Reach sampler seed limit

[zichenwang01]

Hi,
I was calling sampler.seed(seed, wavefront_size) and due to a too large wavefront_size, ran into the following error

File "/home/zw336/IR/bp/integrators/base.py", line 24, in prepare_sampler
    sampler.seed(seed, wavefront_size)
TypeError: seed(): incompatible function arguments. The following argument types are supported:
    1. (self: mitsuba.cuda_ad_rgb.Sampler, seed: int, wavefront_size: int = 4294967295) -> None

Invoked with: IndependentSampler[
  base_seed = 0
  sample_count = 16384
  samples_per_wavefront = 16384
  wavefront_size = 0
], 0, 4294967296

I was rendering a 512*512 image with spp=2**14. Is there any way that I could make spp larger? Thanks!

[njroussel]

Hi @zichenwang01

Unfortunately, no.

OptiX allows launches up to 2**30 samples, and Dr.Jit can do some automatic splitting of larger kernels to accommodate up to 2**32 samples. Anything above that should be handled in multiple passes. For example, Mitsuba will render in multiple passes by default if the kernel is too wide: https://github.com/mitsuba-renderer/mitsuba3/blob/master/src/render/integrator.cpp#L247-L265