8000 Update on "[Array API] Add linalg.vecdot" · pytorch/pytorch@d97e06f · GitHub
[go: up one dir, main page]
Skip to content

Commit d97e06f

Browse files
lezcanolezcano
committed
Update on "[Array API] Add linalg.vecdot"
 This PR adds the function `linalg.vecdot` specified by the [Array API](https://data-apis.org/array-api/latest/API_specification/linear_algebra_functions.html#function-vecdot) For the complex case, it chooses to implement \sum x_i y_i. See the discussion in data-apis/array-api#356 Edit. When it comes to testing, this function is not quite a binopt, nor a reduction opt. As such, we're this close to be able to get the extra testing, but we don't quite make it. Now, it's such a simple op that I think we'll make it without this. Resolves #18027. cc mruberry rgommers pmeier asmeurer leofang AnirudhDagar asi1024 emcastillo kmaehashi [ghstack-poisoned]
2 parents 82bfbdd + 30748e0 commit d97e06f

File tree

2,431 files changed

+169929
-77637
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

2,431 files changed

+169929
-77637
lines changed

‎.bazelrc

Lines changed: 94 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -13,15 +13,100 @@ build:no-tty --curses no
1313
build:no-tty --progress_report_interval 10
1414
build:no-tty --show_progress_rate_limit 10
1515

16-
# Configuration to build with GPU support
17-
build:gpu --define=cuda=true
16+
# Build with GPU support by default.
17+
build --define=cuda=true
18+
# rules_cuda configuration
19+
build --@rules_cuda//cuda:enable_cuda
20+
build --@rules_cuda//cuda:cuda_targets=sm_52
21+
build --@rules_cuda//cuda:compiler=nvcc
22+
build --repo_env=CUDA_PATH=/usr/local/cuda
23+
24+
# Configuration to build without GPU support
25+
build:cpu-only --define=cuda=false
1826
# define a separate build folder for faster switching between configs
19-
build:gpu --platform_suffix=-gpu
27+
build:cpu-only --platform_suffix=-cpu-only
2028
# See the note on the config-less build for details about why we are
21-
# doing this. We must also do it for the "-gpu" platform suffix.
22-
build --copt=-isystem --copt=bazel-out/k8-fastbuild-gpu/bin
29+
# doing this. We must also do it for the "-cpu-only" platform suffix.
30+
build --copt=-isystem --copt=bazel-out/k8-fastbuild-cpu-only/bin
2331
# rules_cuda configuration
24-
build:gpu --@rules_cuda//cuda:enable_cuda
25-
build:gpu --@rules_cuda//cuda:cuda_targets=sm_52
26-
build:gpu --@rules_cuda//cuda:compiler=nvcc
27-
build:gpu --repo_env=CUDA_PATH=/usr/local/cuda
32+
build:cpu-only --@rules_cuda//cuda:enable_cuda=False
33+
34+
# Definition of --config=shell
35+
# interactive shell immediately before execution
36+
build:shell --run_under="//tools/bazel_tools:shellwrap"
37+
38+
# Disable all warnings for external repositories. We don't care about
39+
# their warnings.
40+
build --per_file_copt=^external/@-w
41+
42+
# Set additional warnings to error level.
43+
#
44+
# Implementation notes:
45+
# * we use file extensions to determine if we are using the C++
46+
# compiler or the cuda compiler
47+
# * we use ^// at the start of the regex to only permit matching
48+
# PyTorch files. This excludes external repos.
49+
#
50+
# Note that because this is logically a command-line flag, it is
51+
# considered the word on what warnings are enabled. This has the
52+
# unfortunate consequence of preventing us from disabling an error at
53+
# the target level because those flags will come before these flags in
54+
# the action invocation. Instead we provide per-file exceptions after
55+
# this.
56+
#
57+
# On the bright side, this means we don't have to more broadly apply
58+
# the exceptions to an entire target.
59+
#
60+
# Looking for CUDA flags? We have a cu_library macro that we can edit
61+
# directly. Look in //tools/rules:cu.bzl for details. Editing the
62+
# macro over this has the following advantages:
63+
# * making changes does not require discarding the Bazel analysis
64+
# cache
65+
# * it allows for selective overrides on individual targets since the
66+
# macro-level opts will come earlier than target level overrides
67+
68+
build --per_file_copt='^//.*\.(cpp|cc)$'@-Werror=all
69+
# The following warnings come from -Wall. We downgrade them from error
70+
# to warnings here.
71+
#
72+
# sign-compare has a tremendous amount of violations in the
73+
# codebase. It will be a lot of work to fix them, just disable it for
74+
# now.
75+
build --per_file_copt='^//.*\.(cpp|cc)$'@-Wno-sign-compare
76+
# We intentionally use #pragma unroll, which is compiler specific.
77+
build --per_file_copt='^//.*\.(cpp|cc)$'@-Wno-error=unknown-pragmas
78+
79+
build --per_file_copt='^//.*\.(cpp|cc)$'@-Werror=extra
80+
# The following warnings come from -Wextra. We downgrade them from error
81+
# to warnings here.
82+
#
83+
# unused-parameter-compare has a tremendous amount of violations in the
84+
# codebase. It will be a lot of work to fix them, just disable it for
85+
# now.
86+
build --per_file_copt='^//.*\.(cpp|cc)$'@-Wno-unused-parameter
87+
# missing-field-parameters has both a large number of violations in
88+
# the codebase, but it also is used pervasively in the Python C
89+
# API. There are a couple of catches though:
90+
# * we use multiple versions of the Python API and hence have
91+
# potentially multiple different versions of each relevant
92+
# struct. They may have different numbers of fields. It will be
93+
# unwieldy to support multiple versions in the same source file.
94+
# * Python itself for many of these structs recommends only
95+
# initializing a subset of the fields. We should respect the API
96+
# usage conventions of our dependencies.
97+
#
98+
# Hence, we just disable this warning altogether. We may want to clean
99+
# up some of the clear-cut cases that could be risky, but we still
100+
# likely want to have this disabled for the most part.
101+ B41A
build --per_file_copt='^//.*\.(cpp|cc)$'@-Wno-missing-field-initializers
102+
103+
build --per_file_copt='//:aten/src/ATen/RegisterCompositeExplicitAutograd\.cpp$'@-Wno-error=unused-function
104+
build --per_file_copt='//:aten/src/ATen/RegisterCompositeImplicitAutograd\.cpp$'@-Wno-error=unused-function
105+
build --per_file_copt='//:aten/src/ATen/RegisterMkldnnCPU\.cpp$'@-Wno-error=unused-function
106+
build --per_file_copt='//:aten/src/ATen/RegisterNestedTensorCPU\.cpp$'@-Wno-error=unused-function
107+
build --per_file_copt='//:aten/src/ATen/RegisterQuantizedCPU\.cpp$'@-Wno-error=unused-function
108+
build --per_file_copt='//:aten/src/ATen/RegisterSparseCPU\.cpp$'@-Wno-error=unused-function
109+
build --per_file_copt='//:aten/src/ATen/RegisterSparseCsrCPU\.cpp$'@-Wno-error=unused-function
110+
build --per_file_copt='//:aten/src/ATen/RegisterZeroTensor\.cpp$'@-Wno-error=unused-function
111+
build --per_file_copt='//:torch/csrc/lazy/generated/RegisterAutogradLazy\.cpp$'@-Wno-error=unused-function
112+
build --per_file_copt='//:torch/csrc/lazy/generated/RegisterLazy\.cpp$'@-Wno-error=unused-function

