<syntheticbird> The type of shit ofrn would drop only in a full moon in august during a leap year > <@ofrnxmr> I misse meeting, but i found a 6x speedup in fcmp wallet sync

<ofrnxmr> 10x blocks from 3020000 - 3030000 = 13m stock, 2m14s with 6x speedup (like 50loc)

<ofrnxmr> All of those complaints / worries about wallet sync speed on fcmp can go byebye now

<syntheticbird> @ofrnxmr: wait until monerod aggregate throughput becomes the bottleneck

<jpk68:matrix.org> Doesn't the thing only go from zero

<ofrnxmr> What thing

<ofrnxmr> @syntheticbird: You mean, multiple readers/clients/wallets?

<jpk68:matrix.org> The wallet sync speedup

<ofrnxmr> @jpk68:matrix.org: The 6x one im talking about here, is a bugfix

<ofrnxmr> Not my secret, proprietary one

<ofrnxmr> That one does the sync in 22s 😆

<ofrnxmr> And no, doesnt require from 0

<syntheticbird> yes > <@ofrnxmr> You mean, multiple readers/clients/wallets?

<ofrnxmr> @syntheticbird: tbh, i havent tested multiple but i am using jpk's lock pr

<ofrnxmr> I lie, i did test multiple but dont rememver the results. iirc it was like 40% slower when doing 2 wallets at once, but maybe 1-2% slower now?

<ofrnxmr> This was 2 identical wallets (with history) syncing the full chain at the same time

<jpk68:matrix.org> The database locking PR works a lot better when you also increase the RPC server threads from 2, IIRC

My GPU wallet scanner does sync in 8 seconds on RTX 5070 (full chain scan), try to beat that :P

<ofrnxmr> sech1: Sar, where is the data stored

Preprocessed, on GPU. All required data fits into ~8.5 GB VRAM

<syntheticbird> benchmaxxing

But streaming the data is also possible, it only requires 1 GB/s from PCIe

<syntheticbird> goodmaxxing

<ofrnxmr> Well, i doubt im going to beat your preprocessed gpu numbers - my gpu is an rx570

<ofrnxmr> Missing a T and a 0

<jpk68:matrix.org> @jpk68:matrix.org: Nevermind, I was thinking of the read-write lock one

13:48:47 <br-m> <ofrnxmr> All of those complaints / worries about wallet sync speed on fcmp can go byebye now

I was syncing 100+ wallets at once so I knew it was fast, just... not using monero own code :)

sech1: wen carrot scanning :)