Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See #90357 for more info
and a repro case.
This commit is a mechanical revert of
d5de680e20 from PR #89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)
I have locally verified that this fixes#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
Direct users towards using Rust target feature names in CLI
This PR consists of a couple of changes on how we handle target features.
In particular there is a bug-fix wherein we avoid passing through features that aren't prefixed by `+` or `-` to LLVM. These appear to be causing LLVM to assert, which is pretty poor a behaviour (and also makes it pretty clear we expect feature names to be prefixed).
The other commit, I anticipate to be somewhat more controversial is outputting a warning when users specify a LLVM-specific, or otherwise unknown, feature name on the CLI. In those situations we request users to either replace it with a known Rust feature name (e.g. `bmi` -> `bmi1`) or file a feature request. I've a couple motivations for this: first of all, if users are specifying these features on the command line, I'm pretty confident there is also a need for these features to be usable via `#[cfg(target_feature)]` machinery. And second, we're growing a fair number of backends recently and having ability to provide some sort of unified-ish interface in this place seems pretty useful to me.
Sponsored by: standard.ai
Implement raw-dylib support for windows-gnu
Add support for `#[link(kind = "raw-dylib")]` on windows-gnu targets. Work around binutils's linker's inability to read import libraries produced by LLVM by calling out to the binutils `dlltool` utility to create an import library from a temporary .DEF file; this approach is effectively a slightly refined version of `@mati865's` earlier attempt at this strategy in PR #88801. (In particular, this attempt at this strategy adds support for `#[link_ordinal(...)]` as well.)
In support of #58713.
This was originally introduced in #10916 as a way to remove all landing
pads when performing LTO. However this is no longer necessary today
since rustc properly marks all functions and call-sites as nounwind
where appropriate.
In fact this is incorrect in the presence of `extern "C-unwind"` which
must create a landing pad when compiled with `-C panic=abort` so that
foreign exceptions are caught and properly turned into aborts.
In #79570, `-Z split-dwarf-kind={none,single,split}` was replaced by `-C
split-debuginfo={off,packed,unpacked}`. `-C split-debuginfo`'s packed
and unpacked aren't exact parallels to single and split, respectively.
On Unix, `-C split-debuginfo=packed` will put debuginfo into object
files and package debuginfo into a DWARF package file (`.dwp`) and
`-C split-debuginfo=unpacked` will put debuginfo into dwarf object files
and won't package it.
In the initial implementation of Split DWARF, split mode wrote sections
which did not require relocation into a DWARF object (`.dwo`) file which
was ignored by the linker and then packaged those DWARF objects into
DWARF packages (`.dwp`). In single mode, sections which did not require
relocation were written into object files but ignored by the linker and
were not packaged. However, both split and single modes could be
packaged or not, the primary difference in behaviour was where the
debuginfo sections that did not require link-time relocation were
written (in a DWARF object or the object file).
This commit re-introduces a `-Z split-dwarf-kind` flag, which can be
used to pick between split and single modes when `-C split-debuginfo` is
used to enable Split DWARF (either packed or unpacked).
Signed-off-by: David Wood <david.wood@huawei.com>
Allow loading LLVM plugins with both legacy and new pass manager
Opening a draft PR to get feedback and start discussion on this feature. There is already a codegen option `passes` which allow giving a list of LLVM pass names, however we currently can't use a LLVM pass plugin (as described here : https://llvm.org/docs/WritingAnLLVMPass.html), the only available passes are the LLVM built-in ones.
The proposed modification would be to add another codegen option `pass-plugins`, which can be set with a list of paths to shared library files. These libraries are loaded using the LLVM function `PassPlugin::Load`, which calls the expected symbol `lvmGetPassPluginInfo`, and register the pipeline parsing and optimization callbacks.
An example usage with a single plugin and 3 passes would look like this in the `.cargo/config`:
```toml
rustflags = [
"-C", "pass-plugins=/tmp/libLLVMPassPlugin",
"-C", "passes=pass1 pass2 pass3",
]
```
This would give the same functionality as the opt LLVM tool directly integrated in rust build system.
