[bazel] GPU-support: add @local_config_cuda and @cuda (pytorch#63604)

Summary:
## Context

We take the first step at tackling the GPU-bazel support by adding bazel external workspaces `local_config_cuda` and `cuda`, where the first one has some hardcoded values and lists of files, and the second one provides a nicer, high-level wrapper that maps into the already expected by pytorch bazel targets that are guarded with `if_cuda` macro.

The prefix `local_config_` signifies the fact that we are breaking the bazel hermeticity philosophy by explicitly relaying on the CUDA installation that is present on the machine.

## Testing

Notice an important scenario that is unlocked by this change: compilation of cpp code that depends on cuda libraries (i.e. cuda.h and so on).

Before:
```
sergei.vorobev@cs-sv7xn77uoy-gpu-1628706590:~/src/pytorch4$ bazelisk build --define=cuda=true //:c10
ERROR: /home/sergei.vorobev/src/pytorch4/tools/config/BUILD:12:1: no such package 'tools/toolchain': BUILD file not found in any of the following directories. Add a BUILD file to a directory to mark it as a package.
 - /home/sergei.vorobev/src/pytorch4/tools/toolchain and referenced by '//tools/config:cuda_enabled_and_capable'
ERROR: While resolving configuration keys for //:c10: Analysis failed
ERROR: Analysis of target '//:c10' failed; build aborted: Analysis failed
INFO: Elapsed time: 0.259s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (2 packages loaded, 2 targets configured)
```

After:
```
sergei.vorobev@cs-sv7xn77uoy-gpu-1628706590:~/src/pytorch4$ bazelisk build --define=cuda=true //:c10
INFO: Analyzed target //:c10 (6 packages loaded, 246 targets configured).
INFO: Found 1 target...
Target //:c10 up-to-date:
  bazel-bin/libc10.lo
  bazel-bin/libc10.so
INFO: Elapsed time: 0.617s, Critical Path: 0.04s
INFO: 0 processes.
INFO: Build completed successfully, 1 total action
```

The `//:c10` target is a good testing one for this, because it has such cases where the [glob is different](https://github.com/pytorch/pytorch/blob/075024b9a34904ec3ecdab3704c3bcaa329bdfea/BUILD.bazel#L76-L81), based on do we compile for CUDA or not.

## What is out of scope of this PR

This PR is a first in a series of providing the comprehensive GPU bazel build support. Namely, we don't tackle the [cu_library](https://github.com/pytorch/pytorch/blob/11a40ad915d4d3d8551588e303204810887fcf8d/tools/rules/cu.bzl#L2) implementation here. This would be a separate large chunk of work.

Pull Request resolved: pytorch#63604

Reviewed By: soulitzer

Differential Revision: D30442083

Pulled By: malfet

fbshipit-source-id: b2a8e4f7e5a25a69b960a82d9e36ba568eb64595

.bazelrc

@@ -3,7 +3,11 @@ build --copt=-I.
 build --copt=-isystem --copt bazel-out/k8-fastbuild/bin
 
 # Configuration to disable tty features for environments like CI
-
 build:no-tty --curses no
 build:no-tty --progress_report_interval 10
 build:no-tty --show_progress_rate_limit 10
+
+# Configuration to build with GPU support
+build:gpu --define=cuda=true
+# define a separate build folder for faster switching between configs
+build:gpu --platform_suffix=-gpu

.github/scripts/generate_ci_workflows.py

@@ -497,7 +497,7 @@ def generate_workflow_file(self, workflow_template: jinja2.Template) -> None:
     CIWorkflow(
         arch="linux",
         build_environment="linux-xenial-py3.6-gcc7-bazel-test",
-        docker_image_base=f"{DOCKER_REGISTRY}/pytorch/pytorch-linux-xenial-py3.6-gcc7",
+        docker_image_base=f"{DOCKER_REGISTRY}/pytorch/pytorch-linux-bionic-cuda10.2-cudnn7-py3.9-gcc7",
         test_runner_type=LINUX_CPU_TEST_RUNNER,
         on_pull_request=True,
         ciflow_config=CIFlowConfig(

.jenkins/pytorch/build.sh

@@ -224,7 +224,11 @@ if [[ "$BUILD_ENVIRONMENT" == *-bazel-* ]]; then
 
   get_bazel
 
+  # first build the whole torch for CPU-only
   tools/bazel build --config=no-tty :torch
+  # then build selected set of targets with GPU-support.
+  # TODO: eventually this should converge to building the whole :torch with GPU-support
+  tools/bazel build --config=no-tty --config=gpu :c10
 else
   # check that setup.py would fail with bad arguments
   echo "The next three invocations are expected to fail with invalid command error messages."

WORKSPACE

@@ -1,7 +1,7 @@
 workspace(name = "pytorch")
 
 load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
-load("//tools/rules:workspace.bzl", "new_patched_local_repository")
+load("//tools/rules:workspace.bzl", "new_patched_local_repository", "new_empty_repository")
 
 http_archive(
     name = "bazel_skylib",
@@ -170,3 +170,14 @@ protobuf_deps()
 load("@rules_python//python:repositories.bzl", "py_repositories")
 
 py_repositories()
+
+local_repository(
+    name = "local_config_cuda",
+    path = "third_party/tensorflow_cuda_bazel_build",
+)
+
+# Wrapper to expose local_config_cuda in an agnostic way
+new_empty_repository(
+    name = "cuda",
+    build_file = "//third_party:cuda.BUILD",
+)

third_party/cuda.BUILD

@@ -0,0 +1,43 @@
+"""
+Collect all the CUDA stuff from @local_config_cuda in a single target
+for convenience.
+"""
+
+cc_library(
+    name = "cuda",
+    visibility = ["//visibility:public"],
+    deps = [
+        "@local_config_cuda//cuda:cublas",
+        "@local_config_cuda//cuda:cuda_driver",
+        "@local_config_cuda//cuda:cuda_headers",
+        "@local_config_cuda//cuda:cudart",
+        "@local_config_cuda//cuda:cufft",
+        "@local_config_cuda//cuda:curand",
+    ],
+)
+
+cc_library(
+    name = "cupti",
+    deps = [
+        "@local_config_cuda//cuda:cupti_headers",
+        "@local_config_cuda//cuda:cupti_link",
+    ],
+)
+
+[
+    alias(
+        name = lib,
+        actual = "@local_config_cuda//cuda:{}".format(lib),
+        visibility = ["//visibility:public"],
+    )
+    for lib in [
+        "cublas",
+        "cufft",
+        "cusolver",
+        "cusparse",
+        "curand",
+        "nvrtc",
+        "cuda_driver",
+        "nvToolsExt",
+    ]
+]

third_party/tensorflow_cuda_bazel_build/README.md

@@ -0,0 +1,5 @@
+# Config for CUDA
+
+This is a checked-in copy of the auto-generated config for building CUDA code with bazel. The content of this folder was generated from https://github.com/tensorflow/tensorflow `./configure` execution and then edited manually to fit the pytorch needs.
+
+The LICENSE for the TensorFlow project is APACHE 2. The full LICENSE file could be found here https://github.com/tensorflow/tensorflow/blob/master/LICENSE.

third_party/tensorflow_cuda_bazel_build/WORKSPACE

@@ -0,0 +1 @@
+workspace(name = "local_config_cuda")