Slow mac build rt5


(peter) #1

Hello, after a few weeks of usage, I can’t say I’m totally happy with RT5 on my Mac (2016 iMac with 16 gb Ram). Performance is much slower than the previous build 4.2.1025 that I luckily kept. So I keep on returning to that version, that runs like a breeze.
There’s also the annoying thing that RT5 keeps opening in a small window on starting-up.
So : I hope there will be a fix pretty soon.
Second : maybe it would be a good idea to keep the latest version minus 1 avalable on the downloadpage.
Lately, as an alternative I’ve been playing around with Darktable. Allthough it’s a great piece of software, I’m more familiar with RT over the years and am convinced it can produce better results, specially on highlightrecovery. One thing Darktable does better imho is the startingpoint for my Nikon d5500-files. It seems I can’t make a same starting point in RT. Should be my knowledge or lack of it :slight_smile: If anybody has a profile here, I would be glad to hear.


(Ingo Weyrich) #2

Please tell more detailed where it’s slower:

Is it slower at starting rt?
Is it slower at processing files in queue?
Is it slower in preview?

Ingo


(peter) #3

Performance is significant slower performing all kinds of adjustments/tools. Slower on refreshing preview after an adjustment. Slower on 'open in… ’ other application.


(Ingo Weyrich) #4

Ok, thanks for feedback. Then there must be something wrong with the mac builds because the non-mac builds are not slower at processing than rt4.x.


(Ingo Weyrich) #5

@peterdb Please post the content of AboutThisBuild.txt in your rt5 folder


(peter) #6

It’s the 5,0-r1-gtk3 version.Version: 5.0-r1-gtk3

Branch: releases
Commit: 50114c1a
Commit date: 2017-02-01
Compiler: cc 8.0.0.8000042
Processor: generic x86
System: Apple
Bit depth: 64 bits
Gtkmm: V3.22.0
Build type: Release
Build flags: -stdlib=libc++ -std=c++14 -O3 -mmacosx-version-min=10.9 -std=gnu++11 -mtune=generic -Werror=unused-label -O3 -DNDEBUG
Link flags: -headerpad_max_install_names -mtune=generic
OpenMP support: ON
MMAP support: ON


(Ingo Weyrich) #7

@peterdb Can you try the one from this topic please? Support for Pentax pixel shift files #3489


(Morgan Hardwood) #8

More info needed.


#9

@peterdb, If you use the build from the discussion that @heckflosse suggested, please let me know how slow it is etc.

Also, as @Morgan_Hardwood requested, can you provided a screenshot when you use the one @heckflosse suggested?

TIA,
Partha


New to darktable
(peter) #10

Thanks. Making adjustments seems a bit faster.
Exporting to other application is still terribly slow.
On start-up still the half-size window. Resizing doesn’t change anything on next start-up.
See attached.


(Alberto) #11

@peterdb, can you post the AboutThisBuild for the fast version of RT please? Thanks!


(peter) #12

Version: 5.0-r1-gtk3-304-g209ea35
Branch: dev
Commit: 209ea35
Commit date: 2017-03-28
Compiler: cc 8.0.0.8000042
Processor: generic x86
System: Apple
Bit depth: 64 bits
Gtkmm: V3.22.0
Build type: Release
Build flags: -stdlib=libc++ -std=c++14 -O3 -mmacosx-version-min=10.9 -std=c++11 -mtune=generic -Werror=unused-label -O3 -DNDEBUG
Link flags: -headerpad_max_install_names -mtune=generic
OpenMP support: ON
MMAP support: ON


#13

@peterdb When you say “slow”, can you provide some metric like it, “it takes 5 minutes for the application to start an external editor” etc.?

Also, can you provide a description of your Mac; make, model etc.?

Thanks,
Partha


(Morgan Hardwood) #14

If I understood @peterdb correctly,

Slow:

Version: 5.0-r1-gtk3
Branch: releases
Commit: 50114c1a
Commit date: 2017-02-01
Compiler: cc 8.0.0.8000042
Processor: generic x86
System: Apple
Bit depth: 64 bits
Gtkmm: V3.22.0
Build type: Release
Build flags: -stdlib=libc++ -std=c++14 -O3 -mmacosx-version-min=10.9 -std=gnu++11 -mtune=generic -Werror=unused-label -O3 -DNDEBUG
Link flags: -headerpad_max_install_names -mtune=generic
OpenMP support: ON
MMAP support: ON

Still slow:

Version: 5.0-r1-gtk3-304-g209ea35
Branch: dev
Commit: 209ea35
Commit date: 2017-03-28
Compiler: cc 8.0.0.8000042
Processor: generic x86
System: Apple
Bit depth: 64 bits
Gtkmm: V3.22.0
Build type: Release
Build flags: -stdlib=libc++ -std=c++14 -O3 -mmacosx-version-min=10.9 -std=c++11 -mtune=generic -Werror=unused-label -O3 -DNDEBUG
Link flags: -headerpad_max_install_names -mtune=generic
OpenMP support: ON
MMAP support: ON

Fast:

Branch: master
Version: 4.2.1025
Changeset: dc4bbe906ba92ddc66f98a3c26ce19822bfb99ab
Compiler: clang-mp-3 3.9.0
Processor: generic x86
System: Apple
Bit depth: 64 bits
Gtkmm: V2.24.4
Build type: Release
Build flags: -w -std=gnu++11 -mtune=generic -Werror=unused-label -fopenmp=libomp -Werror=unknown-pragmas -O3 -DNDEBUG
Link flags: -mtune=generic
OpenMP support: ON
MMAP support: ON

@peterdb please report how long it takes to save this raw file to a JPEG using amsterdam.pef.pp3, in both the fast and slow build. While testing, keep the raw file and the output JPEG file on your local hard drive, not on a network drive.


(Morgan Hardwood) #15

@peterdb Please post the contents of AboutThisBuild.txt from 4.2.1025.


(Alberto) #16

Yes, that’s what I meant as well. TIA


(peter) #17

Branch: master
Version: 4.2.1025
Changeset: dc4bbe906ba92ddc66f98a3c26ce19822bfb99ab
Compiler: clang-mp-3 3.9.0
Processor: generic x86
System: Apple
Bit depth: 64 bits
Gtkmm: V2.24.4
Build type: Release
Build flags: -w -std=gnu++11 -mtune=generic -Werror=unused-label -fopenmp=libomp -Werror=unknown-pragmas -O3 -DNDEBUG
Link flags: -mtune=generic
OpenMP support: ON
MMAP support: ON


(peter) #18

Partha, it’s a 21 inch iMac (late 2015) 1,6 ghz intel core i5 with 16 gb ram / intel HD Graphics 6000 1536 MB / 250 gb ssd.


(Alberto) #19

i notice only now that in your slow versions there is no -fopenmp compilation flag. I suspect that is the culprit


(Morgan Hardwood) #20

I updated my post.

@agriggio that looks like it. Ping @partha @ion12

@peterdb please still post the requested timings.