gtbercea updated this revision to Diff 105453.
gtbercea added a comment.
.
https://reviews.llvm.org/D29647
Files:
include/clang/Driver/ToolChain.h
lib/Driver/ToolChain.cpp
lib/Driver/ToolChains/BareMetal.cpp
lib/Driver/ToolChains/BareMetal.h
lib/Driver/ToolChains/Clang.cpp
hfinkel added a comment.
In https://reviews.llvm.org/D29647#795271, @hfinkel wrote:
> LGTM
When you commit this, please make sure to mention in the commit message that
the test cases will be associated with follow-up commits.
Repository:
rL LLVM
https://reviews.llvm.org/D29647
hfinkel accepted this revision.
hfinkel added a comment.
This revision is now accepted and ready to land.
LGTM
Repository:
rL LLVM
https://reviews.llvm.org/D29647
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
gtbercea updated this revision to Diff 104527.
gtbercea added a comment.
Split previous diff into a "device offloading kind" patch (show here) and a
**new** patch which relies on a new compiler flag.
A TODO has been added to signal that the compute capability is to be handled in
the **new**
gtbercea added inline comments.
Comment at: test/Driver/openmp-offload.c:614
+/// Check -march propagates compute capability to device offloading toolchain.
+// RUN: %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda
-save-temps -no-canonical-prefixes
hfinkel added inline comments.
Comment at: test/Driver/openmp-offload.c:614
+/// Check -march propagates compute capability to device offloading toolchain.
+// RUN: %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda
-save-temps -no-canonical-prefixes
gtbercea added inline comments.
Comment at: test/Driver/openmp-offload.c:614
+/// Check -march propagates compute capability to device offloading toolchain.
+// RUN: %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda
-save-temps -no-canonical-prefixes
hfinkel added inline comments.
Comment at: test/Driver/openmp-offload.c:614
+/// Check -march propagates compute capability to device offloading toolchain.
+// RUN: %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda
-save-temps -no-canonical-prefixes
gtbercea marked an inline comment as done.
gtbercea added inline comments.
Comment at: lib/Driver/ToolChains/Cuda.cpp:217
+ std::vector GPUArchNames;
+ // If this is an OpenMP action we need to extract the device architecture
from
+ // the -march option.
gtbercea updated this revision to Diff 104427.
gtbercea added a comment.
Herald added subscribers: aheejin, jgravelle-google.
Updated diff to address comments.
Repository:
rL LLVM
https://reviews.llvm.org/D29647
Files:
include/clang/Driver/ToolChain.h
lib/Driver/ToolChain.cpp
gtbercea added inline comments.
Comment at: lib/Driver/ToolChains/Cuda.cpp:435
+
+// TODO: get the compute capability from offloading arguments when not
+// using the default compute capability of sm_20.
hfinkel wrote:
> gtbercea wrote:
> > hfinkel
hfinkel added inline comments.
Comment at: lib/Driver/ToolChains/Cuda.cpp:435
+
+// TODO: get the compute capability from offloading arguments when not
+// using the default compute capability of sm_20.
gtbercea wrote:
> hfinkel wrote:
> > gtbercea
gtbercea marked an inline comment as done.
gtbercea added inline comments.
Comment at: lib/Driver/ToolChains/Cuda.cpp:435
+
+// TODO: get the compute capability from offloading arguments when not
+// using the default compute capability of sm_20.
hfinkel
hfinkel added inline comments.
Comment at: lib/Driver/ToolChains/Cuda.cpp:435
+
+// TODO: get the compute capability from offloading arguments when not
+// using the default compute capability of sm_20.
gtbercea wrote:
> hfinkel wrote:
> > Why is this a
gtbercea marked an inline comment as done.
gtbercea added inline comments.
Comment at: lib/Driver/ToolChains/Cuda.cpp:435
+
+// TODO: get the compute capability from offloading arguments when not
+// using the default compute capability of sm_20.
hfinkel
hfinkel added a comment.
In general, this patch seems to be missing tests (unless it is actually NFC, or
you can't write tests yet, which, in either case, need to be explained).
Comment at: lib/Driver/ToolChains/Cuda.cpp:217
+ std::vector GPUArchNames;
+ // If this is an
Hahnfeld added a comment.
One minor drive-by comment. I think there is still one outstanding from
Justin...
Comment at: lib/Driver/ToolChains/Cuda.cpp:368
+ assert(!GpuArch.empty() && "Must have an explicit GPU arch.");
+ std::string LibDeviceFile;
You
gtbercea added inline comments.
Comment at: lib/Driver/Tools.cpp:12136
// Obtain architecture from the action.
- CudaArch gpu_arch = StringToCudaArch(JA.getOffloadingArch());
assert(gpu_arch != CudaArch::UNKNOWN &&
jlebar wrote:
> Why does
gtbercea added inline comments.
Comment at: lib/Driver/ToolChains.cpp:4902
+ DeviceOffloadingKind == Action::OFK_Cuda) &&
+ "The offloading kind is not OpenMP or CUDA.");
jlebar wrote:
> Not sure this assertion message helps us much beyond
gtbercea updated this revision to Diff 93181.
gtbercea marked 5 inline comments as done.
gtbercea added a comment.
Herald added subscribers: sbc100, dschuff, jfb, rengolin.
Update patch to reflect latest source code changes.
Repository:
rL LLVM
https://reviews.llvm.org/D29647
Files:
jlebar added inline comments.
Comment at: lib/Driver/ToolChains.cpp:4902
+ DeviceOffloadingKind == Action::OFK_Cuda) &&
+ "The offloading kind is not OpenMP or CUDA.");
Not sure this assertion message helps us much beyond what's already in the
21 matches
Mail list logo
