05188: cottonbm: Game has a major performance loss

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
05188	Gameplay	Major	Always	May 6, 2013, 21:29	May 2, 2018, 22:37

Tester	joe35car	View Status	Public	Platform	MAME (Official Binary)
Assigned To		Resolution	Won't fix	OS	Windows Vista/7/8 (64-bit)
Status [?]	Closed			Driver	sega/stv.cpp
Version	0.148u4	Fixed in Version		Build	64-bit
		Fixed in Git Commit		Github Pull Request #

Summary	05188: cottonbm: Game has a major performance loss
Description	When you start the game it runs maybe at 50% but when you actually play the game it runs full speed. After a level it jumps back to about 50% speed then when the next level starts again it jumps back to full speed.
Steps To Reproduce	Just start the game then actually play it.
Additional Information	http://www.mameworld.info/ubbthreads/showthreaded.php?Cat=&Number=307281 Tafoid "I've researched this some.. it seems to be a valid performance hit and it seems to boil down to some rewrites that happened just after 0.148u2 was released. I don't think simple reverting of code is the answer, however."
Github Commit

Flags
Regression Version	0.148u2
Affected Sets / Systems	cottonbm

Attached Files

No.09471 Fujix Administrator May 7, 2013, 11:10	Looks like the high resolution scenes need more CPU power, it is not special usually. Why do you think it is a regression? Slow down can happen when updating a system.
No.09472 Haze Senior Tester May 7, 2013, 17:57 edited on: May 7, 2013, 17:57	I would consider it a regression, performance on several Saturn based systems has dropped to unacceptable levels. Decathlete, while not working, is even worse. They shouldn't have Naomi-like system requirements.
No.09473 Tafoid Administrator May 7, 2013, 18:08	I've informed Kale about it before - but he's currently inactive in the project. The revision was r21002 as a strong possibility.. replaced a section of code from before/after and noticed most of my speed returned (regarding "get_window_pixel" routine).
No.09476 Haze Senior Tester May 8, 2013, 15:56	adding a couple of inlines might help significantly, depending on what is causing this specific slowdown ;-)
No.09478 Fortuna Tester May 9, 2013, 21:24 edited on: May 9, 2013, 21:24	not only this game, is the driver completly has slow performance in older versions, die hard arcade running in my computer in a full speed, cotton 2 the same and other games to 80-90%, now running 20 to 40%
No.09481 Mamesick Senior Tester May 10, 2013, 14:07	The "get_window_pixel" routine changes in r21002 seems wrong to me... it's the same condition for inside/outside ranges... shouldn't they be different as they were before? Looks like a copy and paste bug....
No.09482 hap Developer May 10, 2013, 19:17	I can confirm that r21002 is the regression, and it's not related to the new stv_vdp2_drawgfx_alpha/stv_vdp2_drawgfx_transpen functions. I'm not sure what Kale's purpose is with changing stv_vdp2_window_process and get_window_pixel, other than having cleaner code. Maybe there's a bugfix too? superKale eyes need to have a look at it :P
No.09927 Kale Developer Oct 28, 2013, 17:08	Again this thing needs a rewrite, but who am I to say it?
No.14996 Kale Developer May 2, 2018, 22:37	Considering how much hard this HW is and people moans everyday about how bad is the Saturn emulation I'm closing this. There's no way we can achieve perfect emulation without breaking some eggs in the process.