Additionally, we can also not specify the `passes` option, and use a plugin which inserts passes in the optimization pipeline, as one could do using clang.
Remove `in_band_lifetimes` from `rustc_codegen_llvm`
See #91867 for more information.
This one took a while. This crate has dozens of functions not associated with any type, and most of them were using in-band lifetimes for `'ll` and `'tcx`.
Apply path remapping to DW_AT_GNU_dwo_name when producing split DWARF
`--remap-path-prefix` doesn't apply to paths to `.o` (in case of packed) or `.dwo` (in case of unpacked) files in `DW_AT_GNU_dwo_name`. GCC also has this bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91888
Use module inline assembly to embed bitcode
In LLVM 14, our current method of setting section flags to avoid
embedding the `.llvmbc` section into final compilation artifacts
will no longer work, see issue #90326. The upstream recommendation
is to instead embed the entire bitcode using module-level inline
assembly, which is what this change does.
I've kept the existing code for platforms where we do not need to
set section flags, but possibly we should always be using the
inline asm approach (which would have to look a bit different for MachO).
r? `@nagisa`
In LLVM 14, our current method of setting section flags to avoid
embedding the `.llvmbc` section into final compilation artifacts
will no longer work, see issue #90326. The upstream recommendation
is to instead embed the entire bitcode using module-level inline
assembly, which is what this change does.
I've kept the existing code for platforms where we do not need to
set section flags, but possibly we should always be using the
inline asm approach.
The default diagnostic handler considers all remarks to be disabled by
default unless configured otherwise through LLVM internal flags:
`-pass-remarks`, `-pass-remarks-missed`, and `-pass-remarks-analysis`.
This behaviour makes `-Cremark` ineffective on its own.
Fix this by configuring a custom diagnostic handler that enables
optimization remarks based on the value of `-Cremark` option. With
`-Cremark=all` enabling all remarks.
Record more artifact sizes during self-profiling.
This PR adds artifact size recording for
- "linked artifacts" (executables, RLIBs, dylibs, static libs)
- object files
- dwo files
- assembly files
- crate metadata
- LLVM bitcode files
- LLVM IR files
- codegen unit size estimates
Currently the identifiers emitted for these are hard-coded as string literals. Is it worth adding constants to https://github.com/rust-lang/measureme/blob/master/measureme/src/rustc.rs instead? We don't do that for query names and the like -- but artifact kinds might be more stable than query names.
The only reason to use `abort_if_errors` is when the program is so broken that either:
1. later passes get confused and ICE
2. any diagnostics from later passes would be noise
This is never the case for lints, because the compiler has to be able to deal with `allow`-ed lints.
So it can continue to lint and compile even if there are lint errors.
Add -Z no-unique-section-names to reduce ELF header bloat.
This change adds a new compiler flag that can help reduce the size of ELF binaries that contain many functions.
By default, when enabling function sections (which is the default for most targets), the LLVM backend will generate different section names for each function. For example, a function `func` would generate a section called `.text.func`. Normally this is fine because the linker will merge all those sections into a single one in the binary. However, starting with [LLVM 12](https://github.com/llvm/llvm-project/commit/ee5d1a04), the backend will also generate unique section names for exception handling, resulting in thousands of `.gcc_except_table.*` sections ending up in the final binary because some linkers like LLD don't currently merge or strip these EH sections (see discussion [here](https://reviews.llvm.org/D83655)). This can bloat the ELF headers and string table significantly in binaries that contain many functions.
The new option is analogous to Clang's `-fno-unique-section-names`, and instructs LLVM to generate the same `.text` and `.gcc_except_table` section for each function, resulting in a smaller final binary.
The motivation to add this new option was because we have a binary that ended up with so many ELF sections (over 65,000) that it broke some existing ELF tools, which couldn't handle so many sections.
