Avisynth 2.6 MT

[Overdrive80]

With last mt avisynth I have troubles to use masktools2. I know that its not trouble of your dll, but could do .dll for avisynth alpha4 with retrocompatibility according to version of date:2012.05.16?

Thanks in advance.

[SEt]

I won't do extra compatibility builds – no one forces you to upgrade, but don't expect any new featured then either.
You want to upgrade – it's time to solve the problems. I'd prefer updated masktools2 to special build of Avisynth.

Expect updated MT version in some days, I'm quite busy atm.

[Overdrive80]

Originally Posted by SEt

I'd prefer updated masktools2 to special build of Avisynth.

Totally agree with you, but Masktools2 hasnt been updated from 2010; for this reason I request this.

Thanks for your answer.

[SEt]

Masktools2 is open source, so it's not like we can't do minor bugfix for it.

[Overdrive80]

Originally Posted by SEt

Masktools2 is open source, so it's not like we can't do minor bugfix for it.

All right, I hope someone who knows how to do it can fix these problems. ^^

[tormento]

Set, do you plan to try an alpha4 compile?

[FLauBerT]

"SetMTMode" SeeSaw does not work. XviD is not a problem, but it does not work at x264. Saving work in "Megui has stopped working" error. Every time you go. MT (parameters) as the between 65-80% When using the processor uses the x264 and I'm having a serious waste of time. Processor Intel Corei5 2300. Waiting for your reply.

[Selur]

"SetMTMode" SeeSaw does not work. XviD is not a problem, but it does not work at x264.

sounds like you are running out of memory, try to lower the lookahead values in x264

[SubJunk]

Try setting SetMemoryMax to lower. Between 400 and 700 is most stable for me.

[FLauBerT]

Originally Posted by SubJunk

Try setting SetMemoryMax to lower. Between 400 and 700 is most stable for me.

Create and save the script in error when I say "stopped working Megui" he says. A problem that occurs only in SeeSaw. SetMTMode scripts (4,4) using the form. A denoise a debanding and using degrain filter. Waiting for your answers.

[Porsche_fan]

I’m having problems running the following script in ffdshow + MPC-HC.

SetMemoryMax(1024)
SetMTMode(3)
ffdshow_source()
SetMTMode(2)
FineSharp(sstr=2.5)
SetMTMode(1)
GetMTMode(false) > 0 ? distributor() : last

After a lot of effort, I used subtitle(string(GetMTMode)) to determine GetMTMode(false) returns "0". So I can get the script to work by changing the last line to...

GetMTMode(false) >= 0 ? distributor() : last

The modified script with the mode=0 is a little unstable but that may be related to other variables like SetMemoryMax. Any idea why Mode is "0" despite setting it to "1" in the previous line?

Here are my ffdshow settings




I recently uninstalled 2.5.8 and replaced it with AVS 2.6.0 Alpha 4 [130114] initially then uninstalled and currently running AVS 2.6.0 Alpha 3 [110525] with 2012.08.28 2.6 MT.

One other question, I've read somewhere else that SEt's avisynth.dll should be placed in the ffdshow folder but I don't think that is correct. Currently there is no avisynth.dll or ffavisynth.dll in my ffdshow folder...is this the correct installation procedure?

Running Windows 7 64 on a i7 860.

[Groucho2004]

Originally Posted by Porsche_fan

I’m having problems running the following script in ffdshow + MPC-HC.

SetMemoryMax(1024)
SetMTMode(3)
ffdshow_source()
SetMTMode(2)
FineSharp(sstr=2.5)
SetMTMode(1)
GetMTMode(false) > 0 ? distributor() : last

After a lot of effort, I used subtitle(string(GetMTMode)) to determine GetMTMode(false) returns "0". So I can get the script to work by changing the last line to...

GetMTMode(false) >= 0 ? distributor() : last

The modified script with the mode=0 is a little unstable but that may be related to other variables like SetMemoryMax. Any idea why Mode is "0" despite setting it to "1" in the previous line?

Here are my ffdshow settings




I recently uninstalled 2.5.8 and replaced it with AVS 2.6.0 Alpha 4 [130114] initially then uninstalled and currently running AVS 2.6.0 Alpha 3 [110525] with 2012.08.28 2.6 MT.

One other question, I've read somewhere else that SEt's avisynth.dll should be placed in the ffdshow folder but I don't think that is correct. Currently there is no avisynth.dll or ffavisynth.dll in my ffdshow folder...is this the correct installation procedure?

Running Windows 7 64 on a i7 860.

1. You're running a single threaded version of Avisynth, that's why GetMTMode() is returning "0".

2. If you're using MPC-HC, you should not be using "Distibutor()" at all, the call is added automatically. So, if you are using Avisynth MT you are unnecessarily increasing the number of threads.

3. "SetMTMode(1)" in your script. Why?

4. The file avisynth.dll (any) should ideally be placed in the Windows or Windows/System32 (sysWOW64 on 64 bit Windows) directory, read the first post in this thread!

[Porsche_fan]

Thanks Groucho2004...I appreciate your reply!

I did read the first post of this thread multiple times and tried to follow it as best as I understood it. I think possibly my last question was misleading. The reason I asked about placing the MT version of avisynth.dll in the ffdshow folder is I read the following from another doom9 member who described one of the steps to setting up 2.6 MT with ffdshow as "- get avisynth 2.6 mt dll and put it in ffdshow directory". As I stated in my first post, I don't think the advice is correct but was asking.

