| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
This patch changes cc1 option for PGO profile use from
-fprofile-instr-use=<path> to -fprofile-instrument-use-path=<path>.
-fprofile-instr-use=<path> is now a driver only option.
In addition to decouple the cc1 option from the driver level option, this patch
also enables IR level profile use. cc1 option handling now reads the profile
header and sets CodeGenOpt ProfileUse (valid values are {None, Clang, LLVM}
-- this is a common enum for -fprofile-instrument={}, for the profile
instrumentation), and invoke the pipeline to enable the respective PGO use pass.
Reviewers: silvas, davidxl
Differential Revision: http://reviews.llvm.org/D17737
llvm-svn: 262515
|
|
In the gcc precompiled header model, one explicitly runs clang with `-x
c++-header` on a .h file to produce a gch file, and then includes the header
with `-include foo.h` and if a .gch file exists for that header it gets used.
This is documented at
http://clang.llvm.org/docs/UsersManual.html#precompiled-headers
cl.exe's model is fairly different, and controlled by the two flags /Yc and
/Yu. A pch file is generated as a side effect of a regular compilation when
/Ycheader.h is passed. While the compilation is running, the compiler keeps
track of #include lines in the main translation unit and writes everything up
to an `#include "header.h"` line into a pch file. Conversely, /Yuheader.h tells
the compiler to skip all code in the main TU up to and including `#include
"header.h"` and instead load header.pch. (It's also possible to use /Yc and /Yu
without an argument, in that case a `#pragma hrdstop` takes the role of
controlling the point where pch ends and real code begins.)
This patch implements limited support for this in that it requires the pch
header to be passed as a /FI force include flag – with this restriction,
it can be implemented almost completely in the driver with fairly small amounts
of code. For /Yu, this is trivial, and for /Yc a separate pch action is added
that runs before the actual compilation. After r261774, the first failing
command makes a compilation stop – this means if the pch fails to build the
main compilation won't run, which is what we want. However, in /fallback builds
we need to run the main compilation even if the pch build fails so that the
main compilation's fallback can run. To achieve this, add a ForceSuccessCommand
that pretends that the pch build always succeeded in /fallback builds (the main
compilation will then fail to open the pch and run the fallback cl.exe
invocation).
If /Yc /Yu are used in a setup that clang-cl doesn't implement yet, clang-cl
will now emit a "not implemented yet; flag ignored" warning that can be
disabled using -Wno-clang-cl-pch.
Since clang-cl doesn't yet serialize some important things (most notably
`pragma comment(lib, ...)`, this feature is disabled by default and only
enabled by an internal driver flag. Once it's more stable, this internal flag
will disappear.
(The default stdafx.h setup passes stdafx.h as explicit argument to /Yc but not
as /FI – instead every single TU has to `#include <stdafx.h>` as first thing it
does. Implementing support for this should be possible with the approach in
this patch with minimal frontend changes by passing a --stop-at / --start-at
flag from the driver to the frontend. This is left for a follow-up. I don't
think we ever want to support `#pragma hdrstop`, and supporting it with this
approach isn't easy: This approach relies on the driver knowing the pch
filename in advance, and `#pragma hdrstop(out.pch)` can set the output
filename, so the driver can't know about it in advance.)
clang-cl now also honors /Fp and puts pch files in the same spot that cl.exe
would put them, but the pch file format is of course incompatible. This has
ramifications on /fallback, so /Yc /Yu aren't passed through to cl.exe in
/fallback builds.
http://reviews.llvm.org/D17695
llvm-svn: 262420
|
|
This patch expands cc1 option -fprofile-instrument= with a new value: -fprofile-instrument=llvm
which enables IR level PGO instrumentation.
Reviewers: davidxl, silvas
Differential Revision: http://reviews.llvm.org/D17622
llvm-svn: 262239
|
|
This patch introduces the -fwhole-program-vtables flag, which enables the
whole-program vtable optimization feature (D16795) in Clang.
Differential Revision: http://reviews.llvm.org/D16821
llvm-svn: 261767
|
|
call to getVFSFromYAML in unittests"
This reverts commit r261552 and r261556 because of failing unittests on
windows:
Failing Tests (4):
Clang-Unit :: Basic/BasicTests.exe/VFSFromYAMLTest.CaseInsensitive
Clang-Unit :: Basic/BasicTests.exe/VFSFromYAMLTest.DirectoryIteration
Clang-Unit :: Basic/BasicTests.exe/VFSFromYAMLTest.MappedFiles
Clang-Unit :: Basic/BasicTests.exe/VFSFromYAMLTest.UseExternalName
llvm-svn: 261613
|
|
The VFS overlay mapping between virtual paths and real paths is done through
the 'external-contents' entries in YAML files, which contains hardcoded paths
to the real files.
