[alexdedalus]

dexter1, thanks for the additional info, next time when I call Indy I will remember to note down the requested version!
Today I make all the changes in the emulator and in the files, to briefly return with the results.

Will these new roms also be placed in the "roms" folder and need to add some option on the command line in the emulator mame?


Another thing, do you think normal indy battery is going to last so long? It's normal?


Thank you!

Add: 

Has the optimization in compilation gone from 3 to 2, does this commence the execution speed, or does it cause a known problem because of the optimization?

I know that under certain circumstances, optimizations may not have the desired effect because of certain peculiarities in the target processor.

Finally tested.

What is the actual mame version?

I prefer the indy system because I have a Indy.

Results: Indigo2 R4400 150 -> 27% (using the i9 PC).
Later I test in my q6600 PC.

The audio is working but with echo.

I'm trying to fix some time now over the weekend to get a look at the mame source code.

Recently I am looking for a 13W3 (male ???) adapter connected directly to the Indy, for a female VGA output. I even made a male 13W3 connector by hand, but it gives a lot of interference and the image does not stay stable colors. A decent adapter is really necessary.

[dexter1]

Will these new roms also be placed in the "roms" folder and need to add some option on the command line in the emulator mame?

These roms go in the "roms" folder, yes.
I added the command for starting mame with the indigo2 emulation, for the indy, use:

Code:

./mame64 indy_4610 -gio64_gfx xl24

Another thing, do you think normal indy battery is going to last so long? It's normal?

The variance in dead Indy clock batteries vary greatly: Some have their Indy battery replaced a decade ago, some are still running on their original Dallas 1386 from 20 years ago. As long as it runs, it runs. Besides, the clock battery rework featured by several people several times is a job of an hour or so and is relatively easy to do.

Has the optimization in compilation gone from 3 to 2, does this commence the execution speed, or does it cause a known problem because of the optimization?

Raising the optimzation from 2 to 3 does nothing in terms of performance and will only increase compilation time: This is because the MAME MIPSIII core, the slowest part of the emulation , doesn't respond well to optimization and is therefore currently the bottleneck.

I know that under certain circumstances, optimizations may not have the desired effect because of certain peculiarities in the target processor.

"Peculiarities" is a nice eufemism for bad compilers :-) Optimization totally depends on how good the compiler can determine and even predict program flow and make sensible assumptions on variable and register volatility. All compilers suffer from bugs in optimization, some more than others...
Also, compiler optimization cannot fix certain choices or errors in program design.

What is the actual mame version?

Freshest "git pull" from https://github.com/mamedev/mame

[TheMogMiner]

So it looks like the R4000 core code in mame is taking about 60% of the total CPU time and if there is room for improvements, that would be the first place to start. I've tried with -drc and -nodrc and it didn't make any difference, so in infer that there is no dynamic recompile code active, which could potentially create lots of speed improvements.

We'll have to see what the future brings.

I'm not expecting any speed improvement, because according to the perverted philosphy of the MAME team, dynamic recompilation, HLE and graphical acceleration are detrimental to "accuracy". On the contrary, I expect the emulation to become even slower, in the name of "increased accuracy".

I'm sorry to be this blunt, but as long as a complex platform like this is only emulated by MAME, it might as well be not emulated at all: even on the fastest machines, it will always be ridiculously slow and unusable. Sure, the MAME team might lull you into a false sense of hope by telling you lies such as "code optimization is not necessary because hardware will eventually become fast enough on its own", but history shows that the sluggishness of MAME has increased much more than hardware speed.

At this point, if this was an Italian emulation newsgroup 10 years ago, MAME zealots would jump at my throat stating that "the purpose of MAME is not to run the software it emulates" and "being able to use the emulated software is a nice side effect". I hope nobody in this forum drinks that Kool-Aid.

This is so wildly inaccurate that I don't even know where to begin, but I'm going to give it a shot.

Prior to the introduction of the new interpreter-only (but more accurate) R4000 core, the 150Mhz R5000 Indy driver ran flat-out at 100% speed on my over 4-year-old desktop. That's with no specific efforts towards optimization.

Speaking for the MAME team, code optimization is absolutely necessary, and with Moore's Law out the window, host machines will absolutely not become fast enough on their own. That is why the long-term plan is to either backport the edge-case fixes from the new R4000 core to the old, fast, dynamic recompiler core, or to simply write a brand new DRC. You're acting as if statements made by the MAME team 15 years ago during the heyday of Moore's Law are somehow still applicable these days. That's ridiculous.

The simple fact of the matter is that there are arcade machines using MIPS III processors clocked even faster than what was used in the Indy, equipped with Voodoo 1 or Voodoo 2 graphics cards, and they manage to run at or better than full speed, so there's no reason why the Indy driver, once it is able to benefit from the DRC system again, wouldn't run at or better than full speed, either.

