Summary:
This patch adds the support to create jobs for the `OffloadBundlingAction` which will invoke the `clang-offload-bundler` tool to unbundle input files.
Unlike other actions, unbundling actions have multiple outputs. Therefore, this patch adds the required changes to have a variant of `Tool::ConstructJob` with multiple outputs.
The way the naming of the results is implemented is also slightly modified so that the same action can use a different offloading prefix for each use by the different offloading actions.
With this patch, it is possible to compile a functional OpenMP binary with offloading support, even with separate compilation.
Reviewers: echristo, tra, jlebar, ABataev, hfinkel
Subscribers: mkuron, whchung, mehdi_amini, cfe-commits, Hahnfeld, andreybokhanko, arpith-jacob, carlo.bertolli, caomhin
Differential Revision: https://reviews.llvm.org/D21857
llvm-svn: 285326
Summary: This patch adds the support to create a job for the `OffloadBundlingAction` which will invoke the `clang-offload-bundler` tool.
Reviewers: echristo, tra, jlebar, ABataev, hfinkel
Subscribers: whchung, mehdi_amini, cfe-commits, Hahnfeld, andreybokhanko, arpith-jacob, carlo.bertolli, caomhin
Differential Revision: https://reviews.llvm.org/D21856
llvm-svn: 285325
Summary:
Each time that offloading support is requested by the user and the input file is not a source file, an action `OffloadUnbundlingAction` is created to signal that the input file may contain bundles, so that the proper tool is then invoked to attempt to extract the components of the bundle. This patch adds the logic to create that action in offload action builder.
The job creation for the new action will be proposed in a separate patch.
Reviewers: echristo, tra, jlebar, ABataev, hfinkel
Subscribers: whchung, mehdi_amini, cfe-commits, Hahnfeld, andreybokhanko, arpith-jacob, carlo.bertolli, caomhin
Differential Revision: https://reviews.llvm.org/D21853
llvm-svn: 285324
Summary:
In order to save the user from dealing with multiple output files (for host and device) while using separate compilation, a new action `OffloadBundlingAction` is used when the last phase is not linking. This action will then result in a job that uses the proposed bundling tool to create a single preprocessed/IR/ASM/Object file from multiple ones.
The job creation for the new action will be proposed in a separate patch.
Reviewers: echristo, tra, jlebar, ABataev, hfinkel
Subscribers: whchung, mehdi_amini, cfe-commits, Hahnfeld, andreybokhanko, arpith-jacob, carlo.bertolli, caomhin
Differential Revision: https://reviews.llvm.org/D21852
llvm-svn: 285323
Summary: This patch adds new logic to create the necessary tool chains to support offloading for OpenMP. The OpenMP related options are checked and the tool chains created accordingly. Diagnostics are emitted in case the options are illegal or express unknown targets.
Reviewers: echristo, tra, jlebar, rsmith, ABataev, hfinkel
Subscribers: whchung, mkuron, mehdi_amini, cfe-commits, Hahnfeld, arpith-jacob, carlo.bertolli, caomhin
Differential Revision: https://reviews.llvm.org/D21843
llvm-svn: 285311
Summary:
This patch replaces the CUDA specific action by a generic offload action. The offload action may have multiple dependences classier in “host” and “device”. The way this generic offloading action is used is very similar to what is done today by the CUDA implementation: it is used to set a specific toolchain and architecture to its dependences during the generation of jobs.
This patch also proposes propagating the offloading information through the action graph so that that information can be easily retrieved at any time during the generation of commands. This allows e.g. the "clang tool” to evaluate whether CUDA should be supported for the device or host and ptas to easily retrieve the target architecture.