When a module compilation crashes, headers are dumped into <name>.cache/vfs
directory and are mapped via the <name>.cache/vfs/vfs.yaml. The script
generated for reproduction uses -ivfsoverlay pointing to file to gather the
mapping between virtual paths and files inside <name>.cache/vfs. Currently, we
are only capable of reproducing such crashes in the same machine as they
happen, because of the hardcoded paths in 'external-contents'.
To be able to reproduce a crash in another machine, this patch introduces a new
option in the VFS yaml file called 'overlay-relative'. When it's equal to
'true' it means that the provided path to the YAML file through the
-ivfsoverlay option should also be used to prefix the final path for every
'external-contents'.
Example, given the invocation snippet "... -ivfsoverlay
<name>.cache/vfs/vfs.yaml" and the following entry in the yaml file:
"overlay-relative": "true",
"roots": [
...
"type": "directory",
"name": "/usr/include",
"contents": [
{
"type": "file",
"name": "stdio.h",
"external-contents": "/usr/include/stdio.h"
},
...
Here, a file manager request for virtual "/usr/include/stdio.h", that will map
into real path "/<absolute_path_to>/<name>.cache/vfs/usr/include/stdio.h.
This is a useful feature for debugging module crashes in machines other than
the one where the error happened.
Differential Revision: http://reviews.llvm.org/D17457
rdar://problem/24499339
llvm-svn: 261552
|
|
The -EHc flag implicitly adds a nothrow attribute to any extern "C"
function when exceptions are enabled.
llvm-svn: 261425
|
|
option. Previously these options could both be used to specify that you were
compiling the implementation file of a module, with a different set of minor
bugs in each case.
This change removes -fmodule-implementation-of, and instead tracks a flag to
determine whether we're currently building a module. -fmodule-name now behaves
the same way that -fmodule-implementation-of previously did.
llvm-svn: 261372
|
|
to trace-bb, but has a different API. We already use the equivalent flag in GCC for Linux kernel fuzzing. We may be able to use this flag with AFL too
llvm-svn: 261159
|
|
specialized code generation for different devices.
This was reverted by r260036, but was not the cause of the problem in the buildbot.
llvm-svn: 260106
|
|
code generation for different devices."
This reverts commit r259985, as it still fails one buildbot.
llvm-svn: 260036
|
|
generation for different devices.
This was reverted due to a failure in a buildbot, but it turned out the failure was unrelated.
llvm-svn: 259985
|
|
generation for different devices.
It triggered some problem in the configuration related with zlib and exposed in the driver.
llvm-svn: 259984
|
|
devices.
Summary:
Different devices may in some cases require different code generation schemes in order to implement OpenMP. This is required not only for performance reasons, but also because it may not be possible to have the current (default) implementation working for these devices. E.g. GPU's cannot implement the same scheme a target such as powerpc or x86b would use, in the sense that it does not have the ability to fork threads, instead all the threads are always executing and need to be managed by the implementation.
This patch proposes a reorganization of the code in the OpenMP code generation to pave the way to have specialized implementation of OpenMP support. More than a "real" patch this is more a request for comments in order to understand if what is proposed is acceptable or if there are better/easier ways to do it.
In this patch part of the common OpenMP codegen infrastructure is moved to a new file under a new namespace (CGOpenMPCommon) so it can be shared between the default implementation and the specialized one. When CGOpenMPRuntime is created, an attempt to select a specialized implementation is done.
In the patch a specialization for nvptx targets is done which currently checks if the target is an OpenMP device and trap if it is not.
Let me know comments suggestions you may have.
Reviewers: hfinkel, carlo.bertolli, arpith-jacob, kkwli0, ABataev
Subscribers: Hahnfeld, cfe-commits, fraggamuffin, caomhin, jholewinski
Differential Revision: http://reviews.llvm.org/D16784
llvm-svn: 259977
|
|
llvm-svn: 259950
|
|
This patch changes cc1 option -fprofile-instr-generate to an enum option
-fprofile-instrument={clang|none}. It also changes cc1 options
-fprofile-instr-generate= to -fprofile-instrument-path=.
The driver level option -fprofile-instr-generate and -fprofile-instr-generate=
remain intact. This change will pave the way to integrate new PGO
instrumentation in IR level.
Review: http://reviews.llvm.org/D16730
llvm-svn: 259811
|
|
from Driver to Frontend.
llvm-svn: 259489
|
|
This is same as GCC behavior (tested with GCC 4.8.2).
