konami/konamigv, konami/konamigq, konami/twinkle: Adjust DMA timings #12375
Conversation
src/mame/konami/konamigq.cpp
Outdated
| } | ||
|
|
||
| if (m_dma_requested && m_dma_size > 0) | ||
| m_dma_timer->adjust(attotime::from_usec(10)); | ||
| m_dma_timer->adjust(attotime::zero); |
There was a problem hiding this comment.
Since this is in a timer callback, using a zero delay here will result in the scheduler calling the callback again before executing another timeslice. You could get more-or-less the same effect with a loop inside this function.
Do you want to leave all the timer adjust calls in and add notes about the timing being unknown, or do you want to change the callbacks to use loops (and maybe still add notes about it)?
FWIW I did try to find information about how DMA timings are generated and got nowhere.
There was a problem hiding this comment.
I removed the timer adjust from scsi_dma_transfer and added a note about finding the proper DMA timings. I think the original situation I was trying to avoid that I have accidentally ran into in the past was a kind of loop situation where scsi_drq would start the DMA transfer, which would then kick off another call to scsi_drq, etc. So the timer calls were used to ensure that the functions would get called after scsi_drq returned so it wouldn't end up in recursion hell. I tested with various konamigv, konamigq, and twinkle games and just using a while loop here works fine.
|
slightly unrelated, but kdeadeye could do with a default / initialized NVRAM (flash) or it will hang on the high score screen as it doesn't have valid data to display. |
The 10usec delay per bytes is too slow and causing issues with games, so this should bring load speeds back closer to what it was with the old code. Should also fix https://mametesters.org/view.php?id=8860.
Previous related commits:
a6a81a6
cb8b90c