| Age | Commit message (Collapse) | Author | Lines |
|---|
|
|
|
|
|
LLVM IR coverage encoding aligns closer to Clang's
I found some areas for improvement while attempting to debug the
SegFault issue when running rust programs compiled using MSVC, with
coverage instrumentation.
I discovered that LLVM's coverage writer was generating incomplete
function name variable names (that's not a typo: the name of the
variable that holds a function name).
The existing implementation used one-up numbers to distinguish
variables, and correcting the names did not fix the MSVC coverage bug,
but the fix in this PR makes the names and resulting LLVM IR easier to
follow and more consistent with Clang's implementation.
I also changed the way the `-Zinstrument-coverage` option is supported in
symbol_export.rs. The original implementation was incorrect, and the
corrected version matches the handling for `-Zprofile-generate`, as it
turns out.
(An argument could be made that maybe `-Zinstrument-coverage` should
automatically enable `-Cprofile-generate`. In fact, if
`-Cprofile-generate` is analagous to Clang's `-fprofile-generate`, as
some documentation implies, Clang always requires this flag for its
implementation of source-based code coverage. This would require a
little more validation, and if implemented, would probably require
updating some of the user-facing messages related to
`-Cprofile-generate` to not be so specific to the PGO use case.)
None of these changes fixed the MSVC coverage problems, but they should
still be welcome improvements.
Lastly, I added some additional FIXME comments in instrument_coverage.rs
describing issues I found with the generated LLVM IR that would be
resolved if the coverage instrumentation is injected with a `Statement`
instead of as a new `BasicBlock`. I describe seven advantages of this
change, but it requires some discussion before making a change like
this.
r? @tmandry
|
|
I found some areas for improvement while attempting to debug the
SegFault issue when running rust programs compiled using MSVC, with
coverage instrumentation.
I discovered that LLVM's coverage writer was generating incomplete
function name variable names (that's not a typo: the name of the
variable that holds a function name).
The existing implementation used one-up numbers to distinguish
variables, and correcting the names did not fix the MSVC coverage bug,
but the fix in this PR makes the names and resulting LLVM IR easier to
follow and more consistent with Clang's implementation.
I also changed the way the `-Zinstrument-coverage` option is supported
in symbol_export.rs. The original implementation was incorrect, and the
corrected version matches the handling for `-Zprofile-generate`, as it
turns out.
(An argument could be made that maybe `-Zinstrument-coverage` should
automatically enable `-Cprofile-generate`. In fact, if
`-Cprofile-generate` is analagous to Clang's `-fprofile-generate`, as
some documentation implies, Clang always requires this flag for its
implementation of source-based code coverage. This would require a
little more validation, and if implemented, would probably require
updating some of the user-facing messages related to
`-Cprofile-generate` to not be so specific to the PGO use case.)
None of these changes fixed the MSVC coverage problems, but they should
still be welcome improvements.
Lastly, I added some additional FIXME comments in instrument_coverage.rs
describing issues I found with the generated LLVM IR that would be
resolved if the coverage instrumentation is injected with a `Statement`
instead of as a new `BasicBlock`. I describe seven advantages of this
change, but it requires some discussion before making a change like
this.
|
|
Also unused since introduction in #35174
|
|
Also introduced in #35174, and immediately unused.
|
|
Added in #35174, this was already unused (and new uses have not been introduced
since then).
|
|
Added in #35174, this was already unused (and new uses have not been introduced
since then).
|
|
Added in b7615389978eae2ae9f3cba9a776fd8da3f743ca, it started out already
unused.
|
|
Some were in librustc_codegen_llvm, but others are not tied to LLVM, so
I put them in a new crate: librustc_codegen_ssa/coverageinfo/ffi.rs
|
|
Found some problems with the coverage map encoding when testing with
more than one counter per function.
While debugging, I realized some better ways to structure the Rust
implementation of the coverage mapping generator. I refactored somewhat,
resulting in less code overall, expanded coverage of LLVM Coverage Map
capabilities, and much closer alignment with LLVM data structures, APIs,
and naming.
This should be easier to follow and easier to maintain.
|
|
rustc now generates the coverage map and can support (limited)
coverage report generation, at the function level.
