I often precalculate the GI with a Incremental add to Current Map prepass. During the times I have either long precalculation times, lots of frames, or am impatient and don't want to wait for one slave to complete the prepass for the entire sequence, I will often split up the job among a couple of slaves, then merge the irradiance maps together using the irradiance map viewer. Pretty standard, I believe.
But, in these cases, I find myself splitting up frame ranges, assigning them to particular servers, and saving out particular irradence map ranges for each slave. Then opening them up and merging them all. Obviously, the more you subdivide a job, theoretically the faster it will finish, but the more overhead you have in preparing the job. So I tend to only split it in half or in thirds, just because I don't care to manage everything. What I am wondering is can't vray / backburner do all this overhead for you? Much the way max automatically manages "Split Scan Lines" in a net render job and assembles the final solution...?
But, in these cases, I find myself splitting up frame ranges, assigning them to particular servers, and saving out particular irradence map ranges for each slave. Then opening them up and merging them all. Obviously, the more you subdivide a job, theoretically the faster it will finish, but the more overhead you have in preparing the job. So I tend to only split it in half or in thirds, just because I don't care to manage everything. What I am wondering is can't vray / backburner do all this overhead for you? Much the way max automatically manages "Split Scan Lines" in a net render job and assembles the final solution...?
Comment