NVIDIA Engineer Continues Working On Proactive Memory Compaction For Linux

[phoronix]

Phoronix: NVIDIA Engineer Continues Working On Proactive Memory Compaction For Linux

NVIDIA's Nitin Gupta continues working on proactive compaction for the Linux kernel's memory management code...

http://www.phoronix.com/scan.php?pag...ve-Mem-Compact

[Shnatsel]

Nitin Gupta is also the author of compcache and zram, so he's got 10+ years of kernel development experience under his belt.

[schmidtbag]

Pretty cool. A little bit of an odd thing for Nvidia to contribute toward, but still very welcome. I'm curious what kind of applications would benefit most from this, since that's not totally clear to me.

[ms178]

I guess managing huge amounts of memory with low latency becomes more important in the future, think of HBM on chiplets and not only for classical graphics but also for GPGPU tasks.

[theriddick]

There has been a long standing issue with NVIDIA drivers allocating memory correctly (or utilizing) for certain games, maybe this is something that will help with it down the line?¿

[dragorth]

Presumably this is helpful torward 3D workflows, such as TB scenes for Movie Studios, 100 GBs of images for AI/Photogammetry and others, and all the other GPU accelerated workloads Nvidia hopes we'll use their GPUs for.

[GreenReaper]

It'd be nice if Linux could improve slab handling with an eye towards fragmentatiin. I always end up with excessive inode tables and resulting memory fragmentation when I do an rsync of a large filesystem, because not all the slabs are returned