Differential Revision: http://reviews.llvm.org/D16365
llvm-svn: 258850
|
|
Summary:
Turns out the variadic function checking added in r258643 was too strict
for some existing users; give them an escape valve. When
-fcuda-allow-variadic-functions is passed, the front-end makes no
attempt to disallow C-style variadic functions. Calls to va_arg are
still not allowed.
Reviewers: tra
Subscribers: cfe-commits, jhen, echristo, bkramer
Differential Revision: http://reviews.llvm.org/D16559
llvm-svn: 258822
|
|
This is part of a new statistics gathering feature for the sanitizers.
See clang/docs/SanitizerStats.rst for further info and docs.
Differential Revision: http://reviews.llvm.org/D16175
llvm-svn: 257971
|
|
of the file name. This is consistent with how other HeaderSearchOptions
are handled.
Due to the other inputs of the module hash (revision number) this is not
really testable in a meaningful way.
llvm-svn: 257520
|
|
llvm-svn: 257291
|
|
Addresses PR4941 and rdar://6756912.
http://reviews.llvm.org/D15195
llvm-svn: 256937
|
|
device codegen.
This patch attempts to fix the regressions identified when the patch was committed initially.
Thanks to Michael Liao for identifying the fix in the offloading metadata generation
related with side effects in evaluation of function arguments.
llvm-svn: 256933
|
|
Addressing review comment in D13221.
Differential Revision: http://reviews.llvm.org/D15882
llvm-svn: 256897
|
|
device codegen.
It was causing two regression, so I'm reverting until the cause is found.
llvm-svn: 256858
|
|
Summary:
In order to offloading work properly two things need to be in place:
- a descriptor with all the offloading information (device entry functions, and global variable) has to be created by the host and registered in the OpenMP offloading runtime library.
- all the device functions need to be emitted for the device and a convention has to be in place so that the runtime library can easily map the host ID of an entry point with the actual function in the device.
This patch adds support for these two things. However, only entry functions are being registered given that 'declare target' directive is not yet implemented.
About offloading descriptor:
The details of the descriptor are explained with more detail in http://goo.gl/L1rnKJ. Basically the descriptor will have fields that specify the number of devices, the pointers to where the device images begin and end (that will be defined by the linker), and also pointers to a the begin and end of table whose entries contain information about a specific entry point. Each entry has the type:
```
struct __tgt_offload_entry{
void *addr;
char *name;
int64_t size;
};
```
and will be implemented in a pre determined (ELF) section `.omp_offloading.entries` with 1-byte alignment, so that when all the objects are linked, the table is in that section with no padding in between entries (will be like a C array). The code generation ensures that all `__tgt_offload_entry` entries are emitted in the same order for both host and device so that the runtime can have the corresponding entries in both host and device in same index of the table, and efficiently implement the mapping.
The resulting descriptor is registered/unregistered with the runtime library using the calls `__tgt_register_lib` and `__tgt_unregister_lib`. The registration is implemented in a high priority global initializer so that the registration happens always before any initializer (that can potentially include target regions) is run.
The driver flag -omptargets= was created to specify a comma separated list of devices the user wants to support so that the new functionality can be exercised. Each device is specified with its triple.
About target codegen:
The target codegen is pretty much straightforward as it reuses completely the logic of the host version for the same target region. The tricky part is to identify the meaningful target regions in the device side. Unlike other programming models, like CUDA, there are no already outlined functions with attributes that mark what should be emitted or not. So, the information on what to emit is passed in the form of metadata in host bc file. This requires a new option to pass the host bc to the device frontend. Then everything is similar to what happens in CUDA: the global declarations emission is intercepted to check to see if it is an "interesting" declaration. The difference is that instead of checking an attribute, the metadata information in checked. Right now, there is only a form of metadata to pass information about the device entry points (target regions). A class `OffloadEntriesInfoManagerTy` was created to manage all the information and queries related with the metadata. The metadata looks like this:
```
!omp_offload.info = !{!0, !1, !2, !3, !4, !5, !6}
!0 = !{i32 0, i32 52, i32 77426347, !"_ZN2S12r1Ei", i32 479, i32 13, i32 4}
!1 = !{i32 0, i32 52, i32 77426347, !"_ZL7fstatici", i32 461, i32 11, i32 5}
!2 = !{i32 0, i32 52, i32 77426347, !"_Z9ftemplateIiET_i", i32 444, i32 11, i32 6}
!3 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 99, i32 11, i32 0}
!4 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 272, i32 11, i32 3}
!5 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 127, i32 11, i32 1}
!6 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 159, i32 11, i32 2}
```
The fields in each metadata entry are (in sequence):
Entry 1) an ID of the type of metadata - right now only zero is used meaning "OpenMP target region".
