Hello,
with OM6.17,
I’m experiencing that the rendering of graphical elements such as editors windows of BPF is much more slower on my Linux-Ubuntu bionic - Mate 8cores 8Go RAM NVIDIA 17’’ laptop than on my old 17’’ macbook pro 4cores 8Go 10.8.5, with same OM workspace, same patches.
Does someone has better performances of graphical rendering with some other Window Manager ?
Or is it a general LW graphical Linux issue ?
Bests.
Dear Fred Voisin,
Please can you send your benchmark tests with a complete specification (cpu, RAM, window managers, etc.) of your working environments for each station in order to apprehend the difference of performance?
Thank you very much
Best regards.
Dear Karim,
Thank you for your attention.
Let say,
A = MacBook Pro 17’’ , Intel Core i7 @ 2.66 GHz,
RAM 8 Go @1067 MHz DDR3
NVIDIA GeForce GT 330M 512 Mo
runs OSX 10.8.5
(23:09 up 153 days, 4:46, 20 users, load averages: 1,32 1,00 0,97)
B: Asus (ROG) 17’’, IntelCore i7-6700HQ 2.6 GHz (3.5 GHz)
RAM 8Go @2133 MHz DDR4
NVIDIA GeForce GTX 950M 2 Go PCI-Express x16 Gen 3 (NVIDIA driver 390.138 latency = 0 Xorg server v.11)
runs Linux 4.15 Ubuntu bionic (18.04) with Cinnamon 3.6.7
(23:10:37 up 1 day, 2:49, 3 users, load average: 0,07, 0,15, 0,17)
The CPU benchmark with same patch same data (MIR from SDIFwithout graphic events),
with A:
User time = 43.764
System time = 0.277
Elapsed time = 43.716
Allocation = 63941173320 bytes
7675 Page faults
Calls to EVAL 211485549
with B:
User time = 33.961
System time = 0.015
Elapsed time = 32.393
Allocation = 67922456624 bytes
0 page faults
Calls to EVAL 222994424
I have morer difficulties to evaluate GPU side or the User graphic events from Window managers.
Let manipulate in OM a MULTI-BPF in a patch with 54 BPFs made of 664 y values each.
On the Mac (A), opening (double-click) and rendering the multi-bpf take less than 2 seconds.
Activating/deactivating lines takes one small second.
On the Linux ROG (B),
Openning/rendering same the multi-bpf takes twice the time.
Activating/deactivating lines takes a big second.
the NVIDIA driver tell me that it’s running pic GPU from 3% to mostly 55% with some peaks up to 75%.
When opening a single BPF with 15000 y values, the Mac takes 4 seconds to render.
The Linux takes 10 seconds and, in the run, apart the BPF window, a curious empty “Untitled Lispworks Interface - #” opens over which can not be closed (cf. screenshot).
And finally, as soon as the BPF windows are opened, zooming takes more a less the same amount of time to be rendered on both A and B.
So, CPU side the newer (linux) is faster than the older (mac).
On GPU side, the older (mac) with smaller GPU is twice faster than the newer (linux) with much bigger and faster GPU (and propriatory driver).
And OM6.17 may crash (without quiting) on Linux but never on Mac (same OM patch).
Does it help ?
Best
K
I forgot to say that OM6.17 was running on the Linux and OM6.15 on the Mac.
With OM6.17 and 0M6.16 on mac, opening the 15000 values on the bpf may takes one short second more than 6.15.
Activating/deactivating lines int the BPF is much more longer (4 to 5 seconds) on OM6.16 and OM6.17 Mac than OM6.15 Mac (1 to 2 seconds).
Not yet tryed OM6.16 on Linux: will I be able to downgrade and upgrade easily with .deb using dpkg ?
Bests.
the BPF exported as an OM instance below attached.
15000BPF.omi (148.0 KB)
Ok it is good to know.
This issue is not related to machines. It is related to ports (linux/mac) and BPF code and will be dealt with. But it is surely not a priority for the time being. It is not a big issue and is “optimizable”. thank you for reporting.
Best
K
You are welcome 
I can manage: CPU on linux and graphics on Mac, as usual 
Thank’s to you.
Bests.
Fred
In case of, at least, my last benchmark with a newer mac, with
C = macbook pro 15’’ Quad-Core Intel Core i7 3,1 GHz
RAM 16 GB 2133 MHz LPDDR3
Intel HD Graphics 630 1536 MB
Catalina 10.15.6
OM6.17
CPU with same patch (MIR) is… 90 times faster than a mac 7 years older (greater than Moore’s law
User time = 0.382
System time = 0.036
Elapsed time = 0.380
Allocation = 239732184 bytes
1052 Page faults
Calls to EVAL 1711562
but GPU a little bit faster than old NVIDIA on Mac (1 short second less to render teh 15000 dots BPF).
Cheers.
