Build fails on ARM64

[carl]

I'm not sure if it's using the right libopenjpeg. Can you go back into the openjpeg source directory and do

Code: Select all

sudo make install | tee log

Then attach log here?

[djkoelkast]

Code: Select all

[ 30%] Built target openjp2_static
[ 60%] Built target openjp2
[ 73%] Built target opj_dump
[ 86%] Built target opj_compress
[100%] Built target opj_decompress
Install the project...
-- Install configuration: "Release"
-- Up-to-date: /usr/local/lib/openjpeg-2.2/OpenJPEGTargets.cmake
-- Up-to-date: /usr/local/lib/openjpeg-2.2/OpenJPEGTargets-release.cmake
-- Up-to-date: /usr/local/lib/openjpeg-2.2/OpenJPEGConfig.cmake
-- Up-to-date: /usr/local/lib/pkgconfig/libopenjp2.pc
-- Up-to-date: /usr/local/include/openjpeg-2.2/opj_config.h
-- Up-to-date: /usr/local/lib/libopenjp2.so.2.2.0
-- Up-to-date: /usr/local/lib/libopenjp2.so.7
-- Up-to-date: /usr/local/lib/libopenjp2.so
-- Up-to-date: /usr/local/lib/libopenjp2.a
-- Up-to-date: /usr/local/include/openjpeg-2.2/openjpeg.h
-- Up-to-date: /usr/local/include/openjpeg-2.2/opj_stdint.h
-- Up-to-date: /usr/local/bin/opj_decompress
-- Up-to-date: /usr/local/bin/opj_compress
-- Up-to-date: /usr/local/bin/opj_dump

[carl]

OK maybe you can try

Code: Select all

LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH dcpomatic2

[djkoelkast]

This seems to work, it's encoding now (a bit slower than I was hoping for at 2.3fps but it works)

[Carsten]

Why did you choose this hardware for your build?

A few years back there were some threads about using a bunch of networked Raspberry PIs for encoding. However, at the time, the Raspberry PI 2B was the fastest PI available, and it was clear that, unless someone just had a few dozen of them lying around already, there was probably little benefit due to CPU/memory vs. network load imbalances.

Now today there are very fast and low cost AMD Ryzen CPUs available that will outperform all these solutions.

You could probably buy a second hand complete basic Ryzen machine on eBay well below 200€/$.

That said, out of sheer interest, I'd still be interested in seeing some benchmark numbers of DCP-o-matic running on a Raspberry PI 5.

- Carsten

[djkoelkast]

Why did you chose this hardware for your build?

A few years back there were some threads about using a bunch of networked Raspberry PIs for encoding. However, at the time, the Raspberry PI 2B was the fastest PI available, and it was clear that, unless someone just had a few dozen of them lying around already, there was probably little benefit due to CPU/memory vs. network load imbalances.

Now today there are very fast and low cost AMD Ryzen CPUs available that will outperform all these solutions.

You could probably buy a second hand complete basic Ryzen machine on eBay well below 200€/$.

That said, out of sheer interest, I'd still be interested in seeing some benchmark numbers of DCP-o-matic running on a Raspberry PI 5.

- Carsten

I had this lying around doing nothing, so I wanted to try it. My main PC is an 11th gen i7, it works great, but it will use 100% cpu for 3-4 hours most of the time (for a full length movie). So I thought what if. It works, but the same build would take over 24 hours.
I haven't measured if there is any difference in power consumption building the same movie, that would be interesting to know. It uses way less power, but longer. My builds aren't really time critical most of the time, I tend to make them 2 weeks up front if I can.

I wanted to try some other ARM64 too, like something with the RK3588, but they are a bit expensive for just a test.
I might be old fashioned or even a fanboy, but I would never trust AMD again after having so much trouble with their cpus in the past. Also AMD was never good to be used on video editing, and as that is what I have always done, I automatically fell to the Intel side.
I don't have room for a full size machine, but some mini PC of like a NUC would be possible. Preferably headless.

[Carsten]

All well - if you don't like AMD - they at least brought Intels high performance workstation cpu prices down to a similar price/performance level.