The code was broken because it printed "llvm-config" instead of the
absolute path to the llvm-config executable, causing Cargo to always
rebuild librustc_llvm if using system LLVM.
Also, it's not the build system's job to rebuild when a system library
changes, so we simply don't emit "rerun-if-changed" if a path to LLVM
was not explicitly provided.
Modify librustc_llvm to pass -DNDEBUG while compiling.
Currently, librustc_llvm builds are not reproducible because the LLVM files it compiles use the debug version of llvm_unreachable, which uses __FILE__. To fix this, we propagate NDEBUG from bootstrap if applicable and use it when compiling librustc_llvm.
r? @alexcrichton
Currently, librustc_llvm builds are not reproducible because the LLVM
files it compiles use the debug version of llvm_unreachable, which
uses __FILE__. To fix this, we propagate NDEBUG from bootstrap if
applicable and use it when compiling librustc_llvm.
on OpenBSD, some architectures relies on libc++ (from LLVM) and some
others on libestdc++ (particular version of libstdc++ from GCC).
sparc64-unknown-openbsd needs libestdc++ and libgcc (as x86_64 some
years ago). Reintroduce the support of them for openbsd, only for
sparc64 arch. Some others architectures on OpenBSD could use them too.
Some -L and -l flags may be needed even when building librustc_llvm,
for example when using static libc++ on Linux we may need to manually
specify the library search path and -ldl -lpthread as additional link
dependencies. We pass LLVM linker flags from config to librustc_llvm
build to make sure these cases are handled.
This commit works around the newly-introduced LLVM shared library.
This is needed such that llvm-config run from
librustc_llvm's build script can correctly locate it's own LLVM, not the
one in stage0/lib. The LLVM build system uses the DT_RUNPATH/RUNPATH
header within the llvm-config binary, which we want to use, but because
Cargo always adds the host compiler's "libdir" (stage0/lib in our
case) to the dynamic linker's search path, we weren't properly finding
the freshly-built LLVM in llvm/lib. By restoring the environment
variable setting the search path to what bootstrap sees, the problem is
resolved and librustc_llvm correctly links and finds the appropriate
LLVM.
Several run-make-fulldeps tests are also updated with similar handling.
Provide the option to use libc++ even on all platforms
This is the default on platforms which use libc++ as the default C++
library but this option allows using libc++ on others as well.
When building a distributed compiler on Linux where we use ThinLTO to
create the LLVM shared object this commit switches the compiler to
dynamically linking that LLVM artifact instead of statically linking to
LLVM. The primary goal here is to reduce CI compile times, avoiding two+
ThinLTO builds of all of LLVM. By linking dynamically to LLVM we'll
reuse the one ThinLTO step done by LLVM's build itself.
Lots of discussion about this change can be found [here] and down. A
perf run will show whether this is worth it or not!
[here]: https://github.com/rust-lang/rust/pull/53245#issuecomment-417015334
When building a distributed compiler on Linux where we use ThinLTO to
create the LLVM shared object this commit switches the compiler to
dynamically linking that LLVM artifact instead of statically linking to
LLVM. The primary goal here is to reduce CI compile times, avoiding two+
ThinLTO builds of all of LLVM. By linking dynamically to LLVM we'll
reuse the one ThinLTO step done by LLVM's build itself.
Lots of discussion about this change can be found [here] and down. A
perf run will show whether this is worth it or not!
[here]: https://github.com/rust-lang/rust/pull/53245#issuecomment-417015334
* Update bootstrap compiler
* Update version to 1.33.0
* Remove some `#[cfg(stage0)]` annotations
Actually updating the version number is blocked on updating Cargo
The issue of passing around SIMD types as values between functions has
seen [quite a lot] of [discussion], and although we thought [we fixed
it][quite a lot] it [wasn't]! This PR is a change to rustc to, again,
try to fix this issue.
The fundamental problem here remains the same, if a SIMD vector argument
is passed by-value in LLVM's function type, then if the caller and
callee disagree on target features a miscompile happens. We solve this
by never passing SIMD vectors by-value, but LLVM will still thwart us
with its argument promotion pass to promote by-ref SIMD arguments to
by-val SIMD arguments.
This commit is an attempt to thwart LLVM thwarting us. We, just before
codegen, will take yet another look at the LLVM module and demote any
by-value SIMD arguments we see. This is a very manual attempt by us to
ensure the codegen for a module keeps working, and it unfortunately is
likely producing suboptimal code, even in release mode. The saving grace
for this, in theory, is that if SIMD types are passed by-value across
a boundary in release mode it's pretty unlikely to be performance
sensitive (as it's already doing a load/store, and otherwise
perf-sensitive bits should be inlined).
