Quote:I didn't mean that exactly. I just said that the PPU recompiler should have top priority, but of course it would be great to improve the interpreter if any optimizations / fixes can be done. However, *redesigning* it like it's proposed here in order to make it faster is going way to far IMO. It's easy: the interpreter should work *perfectly* and be easy to understand and debug, no matter how slow it is (as long as the slowness has some decent explanation).
(Again: this is nothing but my point of view)
No, yeah, I know. I was just mentioning the part where a cached interpreter was mentioned and when you said you believed more focus on the dynarec would be more important. In the meantime I think the interpreter could benefit a little more because apparently the dynarec will take some time.
Yes, exactly what hlide said. That is a pretty good method of cached interpretation, but I guess it might be pointless for now since there seem to be unending focus on the dynamic recompiler and disgaea 3.
(04-29-2014, 06:40 PM)androidlover Wrote: [ -> ]Yes, exactly what hlide said. That is a pretty good method of cached interpretation, but I guess it might be pointless for now since there seem to be unending focus on the dynamic recompiler and disgaea 3.
You could always help out.
(04-30-2014, 12:25 AM)derpf Wrote: [ -> ] (04-29-2014, 06:40 PM)androidlover Wrote: [ -> ]Yes, exactly what hlide said. That is a pretty good method of cached interpretation, but I guess it might be pointless for now since there seem to be unending focus on the dynamic recompiler and disgaea 3.
You could always help out.
Heh ... my help isn't valued here currently. Maybe when the emulator goes later on in development I'll consider.
(04-30-2014, 05:48 PM)androidlover Wrote: [ -> ] (04-30-2014, 12:25 AM)derpf Wrote: [ -> ] (04-29-2014, 06:40 PM)androidlover Wrote: [ -> ]Yes, exactly what hlide said. That is a pretty good method of cached interpretation, but I guess it might be pointless for now since there seem to be unending focus on the dynamic recompiler and disgaea 3.
You could always help out.
Heh ... my help isn't valued here currently. Maybe when the emulator goes later on in development I'll consider.
Please note that all this dynamic recompiler talk is merely from an experimental point of view. RPCS3 is above everything an experiment, so you're always welcome to provide better emulation approaches.
I agree with hlide on many aspects and these ideas are all being studied. Any help is always valued, even if it takes a whole different route than the one currently taken.
I am not sure what you mean by "experimental point of view." Writing a dynamic recompiler needs not be an "experiment" as there's plenty of knowledge on how to do it. RPCS3 isn't the first Power Architecture recompiler/interpreter made, so it's not merely a "guessing" approach.
Unless you mean "experimental" as in "We're doing stuff, testing it, seeing if it works right, etc." You certainly need some experimenting around, since emulating a PS3 is no walk in the part. In that case I don't think "experimental" completely fits the definition of the approach of a PS3 emulator though.
Pulled from Google, "experimental" shortens to:
"based on untested ideas or techniques."
Since there's no fully functioning PS3 emulator like PCSX2, ePSXe, Dolphin, Project64, etc., it's undoubtedly experimenting with a multitude of ways emulating the system can be done. I believe it that the dynamic recompiler is not purely an experiment, even though it can be experimented with, because it's been done before(dynamic recompiling). However, the entirety of PS3 emulation has to be partly experimental. Maybe Hykem meant experimental as in "non-commercial?"
Surely anything unfinished and a WIP can be coined experimental.
(05-01-2014, 07:58 PM)androidlover Wrote: [ -> ]I am not sure what you mean by "experimental point of view." Writing a dynamic recompiler needs not be an "experiment" as there's plenty of knowledge on how to do it. RPCS3 isn't the first Power Architecture recompiler/interpreter made, so it's not merely a "guessing" approach.
Unless you mean "experimental" as in "We're doing stuff, testing it, seeing if it works right, etc." You certainly need some experimenting around, since emulating a PS3 is no walk in the part. In that case I don't think "experimental" completely fits the definition of the approach of a PS3 emulator though.
Yes, I meant the latest. Maybe "experimental" wasn't the best word to describe it since it has such loose meaning.
Anyway,
mushroom made a pretty accurate description of what experimental may mean for RPCS3.
Certainly, writing a dynarec is not an experiment, it has been done countless times before. What I meant is that solutions like dynamic recompilation are being worked on separate forks by a subset of contributors.
Nothing will be blindly merged if it doesn't produce a significant advantage to the emulation process (speed and accuracy wise).
In fact, we're already hitting a barrier related to our current interpreter and a redesign is starting to become a necessity:
https://github.com/DHrpcs3/rpcs3/issues/210 (unknownbrackets pointed out how much the interpreter saturates the CPU).
what about static recompilation? On pcsx2 forums they talk about it being done that way or mixing static with dynamic. Not sure how exactly that works, but could it be easier than dynamic alone?
Pretty sure dynamic itself would be better. Static will be harder to get right, at least from what I heard from some seasoned emu developers. Mixing will be more a bother then anything most likely.