This was a long video, but as someone who is primarily a programmer I found it interesting and covering a lot of topics that I had no knowledge on (and so I had to to pause the video at a lot of...
This was a long video, but as someone who is primarily a programmer I found it interesting and covering a lot of topics that I had no knowledge on (and so I had to to pause the video at a lot of parts to look things up).
Learning that Linux (or whatever OS) doesn't control all of the computer, but instead this tiny part of the computer was news to me.
Cross SoC attacks are a scary and interesting idea.
As a sidenote, I read the description of the video and saw a brief note about Barrelfish, the multikernel OS. Which I thought was a fascinating idea for dealing with scaling operating systems to work with computers that have many many cores which don't have equal bandwidth access to all parts of the RAM.
I forgot I had something to say until you commented. We see a similar trend in high performance flash storage controllers, which are present perhaps only on modern game console SoCs and more...
I forgot I had something to say until you commented.
We see a similar trend in high performance flash storage controllers, which are present perhaps only on modern game console SoCs and more esoteric accelerator card SoCs.
Which is especially interesting as in Linux, F2FS is a great example of leveraging the increasing complexity of flash storage controllers. I don't know if this is a failing of Linux, however. Flash is a commodity and high performance controllers, in addition to their specialized nature, give the manufacturers a way to compete.
This also conflicts with the video, as unlike OS R&D, filesystems are both a historical staple of operating systems and are still an area of red-hot Linux development (and industry too).
Finally, seeing as the video mostly seems to give a call to action but provides few concrete starting points, I'd point out that DPUs are probably the perfect place to experiment with OS R&D for SoCs.
This was a long video, but as someone who is primarily a programmer I found it interesting and covering a lot of topics that I had no knowledge on (and so I had to to pause the video at a lot of parts to look things up).
I forgot I had something to say until you commented.
We see a similar trend in high performance flash storage controllers, which are present perhaps only on modern game console SoCs and more esoteric accelerator card SoCs.
Which is especially interesting as in Linux, F2FS is a great example of leveraging the increasing complexity of flash storage controllers. I don't know if this is a failing of Linux, however. Flash is a commodity and high performance controllers, in addition to their specialized nature, give the manufacturers a way to compete.
This also conflicts with the video, as unlike OS R&D, filesystems are both a historical staple of operating systems and are still an area of red-hot Linux development (and industry too).
Finally, seeing as the video mostly seems to give a call to action but provides few concrete starting points, I'd point out that DPUs are probably the perfect place to experiment with OS R&D for SoCs.