August 22, 2022
Host: Carol Payne
Secretary: Doug Walker
Attendees:
- Rémi Achard (TSC) - DNEG
Mark Boorer (TSC) - Industrial Light & Magic
Mei Chu (TSC) - Sony Pictures Imageworks
Sean Cooper (TSC ACES TAC Rep) - ARRI
Michael Dolan (TSC) - Epic Games
Patrick Hodoul (TSC) - Autodesk
John Mertic - Academy Software Foundation / Linux Foundation
Carol Payne (TSC Chair) - Netflix
Mark Titchener (TSC) - Foundry
Carl Rand (TSC) - Weta Digital
Doug Walker (TSC Chief Architect) - Autodesk
Kevin Wheatley (TSC) - Framestore
- Luan Bach - Foundry
OCIO TSC Meeting Notes
- OCIO Won The Pipeline Tool Award!
- Carol: We were honored to win The Pipeline Tool award out at DigiPro last week! We put together a video that was shown at the event, and we will post that.
- Doug: On my trip home, the TSA was struggling to understand what this heavy object with red, green, and blue glass spheres was. They needed to hand inspect it, LOL. I will send a photo once we receive the plaque for the front of it.
- Carol: We were honored to win The Pipeline Tool award out at DigiPro last week! We put together a video that was shown at the event, and we will post that.
- Siggraph Events
- Carol, Doug, and Thomas presented at the ASWF Open Source Days event, the video will be posted soon.
- Carol: Thomas posted several beta releases of the new Studio Config (the successor to the ACES 1.2 config). Have received some helpful feedback so far and are looking for more testing. Want to ensure it has everything it needs for the official release in October.
- Carol: We had a successful OCIO Meet-up and OpenBeerIO social event. Notes were taken and will be posted soon.
- Linux Foundation course
- Carol: John Mertic suggested we turn our OCIO course from Siggraph 2021 into a Linux Foundation course and I had a meeting with the person at the LF in charge of these. Sounds like we have a good head start relative to some of the other projects they've worked with. The LF has experts to consult on the preparation and review of the material. They have experience in course design we could learn from. They try to prepare the course in a way that it may be added to gradually over time. They suggested extending it with even some real introductory info such as signing the CLA and compiling the code. Doug and I will have a follow-up meeting to discuss further.
- Doug: Would be good to get their perspective on what format is most effective and what should be covered. Probably won't be able to do too much until after OCIO 2.2 is released.
- Carol: John Mertic suggested we turn our OCIO course from Siggraph 2021 into a Linux Foundation course and I had a meeting with the person at the LF in charge of these. Sounds like we have a good head start relative to some of the other projects they've worked with. The LF has experts to consult on the preparation and review of the material. They have experience in course design we could learn from. They try to prepare the course in a way that it may be added to gradually over time. They suggested extending it with even some real introductory info such as signing the CLA and compiling the code. Doug and I will have a follow-up meeting to discuss further.
- Timeline Performance Issue
- Mark: We've run into a performance issue trying to upgrade to OCIO v2 in our timeline-based products. We have a 4k 60fps timeline use-case where there are one or more CDLs changing on each shot, used in connection with one or more 3d-LUTs. The time to build/update the GPU shader and textures are preventing real-time playback at this speed.
- Michael: It would help if the CDL Op used dynamic parameters, which it doesn't currently.
- Doug: Similar use-cases have come up for discussion before, e.g. from Remi. The suggested work-around is to not try to do this as one huge Processor. The client app should separate out the part that is changing per shot (the CDL in this case) from the parts that are not, and build several Processors. The CDL part that is changing should be separate from the 3d-LUT part and thus be very lightweight and fast to update.
- Doug: Regarding the caching, OCIO caches the Processor, CPUProcessor, and GPUProcessor. But it does not try to cache the output of the GPUProcessor (e.g. the shader and textures) since it is more difficult to do that in an application agnostic way. Remi has logged an improvement request relating to caching LUT textures.
- Luan: There is a mem copy with the CDL that is very time-consuming, it involves a 65x65x65 3d-LUT, taking 1 ms. Not an issue on Mac for some reason.
- Kevin/Remi: Perhaps the config is using a 3d-LUT to convert into another color space to apply the CDL? Have seen similar slow mem copy but not sure why it would be taking that much time. Perhaps it is unaligned?
- Michael: A helpful way to debug issues is to convert the Processor into a GroupTransform and then use the write method to save into the CTF ("Color Transform Format") format. That file may be shared as a specific test-case.
- Carol: Please try the work-around Doug suggested and post an example transform as described by Michael.
- Adding getProcessor call for NamedTransforms
- Michael: I posted an issue on GitHub suggesting to add an easier way to create a Processor from a NamedTransform that better handles both directions.
- Doug: Sounds good.
- Studio Config items
- Carol: The Configs Working Group had been meeting weekly to try and hit our Siggraph deliverable, but I will change that back to bi-weekly. Next meeting will be on August 30th.
- Sean: I'm working on an update to my PR for the ARRI transforms. Will discuss more at the next WG meeting. What is the timeline?
- Carol: Let's discuss with Thomas but I think as soon as the ARRI transforms are in we will do another beta release. Want to ensure enough testing happens by October.
- Doug: Received an email from Canon too with an updated CLF. Carol: Let's follow-up after the meeting.