‎.buckconfig.oss

Lines changed: 11 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,15 +1,25 @@
1+
[pt]
2+
is_oss=1
3+
14
[buildfile]
2-
name = BUILD.buck
5+
name = BUCK.oss
6+
includes = //tools/build_defs/select.bzl
37

48
[repositories]
59
bazel_skylib = third_party/bazel-skylib/
10+
ovr_config = .
611

712
[download]
813
in_build = true
914

1015
[cxx]
1116
cxxflags = -std=c++17
1217
should_remap_host_platform = true
18+
cpp = /usr/bin/clang
19+
cc = /usr/bin/clang
20+
cxx = /usr/bin/clang++
21+
cxxpp = /usr/bin/clang++
22+
ld = /usr/bin/clang++
1323

1424
[project]
1525
default_flavors_mode=all

‎.circleci/cimodel/data/pytorch_build_data.py

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -75,6 +75,7 @@ def child_constructor(self):
7575
"vulkan": VulkanConfigNode,
7676
"parallel_tbb": ParallelTBBConfigNode,
7777
"crossref": CrossRefConfigNode,
78+
"dynamo": DynamoConfigNode,
7879
"parallel_native": ParallelNativeConfigNode,
7980
"onnx": ONNXConfigNode,
8081
"libtorch": LibTorchConfigNode,
@@ -179,6 +180,14 @@ def child_constructor(self):
179180
return ImportantConfigNode
180181

181182

183+
class DynamoConfigNode(TreeConfigNode):
184+
def init2(self, node_name):
185+
self.props["is_dynamo"] = node_name
186+
187+
def child_constructor(self):
188+
return ImportantConfigNode
189+
190+
182191
class ParallelNativeConfigNode(TreeConfigNode):
183192
def modify_label(self, label):
184193
return "PARALLELNATIVE=" + str(label)

‎.circleci/cimodel/data/pytorch_build_definitions.py

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -240,6 +240,7 @@ def instantiate_configs(only_slow_gradcheck):
240240
is_xla = fc.find_prop("is_xla") or False
241241
is_asan = fc.find_prop("is_asan") or False
242242
is_crossref = fc.find_prop("is_crossref") or False
243+
is_dynamo = fc.find_prop("is_dynamo") or False
243244
is_onnx = fc.find_prop("is_onnx") or False
244245
is_pure_torch = fc.find_prop("is_pure_torch") or False
245246
is_vulkan = fc.find_prop("is_vulkan") or False
@@ -286,6 +287,9 @@ def instantiate_configs(only_slow_gradcheck):
286287
if is_crossref:
287288
parms_list_ignored_for_docker_image.append("crossref")
288289

290+
if is_dynamo:
291+
parms_list_ignored_for_docker_image.append("dynamo")
292+
289293
if is_onnx:
290294
parms_list.append("onnx")
291295
python_version = fc.find_prop("pyver")

0 commit comments

Comments
 (0)
0