MTR 6 ANNOUNCEMENTS - READ FIRST!

[geezerbuttz]

Hello .

Disk time:
Been building the library , and have just started ti image the original starwars release.
Wow...

These are fully complex releases.
2 region ( 2 and 4 ) 5 languages , multiple angles.
So , in states that is 5 + 2 + region free check. + maximum state count.
that works out over 20 per node.

38 nodes ,... you do the math.
90 PGC in the main menu....

it takes 4 hours per disk to map every last drop of information from it.
Most complex yet.
G

[geezerbuttz]

I have an idea to improve the performance in probe further.
Track the button response , and only execute if:
a) there is no prior operation.
b) the result operation has not yet been processed.

This will avoid repetitive operations over the the same node , with multiple states.

I will look at that now...
G

[geezerbuttz]

As above:
i settled on storing the virtual machine register states after a button operation is complete , but before we start to read the sector data.

Thus: if we have the register set for this button in its store , then there is no reason to continue any further as the end result will be the same:

If not , we add the register set and continue as normal.

I reduced last nights ( actually it took 6 hours and 20 minutes ) down to 2 hours doing this.

Not bad.

The knock on is that less states are generated in existing nodes , as we do not process data generating redundant states.

Of course , all Audio , and regions have to be tested, that in itself takes time.
More soon.

[geezerbuttz]

Well , that was fun while it lasted.
Due to an error in the state calculation , i lost some of the benefit I gained.

I miss-calculated the maximum state count.
Instead of multiplying the total languages but total regions , I had simply added them together ... that was wrong.

Every group of languages needs to pass through the system with an independent region value.
This includes Region Free Value.

So , A disk with 24 region , need to run region 2 , region 4 , region free,
So 5 languages , three region = 15 possible states ( minimum )

not 8.

Well , a side effect is that its pretty well a forensic app now.... just be patient to let it break down all possible navigation paths. ( there can be a lot.... )
G

[geezerbuttz]

a few more optimisations:

I was making some extra reads on the device when copying the virtual machine:
This was costly , as it takes around two seconds plus re-seek time to continue:
Although im not able to bypass all the re reads ( some need the file pointer updated , so they must access the drive )
I have been able to cache current structures , and make a deep copy of them on demand.

I hate making deep copies , as the code is so repeatable and boring...
( but its faster in operation that reading from the disk )

It gets us a little extra time back.

Im on the fence with the next bit.
Currently after a button operation , i check and existing register set , if it exists , we move to the next button.
This does cause a duplicate read on the drive.
my thought is....
If the registers are the same 'before' an explicit button is pressed , would the result of pushing that button be the same?
Buttons cannot in them self support a random function , but the following command tables can.

you cannot check if its a video jump , as a command table will still be executed..

As I said , im on the fence with this one... if possible it the last speed up I can insert.
G

[geezerbuttz]

Re above:

A thought , if i can cache weather a random variable was checked , and store it in the last pre button register state , we can check that and decide weather to miss the next button if we have already been here , or continue , and generate a random condition.


Hmmm...

Let me examine.

[geezerbuttz]

The above makes no difference as we still need to load in the duplicate Virtual machine state.

I have however optimised another read phase out of the equation , reducing disk access even more.

More tomorrow.

[geezerbuttz]

Bit of a crazy day...

I spent the day debugging and changing some code so it can auto detect if it is a deep copy of a structure or a shallow copy.

After spending half a day looking for a single bug ( yes it was a missing '1' added to a struct size prior to division, thus giving the incorrect count of elements prior to memcpy )

I have got the shallow copy of the vm-state to work.

What does this give us?
Well , because a button requires the reloading of the vmState , you can imaging with 18 buttons , and 30 states , thats a lot of reading from the disk we have just eliminated. It can now also bypass used button's very quickly.

I will add a button to the prefs to enable deep copy of the VM if the user wishes.

More soon ( im away for a few days though )

G

[geezerbuttz]

just to justify the insanity of the last few days:

Same disk , same computer same drive:

Deep VM copy: 05:42:00
Shallow VM copy: 00:52:00

( drive dependent but third run was 48 minutes, so speed is conformed now. )

No more to say really.

[rthpjm]

just to justify the insanity of the last few days

No need for justification… you are after all a superbeast

[geezerbuttz]

Running the code ( im back home ) I have a disk , that under the new calculation of states , has Regions 1 - 8 to probe and 8 languages.

All have the possibility to alter navigation.

70 states per node... Its a peach....

