Fix another circular deps link args issue
It turns out that the support in #49316 wasn't enough to handle all cases
notably the example in #48661. The underlying bug was connected to panic=abort
where lang items were listed in the `missing_lang_items` sets but didn't
actually exist anywhere.
This caused the linker backend to deduce that start-group/end-group wasn't
needed because not all items were defined. Instead the missing lang items that
don't actually need to have a definition are filtered out and not considered for
the start-group/end-group arguments
Closes#48661
add emit_debug_gdb_scripts target option and ..
set it to false for no-std targets like ARM Cortex-M and MSP430. For the rationale of this change see the comment in thumb_base.rs
this is a temporary workaround until #44993 is implemented
r? @alexcrichton or @michaelwoerister
Flush executables to disk after linkage
A problem caused by not doing so in Chrome has been reported [here](https://randomascii.wordpress.com/2018/02/25/compiler-bug-linker-bug-windows-kernel-bug/amp/).
`File::sync_all()` calls `FlushFileBuffers()` down the line, causing potentially unflushed buffers on high I/O-load systems to flush and preventing nasty non-reproducible bugs.
Closes#48545
It turns out that the support in #49316 wasn't enough to handle all cases
notably the example in #48661. The underlying bug was connected to panic=abort
where lang items were listed in the `missing_lang_items` sets but didn't
actually exist anywhere.
This caused the linker backend to deduce that start-group/end-group wasn't
needed because not all items were defined. Instead the missing lang items that
don't actually need to have a definition are filtered out and not considered for
the start-group/end-group arguments
Closes#48661
This should be enough and shouldn't require append(true) since we're not
explicitly writing anything so we're not flushing it so we've no risk of
overwriting it
This commit fixes a longstanding issue with the compiler with circular
dependencies between libcore and libstd. The `core` crate requires at least one
symbol, the ability to unwind. The `std` crate is the crate which actually
defines this symbol, but the `std` crate also depends on the `core` crate.
This circular dependency is in general disallowed in Rust as crates cannot have
cycles amongst them. A special exception is made just for core/std, but this is
also unfortunately incompatible with how GNU linkers work. GNU linkers will
process undefined symbols in a left-to-right fashion, only actually linking an
rlib like libstd if there are any symbols used from it. This strategy is
incompatible with circular dependencies because if we otherwise don't use
symbols from libstd we don't discover that we needed it until we're later
processing libcore's symbols!
To fix this GNU linkers support the `--start-group` and `--end-group` options
which indicate "libraries between these markers may have circular dependencies
amongst them. The linker invocation has been updated to automatically pass these
arguments when we're invoking a GNU linker and automatically calculate where the
arguments need to go (around libstd and libcore)
Closes#18807Closes#47074
A problem caused by not doing so in Chrome has been reported here:
https://randomascii.wordpress.com/2018/02/25/compiler-bug-linker-bug-windows-kernel-bug/amp/
File::sync_all() calls FlushFileBuffers() down the line,
causing potentially unflushed buffers on high I/O-load systems to flush
and prevent nasty non-reproducible bugs.
The force-flush is only done on Windows and if the linker exited successfully
Closes#48545
Introduce a TargetTriple enum to support absolute target paths
This PR replaces target triple strings with a `TargetTriple` enum, which represents either a target triple or a path to a JSON target file. The path variant is used if the `--target` argument has a `.json` extension, else the target triple variant is used.
The motivation of this PR is support for absolute target paths to avoid the need for setting the `RUST_TARGET_PATH` environment variable (see rust-lang/cargo#4905 for more information). For places where some kind of triple is needed (e.g. in the sysroot folder), we use the file name (without extension).
For compatibility, we keep the old behavior of searching for a file named `$(target_triple).json` in `RUST_TARGET_PATH` for non-official target triples.
implement minmax intrinsics
This adds the `simd_{fmin,fmax}` intrinsics, which do a vertical (lane-wise) `min`/`max` for floating point vectors that's equivalent to Rust's `min`/`max` for `f32`/`f64`.
It might make sense to make `{f32,f64}::{min,max}` use the `minnum` and `minmax` intrinsics as well.
---
~~HELP: I need some help with these. Either I should go to sleep or there must be something that I must be missing. AFAICT I am calling the `maxnum` builder correctly, yet rustc/LLVM seem to insert a call to `llvm.minnum` there instead...~~ EDIT: Rust's LLVM version is too old :/
Add basic PGO support.
This PR adds two mutually exclusive options for profile usage and generation using LLVM's instruction profile generation (the same as clang uses), `-C pgo-use` and `-C pgo-gen`.
See each commit for details.
Introduce unsafe offset_from on pointers
Adds intrinsics::exact_div to take advantage of the unsafe, which reduces the implementation from
```asm
sub rcx, rdx
mov rax, rcx
sar rax, 63
shr rax, 62
lea rax, [rax + rcx]
sar rax, 2
ret
```
down to
```asm
sub rcx, rdx
sar rcx, 2
mov rax, rcx
ret
```
(for `*const i32`)
See discussion on the `offset_to` tracking issue https://github.com/rust-lang/rust/issues/41079
Some open questions
- Would you rather I split the intrinsic PR from the library PR?
- Do we even want the safe version of the API? https://github.com/rust-lang/rust/issues/41079#issuecomment-374426786 I've added some text to its documentation that even if it's not UB, it's useless to use it between pointers into different objects.
and todos
- [x] ~~I need to make a codegen test~~ Done
- [x] ~~Can the subtraction use nsw/nuw?~~ No, it can't https://github.com/rust-lang/rust/pull/49297#discussion_r176697574
- [x] ~~Should there be `usize` variants of this, like there are now `add` and `sub` that you almost always want over `offset`? For example, I imagine `sub_ptr` that returns `usize` and where it's UB if the distance is negative.~~ Can wait for later; C gives a signed result https://github.com/rust-lang/rust/issues/41079#issuecomment-375842235, so we might as well, and this existing to go with `offset` makes sense.
Pass --strip-debug to GccLinker when building without debuginfo
C.f. #46034
---
This brings a hello-world built by passing rustc no command line options from 2.9M to 592K on Linux.
(This might need to special case MacOS or Windows, not sure if the linkers there support `--strip-debug`, and there is an annoying lack of dependable docs for the linkers there.)
They should at least be that, they usually warn about control flow mismatches,
and or the profile being useless, which looks like at least a warning to me.
Signed-off-by: Emilio Cobos Álvarez <emilio@crisal.io>
Executables crash in the probestack function otherwise... I haven't debugged
much further than that.
Signed-off-by: Emilio Cobos Álvarez <emilio@crisal.io>
