azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | https://github.com/glscopeclient/scopehal-apps | Logs: https://libera.irclog.whitequark.org/scopehal
<d1b2> <johnsel> They should be sending me a mso8k tomorrow-ish, if all goes to plan I can try to set up a more direct line of communication for things like this
<azonenberg> Nice
bvernoux has quit [Read error: Connection reset by peer]
veegee has quit [Quit: My iMac has gone to sleep. ZZZzzz…]
Degi has quit [Ping timeout: 260 seconds]
Degi has joined #scopehal
veegee has joined #scopehal
bvernoux has joined #scopehal
veegee has quit [Quit: My iMac has gone to sleep. ZZZzzz…]
<_whitenotifier-6> [scopehal-apps] vincentbaeten opened issue #597: ngscopeclient: Black screen and unresponsive when starting on Ubuntu 22.04 - https://github.com/glscopeclient/scopehal-apps/issues/597
<_whitenotifier-6> [scopehal-apps] vincentbaeten commented on issue #597: ngscopeclient: Black screen and unresponsive when starting on Ubuntu 22.04 - https://github.com/glscopeclient/scopehal-apps/issues/597#issuecomment-1611203073
<_whitenotifier-6> [scopehal-apps] bvernoux commented on issue #597: ngscopeclient: Black screen and unresponsive when starting on Ubuntu 22.04 - https://github.com/glscopeclient/scopehal-apps/issues/597#issuecomment-1611232075
<_whitenotifier-6> [scopehal-apps] bvernoux edited a comment on issue #597: ngscopeclient: Black screen and unresponsive when starting on Ubuntu 22.04 - https://github.com/glscopeclient/scopehal-apps/issues/597#issuecomment-1611232075
jacekowski has quit [Server closed connection]
jacekowski has joined #scopehal
mandl has joined #scopehal
mandl has quit [Quit: Connection closed]
<_whitenotifier-6> [scopehal-apps] vincentbaeten commented on issue #597: ngscopeclient: Black screen and unresponsive when starting on Ubuntu 22.04 - https://github.com/glscopeclient/scopehal-apps/issues/597#issuecomment-1611597232
bvernoux has quit [Read error: Connection reset by peer]
veegee has joined #scopehal
veegee has quit [Quit: My iMac has gone to sleep. ZZZzzz…]
<azonenberg> So i'm starting to think about writing a multithreaded CSV parser filter lol
<azonenberg> (in addition to optimizing the processing itself)
<azonenberg> basic idea is, chunk the file into N blocks (which may or may not be aligned to a newline)
<azonenberg> once you start a block, if you're not the first one discard any data until the first newline
<azonenberg> then parse the blocks in parallel and concatenate after they all finish
<azonenberg> but more importantly i want to mmap the file and use in-memory parsing rather than a bazillion tiny syscalls
<azonenberg> i think i can probably speed CSV loading by an oom or more
<whitequark> interestingly IIRC ripgrep found it that mmap doesn't actually speed things up
<whitequark> if you're reading everything just once
<azonenberg> whitequark: so in my case i'm in a loop doing fscanf's of single numbers
<azonenberg> i figure if i do one syscall vs millions, i'll save a lot of overhead
<azonenberg> also probably replacing fscanf with std::stof or something so it doesnt have to parse the format string for every data element
<whitequark> oh, nevermind, I have forgotten some details that are crucial here
<whitequark> namely, ripgrep searches many files, which means there is a large fixed mmap overhead per file
<whitequark> and that's not your case
<azonenberg> ah interesting
<azonenberg> yeah my use case here is loading a ~6.5 GB CSV file in a tractable amount of time
<whitequark> it uses mmap() when you're grepping a few files, and read() when you're grepping many
<azonenberg> i'm certain this can be made a lot faster :p
<whitequark> yea for sure
<azonenberg> whitequark: also btw i forget if i mentioned but we fixed a shader bug causing GPU hangs a month or so ago, not sure if you've tried a recent ngscopeclient build
<whitequark> ah thank you, I think that was a friend who was trying it
<whitequark> I'll pass it alon
<whitequark> * I'll pass it along
<azonenberg> there are still known issues on amd integrated cards
<azonenberg> but so far none of the folks who actually work on the rendering have access to such a system to test
<azonenberg> basically the only people doing shader work are me (who has a bunch of discrete nvidia systems as well as a laptop with intel igpu + nvidia optimus) and lain (m1 laptop)
<_whitenotifier-6> [scopehal] 602p opened pull request #780: Fix memory leak in Filter::FindPeaks - https://github.com/glscopeclient/scopehal/pull/780