( shallow state copy is active )

Update:

So , to verify this case , I have run a full deep VM copy..
8 hrs.
Shallow was 2Hrs.

They both found the exact same layout.

I will stop testing probe at this point.

I have to look at the slicer in MTR tomorrow as it keeps giving ( more frequently than I like ) alignment errors.
That will be the last bug/issue I will be fixing ( heard that before..... )

I need to get this out to generate a bit of income to fund stage two , as its a much more complex engine than the original RnR.

G

[TheSystemHasFailed]

Running the code ( im back home ) I have a disk , that under the new calculation of states , has ]

I need to get this out to generate a bit of income to fund stage two , as its a much more complex engine than the original RnR.

G

And we cannot wait!

[geezerbuttz]

ok , slicer is updated , all seem to pass ok , corrected the miss aligned sectors.
Implemented a severity of miss-alignment warning so a few that get corrected we dont post , over 5 its a low level warning .
It goes up from there to a high level warning.


Found that MKV was broken... fixed that , and as an added bonus I reworked the old UI update string parsing routine to be more lightweight and efficient.

Work a treat..

A few more little UI niggles to fix and Im done.

Phew.

[geezerbuttz]

A slight UI issue with the VM command displayed in the MF panel ( for its relevant cell)
The command is longer than the placeholder , it gets truncated , making it hard to see what it is.

To get around this , I added extra flexibility to the right click process in the MF cell panel.

As it was before , a right click shows auxiliary info relevant to the cell .
Now if you hold the qualifier key CTL ( control ) when you right click , it will show the full VM Command rather than the Aux Info.

Of course , if there is no Cell Command , it will still default to the Aux Info.


Now to look at the frame timing calc , as it seems ropey.
G

[geezerbuttz]

Hello All.

I completed the time code alterations.

What changed.

The old code was in Hrs:Mins:Secs:milliSecs...

Not exactly nice to deal in millisSecs.

So now it matches the IFO standard of Hrs:Mins:Seconds.Frames

Had to rework quile a bit , as its dependent on the old code in quite a few places..
Seems to be good now. Testing.

[geezerbuttz]

Time to be unpopular...

I have been here before , so I know how it goes.

I need to raise funding for the next update, and as I suspect , It will be all take , with the usual crowd funding the further development.
So I have an Idea.

Why not make DVDSectorProbe availability depending on your license type.
I can add a new data section to the license that enables Probe to launch , or use a DVDProbe map.

Say charge 10Euro for the added functionality.

( let the cries of outrage begin )

Ideas ? ( none rude please )
G

[geezerbuttz]

Quantise has a bug. Its not thread safe , its locking up in process...

I will get to that in the morning.
G

Side note: probe can be enabled and disabled in the license.
Of course MIF maps can still be saved and loaded with in MTR.

[TheSystemHasFailed]

Why not make DVDSectorProbe availability depending on your license type.
I can add a new data section to the license that enables Probe to launch , or use a DVDProbe map.

Say charge 10Euro for the added functionality.

G

That doesn't seem unreasonable given that I believe [a] most people likely do not use that area of functionality (they hope for their needs it will *just work*) and [b] given the amount of rework you have put into that area of functionlity.

[geezerbuttz]

ok:

The Quantizer is consumer model , with the app main extraction code being the provider.
The QuantEngine is initialised , and waits in the background , waiting to be fed data.

The Quants data buffer size is fixed. When this buffer exceeds a fixed size , the provider waits while the consumer ( the QuantEngine ) consumes the data.

When the buffer size is less than the fixed size ,the Quant Engine waits while the provider tops it up.

The issue was a simple one.

In the checks that verify and watch the Quant buffer , on the provider size i was comparing the buffer size with a "GreaterThanOrEqualTo" , and on the consumer side I was comparing with a "LessThanOrEqualTo"..

If you cannot see , with this when the buffer size == FixedSize , both conditions become true , thus locking the code in a race condition , as neither can relinquish control to the other thread.

Thats fixed.

I also upped the buffer size from 4Mbits to 8Mbits to reduce the read overheads..

Testing now.
G

[geezerbuttz]

ok , quant hit an assertion , no data to refill...

Let me look into that.....

[geezerbuttz]

Turns out the ISO is faulty.

Still , instead of asserting and stopping , we now , jump past the bad data to process the good.

Check your outputs.

Oh last thing for today:

That was a four disk extraction , with quant on each extraction , output to mkv.
All Ok.

