Closed dbz400 closed 8 years ago
@SakataGintokiYT , are you using Nvidia card ? if yes , can you see if you can reproduce.
@kd-11 @raven02 Crash when resizing screen
@raven02 same here (i have a gtx 970)
@SakataGintokiYT https://ci.appveyor.com/project/kd-11/rpcs3/build/1.0.141 NV users can test this. It will very likely crash, but i 'm more interested in RPCS3.log it generates. Thanks.
@kd-11
{rsx::thread} RSX: Window resize request start... {rsx::thread} RSX: Waiting for device to idle {rsx::thread} RSX: Requesting a present surface from the presnentation engine {rsx::thread} RSX: Wait finished. Begin window resize {rsx::thread} RSX: ERROR: [Driver] Code 0 : vkCreateSwapchainKHR: internal drawable creation failed
Log : https://gist.github.com/SakataGintokiYT/dbae18f8fcaaa11271b49fe0257ad775
Thanks. Thats more useful.
@SakataGintokiYT @raven02 Can you test https://ci.appveyor.com/project/kd-11/rpcs3/build/1.0.154 when it's done building? I'm hoping it stops the emulator rather than crashing. I'm assuming when resizing fails the emulator hard crashes (rpcs3 has stopped working). If it only freezes please let me know.
@kd-11 just tried that build that you linked, it still crashes (rpcs3 has stopped working), i have a gtx 970.
@Nezarn log?
@kd-11 now at second run it didn't crash https://gist.github.com/Nezarn/037125eb9c82fdecf700037b95b6d456
·F {rsx::thread} class fmt::exception thrown: vk::swap_chain::init_swapchain(): Assertion failed! Result is C4650B07h (in file c:\rpcs3\rpcs3\emu\rsx\vk\VKHelpers.h:838)
edit: looks like its game dependant, hatsune miku project diva F just crashes without anything unusual in log.
@Nezarn Test this https://ci.appveyor.com/project/kd-11/rpcs3/build/1.0.155 after it finishes building. It will cause a ~2-second freeze when resizing. Try maximizing rpcs3 (not fullscreen) so that I can get a look at what is happening to the window surface.
@kd-11 Yes it works
Log : https://gist.github.com/SakataGintokiYT/396581d971fd105c8f07a32627af1cf7
Seems concurrent access was still happening even after waiting for device idle. Whats odd is that the nvidia driver should in theory be the easiest to resize since it creates a swapchain with 16k x 16k dimensions and handling resizes should be as easy as changing image dimensions.
Anyway, thanks for testing guys. I'll try to force hard sync points at different points and see how we can quickly resize without stalling the CPU thread. Hopefully the next build will work okay now.
Try https://ci.appveyor.com/project/kd-11/rpcs3/build/1.0.156. Hopefully it will block in a better manner than before before trying to resize.
@kd-11 Works good
Log : https://gist.github.com/SakataGintokiYT/4550ec17185faf355e9b8156f647198d
Okay. I'll submit as a new pull request after I clean the code up a bit.
Just tested , it looks a lot better and full screen works prefectly.
This one is affected Nvidia only . AMD works perfectly fine.
@kd-11 , let's track this issue here