This rewrite attempts to implement a fully correct ZBC (Zero Bandwith Clear) mechanism.
The zbc_mutex attempts to mitigate contention by assuring that only threads which hold the mutex can modify the table.
Notify drivers about the index size, I believe some drivers even need the notification.
Only add new entries if a entry was not previously available.
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2501
Reviewed-by: Shinmegumi <[email protected]>
Reviewed-by: MaranBr <[email protected]>
Reviewed-by: crueter <[email protected]>
Co-authored-by: SDK-Chan <[email protected]>
Co-committed-by: SDK-Chan <[email protected]>
This rewrite should improve performance with the buffer history by changing the complexity level to O(1).
Replace std::vector with std::array to ensure that elements are allocated on the stack rather than on the free store.
Avoid expensive resizing at runtime.
Adjust buffer states at the right locations.
Tightly pack the BufferHistoryInfo struct to ensure that it only occupies 28 bytes.
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/528
Co-authored-by: SDK-Chan <[email protected]>
Co-committed-by: SDK-Chan <[email protected]>
Improves object allocation per channel, only allowing max amount of 6 objects contexts per channel.
Previously objects were stored in a heap allocated vector which is sub-optimal for performance reasons.
The new implementation instead uses a stack based array with a O(1) approach.
This should boost performance in games which heavily rely on object context creation.
Co-authored-by: MaranBr <[email protected]>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/333
Reviewed-by: crueter <[email protected]>
Reviewed-by: CamilleLaVey <[email protected]>
Co-authored-by: SDK-Chan <[email protected]>
Co-committed-by: SDK-Chan <[email protected]>
This implementation is basically usable for up to 8GB of DRAM which you can set in the emulator.
It should ensure that the alias or map region for the virtual address space is bigger when requested.
8GB DRAM is the size of Nintendos DRAM sticks in the developers kit.
Going above 8GB DRAM while emulating a game is not recommended.
That is why this implementation.
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/260
Reviewed-by: Shinmegumi <[email protected]>
Co-authored-by: SDK-Chan <[email protected]>
Co-committed-by: SDK-Chan <[email protected]>
During building eden on FreeBSD, it always assumed that FreeBSD due to being part of the unix family, has gamemode support, too.
Since it doesn't the building process failed.
This commit aims to fix that by separating out Unix and Linux for CMake to ensure there are no conflicts.
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/117
Co-authored-by: SDK-Chan <[email protected]>
Co-committed-by: SDK-Chan <[email protected]>
I noticed that AllocGPFIFOEX was missing from eden.
It is the same as GPFIFOEX2, but is called separately, even on the real console.
I updated the struct to match closely the one seen on switchbrew, and changed the STUBS, because according to sources they seem to be complete.
My guess is that the STUBS were remained due to the unknowns back then.
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/115
Co-authored-by: SDK-Chan <[email protected]>
Co-committed-by: SDK-Chan <[email protected]>
FreeBSD doesn't support NVDEC, CUDA, and partially supports VAAPI (mostly for firefox).
Implementing VAAPI for other use cases would be a little bit complicated so, I chose to switch it off for FreeBSD.
This PR ensures that FFmpeg will always default to software decoding on FreeBSD, but should remain the same functionalities for other OS's.
The results are slight CPU increases while decoding in software mode, but still neglectable and they don't really harm performance.
Co-authored-by: MaranBr <[email protected]>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/64
Co-authored-by: SDK-Chan <[email protected]>
Co-committed-by: SDK-Chan <[email protected]>
SDK Chan