You might hope nobody "drinks the Kool-Aid", but meanwhile I hope nobody here listens to your bile, because it's clear you haven't even the slightest clue what you're talking about. I've devoted the past half a year of my life to improving the Indy emulation to the point that the only things left to do are ethernet implementation and optimization. Your ungrateful attitude makes me want to just remove SGI emulation from MAME entirely.

It's real easy to make wild, sweeping generalizations from your position as an end user who doesn't have to lift a finger to do a single shred of work, but at the end of the day all you can do is tear down the work that other people have done. You have a vile and destructive attitude, and it isn't even one backed up with facts or experience. You lob a bunch of conjecture and invective at the MAME team as if it's true simply because it fell out of your empty head. You should be absolutely ashamed of yourself.

[dexter1]

Thank you TheMogMiner, i couldn't have said it better myself.

Maybe my wording in my initial response to Devilmasters post wasn't strong enough, i'm sorry for that. I have a bit of a history for bringing out trolls to forums, so i am cautious to not let that happen again.

That being said, people demeaning other man's valid and respected efforts should shut the hell up and do some proper searching and discussion before posting fake news, wild conjecture and their inflated ego's on the forum.

Everybody should buy TheMogMiner and TyFighter and other MAMEdevs a beer if they meet them. Or a Latte

[Raion]

I've removed DevilMasters post because it offers nothing positive or constructive.

[shrek]

Hey TheMogMiner, the recent emulation advancements you've made are one of the best things to happen in emulation for a while now. I just wanted to say that I really appreciate all the work you've put in. Every online community will always have small disagreements on certain things, though I believe most of the people here are extremely grateful for all of the time and effort you've put in, and for the resulting code. Good stuff man! Keep it up!

[RodneyDavies]

Hey Everyone,

I've managed to get IRIX 6.5.22 on Indy R4600 installed under MAME 210 and running under MAME 212 ...

I had to install using 210 because on my PC (i8700k Win10) both 211 and 212 at the moment do not appear to correctly map the CDROM to SCSI(6), and the PROM thinks it's a 'disk' instead of 'cdrom' (see hinv output).

   

This is the command line being used for this shot (I excluded -cdrom1 option to avoid any possible hangups but the hinv output is the same with or without -cdrom1):

mame64 indy_4613 -gio64_gfx xl24 -cheat -mouse -ui_mouse -hard1 irix36gbHD.6.5_p5086.6.5.22.chd -window -resolution 1280x1024

I have allocated a 36GB HDD and it's running 24-bit graphics.

I have done a full install of IRIX 6.5 (including Freeware and SoftWindows95) and then did a full upgrade (not yet included any extras, just IRIX) to 6.5.22.

It was quite tedious because in MAME 210, the mouse was slow and jerky and then, the left-mouse button did not work however, whenever the mouse moved along the X-axis from left-to-right, a left-button-click event would fire and sometimes hold and this would happen randomly. So, if I wanted to click on a button in a window, I would first need to move the mouse right-to-left to position it just to the left of the button that I wanted to press then, I would slowly drag the mouse left-to-right horizontally and it would magically press the button for me as I dragged the cursor across it ... I installed 6.5.22 under those conditions and it's taken a long time, not to mention the thirty-or-so times that MAME would randomly lock up part way through the install (which in itself took just over 4 hours to complete!).

I will see if I can talk to the MAMEDEV guys about those issues but suffice to say, MAME Indy R4600-133 (Indy_4613) works with Irix 6.5.22!

Cheers,
Rodney

P.S. Big Thanks goes out to all the devs who have made this possible! Thanks heaps!

[TriOx]

Hi TheMogMiner, thank you very much, this is incredible! Also thanks to dexter1 for your tips and the chd with Irix6.2 installed!

On Mame 0.212 for Win10, i9-9800x at 4.8ghz, I was curious about the emulated scsi performance so I setup a few timex dd runs on a 512 blocksize 100mb file and also used a stopwatch to compare timex time to actual time to get the average efficiency.
Write: 3 runs (s) - timex = 10.03, 9.48, 9.22 stopwatch = 66, 64, 62 Average: 15.0%
Read: 3 runs (s) - timex = 8.88, 9.01, 9.02 stopwatch = 48, 48, 48 Average: 18.7%
Average speed for whole session in CMD = 24%

Next step is to make my own CHD from a my r4k/250 Indigo2 Impact. I think I have a handle on a procedure now.

[shrek]

Hey Everyone,

I've managed to get IRIX 6.5.22 on Indy R4600 installed under MAME 210 and running under MAME 212 ...

You would be an absolute saint if you uploaded the chd file somewhere

[RodneyDavies]

Hey Everyone,

I've managed to get IRIX 6.5.22 on Indy R4600 installed under MAME 210 and running under MAME 212 ...

You would be an absolute saint if you uploaded the chd file somewhere

Yep, I can upload it somewhere - it is just over 2.6GB and is setup to hold ~36GB...

@Raion - is there somewhere on IRIX Network I could upload it to via FTP?

Thanks,
Cheers,
Rodney