[FDisk80]

Originally Posted by Gavino

To continue Didée's theme: "slow" is bad. But "crashing" is even worse.

Most source filters don't work multithreaded, so you have to run them in mode 5, before switching to mode 2 for the rest of the script.

Well, it still crashes in mode 5.
The script is just to heavy for realtime 1080p even for a quad core CPU at 3.6Ghz and GPU acceleration + DXVA enabled.
Even at "Very Fast" and "Super Fast" presets.

BTW,
SVP sets "SetMemoryMax(1024)" and "SetMTMode(3,7)" by default, did not see it crash once. CPU load is much much lower and GPU load is the same, about 40%
Some settings are just overkill in InterFrame for realtime.

svp_scheduler=true
SetMemoryMax(1024)
svp_cache_reduce=true
svp_flow_lib="C:\Program Files (x86)\SVP\Plugins\MVTools2\libflowgpu.dll"
multinum=5
multiden=2
algo=13
block=false
ml=0
search=5
sadml=100
sadgamma=1
blend=true
dist=2
pel=2
sharp=0
Recalc=0
blkh=32
blkv=32
overlap=8
overlapv=8
wid_den=1
hei_den=1


LoadPlugin ("C:\Program Files (x86)\SVP\Plugins\MVTools2\mvtools2.dll")
SetMTMode(3,7)
ffdShow_source()
Assert(width==1920 && height==816,"Frame size "+string(width)+"x"+string(height)+" instead of 1920x816")
delim=":"
WriteFileStart("C:\Users\FDisk\AppData\Roaming\SVP\Logs\avsFPS.txt", "FramerateNumerator()", "delim", "FramerateDenominator()")
SetMTMode(2)
src=last
wid_den==1 ? last : \
hei_den==1 ? LanczosResize(int(width/wid_den/2)*2,height) : LanczosResize(int(width/wid_den/2)*2,int(height/hei_den/2)*2)
super=MSuper(pel=pel, hpad=blkh, vpad=blkv, sharp=sharp)
finest=pel==1 ? super : MFinest(super)
backward_vec=MAnalyse(super, isb=true, blksize=blkh, blksizev=blkv, overlap=overlap, overlapv=overlapv, search=search, searchparam=dist, pelsearch=dist, plevel=0)
forward_vec=MAnalyse(super, isb=false, blksize=blkh, blksizev=blkv, overlap=overlap, overlapv=overlapv, search=search, searchparam=dist, pelsearch=dist, plevel=0)
backward_vec = Recalc==0 ? backward_vec : MRecalculate(super, backward_vec, blksize=blkh2, blksizev=blkv2, overlap=int(overlap/4)*2, overlapv=int(overlapv/4)*2, search=search, searchparam=dist)
forward_vec  = Recalc==0 ? forward_vec  : MRecalculate(super, forward_vec, blksize=blkh2, blksizev=blkv2, overlap=int(overlap/4)*2, overlapv=int(overlapv/4)*2, search=search, searchparam=dist)
smooth_video=MSmoothFps(super, backward_vec, forward_vec, finest=finest, num=FramerateNumerator(last)*multinum,den=FramerateDenominator(last)*multiden, algo=algo, ml=ml, sadml=sadml, sadgamma=sadgamma, block=block, blend=blend)
smooth_video


distributor()