[TESTERS NEEDED] cellAtracXdec implementation #15538
Conversation
| ensure(handle.aligned(0x80)); // LLE cellAdec aligns the address to 128 bytes | ||
| ensure(!!notifyAuDone && !!notifyAuDoneArg && !!notifyPcmOut && !!notifyPcmOutArg && !!notifyError && !!notifyErrorArg && !!notifySeqDone && !!notifySeqDoneArg); // These should always be set by LLE cellAdec | ||
|
|
||
| new (handle.get_ptr()) AtracXdecContext(notifyAuDone, notifyAuDoneArg, notifyPcmOut, notifyPcmOutArg, notifyError, notifyErrorArg, notifySeqDone, notifySeqDoneArg, |
There was a problem hiding this comment.
Write out the context using wrote_to_ptr.
| handle, notifyAuDone, notifyAuDoneArg, notifyPcmOut, notifyPcmOutArg, notifyError, notifyErrorArg, notifySeqDone, notifySeqDoneArg, res, spursRes); | ||
|
|
||
| ensure(!!handle && !!res); // Not checked on LLE | ||
| ensure(handle.aligned(0x80)); // LLE cellAdec aligns the address to 128 bytes |
There was a problem hiding this comment.
Then align it down manually.
There was a problem hiding this comment.
I think my comment there was a bit confusing. This function on LLE does not align the address manually or check the alignment. I've rewritten it to make it clearer.
| return handle->send_command<AtracXdecCmdType::decode_au>(ppu, pcmHandle, *auInfo); | ||
| } | ||
|
|
||
| void _CellAdecCoreOpGetVersion_atracx(vm::ptr<std::array<u8, 4>> version) |
There was a problem hiding this comment.
simply use vm::ptr<u8[4]>, I think its unspecifued behavior for std::array to match c array in byte size.
| // For savestates, restore argument | ||
| idm::get<named_thread<ppu_thread>>(static_cast<u32>(atxdec->thread_id))->cmd_list | ||
| ({ | ||
| { ppu_cmd::set_args, 1 }, static_cast<u64>(atxdec.addr()), |
There was a problem hiding this comment.
- Dont check thread_state::state for it, check cpu_flag::again state if it was raised inside exec(). (checking the reason the thread was aborted)
- When ::again flag is set, whats saved in r3 to r6 is ppu_thread::syscall_args, so you rather save to syscall_args[0] instead of this.
| queue_not_empty.wait(queue_mutex, 20000); | ||
| } | ||
|
|
||
| if (!run_thread || thread_ctrl::state() == thread_state::aborting) |
There was a problem hiding this comment.
Rather, check cpu_thread::is_stopped(), raise ::again flag if stopped.
| { | ||
| ensure(!!handle && !!pcmStartAddr); // Not checked on LLE | ||
|
|
||
| std::memcpy(outBuffer.get_ptr(), pcmStartAddr.get_ptr(), handle->decoder.pcm_output_size); |
There was a problem hiding this comment.
There should be an assert before this, checking memory safety with vm::check_addr.
|
|
||
| if (sample >= std::bit_cast<f32>(std::bit_cast<u32>(1.f) - 1)) | ||
| { | ||
| output_f32[sample_idx * decoder.nch_in + ATXDEC_AVCODEC_CH_MAP[decoder.ch_config_idx - 1][channel_idx]] = std::bit_cast<be_t<f32>>(std::array<u8, 4>{ 0x3f, 0x7f, 0xff, 0xff }); // Prevents an unnecessary endian swap |
There was a problem hiding this comment.
I think you can do something like "\x37\x7f\xff\xff"_u32 (resulting in u32)
Known issues
The issue is caused by the game sending invalid data to the decoder. The thread "bnusCoreDecoderAT3PThread" then just exits when it receives the error. I assume this is because of a race condition in the game. I tried slowing down the decoder thread to match realhw as closely as possible, but that didn't improve anything.
I was able to trigger this issue in Tales of Xillia 1+2 and Tekken 6 as well, it's likely other NU engine games are also affected.
Setting PPU Thread Count to 1 still works around the issue with HLE.