11:53 <gruetzkopf> @attie you mentionend in the i2s capture issue that uart is also dropping bytes - i believe i'm observing that- to be true for me (i'm definity triggering on my start condition)

11:54 <gruetzkopf> any hints on where to go hunting?

11:56 <d1b2> <Attie> I believe @marcan addressed it initially by increasing the fifo size... but that won't hold up for longer duration bursts... I wanted to look into controlling the flush to improve the throughput

11:57 <gruetzkopf> ah, that wouldn't work for my infinite duration burst

11:57 <d1b2> <Attie> i.e: only allow flush once n-bytes are received or a timeout if n-frames of idle is reached

11:58 <gruetzkopf> flush every n bytes would be perfectly fine for my case

11:59 <gruetzkopf> (my case being a custom audio bus clocked at 12.288MHz, with /CS high for every two out of 10 clock periods (and CLK not active in that hole)

11:59 <d1b2> <Attie> (both triggers are required to avoid leaving data in the fifo once the bust finishes... but if your burst "doesn't finish", then you should get good results with only the first)

11:59 <gruetzkopf> so exactly 1.2288MBytes/s

12:01 <d1b2> <Attie> that should be well within reach! interested to hear how you get on :)

12:04 <gruetzkopf> yeah i have seen 40+ MByte/s from the bandwidth tests, that was impressive

12:10 <marcan> I think I managed 42 at some point

12:10 <marcan> which is about the same as I'd get on an FT2232H

12:10 <marcan> depends on the kernel and host controller; I remember a Linux upgrade improved performance slightly once

12:14 <gruetzkopf> and on the usb hub, the one in my hp usbc dock is .. interesting

12:14 <gruetzkopf> as in: glasgow works better one further hub down..

12:16 <gruetzkopf> now where have i seen manual flush before

12:20 <d1b2> <Attie> https://github.com/GlasgowEmbedded/glasgow/blob/main/software/glasgow/applet/interface/spi_controller/__init__.py#L328

12:22 <d1b2> <Attie> (and then set flush.eq(1) when you want to push the data to the host)

12:24 <gruetzkopf> aah, thanks!

12:38 <gruetzkopf> clock_domain is "None" on that fifo?

14:14 <d1b2> <DX-MON> no domain specified, so it'll default to sync

20:06 <_whitenotifier-1> [glasgow] electroniceel commented on issue #283: Test 1.2 V capable level shifters 74AXP1T45 - https://git.io/JsEmF