i-robot
8e496e44f0
!22320 support heterogeneous for pynative mode
Merge pull request !22320 from chujinjin/support_heterogeneous_for_pynative
4 years ago
chujinjin
d20c9d93d8
support pynative heterogeneous
4 years ago
limingqi107
312b26080b
unified runtime fix execution timeout and no data source actor of control flow
4 years ago
i-robot
db5ef1136f
!26981 Control flow support function type of formal parameter.
Merge pull request !26981 from gaoyong10/runtime_second12
4 years ago
gaoyong10
080aad5e1d
control flow support function type of formal parameter.
4 years ago
xulei
78b37afc30
fix mindspore mac compile
fix debugger, mindarmour, audio, gnn, grpc, ps/fl.
4 years ago
i-robot
8c5957c387
!26831 fix output tensor num bug
Merge pull request !26831 from caifubi/master-pynative-lazy-build-bug
4 years ago
caifubi
1f25ce5d98
fix output tensor num bug
4 years ago
i-robot
1b0a82fc30
!26752 control flow support call to call
Merge pull request !26752 from gaoyong10/runtime_second12
4 years ago
gaoyong10
a7611d0063
control flow support call to call
4 years ago
i-robot
890a5dbae1
!26734 fix the bug of single thread execution stack overflow
Merge pull request !26734 from limingqi107/bug_fix
4 years ago
i-robot
87b02a8bf4
!26712 unified runtime support the empty kernel graph in control flow
Merge pull request !26712 from limingqi107/new_actor_runtime
4 years ago
limingqi107
79765ceb7f
fix the bug of single thread execution stack overflow
4 years ago
He Wei
41dcac9c49
Replace std::unordered_map/set with robin-hood-hashing
Robin-hood-hashing (https://github.com/martinus/robin-hood-hashing )
is considered faster then std::unordered_map/set,
so we use it to improve mindspore performance.
1. robin_hood head file in `third_party/robin_hood/include`;
2. In `utils/hash_map.h` and `utils/hash_set.h`, we define:
- mindspore::HashMap as an alias of robin_hood::unordered_map;
- mindspore::HashSet as an alias of robin_hood::unordered_set;
3. Replace:
- `#include <unordered_map>` --> `#include "utils/hash_map.h"`;
- `#include <unordered_set>` --> `#include "utils/hash_set.h"`;
- `std::unordered_map` --> `mindspore::HashMap`;
- `std::unordered_set` --> `mindspore::HashSet`;
- `map.insert(std::pair(key, value))` --> `map.emplace(key, value)`;
- `[] (const std::pair<K, V> &p) {..} ` --> `[] (const auto &p) {..} `;
4. Fix issues found by switch to robin_hood:
- AnfNodeConfig hash and equal;
- Fix a bug in `Slice::operator==()`;
- Fix a bug in `CNode::HasPrimalAttr()`;
- Fix map.erase() usage bugs: `map.erase(iter++)` --> `iter = map.erase(iter)`;
- Fix some iterator invalidated problem;
5. Some std::unordered_map/set can not replace by robin_hood:
- As parameter of functions that exposed to python by pybind11;
- Use bad hash that cause robin_hood::map over_flow, such as AbstractBasePtrListHasher;
6. Update cpp unit tests;
7. Add build option '-F' to enable robin_hood, default on.
4 years ago
limingqi107
f016fe0775
unified runtime support the empty kernel graph in control flow
4 years ago
i-robot
1a7a04e4c9
!25132 [PyNative][MindRT][GPU] Op Lazy Build
Merge pull request !25132 from caifubi/master-pynative-mindrt-gpu-async-build
4 years ago
limingqi107
e17a533369
fix the bug of graph and pynative shared the weight
4 years ago
caifubi
38352c1ba8
PyNative MindRT Op Lazy Build
4 years ago
limingqi107
a7f761c215
unified runtime support the control flow of subgraph sink
4 years ago
i-robot
3269c9b881
!26335 Support MindSpore on MacOS
Merge pull request !26335 from xulei/ms_mac_compile_br
4 years ago
i-robot
8e41c6813f
!26401 fix the weight output of exit actor.
