Otherwise, we leave optimization opportunities on the table, and in
some cases even fail to remove marker intrinsics inserted by earlier
optimization.
Background: under -embed-bitcode, compilation happens in two stages: a
"frontend" job that compiles Swift code, generates LLVM IR, optimizes
it, and then emits a .bc file; and a "backend" job that takes that
.bc, converts it to assembly, and emits an object file with the
original bitcode embedded. However, there are actually optimization
passes that run before and during that "convert to assembly" step that
were getting completely skipped.
rdar://problem/34864094
This brings the capability from clang to save remarks in an external YAML files.
YAML files can be viewed with tools like the opt-viewer.
Saving the remarks is activated with the new option -save-optimization-record.
Similarly to -emit-tbd, I've only added support for single-compile mode for now.
In this case the default filename is determined by
getOutputFilenameFromPathArgOrAsTopLevel, i.e. unless explicitly specified
with -save-optimization-record-path, the file is placed in the directory of the
main output file as <modulename>.opt.yaml.
This allows reporting successful and unsuccessful optimizations similar to
clang/llvm.
This first patch adds support for the
options -Rpass=<pass-name-regex> -Rpass-missed=<pass-name-regex>. These allow
reporting successful/unsuccessful optimization on the compiler output for passes
specified by the regex. I've also added one missed and one passed remark type
to the inliner to test the infrastructure.
Clang also has the option of collecting these records in an external YAML data
file. This will be added in a later patch.
A few notes:
* The goal is to use this facility for both user-lever "performance" warnings
and expert-level performance analysis. There will probably be a flag in the
future differentiating the verbosity.
* The intent is match clang/llvm as much as it makes sense. On the other hand I
did make some changes. Unlike in llvm, the emitter is not a pass which
simplifies things. Also the remark class hierarchy is greatly simplified since
we don't derive from DiagnosticInfo. We also don't derive from Diagnostic to
support the streaming API for arbitrary named-value pairs.
* Currently function names are printed mangled which should be fixed.
Allow users to pass `.swiftmodule` files into the Swift driver when
compiling without `-g`. The `.swiftmodule` files are then passed to the
linker via `-add_ast_path` so that LLDB can access their AST
information.
This addresses one of two driver changes suggested in the comments of
https://bugs.swift.org/browse/SR-2660.
A post-commit reviewer on https://github.com/apple/swift/pull/10716
asked that the comments be removed. Remove them, as they're no longer
demarcating code that is internal to Apple.
This option tells the compiler where to find a profdata file. The
information in this file enables PGO. For more information about the PGO
infrastructure, look for the -profile-generate option and for the
llvm-profdata tool [1].
[1] http://llvm.org/docs/CommandGuide/llvm-profdata.html
Also pass flags to disable SIL optimization passes when merging
modules, since that's completely unnecessary.
An evolution test that used to fail with WMO disabled now passes
with this change.
FIxes <rdar://problem/18913977>.
Also pass flags to disable SIL optimization passes when merging
modules, since that's completely unnecessary.
An evolution test that used to fail with WMO disabled now passes
with this change.
FIxes <rdar://problem/18913977>.
Also pass flags to disable SIL optimization passes when merging
modules, since that's completely unnecessary.
An evolution test that used to fail with WMO disabled now passes
with this change.
FIxes <rdar://problem/18913977>.
The etymology of these terms isn't about race, but "black" = "blocked"
and "white" = "allowed" isn't really a good look these days. In most
cases we weren't using these terms particularly precisely anyway, so
the rephrasing is actually an improvement.
Similarly to Clang, the flag enables coverage instrumentation, and links
`libLLVMFuzzer.a` to the produced binary.
Additionally, this change affects the driver logic, and enables the
concurrent usage of multiple sanitizers.
This means it can be emitted during an -emit-module frontend job, which is the
most common place it will be used, so reusing work like this is important for
performance.
For now, this has to happen as part of a single frontend invocation, i.e. -wmo
or -force-single-frontend-invocation.
pass it to the PCH-generation job.
This depends on the merge-modules patch because we don't want merging
modules when optimization is enabled to re-run the optimizer.
(Thanks, Jordan!)
rdar://33541306
Currently we have a number of unsolved disjunctions hard-coded to 5,
which breaks some existing code by terminating shrinking too early.
This patch makes it a command-line option so users have control over
what that threshold can be.
Resolves: rdar://problem/33433595
Typo correction can be particularly expensive, so introduce a
command-line flag to limit the number of typo corrections we will
perform per type-checker instance. Default this limit to 10.
Addresses rdar://problem/28469270 to some extent.
- Reordering the auto link files when linking in static libraries
and object files is required because the ordering of arguments
is important, and the dependant libraries need to come after the
objects/libs that require them. This is not a problem for
libswiftCore.a but can be an issue with libs that sit on top of
it, e.g. libFoundation.a
- Dont add an -rpath to the Swift dynamic libraries if using
-static-stdlib
This updates the rpath linking logic to only add the rpath that points
to the swift dylibs in the case that the libraries are not statically
linked into the binary.
With this patch different sanitizers (tsan/asan) will be enabled or
disabled on the driver level on a particular OS depending on whether
the required library is present.
The current patch only supports Darwin architectures, but Linux support
should not be hard to add.
Adds a test case that the compiler driver will properly insert
-licucore for static standard library builds. This test runs even on
builds that don't build a static standard library, so PR testing can
regressions here.
For the multiple-files mode -emit-pch is still invoked in separate frontend invocation but with using a persistent PCH.
Subsequent frontend invocations use the persistent PCH but they don't need to validate it.
For all-files mode (e.g. WMO) the frontend invocation uses a persistent PCH that it also validates.
Based on recommendations in SE-0160, there are two migration workflows:
- Conservative: Maintain @objc visibility that was inferred in Swift 3
by adding @objc to all declarations that were implicitily visible to
the Objective-C runtime. This is invoked in the migrator by adding the
-migrate-keep-objc-visibility flag.
- Minimal: Only declarations that must be visible to Objective-C based
on their uses (or in cases like dynamic vars) are migrated.
rdar://problem/31876357
