08586: All sets in capcom/cps3.cpp, konami/hornet.cpp, jaleco/ms32.cpp, neogeo/neogeo.cpp, nintendo/nes.cpp, sega/model1.cpp and more: bgfx: Brightness/Contrast/Gamma settings ignored depending on core and random chance

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
08586	Interface	Minor	Random	Mar 22, 2023, 18:04	Dec 24, 2023, 20:20

Tester	ICEknight	View Status	Public	Platform	MAME (Official Binary)
Assigned To		Resolution	Open	OS	Windows 10/11 (64-bit)
Status [?]	Confirmed			Driver
Version	0.252	Fixed in Version		Build	64-bit
		Fixed in Git Commit		Github Pull Request #

Summary	08586: All sets in capcom/cps3.cpp, konami/hornet.cpp, jaleco/ms32.cpp, neogeo/neogeo.cpp, nintendo/nes.cpp, sega/model1.cpp and more: bgfx: Brightness/Contrast/Gamma settings ignored depending on core and random chance
Description	When using the bgfx renderer, the affected cores will randomly ignore any set Brightness/Contrast/Gamma values (see video in comments). If the game uses many screens, this bug may only happen in some of them (see screenshot of `orunners`).
Steps To Reproduce	-Set the video to bgfx, any backend. -Run any systems/games in the list of known affected cores below. You will randomly toggle this bug on/off by: -Entering or exiting full screen (Alt+Enter) -Changing the "Window Screen Effect" chain in the sliders -Hard resetting (Shift+F3)
Additional Information	The affected cores use bitmap_rgb32 instead of bitmap_ind16 in the screen_update function. Known affected cores: ausnz/mbee.cpp capcom/cps3.cpp beehive/microb.cpp ccs/ccs2810.cpp heathkit/h19.cpp jaleco/ms32.cpp konami/hornet.cpp konami/konamigx.cpp midw8080/mw8080bw.cpp mchester/ssem.cpp neogeo/neogeo.cpp nintendo/famibox.cpp nintendo/n64.cpp nintendo/nes.cpp nintendo/snes.cpp nintendo/playch10.cpp skeleton/gimix.cpp sega/mdconsole.cpp sega/megadriv_acbl.cpp sega/megadriv_rad.cpp sega/megaplay.cpp sega/megatech.cpp sega/model1.cpp sega/model2.cpp sega/saturn.cpp sega/segac2.cpp sega/segas32.cpp sega/segasg1000.cpp sega/segasms.cpp toaplan/toaplan1.cpp toaplan/twincobr.cpp toaplan/wardner.cpp Some of the unaffected cores: capcom/mitchell.cpp capcom/gng.cpp capcom/cps1.cpp capcom/cps2.cpp dataeast/cninja.cpp dataeast/karnov.cpp dataeast/supbtime.cpp jaleco/momoko.cpp jaleco/skyfox.cpp kaneko/galpanic.cpp kaneko/kaneko16.cpp kaneko/snowbros.cpp konami/asterix.cpp konami/combatsc.cpp konami/gradius3.cpp konami/parodius.cpp nec/pce.cpp nintendo/n8080.cpp nintendo/popeye.cpp nintendo/vboy.cpp sega/segahang.cpp sega/segaorun.cpp sega/segapico.cpp sega/segapm.cpp sega/segas16a.cpp sega/segas16b.cpp sega/segas16b_isgsm.cpp sega/segas18.cpp sega/segas24.cpp sega/suprloco.cpp sega/system1.cpp sega/system16.cpp sega/turbo.cpp sega/zaxxon.cpp sinclair/spectrum.cpp sinclair/spec128.cpp taito/taitob.cpp technos/ddragon.cpp toaplan/toaplan2.cpp
Github Commit

Flags
Regression Version
Affected Sets / Systems	All sets in capcom/cps3.cpp, konami/hornet.cpp, jaleco/ms32.cpp, neogeo/neogeo.cpp, nintendo/nes.cpp, sega/model1.cpp and more

Attached Files	brightness discrepancy.png (337,936 bytes) Dec 24, 2023, 20:19 Uploaded by ICEknight [orunners] One screen with correct settings, one without.