Merge pull request !26401 from gaoyong10/runtime_second8
4 years ago
i-robot
21ffa1fb7b
!25091 Partial support for multi root graph in online debugger
Merge pull request !25091 from parastooashtari/online_multi_root_graph
4 years ago
xiefangqi
24ac2382ff
mac compile
4 years ago
gaoyong10
c64f62c309
fix the weight output of exit actor.
4 years ago
i-robot
e3988c8e32
!26433 Revert Modification of Creating Output Devcie Address
Merge pull request !26433 from hwjiaorui/mindrt-bug-fix
4 years ago
hwjiaorui
fee98c5f5d
revert modification output device size
4 years ago
i-robot
9e6009dee2
!26355 modify kernel thread num equal to 23
Merge pull request !26355 from 范吉斌/fix_face
4 years ago
i-robot
f010daa31d
!26360 unified runtime optimize control flow
Merge pull request !26360 from limingqi107/new_actor_runtime
4 years ago
i-robot
016f09ed21
!26088 Ascend LaunchKernel for MindRT
Merge pull request !26088 from zhaosida/kernel_by_kernel_mindrt
4 years ago
Parastoo Ashtari
7f682ba2f6
partial support for multi root graph in online debugger
4 years ago
zhaosida
c535efbfae
kernel by kernel mindrt
4 years ago
limingqi107
d8b1b69eab
unified runtime optimize control flow
4 years ago
fan-jibin
0c941ac1eb
fix cpu thread num
4 years ago
gaoyong10
add5551ecf
fix code for testcase
4 years ago
i-robot
390b3c2efa
!26212 Control flow support untail call.
Merge pull request !26212 from gaoyong10/runtime_second8
4 years ago
gaoyong10
f7e3eef63d
control flow support untail call
4 years ago
i-robot
0cf592db7a
!26040 unified runtime support single thread execution and optimize execution performance by cost mode
Merge pull request !26040 from limingqi107/new_actor_runtime
4 years ago
limingqi107
52a182448e
unified runtime support single thread execution and optimize execution performance by cost mode
4 years ago
i-robot
30c0bc5a96
!26169 Add EnableAscendMIndRT Environment Variable
Merge pull request !26169 from hwjiaorui/enable_ascend_mindrt
4 years ago
hwjiaorui
94cc26bf6b
add enable ascend mindrt environment
4 years ago
gaoyong10
b14df4644f
Fix output actor for control flow.
4 years ago
i-robot
1bfedfcb7f
!26007 Ascend Device Context for MindRT
Merge pull request !26007 from hwjiaorui/ascend_context
4 years ago
i-robot
4eef5e5c13
!26045 Delete useless interface in control node parser.
Merge pull request !26045 from gaoyong10/runtime_second8
4 years ago
hwjiaorui
d6b2a34a69
MindRT Ascend Devcie Context
4 years ago
gaoyong10
b5a9588d10
delete useless interface in control node parser
4 years ago
dayschan
cbb84ff580
Move IsRealKernel and IsRealCNodeKernel from AnfAlgo to AnfUtils
the function IsOneOfPrimitive and IsOneOfPrimitiveCNode is useful,
we can move them into anf.cc
4 years ago
gaoyong10
f6bbe2fe03
Add control node scheduler implementation.
4 years ago
i-robot
90b03dbfbb
!25957 Add interface implementation
Merge pull request !25957 from gaoyong10/runtime_second7
4 years ago
i-robot
1509d3f848
!25952 unified runtime support the auto monad in the subgraphs connecting scene
Merge pull request !25952 from limingqi107/new_actor_runtime
4 years ago
gaoyong10
ba28e554d9
Add interface implementation
4 years ago