This is an example of how the action graphs would look like (compilation of a single CUDA file with two GPU architectures)
```
0: input, "cudatests.cu", cuda, (host-cuda)
1: preprocessor, {0}, cuda-cpp-output, (host-cuda)
2: compiler, {1}, ir, (host-cuda)
3: input, "cudatests.cu", cuda, (device-cuda, sm_35)
4: preprocessor, {3}, cuda-cpp-output, (device-cuda, sm_35)
5: compiler, {4}, ir, (device-cuda, sm_35)
6: backend, {5}, assembler, (device-cuda, sm_35)
7: assembler, {6}, object, (device-cuda, sm_35)
8: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {7}, object
9: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {6}, assembler
10: input, "cudatests.cu", cuda, (device-cuda, sm_37)
11: preprocessor, {10}, cuda-cpp-output, (device-cuda, sm_37)
12: compiler, {11}, ir, (device-cuda, sm_37)
13: backend, {12}, assembler, (device-cuda, sm_37)
14: assembler, {13}, object, (device-cuda, sm_37)
15: offload, "device-cuda (nvptx64-nvidia-cuda:sm_37)" {14}, object
16: offload, "device-cuda (nvptx64-nvidia-cuda:sm_37)" {13}, assembler
17: linker, {8, 9, 15, 16}, cuda-fatbin, (device-cuda)
18: offload, "host-cuda (powerpc64le-unknown-linux-gnu)" {2}, "device-cuda (nvptx64-nvidia-cuda)" {17}, ir
19: backend, {18}, assembler
20: assembler, {19}, object
21: input, "cuda", object
22: input, "cudart", object
23: linker, {20, 21, 22}, image
```
The changes in this patch pass the existent regression tests (keeps the existent functionality) and resulting binaries execute correctly in a Power8+K40 machine.
Reviewers: echristo, hfinkel, jlebar, ABataev, tra
Subscribers: guansong, andreybokhanko, tcramer, mkuron, cfe-commits, arpith-jacob, carlo.bertolli, caomhin
Differential Revision: https://reviews.llvm.org/D18171
llvm-svn: 275645
Summary:
Currently our handling of CUDA architectures is scattered all around
clang. This patch centralizes it.
A key advantage of this centralization is that you can now write a C++
switch on e.g. CudaArch and get a compile error if you don't handle one
of the enum values.
Reviewers: tra
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D21867
llvm-svn: 274681
Summary: Also add sm_32, which was missing.
Reviewers: tra
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D21778
llvm-svn: 274680
Summary:
Previously we compiled CUDA device code to PTX assembly and embedded
that asm as text in our host binary. Now we compile to PTX assembly and
then invoke ptxas to assemble the PTX into a cubin file. We gather the
ptx and cubin files for each of our --cuda-gpu-archs and combine them
using fatbinary, and then embed that into the host binary.
Adds two new command-line flags, -Xcuda_ptxas and -Xcuda_fatbinary,
which pass args down to the external tools.
Reviewers: tra, echristo
Subscribers: cfe-commits, jhen
Differential Revision: http://reviews.llvm.org/D16082
llvm-svn: 257809
Summary: This is used by D16082 when it invokes fatbinary.
Reviewers: tra
Subscribers: cfe-commits, jhen, echristo
Differential Revision: http://reviews.llvm.org/D16097
llvm-svn: 257530
Summary:
This makes constructing Action graphs which are DAGs much simpler. It
also just simplifies in general the ownership semantics of Actions.
Depends on D15910.
Reviewers: echristo
Subscribers: klimek, cfe-commits
Differential Revision: http://reviews.llvm.org/D15911
llvm-svn: 257407
Clang needs to know target triple for both sides of compilation so that
preprocessor macros and target builtins from both sides are available.
This change augments Compilation class to carry information about
toolchains used during different CUDA compilation passes and refactors
BuildActions to use it when it constructs CUDA jobs.
Removed DeviceTriple from CudaHostAction/CudaDeviceAction as it's no
longer needed.
Differential Revision: http://reviews.llvm.org/D13144
llvm-svn: 253385
NOTE: reverts r242077 to reinstate r242058, r242065, 242067
and includes fix for OS X test failures.
- Changed driver pipeline to compile host and device side of CUDA
files and incorporate results of device-side compilation into host
object file.
- Added a test for cuda pipeline creation in clang driver.
New clang options:
--cuda-host-only - Do host-side compilation only.
--cuda-device-only - Do device-side compilation only.
--cuda-gpu-arch=<ARCH> - specify GPU architecture for device-side
compilation. E.g. sm_35, sm_30. Default is sm_20. May be used more
than once in which case one device-compilation will be done per
unique specified GPU architecture.
Differential Revision: http://reviews.llvm.org/D9509
llvm-svn: 242085
The tests were failing on OS X.
Revert "[cuda] Driver changes to compile and stitch together host and device-side CUDA code."
Revert "Fixed regex to properly match '64' in the test case."
Revert "clang/test/Driver/cuda-options.cu REQUIRES clang-driver, at least."
llvm-svn: 242077
- Changed driver pipeline to compile host and device side of CUDA
files and incorporate results of device-side compilation into host
object file.
