Tomas Vondra / is on Nostr

bug developer, technical debt contributor

Last Notes

2024-05-26 09:34:44

- reply

@npub1809…nd7t Ah, right, I switched the core numbers. Anyway, it seems 10 is somehow exceptional. Could it be that it's a CPU manufacturing issue but it was good enough to pass the acceptance tests after manufacturing?

2024-05-26 09:34:44

- reply

2024-05-26 09:22:47

- reply

@npub1809…nd7t So all other cores perform like 10, and 11 is slower?

2024-05-26 09:22:47

- reply

@npub1809…nd7t So all other cores perform like 10, and 11 is slower?

2024-05-26 09:13:32

- reply

@npub1809…nd7t It's interesting the example is for core 10 -> 11, and this is a 10 core CPU. I wonder if that's related, somehow. Does the same thing happen for other core #?

2024-05-26 09:13:32

- reply

@npub1809…nd7t It's interesting the example is for core 10 -> 11, and this is a 10 core CPU. I wonder if that's related, somehow. Does the same thing happen for other core #?

2024-03-02 17:44:41

Sometimes I wonder if someone in the Google Sheets admin team is looking at monitoring and wondering "WTF is this looney doing, loading CSVs with 300k rows into a sheet and trying to do a pivot table on it?"

2024-01-07 12:22:30

- reply

@npub1m26…2jhh That seems to recommend to *disable* prefetch, which may be appropriate for some workloads, but it's the opposite of what I'm trying to achieve.

2024-01-07 11:06:17

Any ZFS connoisseurs here? What's the right way to enable & tune read-ahead / prefetch? Something like "blockdev --setra" on Linux. Doing some benchmarking and can't get it close to the traditional filesystems for sequential reads ...