aeva

aeva, 2 hours ago

@inlovewithpda what a gorgeous laptop :O

aeva, 2 hours ago

@inlovewithpda ^_^

aeva, 2 hours ago

@gkrnours night in the woods

aeva, 2 hours ago

@gkrnours it's a fantastic game imo

aeva, 1 hour ago

I'm still cracking up about this.

My laptop has one of those newer Intel CPUs with a mix of [P]erformance cores (fast, and has hyperthreading) and [E]conolodge cores (low power use, no hyperthreading) to a total of 20 logical cores. The idea is the CPU always scales down to meet the current workload perfectly. The scheduler picks the cores most suited for the current tasks, and then parks the rest of them to save power. So what happens when the current workload is "everything you've got?"

aeva, 1 hour ago

Shrill, vaguely worrying jet engine sounds is what happens.

aeva, 1 hour ago

Conceptually there's nothing wrong with this wonderful asymmetrical CPU. It seems to do a fine job at scaling up to full utilization. And the theory makes sense for laptops if a normal application flow is largely asynchronous and largely idle (eg, short processing job happens in reaction to the user clicking on something).

it just kinda fails to take in account that some users want to play games, and small laptops have small irritating fans...

aeva, 1 hour ago

Now, as it happens I've built a rendering architecture that's able to scale perfectly to the available CPU resources, which creates the interesting problem of how to achieve and maintain the maximum computational throughput that doesn't also melt dainty aluminum laptops.

aeva, 35 minutes ago

@oblomov "the solution is obviously to monitor the cooling fan RPMs" ok so your post went in a different direction, but listening to the fans is how I judge the active workload for some of my computers

aeva, 34 minutes ago

@brouhaha no! I paid a lot of money for this laptop.

aeva, 25 minutes ago

@Setzer22 so I did see a significant perf gain from splitting up the ray tracing work to run asynchronously to the main thread in wide thread pool, but keep in mind that this rendering system is designed to be aggressively parallelized, so this was the expected outcome.

the grain size I picked for the parallel experiment which gets a similarly responsive update time is much finer

aeva, 21 minutes ago

@Setzer22 for a direct comparison of C# and C++, an idiomatic C# program will always come in behind in a direct comparison to an equivalent idiomatic C++ program, because C# has more runtime overhead.

However, C# has a lot of nice high level features for multithreading that elegantly fit a lot of different common concurrency paradigms, so I'm inclined to say that concurrent programming is much easier in C#.

aeva, 17 minutes ago

@Setzer22 in my opinion, the biggest pain point for C++ concurrency programming is that deleting resources that may be shared by multiple threads is currently an unsolved problem. C# simply does not have that problem, and I think that's great.

aeva, 2 minutes ago

ok, I think maybe the better conceptual framing of this problem is how best to throttle the application if the user is on a laptop and they turned off power saving mode, because otherwise it behaves fine without the extra sleeps etc.