The implementation here is basically a big wad of C++. It was largely
copied from LLVM's own argument promotion pass, only doing the reverse.
In local testing this...
Closes#50154Closes#52636Closes#54583Closes#55059
[quite a lot]: https://github.com/rust-lang/rust/pull/47743
[discussion]: https://github.com/rust-lang/rust/issues/44367
[wasn't]: https://github.com/rust-lang/rust/issues/50154
This commit updates our Fat LTO logic to tweak our custom wrapper around LLVM's
"link modules" functionality. Previously whenever the
`LLVMRustLinkInExternalBitcode` function was called it would call LLVM's
`Linker::linkModules` wrapper. Internally this would crate an instance of a
`Linker` which internally creates an instance of an `IRMover`. Unfortunately for
us the creation of `IRMover` is somewhat O(n) with the input module. This means
that every time we linked a module it was O(n) with respect to the entire module
we had built up!
Now the modules we build up during LTO are quite large, so this quickly started
creating an O(n^2) problem for us! Discovered in #48025 it turns out this has
always been a problem and we just haven't noticed it. It became particularly
worse recently though due to most libraries having 16x more object files than
they previously did (1 -> 16).
This commit fixes this performance issue by preserving the `Linker` instance
across all links into the main LLVM module. This means we only create one
`IRMover` and allows LTO to progress much speedier.
From the `cargo-cache` project in #48025 a **full build** locally when from
5m15s to 2m24s. Looking at the timing logs each object file was linked in in
single-digit millisecond rather than hundreds, clearly being a nice improvement!
Closes#48025
Building on the work of # 45684 this commit updates the compiler to
unconditionally load the `rustc_trans` crate at runtime instead of linking to it
at compile time. The end goal of this work is to implement # 46819 where rustc
will have multiple backends available to it to load.
This commit starts off by removing the `extern crate rustc_trans` from the
driver. This involved moving some miscellaneous functionality into the
`TransCrate` trait and also required an implementation of how to locate and load
the trans backend. This ended up being a little tricky because the sysroot isn't
always the right location (for example `--sysroot` arguments) so some extra code
was added as well to probe a directory relative to the current dll (the
rustc_driver dll).
Rustbuild has been updated accordingly as well to have a separate compilation
invocation for the `rustc_trans` crate and assembly it accordingly into the
sysroot. Finally, the distribution logic for the `rustc` package was also
updated to slurp up the trans backends folder.
A number of assorted fallout changes were included here as well to ensure tests
pass and such, and they should all be commented inline.
The main goal here is to use FreeBSD's normal libc++, instead of
statically linking the libstdc++ packaged with GCC, because that
libstdc++ has bugs that cause rustc to deadlock inside LLVM.
But the easiest way to use libc++ is to switch the build from GCC to
Clang, and the Clang package in the Ubuntu image already knows how to
cross-compile (given a sysroot and preferably cross-binutils), so the
toolchain script now uses that instead of building a custom compiler.
This also de-duplicates the `build-toolchain.sh` script.
Assume at least LLVM 3.9 in rustllvm and rustc_llvm
We bumped the minimum LLVM to 3.9 in #45326. This just cleans up the conditional code in the `rustllvm` C++ wrappers to assume that minimum, and similarly cleans up the `rustc_llvm` build script.
Remove support for the PNaCl target (le32-unknown-nacl)
This removes support for the `le32-unknown-nacl` target which is currently supported by rustc on tier 3. Despite the "nacl" in the name, the target doesn't output native code (x86, ARM, MIPS), instead it outputs binaries in the PNaCl format.
There are two reasons for the removal:
* Google [has announced](https://blog.chromium.org/2017/05/goodbye-pnacl-hello-webassembly.html) deprecation of the PNaCl format. The suggestion is to migrate to wasm. Happens we already have a wasm backend!
* Our PNaCl LLVM backend is provided by the fastcomp patch set that the LLVM fork used by rustc contains in addition to vanilla LLVM (`src/llvm/lib/Target/JSBackend/NaCl`). Upstream LLVM doesn't have PNaCl support. Removing PNaCl support will enable us to move away from fastcomp (#44006) and have a lighter set of patches on top of upstream LLVM inside our LLVM fork. This will help distribution packagers of Rust.
Fixes#42420
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
