Avisynth+

[StainlessS]

Yep:- http://www.mediafire.com/file/r37y2m...o/SVP.zip/file
(32 bit only)

EDIT: Maybe I imagine it, but did SVP start requiring their own version of AVS+ at some point (I never really paid much attention).

[Selur]

Sadly those files also give me:
---------------------------
SVSmoothFps: unable to init GPU-based renderer [code 0x10000]
---------------------------
I used:

LoadCPlugin("I:\Hybrid\32bit\AVISYN~1\ffms2.dll")
LoadPlugin("I:\Hybrid\32bit\AVISYN~1\svpflow1.dll")
LoadPlugin("I:\Hybrid\32bit\AVISYN~1\svpflow2.dll")
Import("I:\Hybrid\32bit\avisynthPlugins\InterFrame2.avsi")
SetFilterMTMode("DEFAULT_MT_MODE", MT_MULTI_INSTANCE)
# loading source: F:\TestClips&Co\files\test.avi
#  input color sampling YV12
#  input luminance scale tv
FFVideoSource("F:\TESTCL~1\files\test.avi",cachefile="E:\Temp\avi_078c37f69bb356e7b5fa040c71584c40_41_1_0.ffindex",fpsnum=25)
# current resolution: 640x352
InterFrame(GPU=true,NewNum=60,NewDen=1,Cores=32)
# filtering
PreFetch(16)
return last

When using AvisynthMT with:

SetMemoryMax(768)
SetMTMode(5,16) # changing MT mode
LoadCPlugin("I:\Hybrid\32bit\AVISYN~1\ffms2.dll")
LoadPlugin("I:\Hybrid\32bit\AVISYN~1\svpflow1.dll")
LoadPlugin("I:\Hybrid\32bit\AVISYN~1\svpflow2.dll")
Import("I:\Hybrid\32bit\avisynthPlugins\InterFrame2.avsi")
# loading source: F:\TestClips&Co\files\test.avi
#  input color sampling YV12
#  input luminance scale tv
FFVideoSource("F:\TESTCL~1\files\test.avi",cachefile="E:\Temp\avi_078c37f69bb356e7b5fa040c71584c40_41_1_0.ffindex",fpsnum=25)
# current resolution: 640x352
SetMTMode(2) # changing MT mode
InterFrame(GPU=true,NewNum=60,NewDen=1,Cores=32)
distributor()
return last

everything works. (Side note: when using 'GPU=False' it works in both)

[StainlessS]

SubJunk has later version script (v2.8.2):- https://forum.doom9.org/showthread.php?t=160226

EDIT: And I've never used it much, and never with GPU=True.

[Selur]

Tried that one with the binaries it comes with.
Same result, GPU=True works fine with AvisynthMT, doesn't with Avisynth+.

never with GPU=True.

That explains why it works for you.

[wonkey_monkey]

Originally Posted by StainlessS

Is there some clip/colorspace property that denotes X and Y granularity for cropping, eg YV12 may crop on even boundaries both x and y,
and for YV411 multiple of 4 for x and 1 for y ?

vi.GetPlaneHeightSubsampling(PLANAR_U);
vi.GetPlaneWidthSubsampling(PLANAR_V);

Is that it?

[pinterf]

Originally Posted by StainlessS

Is there some clip/colorspace property that denotes X and Y granularity for cropping, eg YV12 may crop on even boundaries both x and y,
and for YV411 multiple of 4 for x and 1 for y ?

Or does one have to inquire eg "Is420" and deduce that has same modulo requirements as YV12.

EDIT: Or extract Y plane and U plane and calculate it manually (last resort).

No. I supposed that GetPlaneWidthSubSampling was available but is wasn't.

[StainlessS]

Thanks Wonkey, I know how to do that in C, was wantin' script solution.
EDIT: I usually work it out manually in C, as GetPlaneWidthSubSampling() not available in v2.58 (I think).

I have this in RT_stats

RT_ColorSpaceXMod(clip)
 Return int, the natural cropping XMod for clip colorspace, eg YV411=4, YV12=2, YUY2=2, RGB=1
 v2.5 plugin dll limited to v2.58 colorspaces.

***
***
***

RT_ColorSpaceYMod(clip,bool "Laced"=true)
 Return int, the natural cropping YMod for clip colorspace, eg YV411=1, YV12=2, YUY2=1, RGB=1
 Laced, bool. Default true. If Laced==true, returns doubled YMod.
 v2.5 plugin dll limited to v2.58 colorspaces.

EDIT: Thanx Pinterf, I guess its coming

[pinterf]

Originally Posted by Selur

Sadly those files also give me:
---------------------------
SVSmoothFps: unable to init GPU-based renderer [code 0x10000]
---------------------------

Setting the same SetMemoryMax(768) for the avs+ script helps?

