DraStic

With the latest build onscreen controls were implemented and Gradius V is actually quite playable on Nexus 9: https://youtu.be/2q5jpQZE-1w

Reminds me of Gradius 3 on SNES I bet playing at this speed makes the game a whole lot easier. That is, if you have the patience. I'd probably max out speedup, something I'd never come close to in a Gradius game at normal speed.

Interesting how well close the loads are on the two CPU cores. And that neither ever seems to be pegging at 100%; I'm guessing that it's actually GPU limited instead.

Exophase, Yea I was surprised actually. Especially after measuring thermal throttling on Exynos 7 while running Dolphin (this probably happens because of the CPU core is waiting on the video thread, which will often happen due to video driver stalling) 4 cores displayed are the A53s
https://youtu.be/Zq4NPA_zxWk

Interesting. The CPU load being consistently below 100% does seem to rule out the possibility of throttling here. Again I'm surprised that, in gameplay, the CPU loads is so well distributed between the cores. AFAIK Dolphin will run a CPU thread, a GPU thread, and a DSP thread. I wonder how much of what we see in CPU load is GPU driver overhead.

Have you done a CPU load monitor video with DraStic on SGS6 and/or Nexus 9? Very curious to see.

In before ppl start complaining about ps2 dvd iso's taking up 1-4+ GB. I mean it is not like anyone complains about nintendo DS game sizes taking up too much space.....

Exophase, I'm away now so can't really make a video, but I tried DraStic yesterday and CPU load is pretty much on par with Play! on Nexus 9. I'd say it's identical and it mostly uses 1.5Ghz frequency (that's with enabled Hi-res, no frameskip and Multi-Threaded rendering on). The situation is much different on S4 Pro, there you have frequencies jumping on all 4 cores, some cores runs @486mhz and others @1.2Ghz, I never seen it got maximized to 1.5Ghz. I wonder if it has something to do with specifics of Denver's architecture which is IOE vs OoOE on S4 Pro

This is a very interesting project. I'm thinking of contributing, however I'll probably have to stick to working on the Android UI, as I know a lot more Java than I do C++.

There's much to do, especially in the UI so you're welcomed

zxcvbad wrote:Exophase, I'm away now so can't really make a video, but I tried DraStic yesterday and CPU load is pretty much on par with Play! on Nexus 9. I'd say it's identical and it mostly uses 1.5Ghz frequency (that's with enabled Hi-res, no frameskip and Multi-Threaded rendering on). The situation is much different on S4 Pro, there you have frequencies jumping on all 4 cores, some cores runs @486mhz and others @1.2Ghz, I never seen it got maximized to 1.5Ghz. I wonder if it has something to do with specifics of Denver's architecture which is IOE vs OoOE on S4 Pro

Or the Transmeta-style code morphing scheme for that matter.

zxcvbad wrote:Exophase, I'm away now so can't really make a video, but I tried DraStic yesterday and CPU load is pretty much on par with Play! on Nexus 9. I'd say it's identical and it mostly uses 1.5Ghz frequency (that's with enabled Hi-res, no frameskip and Multi-Threaded rendering on). The situation is much different on S4 Pro, there you have frequencies jumping on all 4 cores, some cores runs @486mhz and others @1.2Ghz, I never seen it got maximized to 1.5Ghz. I wonder if it has something to do with specifics of Denver's architecture which is IOE vs OoOE on S4 Pro

The difference here is that on Denver the CPU cores most likely physically can't be clocked separately. This is true on most SoCs. It's not true on S4 Pro.

In your Play! video it looks like the core clock is between 2.1 to 2.3GHz most of the time, and of course it's nowhere close to full speed. It's actually a good thing if the CPU load is pretty high if there's a moderate constant workload going on, that means that it's appropriately lowering the clock speed. Generally it's better for power consumption to have a lower clock speed with less time spent idle than the reverse.