No.21186 Robbbert Moderator Mar 23, 2023, 03:32	Unable to replicate this with current git.
No.21189 cuavas Administrator Mar 23, 2023, 09:02	Which render module? If bgfx, which backend? If d3d, is hlsl on or off? Windowed or fullscreen? What GPU and driver version?
No.21190 ICEknight Tester Mar 23, 2023, 15:58	> Which render module? bgfx. > If bgfx, which backend? Any. > Windowed or fullscreen? Both. > What GPU and driver version? Any, tested with different cards and computers with Windows 10. tl;dr: Happens when using bgfx, any combination.
No.21192 Robbbert Moderator Mar 23, 2023, 23:49	With bgfx it is confirmed. >mame -video bgfx nes smb
No.21196 cuavas Administrator Mar 24, 2023, 01:35	What do the affected systems have in common which is different from the unaffected systems?
No.21197 Robbbert Moderator Mar 24, 2023, 04:28	I assume that is a rhetorical question, as that is a job for the assigned dev to determine. However, I did more testing, and established 3 groups of systems. 1. The sliders don't work at all - examples are mbee, ssem, ccs2422 (they work with d3d) 2. The sliders work correctly - most arcade games 3. The bug reported here occurs - anything with a software list item, plus some other random things such as h19, gimix, invaders, dm3270.
No.21198 cuavas Administrator Mar 24, 2023, 08:27	No it's a serious question. There's clearly some difference between the systems where it happens, probably somehow related to screen configuration. Determining what ths difference is would be the first step in working out what's going wrong. I'm not going to believe the presence of software list items triggers the bug - they don't affect the render stack at all. I mean are you saying it happens with slotted MVS (e.g. mame neogeo kof99) but not the fixed MVS configurations (e.g. mame kof99)?
No.21199 Robbbert Moderator Mar 24, 2023, 08:41	I tested 3 versions of kof99 1. kof99 - all is good 2. aes kof99 - sliders don't work 3. neogeo kof99 - all is good I would assume they all use exactly the same screen coding, so what do you make of that?
No.21201 cuavas Administrator Mar 24, 2023, 13:18	Just tested all three kof99 variants and a couple of NES games with a recent build, and the screen brightness and contrast sliders are non-functional for all of them with BGFX. I can’t get the behaviour ICEknight described at all. Maybe it’s something triggering undefined behaviour, and the “certain systems” part is a red herring?
No.21202 cuavas Administrator Mar 24, 2023, 13:33	Unfortunately I don’t have a new enough version of valgrind here to deal with BGFX (the version I have can’t handle “vcvtph2ps ymm5,xmm5” instructions), so I can’t find check for uninitialised variables that way.
No.21203 Robbbert Moderator Mar 24, 2023, 14:03	I just updated to the latest git, and some behaviour has changed. I'm only adjusting Brightness (but the Contrast and Gamma have the same issues). schaser, invaders and invadpt2 all behave the same now, the sliders have no effect, but if you leave (for example the brightness all the way up), then quit and restart, suddenly the brightness is all the way up. So the setting is being saved and restored correctly, but not being applied "live". Yet, kof99 and captcomm, the sliders do work live, there's no bug at all with them. And nes smb, it's exactly how iceknight says - it applies, up and down, but while the Del key returns it to 1.0, the screen doesn't follow suit. So yes, either undefined behaviour, a straight-out bug, or the dreaded undefined variable.
No.21228 Kale Developer Mar 29, 2023, 16:16	On my experience seldomly brightness/contrast/gamma don't pickup properly, and I have brightness and gamma to non-1. Windows 10, toaplan1.cpp games (outzone). brightness 0.95 contrast 1.0 gamma 0.95 [...] bgfx_path bgfx bgfx_backend auto bgfx_debug 0 bgfx_screen_chains unfiltered bgfx_shadow_mask slot-mask.png bgfx_lut bgfx_avi_name auto
No.21500 ICEknight Tester Jun 2, 2023, 19:57	Just uploaded a new video showing how bug report #8621 also happens at random when changing between bgfx Screen Effects:
No.21501 cuavas Administrator Jun 2, 2023, 20:02	If you want to help, run the code in a memory analyser and work out where it’s misbehaving. Bumping with irrelevant videos is just going to annoy people. If you actually thought about how software works, it would be obvious why MT08621 is a duplicate of this.
No.21502 ICEknight Tester Jun 2, 2023, 20:49	I'm just a user reporting issues as I come across them, which is what this platform is for. If somebody moderating a bug tracker finds it annoying that regular users are not tracing and finding the source of the bugs themselves, I really don't know what to say.
No.21503 cuavas Administrator Jun 2, 2023, 21:04	There’s a difference between reporting issues and your constant stream of argumentative comments. You need to learn to listen and think. We’ve established that screen adjustments aren’t always applied with BGFX with no clear pattern for when it does or doesn’t happen. Your video adds nothing except making the comments take longer to load and sending web traffic to Google every time this bug is viewed. If you don’t have the skills to further isolate the issue, there’s probably nothing you can add.
No.21715 ICEknight Tester Aug 29, 2023, 03:24	Just did some more testing and added a list of known cores which show this problem. There's many more than the console-related systems we already knew of, even including the CPS3.
No.21719 ICEknight Tester Aug 29, 2023, 17:03 edited on: Dec 24, 2023, 22:24	Osso pointed out that the cores affected by this bug use bitmap_rgb32 instead of bitmap_ind16 in the screen_update function. This indeed seems to be the common denominator.