Here's our old binary:
```
$ readelf --sections old.elf | head -1
There are 71746 section headers, starting at offset 0x2a246508:
$ readelf --sections old.elf | grep shstrtab
[71742] .shstrtab STRTAB 0000000000000000 2977204c ad44bb 00 0 0 1
```
That's an 11MB+ string table. Here's the new binary using this option:
```
$ readelf --sections new.elf | head -1
There are 43 section headers, starting at offset 0x29143ca8:
$ readelf --sections new.elf | grep shstrtab
[40] .shstrtab STRTAB 0000000000000000 29143acc 0001db 00 0 0 1
```
The whole binary size went down by over 20MB, which is quite significant.
This change adds a new compiler flag that can help reduce the size of
ELF binaries that contain many functions.
By default, when enabling function sections (which is the default for most
targets), the LLVM backend will generate different section names for each
function. For example, a function "func" would generate a section called
".text.func". Normally this is fine because the linker will merge all those
sections into a single one in the binary. However, starting with LLVM 12
(llvm/llvm-project@ee5d1a0), the backend will
also generate unique section names for exception handling, resulting in
thousands of ".gcc_except_table.*" sections ending up in the final binary
because some linkers don't currently merge or strip these EH sections.
This can bloat the ELF headers and string table significantly in
binaries that contain many functions.
The new option is analogous to Clang's -fno-unique-section-names, and
instructs LLVM to generate the same ".text" and ".gcc_except_table"
section for each function, resulting in smaller object files and
potentially a smaller final binary.
Implement `#[link_ordinal(n)]`
Allows the use of `#[link_ordinal(n)]` with `#[link(kind = "raw-dylib")]`, allowing Rust to link against DLLs that export symbols by ordinal rather than by name. As long as the ordinal matches, the name of the function in Rust is not required to match the name of the corresponding function in the exporting DLL.
Part of #58713.
This largely involves implementing the options debug-info-for-profiling
and profile-sample-use and forwarding them on to LLVM.
AutoFDO can be used on x86-64 Linux like this:
rustc -O -Cdebug-info-for-profiling main.rs -o main
perf record -b ./main
create_llvm_prof --binary=main --out=code.prof
rustc -O -Cprofile-sample-use=code.prof main.rs -o main2
Now `main2` will have feedback directed optimization applied to it.
The create_llvm_prof tool can be obtained from this github repository:
https://github.com/google/autofdoFixes#64892.
Fix clippy lints
I'm currently working on allowing clippy to run on librustdoc after a discussion I had with `@Mark-Simulacrum.` So in the meantime, I fixed a few lints on the compiler crates.
Fix use after drop in self-profile with llvm events
self-profile with `-Z self-profile-events=llvm` have failed with a segmentation fault due to this use after drop.
this type of events can be more useful now that the new passmanager is the default.
The new pass manager is enabled by default in clang since
Clang/LLVM 13. While the discussion about this is still ongoing
(https://lists.llvm.org/pipermail/llvm-dev/2021-August/152305.html)
it's expected that support for the legacy pass manager will be
dropped either in LLVM 14 or 15.
This switches us to use the new pass manager if LLVM >= 13 is used.
This does not yet support #[link_name] attributes on functions, the #[link_ordinal]
attribute, #[link(kind = "raw-dylib")] on extern blocks in bin crates, or
stdcall functions on 32-bit x86.
This doesn't seem to be necessary anymore, although I don't know
at which point or why that changed.
Forcing -O1 makes some tests fail under NewPM, because NewPM also
performs inlining at -O1, so it ends up performing much more
optimization in practice than before.
This commit implements both the native linking modifiers infrastructure
as well as an initial attempt at the individual modifiers from the RFC.
It also introduces a feature flag for the general syntax along with
individual feature flags for each modifier.
Use FromStr trait for number option parsing
Replace `parse_uint` with generic `parse_number` based on `FromStr`.
Use it for parsing inlining threshold to avoid casting later.
