SH3 Blitter Delay Discussion Thread

[KAI]

All I hear is bitching about how much the SH-3 emulation sucks and no one saying something helpful to improve it, so I think this needs a serious discussion.

So far I've found this regarding the blitter delay:

Futari: around 62-66%
Futari BL: unplayable (IMO)
Deathsmiles: 63% (equal to the 360 port, apparently)

ESPgaluda 2 needs more slowdowns, and IBARA/PINKSWEETS/MMP work "fine" for me.

[shmuppyLove]

How is this set in MAME?

(143u9 if it makes a difference)

[KAI]

Assuming you are talking about slowpoke MAME:
in game - Tab>Slider Control

[Obscura]

I just set it to 61% for everything.

[z0mbie90]

All I hear is bitching about how much the SH-3 emulation sucks and no one saying something helpful to improve it, so I think this needs a serious discussion.

Well they not in the newer MAME releases and they said they wont work on it on a while. Games to new and most of them have 360 port.
I tried some slowpoke when they where first released and it was unplayable.

So if you want the games to have correct slowdowns and be as near the Arcade originals your best choice is the 360 ports.

[trap15]

All I hear is bitching about how much the SH-3 emulation sucks and no one saying something helpful to improve it, so I think this needs a serious discussion.

There's no need for discussion because it can't be improved. MAME should not be used for CV1000 games.

Stop disrespecting CAVE

[KAI]

OK, then I'll buy a 360 just to play them you fucking rickboys ;P

[z0mbie90]

OK, then I'll buy a 360 just to play them you fucking rickboys ;P

Well 360's aint that expensive anymore. Sure import a ntsc-j 360 may cost a bit more.
Pal have a few ports, DDP-DFK, Deathsmiles, Akai Katana and Mushi Futari, Espgaluda 2 and MMP/Pink Sweets are region free.
And if you have money to get a pc that run sh-3 emulation fine then you should have money to buy a 360.
Don't be so fucking cheap, and support the game company you like!!! Saving money so you can buy a console is a start, maybe ask Mom and Dad!

[Nasirosuchus]

I've cleared several of the SH3 games using slowpoke MAME and I don't see anything to suggest that any of the games are unplayable. I don't see any freezing, crashing, or graphics glitches such as enemies and bullets not being rendered properly.

Is the slowdown 100% accurate to the PCB? No, but neither are the ports.

[Vyxx]

Is the slowdown 100% accurate to the PCB? No, but neither are the ports.

So neither is as good? Check.

Seriously there's nothing wrong with playing these games with MAME or anything but trying to make them on par with PCBs is quite humorous.

Either they are 100% perfect or they're not, gauging some in between percentage is futile.

[moh]

OK, then I'll buy a 360 just to play them you fucking rickboys ;P

Well 360's aint that expensive anymore. Sure import a ntsc-j 360 may cost a bit more.
Pal have a few ports, DDP-DFK, Deathsmiles, Akai Katana and Mushi Futari, Espgaluda 2 and MMP/Pink Sweets are region free.
And if you have money to get a pc that run sh-3 emulation fine then you should have money to buy a 360.
Don't be so fucking cheap, and support the game company you like!!! Saving money so you can buy a console is a start, maybe ask Mom and Dad!

I completely agree, but I also like to support the MAME community as well. Judging by the 360 build quality, it wont be long before all those ports become useless

[nZero]

Please don't turn this into yet another PCB vs MAME vs (360) ports thread.

[shmuppyLove]

Assuming you are talking about slowpoke MAME:
in game - Tab>Slider Control

So the setting's only exposed in the slowpoke build I guess -- I don't see it in the baseline MAMEUI32 build I use.

Bummer.

[ninn]

what does "blitter delay" do?

[BPzeBanshee]

what does "blitter delay" do?

My explanation isn't the best, but basically the video blitter has a set amount of refreshes or something it can do every frame. If it gets overloaded, the game slows down. No one including MAME/Demul dev apparently have any idea what the blitter refresh rate is precisely so in Slowpoke MAME there's an option to manually set it. The higher it is, the slower it is, thus 'Blitter delay'.

For the record, I find Deathsmiles on par with the 360 port at 63% and this is the setting I use when not using outright defaults for the other games. I said this much back in the shitpile that is the TCFC thread which probably where you heard it from. As for rendering, the only glitches you'll notice in terms of things like layers missing is due to Slowpoke MAME's handling of trying to do things when it's not running 100%, which only MAMEdev/Demul know about since the source wasn't made public (along with the Cease and Desist letter that sent this underground anyway, and frankly I don't think ever existed).

[z0mbie90]

I completely agree, but I also like to support the MAME community as well. Judging by the 360 build quality, it wont be long before all those ports become useless

Well I like MAME to, and use it alot, but mostly for older and games that don't have a console release.
And Im sorry if I sounded a bit of a shithead, I do understand people dont wanna waste money on a console that maybe will break.
My 360 died after 2 years, but I think the newer ones should be a bit better.

[trap15]

(along with the Cease and Desist letter that sent this underground anyway, and frankly I don't think ever existed).

Considering I got a Cease and Desist for something far less harmful to their business, I think it's pretty obvious that it did exist, at least this time around.

[BPzeBanshee]

(along with the Cease and Desist letter that sent this underground anyway, and frankly I don't think ever existed).

Considering I got a Cease and Desist for something far less harmful to their business, I think it's pretty obvious that it did exist, at least this time around.

I certainly wouldn't put it past Cave to send one, I have no doubt they'll at least try to seem like they're taking action, but I also certainly wouldn't put it past MAMEdev to forge the whole thing as a F-U statement to the freebie players - MAMEdev has after all liked to keep their secrets in the past, and frankly combined with Haze's reaction to Demul 0.57's official release etc I think the whole damn incident smells fishy.

