For parameters and return types of functions with the C calling
convention, the LLVM backend now has a special lowering for the function
type that makes the function adhere to the C ABI. The AIR instruction
lowerings for call, ret, and ret_load are adjusted to bitcast the real
type to the ABI type if necessary.
More work on this will need to be done, however, this improvement is
enough that stage3 now passes all the same behavior tests that stage2
passes - notably, translate-c no longer has a segfault due to C ABI
issues with Zig's Clang C API wrapper.
This makes stage2 and stage3 have different cache namespaces, so that
building something with stage3 does not try to reuse the same cached
artifacts as were produced by stage2. This makes sense since the code
of stage3 is produced by the self-hosted compiler, whereas the code of
stage2 is produced by the bootstrap compiler. Note also that stage4 and
stage3 will share the same zig_backend, end hence cache namespace.
Ideally stage4 and stage3 are identical binaries, so this checks out.
Rather than allocating Decl objects with an Allocator, we instead allocate
them with a SegmentedList. This provides four advantages:
* Stable memory so that one thread can access a Decl object while another
thread allocates additional Decl objects from this list.
* It allows us to use u32 indexes to reference Decl objects rather than
pointers, saving memory in Type, Value, and dependency sets.
* Using integers to reference Decl objects rather than pointers makes
serialization trivial.
* It provides a unique integer to be used for anonymous symbol names,
avoiding multi-threaded contention on an atomic counter.
While this code probably could do with some love and a redesign,
this commit fixes the allocations by making sure we explicitly
pass an allocator where required, and we use arenas for temporary
or narrowly-scoped objects such as a `Die` (for `Die` in particular,
not every `FormValue` will be allocated - we could duplicate, or
we can use an arena which is the proposal of this commit).
So that people can start experimenting with compiling their projects
with the self-hosted compiler.
I expect this commit to be reverted after #89 is closed.
When the last instruction is a debug instruction, the type of it is void.
Similarly for 'noreturn' emit an 'unreachable' instruction to tell the wasm-validator
the path cannot be reached.
Also respect the '--strip' flag in the self-hosted wasm linker and not emit a 'name' section
when the flag is set to `true`.
This change adds support for locating the Zig executable and the library
and global cache directories, based on looking in the fixed "/zig" and
"/cache" directories.
Since our argv[0] on WASI is just the basename (any absolute/relative
path information is deleted by the runtime), there's very limited
introspection we can do on WASI, so we rely on these fixed directories.
These can be provided on the command-line using `--mapdir`, as follows:
```
wasmtime --mapdir=/cwd::. --mapdir=/cache::"$HOME/.cache/zig" --mapdir=/zig::./zig-out/ ./zig-out/bin/zig.wasm
```
According to Apple docs, the long double type is a double precision
IEEE754 binary floating-point type, which makes it identical to the
double type. This behavior contrasts to the standard specification,
in which a long double is a quad-precision, IEEE754 binary,
floating-point type.
Thus, we need to take this into account when using the compiler
intrinsics so that we select the correct function version for
FloatMulAdd.
Previously, the data segments were being aligned twice.
This caused us to overalign the segment and therefore allocate a much larger
size for each segment than was required. This fix ensures we align and set the size
just once, ensuring semantically correct binaries as well as smaller binaries.