Run LLVM coverage instrumentation passes before optimization passes
This matches the behavior of Clang and allows us to remove several
hacks which were needed to ensure functions weren't optimized away
before reaching the instrumentation pass.
Fixes#83429
cc `@richkadel`
r? `@tmandry`
This matches the behavior of Clang and allows us to remove several
hacks which were needed to ensure functions weren't optimized away
before reaching the instrumentation pass.
Adjust `-Ctarget-cpu=native` handling in cg_llvm
When cg_llvm encounters the `-Ctarget-cpu=native` it computes an
explciit set of features that applies to the target in order to
correctly compile code for the host CPU (because e.g. `skylake` alone is
not sufficient to tell if some of the instructions are available or
not).
However there were a couple of issues with how we did this. Firstly, the
order in which features were overriden wasn't quite right – conceptually
you'd expect `-Ctarget-cpu=native` option to override the features that
are implicitly set by the target definition. However due to how other
`-Ctarget-cpu` values are handled we must adopt the following order
of priority:
* Features from -Ctarget-cpu=*; are overriden by
* Features implied by --target; are overriden by
* Features from -Ctarget-feature; are overriden by
* function specific features.
Another problem was in that the function level `target-features`
attribute would overwrite the entire set of the globally enabled
features, rather than just the features the
`#[target_feature(enable/disable)]` specified. With something like
`-Ctarget-cpu=native` we'd end up in a situation wherein a function
without `#[target_feature(enable)]` annotation would have a broader
set of features compared to a function with one such attribute. This
turned out to be a cause of heavy run-time regressions in some code
using these function-level attributes in conjunction with
`-Ctarget-cpu=native`, for example.
With this PR rustc is more careful about specifying the entire set of
features for functions that use `#[target_feature(enable/disable)]` or
`#[instruction_set]` attributes.
Sadly testing the original reproducer for this behaviour is quite
impossible – we cannot rely on `-Ctarget-cpu=native` to be anything in
particular on developer or CI machines.
cc https://github.com/rust-lang/rust/issues/83027 `@BurntSushi`
When cg_llvm encounters the `-Ctarget-cpu=native` it computes an
explciit set of features that applies to the target in order to
correctly compile code for the host CPU (because e.g. `skylake` alone is
not sufficient to tell if some of the instructions are available or
not).
However there were a couple of issues with how we did this. Firstly, the
order in which features were overriden wasn't quite right – conceptually
you'd expect `-Ctarget-cpu=native` option to override the features that
are implicitly set by the target definition. However due to how other
`-Ctarget-cpu` values are handled we must adopt the following order
of priority:
* Features from -Ctarget-cpu=*; are overriden by
* Features implied by --target; are overriden by
* Features from -Ctarget-feature; are overriden by
* function specific features.
Another problem was in that the function level `target-features`
attribute would overwrite the entire set of the globally enabled
features, rather than just the features the
`#[target_feature(enable/disable)]` specified. With something like
`-Ctarget-cpu=native` we'd end up in a situation wherein a function
without `#[target_feature(enable)]` annotation would have a broader
set of features compared to a function with one such attribute. This
turned out to be a cause of heavy run-time regressions in some code
using these function-level attributes in conjunction with
`-Ctarget-cpu=native`, for example.
With this PR rustc is more careful about specifying the entire set of
features for functions that use `#[target_feature(enable/disable)]` or
`#[instruction_set]` attributes.
Sadly testing the original reproducer for this behaviour is quite
impossible – we cannot rely on `-Ctarget-cpu=native` to be anything in
particular on developer or CI machines.
Set path of the compile unit to the source directory
As part of the effort to implement split dwarf debug info, we ended up
setting the compile unit location to the output directory rather than
the source directory. Furthermore, it seems like we failed to remap the
prefixes for this as well!
The desired behaviour is to instead set the `DW_AT_GNU_dwo_name` to a
path relative to compiler's working directory. This still allows
debuggers to find the split dwarf files, while not changing the
behaviour of the code that is compiling with regular debug info, and not
changing the compiler's behaviour with regards to reproducibility.
