What is the biggest bottleneck in emulation today?

[arbuzixx]

Hi,
I think the biggest advantage of emulator vs hardware console is possibility of beautify graphics (mainly by resolution increase). Disadvantage - problems with smooth gameplay.
Over year ago I was wondering if is possible to get around this problems, at least for Gamecube. That time I though main problem is with emulating PowerPC CPU. I am interested in FPGA chips so I started working on gekko-flipper bus sniffing tool (to make co-operation of hardware and emulation).
Nowadays my project is on significant stage. Two FPGAs, FIFO to USB3.0 bridge, RAM chips. All 108 bus signals connected. I already have a few samples of data, for example full GC boot. But I started wondering again - is sense of make this? I decided to make simply test scenario. Episode from game Zelda Twilight Princess on Gamecube. After lakebed temple boss (save from here: https://gamefaqs.gamespot.com/gamecube/9...cess/saves). My emulation speed - 30% (look screenshoot). Later I started emulation without graphics (backend "NULL"). Speed ~135%. Over 4x more!
Soo... does this means that my project is useless? What is the biggest bottleneck in emulation today?

[benjewmin]

Hi,
I think the biggest advantage of emulator vs hardware console is possibility of beautify graphics (mainly by resolution increase). Disadvantage - problems with smooth gameplay.
Over year ago I was wondering if is possible to get around this problems, at least for Gamecube. That time I though main problem is with emulating PowerPC CPU. I am interested in FPGA chips so I started working on gekko-flipper bus sniffing tool (to make co-operation of hardware and emulation).
Nowadays my project is on significant stage. Two FPGAs, FIFO to USB3.0 bridge, RAM chips. All 108 bus signals connected. I already have a few samples of data, for example full GC boot. But I started wondering again - is sense of make this? I decided to make simply test scenario. Episode from game Zelda Twilight Princess on Gamecube. After lakebed temple boss (save from here: https://gamefaqs.gamespot.com/gamecube/9...cess/saves). My emulation speed - 30% (look screenshoot). Later I started emulation without graphics (backend "NULL"). Speed ~135%. Over 4x more!
Soo... does this means that my project is useless? What is the biggest bottleneck in emulation today?

I was JUST about to look into this! I've been inspired by HAX$ interest in moving to LCD. I've been educating myself on ISA lately, been in SI and PCB design for a few years. I'd love to understand your quest.
What is the graphics backend?... Please link me to some reading material.
From what I understand, you just ARE NOT going to get what you need speedwise out of an FPGA... There are gross physical/chemical/EM optimizations in mass-produced silicon; we take them for granted. That does not mean your work is useless. I think on unlike architecture, the issues are the translation layers (idk if right term). The compiled code simply MUST run through an emulator on x86... Unless ? there was a way to achieve partial native processing (even if only ISA extensions ran through an FPGA... while main functionality was handled natively... I'm major spit balling here, I don't know a lot...)
I'm looking into shared ISA between PowerPC and extensions to modern processors (ARMv8,RISCV,etc.) I have a lot of ground to make up, it'd be awesome to see if you or anyone else is interested in modernizing and eternalizing the NintendoGC. (hopefully WAY brighter than me).

[benjewmin]

As of the end of summer (August 2019); IBM apparently released the PowerPC architecture as open source...
This may be perfectly timed to integrate with the work you have done, thus-far...

The community would never go for running Melee as emulation... but what if we could make it native... is anyone out there? lol

[nsm0220]

i think its these architectures like they are using in the Sega Saturn for example