[StainlessS]

I usually use InterFrame (seldomly) via a much larger script, and with GPU=False and cores=1, [some machines I use with same script are single core without GPU]
however, just tried

AviSource("D:\B.avi")
Interframe(GPU=True,NewNum=60000,NewDen=1001,cores=32)

works ok.
I only have quad core(Core 2 Q9550) , and crap GPU (Nvidia GT 520, my Dell Optiplex 780 limited to 6.5 inch cards, and total PSU only 300Watt, GT 520 5.5 inch, full load 29Watts).
No idea if it is actually using the GPU, or not.

EDIT: OK, we do seem to be using the GPU


EDIT: For me, CPU is about 40% faster than GPU.
EDIT: In above graphic, shows GPU load at 8.0%, also memory controller load=11.0%, both seem a bit underused for some reason.
EDIT: And CPU usage about same whether or not GPU=True.

[wonkey_monkey]

Since we're talking about GetPlaneWidthSubsampling, it looks like there's an error on the Wiki (http://avisynth.nl/index.php/Filter_...ghtSubsampling). YV16 should be 1/1 and YV411 should be 2/0. I think there are errors in both the table and the text.

I'd create an account and fix it myself but HTTPS doesn't work (which it really should in this day and age).

[pinterf]

Originally Posted by wonkey_monkey

Since we're talking about GetPlaneWidthSubsampling, it looks like there's an error on the Wiki (http://avisynth.nl/index.php/Filter_...ghtSubsampling). YV16 should be 1/1 and YV411 should be 2/0. I think there are errors in both the table and the text.

I'd create an account and fix it myself but HTTPS doesn't work (which it really should in this day and age).

YV16 is 1/0

[StainlessS]

YV16 should be 1/1

YV16 should be 1/0 [text is back to front, but table is correct, methinks]

YV411 should be 2/0

agreed, table is wrong

EDIT:

Text is back to front

YV16: GetPlaneWidthSubsampling(PLANAR_U) = 0 // since there is no horizontal subsampling on a chroma plane
YV16: GetPlaneHeightSubsampling(PLANAR_U) = 1 // since vertically there are two times less samples on a chroma plane compared to a plane which is not subsampled

[Selur]

Setting the same SetMemoryMax(768) for the avs+ script helps?

Sadly no.

works ok.

simplifying the script to

LoadCPlugin("I:\Hybrid\32bit\AVISYN~1\ffms2.dll")
LoadPlugin("I:\Hybrid\32bit\AVISYN~1\svpflow1.dll")
LoadPlugin("I:\Hybrid\32bit\AVISYN~1\svpflow2.dll")
Import("I:\Hybrid\32bit\avisynthPlugins\InterFrame2.avsi")
FFVideoSource("F:\TESTCL~1\files\test.avi",cachefile="E:\Temp\avi_078c37f69bb356e7b5fa040c71584c40_41_1_0.ffindex",fpsnum=25)
InterFrame(GPU=True,NewNum=60,NewDen=1,Cores=32)

doesn't help. Only thing that works is switching from Avisynth+ to old AvisynthMT. First thought it to be a driver issue, but threw that assumption out once I realized that it works with 32bit Avisynth MT.

[wonkey_monkey]

Originally Posted by pinterf

YV16 is 1/0

Well I blame my error on the whole stupid 4:2:2 system!

[Groucho2004]

Originally Posted by Selur

Sadly no.

simplifying the script to

LoadCPlugin("I:\Hybrid\32bit\AVISYN~1\ffms2.dll")
LoadPlugin("I:\Hybrid\32bit\AVISYN~1\svpflow1.dll")
LoadPlugin("I:\Hybrid\32bit\AVISYN~1\svpflow2.dll")
Import("I:\Hybrid\32bit\avisynthPlugins\InterFrame2.avsi")
FFVideoSource("F:\TESTCL~1\files\test.avi",cachefile="E:\Temp\avi_078c37f69bb356e7b5fa040c71584c40_41_1_0.ffindex",fpsnum=25)
InterFrame(GPU=True,NewNum=60,NewDen=1,Cores=32)

doesn't help. Only thing that works is switching from Avisynth+ to old AvisynthMT. First thought it to be a driver issue, but threw that assumption out once I realized that it works with 32bit Avisynth MT.

I can't reproduce this problem. Can you upload the exact files you're using (ffms2.dll, svpflow*.dll, InterFrame2.avsi)?

[Selur]

here you go: https://drive.google.com/open?id=1w3...f5V4r8EA9goez8
using a GeForce GTX 1070 Ti with 418.81 drivers and a Amd Ryzen 7 1800X.

