Longhorn / is on Nostr

Kernel/hypervisor engineer, Amazon EC2. 

Blog: https://threedots.ovh/blog/

Last Notes

2024-08-26 06:32:53

- reply

now that's funny - u-boot on X Elite should be funny too

2024-08-26 06:31:31

how often do you all rely on nested virtualisation? Is it to make things easier or is it an actual dependency for your workflow?
And is it for client/development machine usage or for some production use cases too?

2024-08-25 01:53:32

hot take: I think launching the original Snapdragon 7c on a Windows platform was a bad idea, especially with 4GB of RAM devices

2024-08-21 04:56:35

question: is an ecore clocked at 5GHz still an ecore

2024-08-19 21:36:47

- reply

if you had to choose between edk2 or RTKit what would you pick

2024-08-19 21:13:52

uefi as an operating system on cortex-m with preemptive scheduling

2024-08-11 17:39:47

so
is 50% return rates good
for snapdragon windows laptops

2024-08-10 18:46:37

Making RSA faster on Graviton2: https://www.amazon.science/blog/formal-verification-makes-rsa-faster-and-faster-to-deploy

2024-06-02 22:16:22

- reply

oh, and as a related reminder, NV GPUs have no blockers on apple silicon macs when running Linux.
Apart from that Vulkan quirk, SW issues with PCIe non-mappable as normal NC don't cause issues on NVIDIA hw for those systems.

2024-06-02 22:13:50

gathered from x.com:
nvidia actually exposes a GPU upload heap (mapped BAR to user-mode) in Vulkan on Ampere Altra in their drivers.
This caused app breakage for Tellusim as Ampere Altra forces mapping BARs as device mem, instead of normal NC, for things to work properly. Device mem doesn't allow unaligned accesses.
It's a quirk that Apple Silicon Macs also have, and I'm tempted to say that perhaps NV should just not expose a GPU upload heap on Altra. After all, D3D12 didn't need it for ages.