Hey folks, I just recently started making myself familiar with gpu work via Niagara when I stumbled upon an error I wasn’t able to find a lot of information about.
LogNiagara: Error: Particle read DI reading self 'NiagaraSystem.Emitter' on stage 'StageName' is unsafe, please disable partial writes on the stage.
LogNiagara: Warning: [AssetLog] NiagaraSystem.uasset: Simulation stage 'StageName' is incompatible with particle state iteration due to killing particles or disabling particle updates. -
What I found out by analyzing the logs, looking a bit into source code and testing various combinations within the niagara system is that there seems to be a clash between subsetting particles in a simulation stage with “Particle Iteration State Enabled” and using a Particle Attribute Reader in the same emitter (does not have to be the same simulation stage tho).
I assume it has to do with the parallel processing nature of the gpu and reading/writing to the same buffer however I was not able to get a satisfying explanation yet of why this error is thrown.
Would love to get some insight from gpu experts or people who know how this particle subsetting works from a technical POV and why this process might cause errors down the road. Maybe there are ways to circumvent this error, maybe epic is working on a solution already but it could also be that this is just how gpus work and nothing we/epic could do about it.
Thanks, if you need more input for how you could reproduce this error let me know 