[pinterf]

Meanwhile a question regarding "Layer". I have the YUV stuff made ready but I experienced one little problem.

Layer mode: darken and ligthen is giving different results for RGB32 and YUY2 when Threshold>0

Even classic Avisynth 2.6.0.5 is giving different results for me.

Wiki:
http://avisynth.nl/index.php/Layer

E.g.: lighten:
"Copy overlay_clip over base_clip in areas where overlay_clip is lighter by threshold. Performs the same operation as add, but only when overlay_clip is BRIGHTER than base_clip."

And the relevant code from Avisynth+

Rgb32:

      if constexpr(mode == LIGHTEN) {
        alpha = luma_ovr > thresh + luma_src ? alpha : 0;
      } else {
        alpha = luma_ovr < thresh + luma_src ? alpha : 0
      }

YUV:

      if constexpr (mode == LIGHTEN) {
        alpha_mask = (thresh + ovr) > src ? level : 0;
      }
      else {
        alpha_mask = (thresh + src) > ovr ? level : 0;
      }

The variable names are different a bit but one can see the difference. Sure, they are equivalent when thresh is 0.

Which computation covers the text definition?

Script to reproduce:

x=some source
x = x.Spline16Resize(512,256)
ovr=ColorbarsHD().Spline16Resize(x.width,x.height)
clipRgb = x.ConvertToRGB32(matrix = "PC.709").ResetMask()
ovrRgb = ovr.ConvertToRGB32(matrix = "PC.709").ResetMask()
clipYuy2 = x.ConvertToYUY2() 
ovrYuy2 = ovr.ConvertToYUY2() 

threshold=255 # 0 is O.K.
# which one works fine?
testRgb = Layer(clipRgb, ovrRgb, "lighten", level=128, Threshold=threshold)
testYuy2= Layer(clipYuy2, ovrYuy2, "lighten", level=128, Threshold=Threshold)

StackVertical(testRgb.ConvertToYV24(matrix = "PC.709"), testYUY2.ConvertToYV24())

[Groucho2004]

Originally Posted by Selur

here you go: https://drive.google.com/open?id=1w3...f5V4r8EA9goez8
using a GeForce GTX 1070 Ti with 418.81 drivers and a Amd Ryzen 7 1800X.

Using your files I can reproduce the problem on Win7 (in a VM -> VMWare generic graphics driver). Also, it crashes with every (2.6) Avisynth version, not just AVS+.

It works fine on XP/XP64 (GTX750, 350.12 driver).

[Selur]

Okay, for me it doesn't crash with Avisynth MT 2.6 (https://drive.google.com/open?id=1C0...Y7On5-_cLCaMj4), strange it crashes for all Avisynth 2.6 for you.

Cu Selur

[StainlessS]

Which computation covers the text definition?

"Copy overlay_clip over base_clip in areas where overlay_clip is lighter by threshold.

I would say this one.

// Rgb32:

      if constexpr(mode == LIGHTEN) {
        alpha = luma_ovr > thresh + luma_src ? alpha : 0;
      } else {
        alpha = luma_ovr < thresh + luma_src ? alpha : 0   // EDIT: Although I think this also wrong, see later.
      }

EDIT:

isn't this for YUV lighten

        alpha_mask = (thresh + ovr) > src ? level : 0;

equiv to this [if so then methinks obviously wrong]

        alpha_mask = ovr > (src - thresh) ? level : 0;

EDIT: And is this correct for darken

Copy overlay_clip over base_clip in areas where overlay_clip is darker by threshold.

// Rgb32:

      if constexpr(mode == LIGHTEN) {
        alpha = luma_ovr > thresh + luma_src ? alpha : 0;
      } else {
        alpha = luma_ovr < thresh + luma_src ? alpha : 0

      }

Should it be

alpha = luma_ovr <  (luma_src - thresh) ? alpha : 0

EDIT: Or in full as here: (3 out of the 4 original lines look messed up to me)

//RGB:
      if constexpr(mode == LIGHTEN) {
        alpha = luma_ovr > (luma_src + thresh) ? alpha : 0;    // As original
      } else { // EDIT: DARKEN
        alpha = luma_ovr < (luma_src - thresh) ? alpha : 0
      }


//YUV:
      if constexpr (mode == LIGHTEN) {
        alpha_mask = ovr > (src + thresh) ? level : 0;
      }
      else { // EDIT: DARKEN
        alpha_mask = ovr < (src - thresh) ? level : 0;
      }

EDIT: I remember in some thread, that RaffRiff42 got some weird results from Layer, and could not figure out why, maybe this was it.