Entry 2) a unique ID of the device where the input source file that contain the target region lives.
Entry 3) a unique ID of the file where the input source file that contain the target region lives.
Entry 4) a mangled name of the function that encloses the target region.
Entries 5) and 6) line and column number where the target region was found.
Entry 7) is the order the entry was emitted.
Entry 2) and 3) are required to distinguish files that have the same function name.
Entry 4) is required to distinguish different instances of the same declaration (usually templated ones)
Entries 5) and 6) are required to distinguish the particular target region in body of the function (it is possible that a given target region is not an entry point - if clause can evaluate always to zero - and therefore we need to identify the "interesting" target regions. )
This patch replaces http://reviews.llvm.org/D12306.
Reviewers: ABataev, hfinkel, tra, rjmccall, sfantao
Subscribers: FBrygidyn, piotr.rak, Hahnfeld, cfe-commits
Differential Revision: http://reviews.llvm.org/D12614
llvm-svn: 256842
|
|
r255281.
llvm-svn: 256396
|
|
The /Brepro flag controls whether or not the compiler should embed
timestamps into the object file. Object files which do not embed
timestamps are not suitable for incremental linking but are suitable for
hermetic build systems and staged self-hosts of clang.
A normal clang spelling of this flag has been added,
-mincremental-linker-compatible.
llvm-svn: 256204
|
|
Reapplies r256063, except instead of frugally re-using an LLVM enum,
we define a Clang enum, to avoid exposing too much LLVM interface.
Differential Revision: http://reviews.llvm.org/D15650
llvm-svn: 256078
|
|
llvm-svn: 256066
|
|
Differential Revision: http://reviews.llvm.org/D15650
llvm-svn: 256063
|
|
Clang-side cross-DSO CFI.
* Adds a command line flag -f[no-]sanitize-cfi-cross-dso.
* Links a runtime library when enabled.
* Emits __cfi_slowpath calls is bitset test fails.
* Emits extra hash-based bitsets for external CFI checks.
* Sets a module flag to enable __cfi_check generation during LTO.
This mode does not yet support diagnostics.
llvm-svn: 255694
|
|
nested) for all the platforms except PS4.
For PS4, generate explicit import for anonymous namespaces and mark it by DW_AT_artificial attribute.
Differential Revision: http://reviews.llvm.org/D12624
llvm-svn: 255281
|
|
Summary:
Adds new option -fthinlto-index=<file> to invoke the LTO pipeline
along with function importing via clang using the supplied function
summary index file. This supports invoking the parallel ThinLTO
backend processes in a distributed build environment via clang.
Additionally, this causes the module linker to be invoked on the bitcode
file being compiled to perform any necessary promotion and renaming of
locals that are exported via the function summary index file.
Add a couple tests that confirm we get expected errors when we try to
use the new option on a file that isn't bitcode, or specify an invalid
index file. The tests also confirm that we trigger the expected function
import pass.
Depends on D15024
Reviewers: joker.eph, dexonsmith
Subscribers: joker.eph, davidxl, cfe-commits
Differential Revision: http://reviews.llvm.org/D15025
llvm-svn: 254927
|
|
This flag causes all files that were read by the compilation to be embedded
into a produced module file. This is useful for distributed build systems that
use an include scanning system to determine which files are "needed" by a
compilation, and only provide those files to remote compilation workers. Since
using a module can require any file that is part of that module (or anything it
transitively includes), files that are not found by an include scanner can be
required in a regular build using explicit modules. With this flag, only files
that are actually referenced by transitively-#included files are required to be
present on the build machine.
llvm-svn: 253950
|
|
(Re-apply patch after bug fixing)
This diff makes sure that the driver does not pass
-fomit-frame-pointer or -momit-leaf-frame-pointer to
the frontend when -pg is used. Currently, clang gives
an error if -fomit-frame-pointer is used in combination
with -pg, but -momit-leaf-frame-pointer was forgotten.
Also, disable frame pointer elimination in the frontend
when -pg is set.
Patch by Stefan Kempf.
llvm-svn: 253886
|
|
llvm-svn: 253851
|
|
This diff makes sure that the driver does not pass
-fomit-frame-pointer or -momit-leaf-frame-pointer to
the frontend when -pg is used. Currently, clang gives
an error if -fomit-frame-pointer is used in combination
with -pg, but -momit-leaf-frame-pointer was forgotten.
Also, disable frame pointer elimination in the frontend
when -pg is set.
Patch by Stefan Kempf.
llvm-svn: 253846
|
|
This reverts commit r253269.
