Age | Commit message (Collapse) | Author |
|
Signed-off-by: Jim Flynn <jim.flynn@arm.com>
|
|
* Adding BackendProfilingContext to the MockBackend
* Made IBackendProfilingContext pure Virtual
* Added UnitTest using MockBackend for testing Backend Counter Registration
* Moved Registry of backend counters from Initialize() to AddBackendProfilingContext()
* Added m_MaxGlobalCounterId to ProfilingService
* Removed automatic registration of MockBack in BackendRegistry()
Signed-off-by: David Monahan <david.monahan@arm.com>
Change-Id: Ie1c6c31e56d1ac7079d6116ecad041961014aedc
|
|
!android-nn-driver:2572
Change-Id: I8fe52ceb09987b3d05c539409510f535165455cc
Signed-off-by: Derek Lamberti <derek.lamberti@arm.com>
|
|
The current algorithm in SubgraphViewSelector has a bug that can lead to
it producing subgraphs which have a dependency cycle (see the newly
added test case 'ValidMerge' for a repro). It also fails to merge
subgraphs in some cases where it could, which leads to smaller subgraphs.
In the case of FSRCNN, the NPU cannot support these smaller subgraphs and
so this is blocking us from supporting that network.
This commit changes the algorithm to fix the dependency bug and
also make it so that subgraphs are merged in the cases that were missed
before. It also adds some unit tests to cover cases that were problematic
before, and to extend coverage for the new algorithm.
The new algorithm has two downsides compared to the previous one:
1. Disjoint subgraphs are not merged. This can never lead to a failed
compilation by the NPU and so I believe this is less of an issue than
the previous algorithm's "missed merges". This could however lead to a
runtime performance loss in some cases as the NPU will be unable
to parallelise as many operations. There are some unit tests that cover
this which I have disabled.
2. The performance is worse. I have spent some time analysing this and
for a graph with ~1000 layers the new algorithm takes 20ms vs. the
old algorithm's 4ms (on my desktop PC). I believe the performance is
still within acceptable limits. I also compared inception V3 (which was
the network which caused performance issues with the original version of
the splitting algorithm) and this new algorithm has not regressed there
(200-300us in both cases).
Change-Id: I1dd64a779f272723621e04d203b5a2752a6af2ef
Signed-off-by: Robert Hughes <robert.hughes@arm.com>
|
|
* Moved to BackendRegistry.hpp include/armnn
* Updated makefiles and sources accordingly
Signed-off-by: Matteo Martincigh <matteo.martincigh@arm.com>
Change-Id: I4d83abb581d523218a880c879fcf30c9611f7fd7
|
|
* Added OptimizeSubgraphViewTests file covering a number of
use cases for the Optimization API
* Fixed a bug in the sub-graph selector algorithm that skipped the
first layer in a sub-graph if it wasn't an input layer
* Changed the graph splitting logic to make use of maps instead of
unordered_maps to keep the split sub-graphs in consistent order
between executions
* Added more common unit test utils
* Minor fixes to comply to the include file conventions
Change-Id: Iad464eaedd004109e5ef41aa487cea3ad86177d3
Signed-off-by: Matteo Martincigh <matteo.martincigh@arm.com>
|