mirror of
https://github.com/pytorch/pytorch.git
synced 2026-11-20 09:04:25 +08:00
c11b301bcd
This PR is the first step towards refactors the build for nvfuser in order to have the coegen being a standalone library.
Contents inside this PR:
1. nvfuser code base has been moved to `./nvfuser`, from `./torch/csrc/jit/codegen/cuda/`, except for registration code for integration (interface.h/interface.cpp)
2. splits the build system so nvfuser is generating its own `.so` files. Currently there are:
- `libnvfuser_codegen.so`, which contains the integration, codegen and runtime system of nvfuser
- `nvfuser.so`, which is nvfuser's python API via pybind. Python frontend is now exposed via `nvfuser._C.XXX` instead of `torch._C._nvfuser`
3. nvfuser cpp tests is currently being compiled into `nvfuser_tests`
4. cmake is refactored so that:
- nvfuser now has its own `CMakeLists.txt`, which is under `torch/csrc/jit/codegen/cuda/`.
- nvfuser backend code is not compiled inside `libtorch_cuda_xxx` any more
- nvfuser is added as a subdirectory under `./CMakeLists.txt` at the very end after torch is built.
- since nvfuser has dependency on torch, the registration of nvfuser at runtime is done via dlopen (`at::DynamicLibrary`). This avoids circular dependency in cmake, which will be a nightmare to handle. For details, look at `torch/csrc/jit/codegen/cuda/interface.cpp::LoadingNvfuserLibrary`
Future work that's scoped in following PR:
- Currently since nvfuser codegen has dependency on torch, we need to refactor that out so we can move nvfuser into a submodule and not rely on dlopen to load the library. @malfet
- Since we moved nvfuser into a cmake build, we effectively disabled bazel build for nvfuser. This could impact internal workload at Meta, so we need to put support back. cc'ing @vors
Pull Request resolved: https://github.com/pytorch/pytorch/pull/89621
Approved by: https://github.com/davidberard98
JIT C++ Tests
Adding a new test
First, create a new test file. Test files should have be placed in this
directory, with a name that starts with test_, like test_foo.cpp.
In general a single test suite
Add your test file to the JIT_TEST_SRCS list in test/cpp/jit/CMakeLists.txt.
A test file may look like:
#include <gtest/gtest.h>
using namespace ::torch::jit
TEST(FooTest, BarBaz) {
// ...
}
// Append '_CUDA' to the test case name will automatically filter it out if CUDA
// is not compiled.
TEST(FooTest, NeedsAGpu_CUDA) {
// ...
}
// Similarly, if only one GPU is detected, tests with `_MultiCUDA` at the end
// will not be run.
TEST(FooTest, NeedsMultipleGpus_MultiCUDA) {
// ...
}
Building and running the tests
The following commands assume you are in PyTorch root.
# ... Build PyTorch from source, e.g.
python setup.py develop
# (re)build just the binary
ninja -C build bin/test_jit
# run tests
build/bin/test_jit --gtest_filter='glob_style_filter*'