Dear all, I am working on a project based on a A53 SoC. The goal is for the CPU to receive and process data. I am running vanilla 5.4.9. Benchmarking memory performance shows improvements when using HugeTLBs. I managed to run with 4K base pages and 2MB Huge Pages specifying on the kernel command line how many pages to allocate, i.e. hugepages=512 default_hugepagesz=2m hugepagesz=2m. The ARM64 when using 4K pages should also support 1G Huge Pages. However, when specifying this: hugepages=1 default_hugepagesz=1G hugepagesz=1G I see the size properly taken into account in /proc/meminfo, but it does not manage to allocate them: # cat /proc/meminfo ….. HugePages_Total: 0 HugePages_Free: 0 HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize: 1048576 kB Hugetlb: 0 kB There is enough memory on the board. For instance if I declare a CMA area of 1G in the device tree it manges to reserve it properly at 0x40000000. Of course I remove CMA reservation when defining 1G Huge Pages. I am looking for the reason why this would fail, but I am not successful. Any hints will be highly appreciated. Thanks a lot and best regards Andreas