This leads to assert / segfault triggering on the following reduced example:
float foo(float U, float base, float cell) { return (U = 2 * base) - cell; }
llvm-svn: 253337
|
|
Differential Revision: D14200
llvm-svn: 253269
|
|
This failed to solve the problem it was aimed at, and introduced just as many
issues as it resolved. Realistically, we need to deal with the possibility that
multiple modules might define different internal linkage symbols with the same
name, and this isn't a problem unless two such symbols are simultaneously
visible.
The case where two modules define equivalent internal linkage symbols is
handled by r252063: if lookup finds multiple sufficiently-similar entities from
different modules, we just pick one of them as an extension (but we keep them
separate).
llvm-svn: 252957
|
|
The -meabi flag to control LLVM EABI version.
Without '-meabi' or with '-meabi default' imply LLVM triple default.
With '-meabi gnu' sets EABI GNU.
With '-meabi 4' or '-meabi 5' set EABI version 4 and 5 respectively.
A similar patch was introduced in LLVM.
Patch by Vinicius Tinti.
llvm-svn: 252463
|
|
rdar://problem/23415863
llvm-svn: 252187
|
|
Introduce the notion of a module file extension, which introduces
additional information into a module file at the time it is built that
can then be queried when the module file is read. Module file
extensions are identified by a block name (which must be unique to the
extension) and can write any bitstream records into their own
extension block within the module file. When a module file is loaded,
any extension blocks are matched up with module file extension
readers, that are per-module-file and are given access to the input
bitstream.
Note that module file extensions can only be introduced by
programmatic clients that have access to the CompilerInvocation. There
is only one such extension at the moment, which is used for testing
the module file extension harness. As a future direction, one could
imagine allowing the plugin mechanism to introduce new module file
extensions.
llvm-svn: 251955
|
|
Linking options for particular file depend on the option that specifies the file.
Currently there are two:
* -mlink-bitcode-file links in complete content of the specified file.
* -mlink-cuda-bitcode links in only the symbols needed by current TU.
Linked symbols are internalized. This bitcode linking mode is used to
link device-specific bitcode provided by CUDA.
Files are linked in order they are specified on command line.
-mlink-cuda-bitcode replaces -fcuda-uses-libdevice flag.
Differential Revision: http://reviews.llvm.org/D13913
llvm-svn: 251427
|
|
Previously, __weak was silently accepted and ignored in MRC mode.
That makes this a potentially source-breaking change that we have to
roll out cautiously. Accordingly, for the time being, actual support
for __weak references in MRC is experimental, and the compiler will
reject attempts to actually form such references. The intent is to
eventually enable the feature by default in all non-GC modes.
(It is, of course, incompatible with ObjC GC's interpretation of
__weak.)
If you like, you can enable this feature with
-Xclang -fobjc-weak
but like any -Xclang option, this option may be removed at any point,
e.g. if/when it is eventually enabled by default.
This patch also enables the use of the ARC __unsafe_unretained qualifier
in MRC. Unlike __weak, this is being enabled immediately. Since
variables are essentially __unsafe_unretained by default in MRC,
the only practical uses are (1) communication and (2) changing the
default behavior of by-value block capture.
As an implementation matter, this means that the ObjC ownership
qualifiers may appear in any ObjC language mode, and so this patch
removes a number of checks for getLangOpts().ObjCAutoRefCount
that were guarding the processing of these qualifiers. I don't
expect this to be a significant drain on performance; it may even
be faster to just check for these qualifiers directly on a type
(since it's probably in a register anyway) than to do N dependent
loads to grab the LangOptions.
rdar://9674298
llvm-svn: 251041
|
|
Add -fcoroutines flag (just for -cc1 for now) to enable the feature. Early
indications are that this will be part of -std=c++1z.
llvm-svn: 250980
|
|
The ELF symbol visibilities are:
- internal: Not visibile across DSOs, cannot pass address across DSOs
- hidden: Not visibile across DSOs, can be called indirectly
- default: Usually visible across DSOs, possibly interposable
- protected: Visible across DSOs, not interposable
LLVM only supports the latter 3 visibilities. Internal visibility is in
theory useful, as it allows you to assume that the caller is maintaining
a PIC register for you in %ebx, or in some other pre-arranged location.
As far as LLVM is concerned, this isn't worth the trouble. Using hidden
visibility is always correct, so we can just do that.
Resolves PR9183.
llvm-svn: 250954
|
|
Summary: It breaks the build for the ASTMatchers
Subscribers: klimek, cfe-commits
Differential Revision: http://reviews.llvm.org/D13893
llvm-svn: 250827
|