Fixes#82074
cc `@alexcrichton` `@davidtwco`
In the backend we may want to remove certain temporary files, but in
certain other situations these files might not be produced in the first
place. We don't exactly care about that, and the intent is really that
these files are gone after a certain point in the backend.
Here we unify the backend file removing calls to use `ensure_removed`
which will attempt to delete a file, but will not fail if it does not
exist (anymore).
The tradeoff to this approach is, of course, that we may miss instances
were we are attempting to remove files at wrong paths due to some bug –
compilation would silently succeed but the temporary files would remain
there somewhere.
As part of the effort to implement split dwarf debug info, we ended up
setting the compile unit location to the output directory rather than
the source directory. Furthermore, it seems like we failed to remap the
prefixes for this as well!
The desired behaviour is to instead set the `DW_AT_GNU_dwo_name` to a
path relative to compiler's working directory. This still allows
debuggers to find the split dwarf files, while not changing the
behaviour of the code that is compiling with regular debug info, and not
changing the compiler's behaviour with regards to reproducibility.
Fixes#82074
This commit adds a new stable codegen option to rustc,
`-Csplit-debuginfo`. The old `-Zrun-dsymutil` flag is deleted and now
subsumed by this stable flag. Additionally `-Zsplit-dwarf` is also
subsumed by this flag but still requires `-Zunstable-options` to
actually activate. The `-Csplit-debuginfo` flag takes one of
three values:
* `off` - This indicates that split-debuginfo from the final artifact is
not desired. This is not supported on Windows and is the default on
Unix platforms except macOS. On macOS this means that `dsymutil` is
not executed.
* `packed` - This means that debuginfo is desired in one location
separate from the main executable. This is the default on Windows
(`*.pdb`) and macOS (`*.dSYM`). On other Unix platforms this subsumes
`-Zsplit-dwarf=single` and produces a `*.dwp` file.
* `unpacked` - This means that debuginfo will be roughly equivalent to
object files, meaning that it's throughout the build directory
rather than in one location (often the fastest for local development).
This is not the default on any platform and is not supported on Windows.
Each target can indicate its own default preference for how debuginfo is
handled. Almost all platforms default to `off` except for Windows and
macOS which default to `packed` for historical reasons.
Some equivalencies for previous unstable flags with the new flags are:
* `-Zrun-dsymutil=yes` -> `-Csplit-debuginfo=packed`
* `-Zrun-dsymutil=no` -> `-Csplit-debuginfo=unpacked`
* `-Zsplit-dwarf=single` -> `-Csplit-debuginfo=packed`
* `-Zsplit-dwarf=split` -> `-Csplit-debuginfo=unpacked`
Note that `-Csplit-debuginfo` still requires `-Zunstable-options` for
non-macOS platforms since split-dwarf support was *just* implemented in
rustc.
There's some more rationale listed on #79361, but the main gist of the
motivation for this commit is that `dsymutil` can take quite a long time
to execute in debug builds and provides little benefit. This means that
incremental compile times appear that much worse on macOS because the
compiler is constantly running `dsymutil` over every single binary it
produces during `cargo build` (even build scripts!). Ideally rustc would
switch to not running `dsymutil` by default, but that's a problem left
to get tackled another day.
Closes#79361
remove unused return type of dropck::check_drop_obligations()
don't wrap return type in Option in get_macro_by_def_id() since we would always return Some(..)
remove redundant return type of back::write::optimize()
don't Option-wrap return type of compute_type_parameters() since we always return Some(..)
don't return empty Result in assemble_generator_candidates()
don't return empty Result in assemble_closure_candidates()
don't return empty result in assemble_fn_pointer_candidates()
don't return empty result in assemble_candidates_from_impls()
don't return empty result in assemble_candidates_from_auto_impls()
don't return emtpy result in assemble_candidates_for_trait_alias()
don't return empty result in assemble_builtin_bound_candidates()
don't return empty results in assemble_extension_candidates_for_traits_in_scope() and assemble_extension_candidates_for_trait()
remove redundant wrapping of return type of StripItem::strip() since it always returns Some(..)
remove unused return type of assemble_extension_candidates_for_all_traits()
llvm-dwp concatenates `DW_AT_comp_dir` with `DW_AT_GNU_dwo_name` (only
when `DW_AT_comp_dir` exists), which can result in it failing to find
the DWARF object files.
