Comments (2)
I can't reproduce this bug anymore and there are no more monitor tests failing in the test suite. Closing.
from brian2cuda.
Looks like this bug is still around, but only becomes apparent for large enough recordings. Because of warp-level SIMT, atomicAdd
during thresholding will lead to neuron IDs being sorted if their indices are withing a single warp (group of 32 neurons). Since most tests use small neurongroups (typically less than 32 neurons), the data ends up being sorted. But the order of warps in the eventspace is random, hence for recordings of > 32 neurons, the eventspace is not sorted and the spikemonitor just copies from the eventspace is not sorted either.
Solution: Perform on copy operation per time step on the recorded data (either via thrust::sort or via custom sorting code in the spikemonitor kernel). Or find a way to sort the data in the end of the simulation only (maybe even in parallel on the GPU by storing the start indices for each time step during recording and then perform parallel sorting per time step in one kernel call at the end?).
For now this is a known issue and remains as such.
from brian2cuda.
Related Issues (20)
- Check if storing the size of synapse groups is necessary? HOT 1
- Needs patch to run with Brian 2.4.2 HOT 2
- Optimize `StateMonitor`
- Impelement brian2cuda preference file support
- Copy all eventspace counters to host efficiently at each time step
- Investigate and document performance effects when working with `Subgroup`s HOT 1
- Consider partitioning eventspaces when using `Subgroup`s HOT 4
- Optimize `PopulationRateMonitor`
- Fix `SpikeMonitor` for `Subgroup`s HOT 1
- Optimize our `SpikeMonitor` for `Subgroups`
- Refactor test suite scripts
- Fix `ReferencError` in spatialneuron tests HOT 3
- Fix memory leak when having multiple `run` calls
- Spikes are lost when changing delays between `run` calls HOT 3
- Recent Brian2 update PR broke benchmark scripts HOT 1
- Brian2Cuda Uninstalls Brian2 2.5.1 and Installs 2.4.2 Which Won't Import HOT 3
- Brian2Hears and Brian2CUDA HOT 3
- Optimise brian2CUDA HOT 2
- Bypass nvidia-smi when using a Jetson platform HOT 6
- Why the volatile GPU utilization is so low? I run the same code respectively on the CPU(cpp_standalone) and GPU(brian2cuda), and i expect that it can run quickly by using brian2ccuda. However, it runs much slower... Can someone explain this? HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from brian2cuda.