- Added a test for cuda pipeline creation in clang driver.
New clang options:
--cuda-host-only - Do host-side compilation only.
--cuda-device-only - Do device-side compilation only.
--cuda-gpu-arch=<ARCH> - specify GPU architecture for device-side
compilation. E.g. sm_35, sm_30. Default is sm_20. May be used more
than once in which case one device-compilation will be done per
unique specified GPU architecture.
Differential Revision: http://reviews.llvm.org/D9509
llvm-svn: 242058
This reapplies r224503 along with a fix for compiling Fortran by having the
clang driver invoke gcc (see r224546, where it was reverted). I have added
a testcase for that as well.
Original commit message:
It is often convenient to use -save-temps to collect the intermediate
results of a compilation, e.g., when triaging a bug report. Besides the
temporary files for preprocessed source and assembly code, this adds the
unoptimized bitcode files as well.
This adds a new BackendJobAction, which is mostly mechanical, to run after
the CompileJobAction. When not using -save-temps, the BackendJobAction is
combined into one job with the CompileJobAction, similar to the way the
integrated assembler is handled. I've implemented this entirely as a
driver change, so under the hood, it is just using -disable-llvm-optzns
to get the unoptimized bitcode.
Based in part on a patch by Steven Wu.
rdar://problem/18909437
llvm-svn: 224688
This reverts commit r224503.
It broke compilation of fortran through the Clang driver. Previously
`clang -c t.f` would invoke `gcc t.f` and `clang -cc1as`, but now it
tries to call `clang -cc1 t.f` which fails for obvious reasons.
llvm-svn: 224546
It is often convenient to use -save-temps to collect the intermediate
results of a compilation, e.g., when triaging a bug report. Besides the
temporary files for preprocessed source and assembly code, this adds the
unoptimized bitcode files as well.
This adds a new BackendJobAction, which is mostly mechanical, to run after
the CompileJobAction. When not using -save-temps, the BackendJobAction is
combined into one job with the CompileJobAction, similar to the way the
integrated assembler is handled. I've implemented this entirely as a
driver change, so under the hood, it is just using -disable-llvm-optzns
to get the unoptimized bitcode.
Based in part on a patch by Steven Wu.
rdar://problem/18909437
llvm-svn: 224503
It hits a limit when we reach ActionList, which is used for dynamic
conditional ownership, so we lose type safety there.
This did expose at least one caller "lying" about ownership (passing
ownership to an Action, then updating the Action to specify that it
doesn't actually own the thing that was passed) - changing this to
unique_ptr just makes that oddity more obvious.
llvm-svn: 216713
Use the verify hook rather than the compile hook to represent the
-verify-pch action, and move the exising --verify-debug-info action
into its own subclass of VerifyJobAction. Incidentally change the name
printed by -ccc-print-phases for --verify-debug-info.
llvm-svn: 200938
The big changes are:
- Deleting Driver/(Arg|Opt)*
- Rewriting includes to llvm/Option/ and re-sorting
- 'using namespace llvm::opt' in clang::driver
- Fixing the autoconf build by adding option everywhere
As discussed in the review, this change includes using directives in
header files. I'll make follow up changes to remove those in favor of
name specifiers.
Reviewers: espindola
Differential Revision: http://llvm-reviews.chandlerc.com/D975
llvm-svn: 183989
uncovered.
This required manually correcting all of the incorrect main-module
headers I could find, and running the new llvm/utils/sort_includes.py
script over the files.
I also manually added quite a few missing headers that were uncovered by
shuffling the order or moving headers up to be main-module-headers.
llvm-svn: 169237
the new Objective-C NSArray/NSDictionary/NSNumber literal syntax.
This introduces a new library, libEdit, which provides a new way to support
migration of code that improves on the original ARC migrator. We now believe
that most of its functionality can be refactored into the existing libraries,
and thus this new library may shortly disappear.
llvm-svn: 152141
output on darwin so is hard coded there.
As a note this will need a little bit of refactoring in the class
hierarchy to separate it out for different verifiers based on input type.
Fixes rdar://8256258.
llvm-svn: 138343
earlier.
- This gives us a simple ownership model, and allows clients access
to more information should they ever want it.
- We now free Actions correctly.
llvm-svn: 67158
Simon Pilgrim