[Nasirosuchus]

Is the slowdown 100% accurate to the PCB? No, but neither are the ports.

So neither is as good? Check.

Seriously there's nothing wrong with playing these games with MAME or anything but trying to make them on par with PCBs is quite humorous.

Either they are 100% perfect or they're not, gauging some in between percentage is futile.

Did I suggest that they were on par with the PCBs? No, I didn't. The point that I was making is that just because they aren't 100% to the PCBs doesn't make them unplayable because they ports aren't either.

PCBs weren't meant for home use, they were meant for arcades. Considering that arcades are dead in the west and that most people aren't willing to spend tons of money on PCBs and arcade setups, the PCBs aren't relevant anyway.

[ninn]

What are the PCBs not relevant for?

Was that an explanation or an excuse?

[Nasirosuchus]

What are the PCBs not relevant for?

Was that an explanation or an excuse?

They are relevant if you live in an area where arcades aren't dead and most of us don't.

PCBs aren't relevant because most of use will never have the opportunity to play them anyway and will have to rely on ports or MAME to play these games.

[Muchi Muchi Spork]

You're broke and your arcade went out of business so pcbs aren't relevant so thus mame should run like a piece of sheep shit and it's fine, because we will pretend there is nothing to compare it to. Run it how in the hell ever. The pcb is not relevant. Whatever mame says is cool. And mamedevs made a conspiracy to fuck over users with a BS C&D story. Seriously why didn't EOJ let someone else take over cave-stg.

[emphatic]

Seriously why didn't EOJ let someone else take over cave-stg.

[Nasirosuchus]

You're broke and your arcade went out of business so pcbs aren't relevant so thus mame should run like a piece of sheep shit and it's fine, because we will pretend there is nothing to compare it to. Run it how in the hell ever. The pcb is not relevant. Whatever mame says is cool. And mamedevs made a conspiracy to fuck over users with a BS C&D story. Seriously why didn't EOJ let someone else take over cave-stg.

I'm not one for internet memes, but: U mad, bro?

If you are, I suppose your anger is directed at most of the STG community considering 99% of those involved:

A) Don't have arcades in their areas. As for those that do, their arcades are probably not going to have the latest shmups.

B) Don't see PCB's and arcade hardware as a worthwhile investment. Just because someone doesn't want to spend their money on a setup that was meant to be played in arcades doesn't mean that he or she is broke.

Seriously, without MAME, this site and the whole community probably wouldn't exist because MAME and other emulators are the only way to play many of these games. MAME isn't perfect, but it runs nearly everything to the point that you can play from beginning to the end with no issues.

[Muchi Muchi Spork]

We're not trying to censor people, just improve the signal to noise ratio that many members have been concerned about.

[AntiFritz]

Seriously why didn't EOJ let someone else take over cave-stg.

[KAI]

SH-3 blitter delay discussion please.

[Ex-Cyber]

As I mentioned in some old-ass thread that touched on this topic, it seems likely to me that "the blitter delay" is not a fixed rate but rather varies depending on the type of operation (e.g. semitransparent vs. 1-bit transparent) and/or the workload in a given frame (e.g. the blitter and/or CPU might encounter more wait states if more sprites are rendered in a given frame). Then each game would be expected to have a different apparent blitter speed because they don't all use the blitter identically.

[BPzeBanshee]

SH-3 blitter delay discussion please.

QFT

@Ex-Cyber: Yeah, I recall MetalliC saying something around those lines. If memory serves there was also mentioning that since it's current-gen enough there could even be intricacies in stuff like the *RAM speed* which would have an affect on accuracy. That kind of detail is where my memory goes foggy though.

[xMetalliCx]

As I mentioned in some old-ass thread that touched on this topic, it seems likely to me that "the blitter delay" is not a fixed rate but rather varies depending on the type of operation (e.g. semitransparent vs. 1-bit transparent)

I dont think so.

In my experiments I've counted "normal" "alpha blended" and "use transparent" pixels blitted per frame, and the results is:
in the one games (like Futari) A LOT of blending used, but it works at more-less correct speed;
in another (like Ibara and other games on that engine) - most of blitting done without blending, BUT emulated game lacks a lot of slowdowns (with the same 'blitter delay" value as in Futari).
all that doesnt look like blitter works slower in "heavy blending" modes and faster without blending.

so, my current guess - blitter's math-unit works as kind of fixed pipeline, so whatever blit-mode was used, the math-unit will operate in the same way and with the same speed, we just do not know that exactly and how affects generall blitting performance.

Then each game would be expected to have a different apparent blitter speed because they don't all use the blitter identically.

do you mean "each game use different FPGA/blitter microcode" ?
the truth is all games use only 3 FPGA firmwares, and I think its no more whan just bugfixes.
for example DS, Futari, FutariBL, IbaraBL, MMP, PS (and I bet all recent games too) uses the same blitter firmware.

and/or the workload in a given frame (e.g. the blitter and/or CPU might encounter more wait states if more sprites are rendered in a given frame).

the number of "sprites" (actually number of blits in blit-list) definitelly must affect the speed, but my current experiments says that it must be very very low value, so it can be ignored, at least for now.

but you can be right about "CPU wait states", my latest guess is - during blitting the CPU is partially or completely stopping. that can be good answer on the question "why Cave use 133Mhz CPU clocked at 102,4Mhz, and why games usualy use only about 30% of its power ?"
but assume easier than to check , because the implementation of memory latency in emulators is not an easy task, maybe in the future I will have enough free time to test it.