In earlier testing, `DW_AT_comp_dir` wasn't present in the final
object and the current directory was the output directory.
When running tests through compiletest, the working directory of the
compilation is different from output directory and that resulted in
`DW_AT_comp_dir` being in the object file (and set to the current
working directory, rather than the output directory), and
`DW_AT_GNU_dwo_name` being set to the full path (rather than just
the filename), so llvm-dwp was failing.
This commit changes the compilation directory provided to LLVM to match
the output directory, where DWARF objects are output; and ensures that
only the filename is used for `DW_AT_GNU_dwo_name`.
Signed-off-by: David Wood <david@davidtw.co>
This commit implements Split DWARF support, wiring up the flag (added in
earlier commits) to the modified FFI wrapper (also from earlier
commits).
Signed-off-by: David Wood <david@davidtw.co>
This commit removes the `TargetMachineFactory` struct and adds a
`TargetMachineFactoryFn` type alias which is used everywhere that the
previous, long type was used.
Signed-off-by: David Wood <david@davidtw.co>
This commit modifies the FFI bindings to LLVM required for Split DWARF
support in rustc. In particular:
- `addPassesToEmitFile`'s wrapper, `LLVMRustWriteOutputFile` now takes
a `DwoPath` `const char*`. When disabled, `nullptr` should be provided
which will preserve existing behaviour. When enabled, the path to the
`.dwo` file should be provided.
- `createCompileUnit`'s wrapper, `LLVMRustDIBuilderCreateCompileUnit`
now has two additional arguments, for the `DWOId` and to enable
`SplitDebugInlining`. `DWOId` should always be zero.
- `createTargetMachine`'s wrapper, `LLVMRustCreateTargetMachine` has an
additional argument which should be provided the path to the `.dwo`
when enabled.
Signed-off-by: David Wood <david@davidtw.co>
This is useful for embedded targets where small code size is desired.
For example, on my project (thumbv7em-none-eabi) this yields a 0.6% code size reduction.
This commit grepped for LLVM_VERSION_GE, LLVM_VERSION_LT, get_major_version and
min-llvm-version and statically evaluated every expression possible
(and sensible) assuming that the LLVM version is >=9 now
with an eye on merging `TargetOptions` into `Target`.
`TargetOptions` as a separate structure is mostly an implementation detail of `Target` construction, all its fields logically belong to `Target` and available from `Target` through `Deref` impls.
This lets rustc users tweak whether the linker should relax ELF relocations,
namely whether it should emit R_X86_64_GOTPCRELX relocations instead of
R_X86_64_GOTPCREL, as the former is allowed by the ABI to be further
optimised. The default value is whatever the target defines.
This lets rustc users tweak whether all functions should be put in their own
TEXT section, using whatever default value the target defines if the flag
is missing.
Set .llvmbc and .llvmcmd sections as allocatable
This marks both sections as allocatable rather than excluded, which matches what
clang does with the equivalent `-fembed-bitcode` flag.
Preparation for a subsequent change that replaces
rustc_target::config::Config with its wrapped Target.
On its own, this commit breaks the build. I don't like making
build-breaking commits, but in this instance I believe that it
makes review easier, as the "real" changes of this PR can be
seen much more easily.
Result of running:
find compiler/ -type f -exec sed -i -e 's/target\.target\([)\.,; ]\)/target\1/g' {} \;
find compiler/ -type f -exec sed -i -e 's/target\.target$/target/g' {} \;
find compiler/ -type f -exec sed -i -e 's/target.ptr_width/target.pointer_width/g' {} \;
./x.py fmt
Use llvm::computeLTOCacheKey to determine post-ThinLTO CGU reuse
During incremental ThinLTO compilation, we attempt to re-use the
optimized (post-ThinLTO) bitcode file for a module if it is 'safe' to do
so.