Next release ( hahahaha ) im planning to implement basic FFMEG settings into batch.

[geezerbuttz]

Updated the Quant a little bit more...

The data to save is handed out to an OperationQueue.

When this Queue gets above 8 operations , I was sleeping the main provider for 200m/s while we wait for the queue to unload.

This has been changed to use NSCondition.

The Operation Queue count is registered for observation. When the queue count is greater than 8 ops , we issue a wait condition. This stops the thread in its tracks , wasting no CPU cycles.

When the observer reports that the queue is less than 4 ops , we send a signal to release the wait condition...

Work very nicely.. really reduces operation load , and disk mashing on multiple writes.


I have seen an issue once or twice that the app reports the title data as corrupted , when it is not , and no UI will load,
I have to unload the disk , and reload it.

Im looking to see if I can replicate the procedure to make it happen..
Stay tuned...

[geezerbuttz]

Above issue is a no brainer , its done.

[geezerbuttz]

ok , new licensing handler is in for DVDSectorProbe.

Its drag and drop from within the app. you can request a license from with in the app.
Just drag the license into the drop view , and it will install it.

I plan over the next few days to remove the Mac Address requirement for the MTR license , and use a more stable non changing internal hardcoded value to lock to.( i have tested in the new code ,and its ok... )
That should end the need for a license change when the mac decides to switch interfaces.

I may implement drag and drop for MTR license installation as well..... I will see.




G

[geezerbuttz]

ok , the old license system has been completely overhauled.

Its now locked onto a single stable variable inside the machine. So hopefully no more licenses being made invalid.

Next , is to change the old window/button based license requester/installer to to a single button request , with a drag drop installer , and maybe remove the relaunch phase.

That is tomorrow...

Lots of testing on this new license code , as it needs to be rock solid.
More soon.

[geezerbuttz]

Well there we go:

I wrote an injectable DropView controller.
We can pass in different blocks to the same class to alter its behaviour.
Same class can license MTR , and be swizzled to license DVDSectorProbe.

Works nicely, full drag and drop , no relaunch of the app as promised.

Im just going to go though it cleaning up some unused code that is floating around....
Already doing that.

More soon.

[geezerbuttz]

Well . after a very long development time , phase one is complete.

It took a long time , Im not so quick these days , but I have spent a huge amount of time reading over the code ,
Optimising , changing things Im not happy with etc etc etc....

It did feel like it would never get to a point that I would be happy with it.

I spent today force inlining some code ( it was being stubborn) , checking over the Graphics , just to be sure.
I may re render the Icon and splash screen to match the new User look...

That for this weekend , ,,, Thats relaxing.

I am pleased with the license system , it is much more hands on , and very simplistic.
Get a license , drag it to the window , job done. Should make users a lot happier.
That with the changed lock ident , and we should remove 80% of the license issues.
Change you machine , change your license , thats the only time.

Right I need to sleep a bit.

More soon.

[TheSystemHasFailed]

Congrats...! Enjoy some well-deserved sleep!

[geezerbuttz]

one small thing I can address.
The save location for the IMF results can be set in both apps.
Im thinking that it would be better to share the location so it is synchronised between both apps , leading to less issues in the long run.

Just one less thing for the end user to worry about.
G

[geezerbuttz]

ok ,

All Icons refreshed , and redesigned.
Created new Icons for the .IMF and .MIF file format.

Took a bit longer than normal as I used to use Modo , but the drip feed of functionality from The Foundary that they call progress , and the rather steep upgrade price for it is no longer to my liking.

Goodbye Modo... thats the end of my long journey with Allen Hastings and Stuart Ferguson.
They , If you do not know were the original developers of a little program known as Lightwave, that I learned back in the early 90's on an Amiga B2000 ( with 060 whooHooo)

They later departed and formed Luxology , and gave life to Modo.
It was a welcome departure from the nuts and bolts that Lightwave had become. Alas , they moved on and sold out , and Modo is now a shadow of its former self.

So , Its Blender , a program that I simply kept away from after trying it out in its earlier forms.
I never could adapt to the crazy UI it has in the old days..

These days it is a vastly different story. Im up to speed , ( nearly , still afew things I got to get my old brain around ) and I have to say , wow.

The bloated , money hungry companied should watch out... It does everything and more that you can payy hundreds , if thousands for.

Needless to say , I modeled and rendered everything in Blender for the Icons.
It was fun...

Just need to check it all the changes propogate down the line, and when Im happy they do , I will package it up.

G