yes, i know about this, i wrote a miner for my own custom proof of work on a shitcoin upgrade chain and when i ran it as goroutines (these are not strictly parallelised) it was 20% slower than when i set up child processes to mine on a single kernel thread
20% was the difference, this is back in 2019, i'm sure that Go's coroutine scheduler is better than this now but i doubt it is as good as kernel threads
this is something you can do well with any language that has decent optimization and (heavy) kernel threads that do bulk work
coroutines give you low latency, not high throughput
mleku / mleku™️
npub1fj…mleku
2024-01-17 19:44:31
in reply to nevent1q…2lhl
Author Public Key
npub1fjqqy4a93z5zsjwsfxqhc2764kvykfdyttvldkkkdera8dr78vhsmmlekuPublished at
2024-01-17 19:44:31Event JSON
{
"id": "7aa2741c1a55d78cdf1fa4b5507cc19c144a570a71f122ec6390effc64250ca9",
"pubkey": "4c800257a588a82849d049817c2bdaad984b25a45ad9f6dad66e47d3b47e3b2f",
"created_at": 1705520671,
"kind": 1,
"tags": [
[
"e",
"804964a3218bfa004b3ba77ac9cfd430448aaf0bc1f8b4a598eb0da7ad6171db",
"",
"reply"
],
[
"p",
"0f28196ced1024be7ceb3d7b194337bf8f21e7ad0a8be803b1ec0a7c52e926de"
]
],
"content": "yes, i know about this, i wrote a miner for my own custom proof of work on a shitcoin upgrade chain and when i ran it as goroutines (these are not strictly parallelised) it was 20% slower than when i set up child processes to mine on a single kernel thread\n\n20% was the difference, this is back in 2019, i'm sure that Go's coroutine scheduler is better than this now but i doubt it is as good as kernel threads\n\nthis is something you can do well with any language that has decent optimization and (heavy) kernel threads that do bulk work\n\ncoroutines give you low latency, not high throughput",
"sig": "88e52ec83d096726a37dc759281065913adeae13d2c99a3af80b3e705e57d52c935773b30565564957ba95c9d6a72fe8972dcbc33ab41d0a36fc7fb39bf5c65f"
}