Example:
$ BUILD=$HOME/rust/build/x86_64-unknown-linux-gnu
$ $BUILD/stage1/bin/rustc -Zinstrument-coverage \
$HOME/rust/src/test/run-make-fulldeps/instrument-coverage/main.rs
$ LLVM_PROFILE_FILE="main.profraw" ./main
called
$ $BUILD/llvm/bin/llvm-profdata merge -sparse main.profraw -o main.profdata
$ $BUILD/llvm/bin/llvm-cov show --instr-profile=main.profdata main
1| 1|pub fn will_be_called() {
2| 1| println!("called");
3| 1|}
4| |
5| 0|pub fn will_not_be_called() {
6| 0| println!("should not have been called");
7| 0|}
8| |
9| 1|fn main() {
10| 1| let less = 1;
11| 1| let more = 100;
12| 1|
13| 1| if less < more {
14| 1| will_be_called();
15| 1| } else {
16| 1| will_not_be_called();
17| 1| }
18| 1|}
|
|
debuginfo: Mangle tuples to be natvis friendly, typedef basic types
These changes are meant to unblock rust-lang/rust#70052 "Update hashbrown to 0.8.0" by allowing the use of `tuple<u64, u64>` as a .natvis expression in MSVC style debuggers (MSVC, WinDbg, CDB, etc.)
* f8eb81b does the actual mangling of `(u64, u64)` -> `tuple<u64, 64>`
* 24a728a allows `u64` to resolve (fixing `$T1` / `$T2` when used to visualize `HashMap<u64, u64, ...>`)
|
|
|
|
PDB debug information doesn't appear to be emitted for basic types.
By defining u32 as a typedef for unsigned __int32 when targeting MSVC,
we allow CDB and other debuggers to recognize "u32" as a type/expression.
This in turn unblocks rust-lang#70052 "Update hashbrown to 0.8.0" by
allowing $T1 ..= $T3 to resolve, which would otherwise fail to resolve
when builtin types fail to parse.
|
|
Diagnose use of incompatible sanitizers
Emit an error when incompatible sanitizer are configured through command
line options. Previously the last one configured prevailed and others
were silently ignored.
Additionally use a set to represent configured sanitizers, making it
possible to enable multiple sanitizers at once. At least in principle,
since currently all of them are considered to be incompatible with
others.
|
|
This initial version only injects counters at the top of each function.
Rust Coverage will require injecting additional counters at each
conditional code branch.
|
|
Emit an error when incompatible sanitizer are configured through command
line options. Previously the last one configured prevailed and others
were silently ignored.
Additionally use a set to represent configured sanitizers, making it
possible to enable multiple sanitizers at once. At least in principle,
since currently all of them are considered to be incompatible with
others.
|
|
Enable AVR as a Tier 3 target upstream
Tracking issue: #44052.
Things intentionally left out of the initial upstream:
* The `target_cpu` flag
I have made the cleanup suggestions by @jplatte and @jplatte in https://github.com/avr-rust/rust/commit/043550d9db0582add42e5837f636f61acb26b915.
Anybody feel free to give the branch a test and see how it fares, or make suggestions on the code patch itself.
|
|
|
|
|
|
|
|
|
|
|
|
Cleanup and document `-C code-model`
r? @Amanieu
|
|
Introduce `enum CodeModel` instead.
|
|
Ensure that inliner inserts lifetime markers if they have been emitted during
codegen. Otherwise if allocas from inlined functions are merged together,
lifetime markers from one function might invalidate load & stores performed
by the other one.
|
|
rustllvm: Use .init_array rather than .ctors
LLVM TargetMachines default to using the (now-legacy) .ctors
representation of init functions. Mixing .ctors and .init_array
representations can cause issues when linking with lld.
This happens in practice for:
* Our profiling runtime which is currently implicitly built with
.init_array since it is built by clang, which sets this field.
* External C/C++ code that may be linked into the same process.
Fixes: #71233
|
|
Store LLVM bitcode in object files, not compressed
This commit is an attempted resurrection of #70458 where LLVM bitcode
emitted by rustc into rlibs is stored into object file sections rather
than in a separate file. The main rationale for doing this is that when
rustc emits bitcode it will no longer use a custom compression scheme
which makes it both easier to interoperate with existing tools and also
cuts down on compile time since this compression isn't happening.
The blocker for this in #70458 turned out to be that native linkers
didn't handle the new sections well, causing the sections to either
trigger bugs in the linker or actually end up in the final linked
artifact. This commit attempts to address these issues by ensuring that
native linkers ignore the new sections by inserting custom flags with
module-level inline assembly.
