onnxruntime
Support multi-loop parallel sections, use multi-loop sections in GRU
#5602
Merged
Merged
Support multi-loop parallel sections, use multi-loop sections in GRU #5602
tlh20
merged
69 commits into
master
from
tiharr/dev-threading-2
Pause in spin loops, use static methods for thread pool API
3f3bada4
Fix should-parallelize test
f8c8612c
Use static-method API to thread pool
2ee43c4c
Fix spacing
3894888a
Remove warning in OpenMP builds
12371ec4
Fix WIN32 test case
d88b6b67
Fix ORT training thread pool use
b5f25937
Do not call fn(0,0) when 0 iterations expected
081c4fc5
Fix Windows test
5eaf73c4
Correct if-def for MSVC
2192b2d2
Merge remote-tracking branch 'origin/master' into tiharr/dev-threading-1
117e8400
Allow explicit use of thread pools with OpenMP builds
133f22bf
MSVC build
86d2ada6
MSVC build
a9920304
Multi-loop sections
52039e81
Multi-loop sections
b765a75a
Multi-loop sections
d11ca159
tlh20
changed the title
Fix OpenMP build
3df0fca6
Fix use of std::max
ec56a7f9
Use SpinPause() rather than _mm_pause intrinsic directly
b720e7cd
Add test cases for multi-loop sections
dfdd3606
merge
de0cf4fa
RAII API for parallel sections
7eb7c58b
Fix OpenMP build
b46bb7a1
Fix unused field warning
c8838ad8
Merge remote-tracking branch 'origin/master' into tiharr/dev-threading-2
52662c23
Make loop counter _num_shards const
cbe07c30
Merge remote-tracking branch 'origin/master' into tiharr/dev-threading-2
dad66f23
Pre re-factoring of per-section state
f796282f
Refactor thread local storage
236b42a9
Merge remote-tracking branch 'origin/master' into tiharr/dev-threading-2
907a1d46
Improve comments
8b9ed8ce
Use decltype, not typeof, for MSVC build
1a276fa0
Fix MSVC build
1d7d260e
Fix MSVC build
82e63942
Fix OS X build
86e783f6
Fix OS X build
19d243d9
Fix OS X build
d8d05327
unique_ptr for parallel sections
938bb38d
Fix non-OpenMP builds without thread pool
e154034c
Add test cases for threadpool=nullptr
e9546b09
Merge remote-tracking branch 'origin/master' into tiharr/dev-threading-2
ba65dcfb
Limit loop degree-of-parallelism to number of blocks available
5f6b4f32
Remove unused field
1e1dba6e
Fix microbenchmark build
72ae87c9
Merge remote-tracking branch 'origin/master' into tiharr/dev-threading-2
d5ee2ec0
Fix microbenchmarks
3a1c21f6
Refactor prior to single-loop special case
1940b3d7
Single-loop special case
6d12df72
Commit pre refactor
ff19e107
Refactor parallel section implementation
6726cc55
Remove white-space changes
1f320a5c
Remove C++14 lambda initializer capture
707c5a42
Fix windows build
0188e176
Remove unneeded include
97641d8a
Tidy code, extend comments
23b49018
Update comments
5fc273b2
Remove unneeded lambda
6fd16a76
tlh20
changed the title 
Merge remote-tracking branch 'origin/master' into tiharr/dev-threading-2
2298fc19
Avoid code analysis warnings
1c9fa9b9
Merge branch 'tiharr/dev-threading-2' of https://github.com/microsoft…
f04ac287
Tidy return types, remove unneeded ORT_ENFORCE
984f3f0a
Use relaxed store in LoopCounter
ebb71d1d
Avoid repeated per-thread fetches
19a30d10
Use assert for internal checks
cb57604e
Remove unused function
59056829
Use by-ref where possible (avoiding ptr-typed parameters, and passing…
6dd7497a
PR comments, explicit constructor, _X -> X_ field names
865b70e1
Merge remote-tracking branch 'origin/master' into tiharr/dev-threading-2
b2e39421
pranavsharma
approved these changes
on 2020-11-10
Login to write a write a comment.
Login via GitHub