Up until now, 'safe' has meant that the set of modules that our current
modules imports from/exports to is unchanged from the previous
compilation session. See PR #67020 and PR #71131 for more details.
However, this turns out be insufficient to guarantee that it's safe
to reuse the post-LTO module (i.e. that optimizing the pre-LTO module
would produce the same result). When LLVM optimizes a module during
ThinLTO, it may look at other information from the 'module index', such
as whether a (non-imported!) global variable is used. If this
information changes between compilation runs, we may end up re-using an
optimized module that (for example) had dead-code elimination run on a
function that is now used by another module.
Fortunately, LLVM implements its own ThinLTO module cache, which is used
when ThinLTO is performed by a linker plugin (e.g. when clang is used to
compile a C proect). Using this cache directly would require extensive
refactoring of our code - but fortunately for us, LLVM provides a
function that does exactly what we need.
The function `llvm::computeLTOCacheKey` is used to compute a SHA-1 hash
from all data that might influence the result of ThinLTO on a module.
In addition to the module imports/exports that we manually track, it
also hashes information about global variables (e.g. their liveness)
which might be used during optimization. By using this function, we
shouldn't have to worry about new LLVM passes breaking our module re-use
behavior.
In LLVM, the output of this function forms part of the filename used to
store the post-ThinLTO module. To keep our current filename structure
intact, this PR just writes out the mapping 'CGU name -> Hash' to a
file. To determine if a post-LTO module should be reused, we compare
hashes from the previous session.
This should unblock PR #75199 - by sheer chance, it seems to have hit
this issue due to the particular CGU partitioning and optimization
decisions that end up getting made.
Secure entry functions do not support if arguments are passed on the
stack. An "unsupported" diagnostic will be emitted by LLVM if that is
the case.
This commits adds support in Rust for that diagnostic so that an error
will be output if that is the case!
Signed-off-by: Hugues de Valon <hugues.devalon@arm.com>
During incremental ThinLTO compilation, we attempt to re-use the
optimized (post-ThinLTO) bitcode file for a module if it is 'safe' to do
so.
Up until now, 'safe' has meant that the set of modules that our current
modules imports from/exports to is unchanged from the previous
compilation session. See PR #67020 and PR #71131 for more details.
However, this turns out be insufficient to guarantee that it's safe
to reuse the post-LTO module (i.e. that optimizing the pre-LTO module
would produce the same result). When LLVM optimizes a module during
ThinLTO, it may look at other information from the 'module index', such
as whether a (non-imported!) global variable is used. If this
information changes between compilation runs, we may end up re-using an
optimized module that (for example) had dead-code elimination run on a
function that is now used by another module.
Fortunately, LLVM implements its own ThinLTO module cache, which is used
when ThinLTO is performed by a linker plugin (e.g. when clang is used to
compile a C proect). Using this cache directly would require extensive
refactoring of our code - but fortunately for us, LLVM provides a
function that does exactly what we need.
The function `llvm::computeLTOCacheKey` is used to compute a SHA-1 hash
from all data that might influence the result of ThinLTO on a module.
In addition to the module imports/exports that we manually track, it
also hashes information about global variables (e.g. their liveness)
which might be used during optimization. By using this function, we
shouldn't have to worry about new LLVM passes breaking our module re-use
behavior.
In LLVM, the output of this function forms part of the filename used to
store the post-ThinLTO module. To keep our current filename structure
intact, this PR just writes out the mapping 'CGU name -> Hash' to a
file. To determine if a post-LTO module should be reused, we compare
hashes from the previous session.
This should unblock PR #75199 - by sheer chance, it seems to have hit
this issue due to the particular CGU partitioning and optimization
decisions that end up getting made.