Note that this does not currently change the API of the compiler at all.
The pre-existing `-C bitcode-in-rlib` flag is co-opted to indicate
whether the bitcode should be present in the object file or not.
Finally, note that an important consequence of this commit, which is also
one of its primary purposes, is to enable rustc's `-Clto` bitcode
loading to load rlibs produced with `-Clinker-plugin-lto`. The goal here
is that when you're building with LTO Cargo will tell rustc to skip
codegen of all intermediate crates and only generate LLVM IR. Today
rustc will generate both object code and LLVM IR, but the object code is
later simply thrown away, wastefully.
|
|
LLVM TargetMachines default to using the (now-legacy) .ctors
representation of init functions. Mixing .ctors and .init_array
representations can cause issues when linking with lld.
This happens in practice for:
* Our profiling runtime which is currently implicitly built with
.init_array since it is built by clang, which sets this field.
* External C/C++ code that may be linked into the same process.
To support legacy systems which may use .ctors, targets may now specify
that they use .ctors via the use_ctors attribute which defaults to
false.
For debugging and manual control, -Z use-ctors-section=yes/no will allow
manual override.
Fixes: #71233
|
|
This commit is an attempted resurrection of #70458 where LLVM bitcode
emitted by rustc into rlibs is stored into object file sections rather
than in a separate file. The main rationale for doing this is that when
rustc emits bitcode it will no longer use a custom compression scheme
which makes it both easier to interoperate with existing tools and also
cuts down on compile time since this compression isn't happening.
The blocker for this in #70458 turned out to be that native linkers
didn't handle the new sections well, causing the sections to either
trigger bugs in the linker or actually end up in the final linked
artifact. This commit attempts to address these issues by ensuring that
native linkers ignore the new sections by inserting custom flags with
module-level inline assembly.
Note that this does not currently change the API of the compiler at all.
The pre-existing `-C bitcode-in-rlib` flag is co-opted to indicate
whether the bitcode should be present in the object file or not.
Finally, note that an important consequence of this commit, which is also
one of its primary purposes, is to enable rustc's `-Clto` bitcode
loading to load rlibs produced with `-Clinker-plugin-lto`. The goal here
is that when you're building with LTO Cargo will tell rustc to skip
codegen of all intermediate crates and only generate LLVM IR. Today
rustc will generate both object code and LLVM IR, but the object code is
later simply thrown away, wastefully.
|
|
|
|
Introduce `enum RelocModel` instead.
|
|
* Adds either an MD5 or SHA1 hash to the debug info.
* Adds new unstable option `-Z src-hash-algorithm` to control the hashing algorithm.
|
|
asm! is left as a wrapper around llvm_asm! to maintain compatibility.
|
|
librustc_codegen_llvm: Replace deprecated API usage
|
|
|
|
|
|
|
|
|
|
Changed functions:
* LLVMRustGetOrInsertFunction
* LLVMRustGetNamedValue
* LLVMRustBuildCall (removed unused name argument)
* LLVMRustInlineAsm
* LLVMRustInlineAsmVerify
* LLVMRustAppendModuleInlineAsm
|
|
Additionally whenever possible match C API provided by the LLVM.
|
|
No functional changes intended.
|
|
|
|
|
|
Support new LLVM pass manager
Add support for the new LLVM pass manager behind a `-Z new-llvm-pass-manager=on` option. Both the pre-link optimization and LTO pipelines use the new pass manager. There's some bits that are not supported yet:
* `-C passes`. NewPM requires an entirely different way of specifying custom pass pipelines. We should probably expose that functionality, but it doesn't directly map to what `-C passes` does.
* NewPM has no support for custom inline parameters right now. We'd have to add upstream support for that first.
* NewPM does not support PGO at O0 in LLVM 9 (which is why those tests fail with NewPM enabled). This is supported in LLVM 10.
* NewPM does not support MergeFunctions in LLVM 9. I've landed this upstream just before the cut, so we'll be able to re-enable that with LLVM 10.
Closes #64289.
r? @ghost
|
|
The new pass manager can be enabled using
-Z new-llvm-pass-manager=on.
|
|
To avoid creating memsets with outdated signature. For some reason
SROA chokes on this when using NewPM.
|
|
See https://github.com/servo/servo/issues/25550.
|
|
|
