Considering all possible features are known by the linker during
compile-time, we can create arrays on the stack instead of
dynamically allocating hash maps. We use a simple bitset to determine
whether a feature is enabled or not, and from which object file
it originates. This allows us to make feature validation slightly
faster and use less runtime memory.
In the future this could be enhanced further by having a single
array instead with a more sophisticated bitset.
The list of features a Wasm object/binary file can emit can differ
from the list of cpu features. The reason for this is because the
"target_features" section also contains linker features. An example
of this is the "shared-mem" feature, which is a feature for the linker
and not that of the cpu target as defined by LLVM.
Adds a test for inferring features based on a different object file.
Also provides a test case where cpu features are explicitly set on
a library where the end result will output said target features.
When an object file or binary contains the target_features section
we can now parse and then dump its contents in string format so
we can use them in our linker tests to verify the features section.
When the result is not being stripped, we emit the `target_features`
section based on all the used features. This includes features
inferred from linked object files.
Considering we know all possible features upfront, we can use an
array and therefore do not have to dynamically allocate memory.
Using this trick we can also easily order all features based
the same ordering as found in `std.Target.wasm` which is the same
ordering used by LLVM and the like.
Verifies disallowed and used/required features. After verifying,
all errors will be emit to notify the user about incompatible
features. When the user did not define any featureset, we infer
the features from the linked objects instead.
These ifs were missing a case for f80 which should have shifted by one,
but we can just compute the correct value instead. Also, we want the
fractional bits to be a multiple of four, not the mantissa bits, since
the mantissa could have a leading one which we want to be separated.
This makes it easier to understand how control flow should happen in
various cases; already just by doing this it is revealed that
UndefinedSymbol and UndefinedSymbolReference should be merged, and that
MissingMainEntrypoint should be removed in favor of the ErrorFlags
mechanism thath we already have for missing the main entrypoint.
The main motivation for this change, however, is preventing a compile
error when there is conditional compilation inside linker
implementations, causing the flush() error set to depend on compilation
options. With this change, the error set is fixed, and, notably, the
`-Donly-c` flag no longer has compilation errors due to this error set.
The larger alignment on this platform means that long double reports
a sizeof 16 bytes, but it's underlying size is really just the 10
bytes of `f80`
C doesn't give us a way to see the "underlying" size of a type, so
this has to be caught by hand or by monitoring runtime memory. Luckily,
x86 and x86-64 are the only platforms that seem to use a non-power-of-two
type like this.
This option can be used to produce a C backend build of the self-hosted
compiler, which only has the C backend enabled. Once the C backend is
capable of self-hosting, this will be a way for us to replace our stage1
codebase with a C backend build of self-hosted, which we can then use
for bootstrapping. See #5246 for more details.
Using this option right now results in a crash because the C backend is
not yet passing all the behavior tests.
CMake recognizes the CMAKE_PREFIX_PATH environment variable for some
things, and also the CMAKE_PREFIX_PATH cache variable for other things.
However, it does not relate these two things, i.e. if the environment
variable is set, CMake does not populate the cache variable in a
corresponding manner. Some package systems, such as Homebrew, set the
environment variable but not the cache variable. Furthermore, the
environment variable follows the system path separator, such as ':' on
POSIX and ';' on Windows, but the cache variable follows CMake's array
behavior, i.e. always ';' for a separator.
Closes#13242
This value corresponds to clang/gcc's `__alignof` (rather than
`_Alignof` which reports the minimum alignment). We don't use this
information yet, but it might be useful for implementing ABIs so it
is included here.
The Zig LLVM backend emits calls to softfloat methods with the "standard
compiler-rt" names. Rather than add complexity to the backend and
have to synchronize the naming scheme across all targets, the simplest
fix is just to export these symbols under both the "standard" and the
platform-specific naming convention.
This documents status of routines and adds the next work item
"Decimal float library routines", which are only recommended for
binary data. Complete absence of tests is also documented.
This does not document the various aliases, e.g. those for ARM.
Missing Integer library routines:
- __addvsi3
- __addvdi3
- __addvti3
- __addvdi3
- __addvti3
- __subvsi3
- __subvdi3
- __subvti3
- __subvdi3
- __subvti3
- __mulvsi3
- __mulvdi3
- __mulvti3
- __mulvdi3
- __mulvti3
Missing floating library routines:
- __powisf2
- __powidf2
- __powitf2
- __powixf2
Missing routines for symbol-level compatibility to gcc:
- __ashlsi3
- __ashrsi3
- __lshrsi3
cmp.zig was accidently being referenced twice, rather than importing
memcmp.zig. This means that its symbols were also not included in
the generated compiler-rt output.
1. If an object file was not compiled with `MH_SUBSECTIONS_VIA_SYMBOLS`
such a hand-written ASM on x86_64, treat the entire object file as
not suitable for dead code stripping aka a GC root.
2. If there are non-extern relocs within a section, treat the entire
section as a root, at least temporarily until we work out the exact
conditions for marking the atoms live.
