directory, the
global cache directory, or the local cache directory. As a fallback, we
use absolute (or cwd-relative on WASI) paths. This is kind of similar to
`std.Build.Cache.Path` with a pre-defined list of possible
`std.Build.Cache.Directory`, but has stricter canonicalization rules
based on path resolution to make sure deduplicating files works
properly. A `Compilation.Path` can be trivially converted to a
`std.Build.Cache.Path` from a `Compilation`, but is smaller, has a
canonical form, and has a digest which will be consistent across
different compiler processes with the same lib and cache directories
(important when we serialize incremental compilation state in the
future). `Zcu.File` and `Zcu.EmbedFile` both contain a
`Compilation.Path`, which is used to access the file on-disk;
module-relative sub paths are used quite rarely (`EmbedFile` doesn't
even have one now for simplicity).
After the AstGen workers all complete, we know that any file which might
be imported is definitely in `import_table` and up-to-date. So, we
perform a single-threaded graph traversal; similar to what
`resolveReferences` plays for `AnalUnit`s, but for files instead. We
figure out which files are alive, and which module each file is in. If a
file turns out to be in multiple modules, we set a field on `Zcu` to
indicate this error. If a file is in a different module to a prior
update, we set a flag instructing `updateZirRefs` to invalidate all
dependencies on the file. This traversal also discovers "import errors";
these are errors associated with a specific `@import`. With Zig's
current design, there is only one possible error here: "import outside
of module root". This must be identified during this traversal instead
of during AstGen, because it depends on which module the file is in. I
tried also representing "module not found" errors in this same way, but
it turns out to be much more useful to report those in Sema, because of
use cases like optional dependencies where a module import is behind a
comptime-known build option.
For simplicity, `failed_files` now just maps to `?[]u8`, since the
source location is always the whole file. In fact, this allows removing
`LazySrcLoc.Offset.entire_file` completely, slightly simplifying some
error reporting logic. File-level errors are now directly built in the
`std.zig.ErrorBundle.Wip`. If the payload is not `null`, it is the
message for a retryable error (i.e. an error loading the source file),
and will be reported with a "file imported here" note pointing to the
import site discovered during the single-threaded file traversal.
The last piece of fallout here is how `Builtin` works. Rather than
constructing "builtin" modules when creating `Package.Module`s, they are
now constructed on-the-fly by `Zcu`. The map `Zcu.builtin_modules` maps
from digests to `*Package.Module`s. These digests are abstract hashes of
the `Builtin` value; i.e. all of the options which are placed into
"builtin.zig". During the file traversal, we populate `builtin_modules`
as needed, so that when we see this imports in Sema, we just grab the
relevant entry from this map. This eliminates a bunch of awkward state
tracking during construction of the module graph. It's also now clearer
exactly what options the builtin module has, since previously it
inherited some options arbitrarily from the first-created module with
that "builtin" module!
The user-visible effects of this commit are:
* retryable file errors are now consistently reported against the whole
file, with a note pointing to a live import of that file
* some theoretical bugs where imports are wrongly considered distinct
(when the import path moves out of the cwd and then back in) are fixed
* some consistency issues with how file-level errors are reported are
fixed; these errors will now always be printed in the same order
regardless of how the AstGen pass assigns file indices
* incremental updates do not print retryable file errors differently
between updates or depending on file structure/contents
* incremental updates support files changing modules
* incremental updates support files becoming unreferenced
Resolves: #22696
Textual PTX is just assembly language like any other. And if we do ever add
support for emitting PTX object files after reverse engineering the bytecode
format, we'd be emitting ELF files like the CUDA toolchain. So there's really no
need for a special ObjectFormat tag here, nor linker code that treats it as a
distinct format.
Matthew Lugg