Twitter | Pretraživanje | |
Alexander Graf
142
Tweetovi
89
Pratim
174
Osobe koje vas prate
Tweetovi
Alexander Graf 2. velj
Odgovor korisniku/ci @fagiolinux @openlabbott
It might be rainy, but it's surprisingly warm! No slush this year around!
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 2. velj
Odgovor korisniku/ci @apellegr
That means you get reimbursed now, thanks to EU regulations. Oh.
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 31. sij
Odgovor korisniku/ci @Intel_Brian @AMI_PR
Awesome, see you there then!
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 31. sij
Odgovor korisniku/ci @Intel_Brian @AMI_PR
Are you going to FOSDEM?
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 30. sij
Odgovor korisniku/ci @_msw_
😢
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 29. sij
Odgovor korisniku/ci @haerwu
All the Pine64 (and Rock64) products except for the very first one have SPI flashes, they are just empty. How about you foster change and provide UEFI enabled U-Boot images and flashing tools for them?
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 28. sij
Odgovor korisniku/ci @jonmasters @haerwu i 3 ostali
I'm sure as a VP at Arm you would quickly find yourself without a job then, because there is no money left to pay you 😉
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 27. pro
Odgovor korisniku/ci @olofj @_msw_ @cperciva
Yes :)
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 23. pro
Odgovor korisniku/ci @never_released
It's supposedly an option that's viable enough for commercial compilers even:
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 22. pro
Odgovor korisniku/ci @blueCmd @OrangeCMS
With these high perf things, it's rarely about legacy. Most bits rather are that way because of cache line granularities, DMA request dependencies (read: latency) and efficient multi-queuing (many CPUs, many memory channels). ASIC limits apply to you too, no?
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 21. pro
Odgovor korisniku/ci @blueCmd @OrangeCMS
Looking forward to it :). You may also just for the sake of it explore a few high perf network ring designs, such as i40e, mlx5 and ena. Always good to learn from mistakes others already made.
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 21. pro
Odgovor korisniku/ci @blueCmd @OrangeCMS
Usually, yes. You also try to limit the number of individual (non-consecutive) DMA requests. For offloading, you also may need additional smartness, like writing metadata to a different target range than the actual payload.
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 21. pro
Odgovor korisniku/ci @blueCmd @OrangeCMS
How fast is your NIC? 100 entries will overflow with high perf network I/O.
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 13. pro
Odgovor korisniku/ci @WhatAintInside @never_released
You don't need one. Apfs comes with its uefi driver binary readily included:
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 5. pro
Odgovor korisniku/ci @WingsOfMimoja
Your Africa picture is missing some game. Let me fix that for you :)
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 7. stu
Odgovor korisniku/ci @fagiolinux
So after that comes secure nested virtualization? 😅
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 23. lis
Odgovor korisniku/ci @WhatAintInside
Just limit the memory map to the first GB to check?
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 10. lis
Odgovor korisniku/ci @bruggems @WhatAintInside i 3 ostali
With metal instances you can run ESX for ARM on AWS ;).
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 28. ruj
Odgovor korisniku/ci @AviKivity
Probably just yet another case of "we control it or we try to kill it"
Reply Retweet Označi sa "sviđa mi se"
Alexander Graf 11. ruj
Odgovor korisniku/ci @hugelgupf @winocm
Just do it? :)
Reply Retweet Označi sa "sviđa mi se"