1. You're running a single threaded version of Avisynth, that's why GetMTMode() is returning "0".

4. The file avisynth.dll (any) should ideally be placed in the Windows or Windows/System32 (sysWOW64 on 64 bit Windows) directory, read the first post in this thread!

As mentioned I am able to get multi-threading to work after modifing the script to GetMTMode(false) >= 0 ? distributor() : last. So based on this I believe I am running a MT version of avisynth.

However, I realize that mode=0 indicates the state of no multi-threading so possibly what you said has some truth. From the instructions I did the following...installed AVS 2.6.0 Alpha 3 [110525] then moved the avisynth.dll file from 2012.08.28 into syswow64 replacing existing avisynth.dll from the initial installation. If I did this incorrectly what should I have done?

2. If you're using MPC-HC, you should not be using "Distibutor()" at all, the call is added automatically. So, if you are using Avisynth MT you are unnecessarily increasing the number of threads.

3. "SetMTMode(1)" in your script. Why?

I have tried it without the distributor() call but couldn't get multi-threading to work with ffdshow in MPC-HC until it was added.

The reason I use the SetMTMode(1) and distributor() is because SEt said it was necessary on the first post of this thread and also in this post.

Appreciate the help!

[Groucho2004]

Originally Posted by Porsche_fan

From the instructions I did the following...installed AVS 2.6.0 Alpha 3 [110525] then moved the avisynth.dll file from 2012.08.28 into syswow64 replacing existing avisynth.dll from the initial installation. If I did this incorrectly what should I have done?

That sounds right to me.

Originally Posted by Porsche_fan

The reason I use the SetMTMode(1) and distributor() is because SEt said it was necessary on the first post of this thread and also in this post.

In that case ignore points 2. and 3. from my previous post. I have little experience with ffdshow so maybe someone else may be more helpful.

Either way, if GetMTMode() from your script returns "0", you are not using the correct (MT) Avisynth DLL. Make sure there is only one avisynth.dll on your system (in your syswow64 directory).

Edit: Do you really need Avisynth MT for FineSharp? As far as I know, it's quite fast.

[Porsche_fan]

Originally Posted by Groucho2004

Do you really need Avisynth MT for FineSharp? As far as I know, it's quite fast.

I agree I find my experience using singled threaded FineShap puzzling as you are right it is pretty fast. I always resize everything prior to applying FineSharp so that is an added factor.

DVD playback while using avisynth spline64 resizing before FineSharp has been trouble free. However, I have had differening results using FineSharp single threading with x.264 720/1080p with some instances of audio and video desynching. Desyncing seems to usually occur with x.264 encoded 720p @30fps...so far 24fps hasn't been an issue.

In some cases I've been able to solve the desync problem by having ffdshow decode but I prefer using the MPC AVC codec. I know ffdshow is suppose to be less CPU intensive when decoding but I can't think it should make that big a difference given the i7 processor.

I also had a strange issue with AVC in an avi container that had sync issues but it was fixed when I put it in a mkv container. Anyways, I've found that 2.6 MT allows the overhead/flexibility to use any codec without desync.

The really odd part to me is the graph pattern with or without multi-threading is the same in resource monitor for my i7 860 @2.8Ghz. In both cases 4 cores are parked and the other 4 cores that are active distribute the load equally and are at about 40% utilization. It could be that I have an issue with the way my computer is set up to process information. Would be interested in others thoughts.

EDIT: Don't know if this is the same for everyone else but after I uninstalled 2.5.8 I deleted the remaining folder labeled "AviSynth 2.5". I then installed 2.6 Alpha 3 but the new AviSynth 2.6 program is in a folder in Program Files (86) still labeled "AviSynth 2.5" not "AviSynth 2.6". I did use Version() to verify it is the 2.6 Alpha 3 build.

[LigH]

Some extremely annoying semi-freeware converters (ERightSoft SUPER, FormatFactory) are able to bend the hook to an own copy of an AviSynth DLL. But I believe the repeated re-installation of AviSynth versions should have repaired it, if that was a possible reason; still, better scan all your harddisk partitions for any copy.

[ryrynz]

Originally Posted by rean

Do not use SetMTMode(3). Use SetMTMode(3,4)!
mt_lutxy, mt_lut, FFmpegSource use a lot of memory. 8 threads will eat your memory faster.

Also 8 threads on i7 is useless. You will get slow execution, because most time the processor
will switch threads and wait for spin locks.

I use GraphstudioNext to benchmark my scripts (Psharpen, Awarpsharp, Vaguedenoiser etc) and on my 2600K with Hyperthreading I get a considerable performance improvement having 8 threads over 4.
My recommendation would be to test this for yourself.

[CraziFuzzy]

How close are you to rolling out an MT version of 2.6 Beta-4?

[LigH]

Last asked in #607; no reply from SEt yet.

[SEt]

It took longer than I expected, but here we go – updated version. Can't say I've tested everything, so let's keep 'stable' status on previous one for some time.

rean, please don't write nonsense. i7-930 with 8 threads is my work environment for the last 3 years and so you can say "native mode" of these builds. Also, if you are doing things right, FFmpegSource will consume exactly the same amount of memory regardless of number of threads used.