I am getting a roughly a 3.5X slowdown on render when I enable GPU processing in kdenlive under
Settings -> configure kdenlive -> playback
"Use GPU processing (Movit library)"

Somehow this doesn't seem right? What am I doing wrong?

Do I need to recompile something to make use of something or other?

A few details:
I am running:
Ubuntu 17.10
Kdenlive 17.08.2
MLT version 6.4.1
libmovit7: 1.5.3-1
nVidia binary driver: 384.90

From: sudo lshw - short

H/W path Device Class Description
====================================================
system MS-7641 (To Be Filled By O.E.M.)
/0 bus 760GM-P23(FX) (MS-7641)
/0/0 memory 64KiB BIOS
/0/4 processor AMD FX(tm)-6300 Six-Core Processor
/0/4/5 memory 288KiB L1 cache
/0/4/6 memory 6MiB L2 cache
/0/4/7 memory 8MiB L3 cache
/0/25 memory 8GiB System Memory
/0/25/0 memory 8GiB DIMM Synchronous 1600 MHz (0.6 ns)
/0/25/1 memory [empty]
/0/100 bridge RS780 Host Bridge
/0/100/2 bridge RS780 PCI to PCI bridge (ext gfx port 0)
/0/100/2/0 display GP107 [GeForce GTX 1050]
[snipage]

Hello,
GPU processing can be very efficient... provided that memory transfers (very time costly) are limited to minimal, requiring very careful design.
I believe Dan explained on MLT list that frames are copied between CPU/GPU as long as there needs a colorspace change (YUV<->RGB), or any CPU effect or transition (eg compositions via qtblend)...
If you are not in the optimal case, then GPU transfers actually degrade performance.
MLT potentially can benefit from Movit & other FFmpeg GPU operations, but it is currently very tricky.
Reworking our rendering pipe (enforcing safe conditions) has been in discussion, but it requires skills and time that we don't have at the moment