Tuesday, 2019-05-14

*** tpb has joined #tomu00:00
*** deltab has quit IRC00:08
*** deltab has joined #tomu00:15
xobssldkfjsldkf: you could run "arm-none-eabi-size" on it, or "arm-none-eabi-nm"00:34
*** futarisIRCcloud has joined #tomu01:31
xobsWell that's a good sign.  I've been continuously transferring random data back and forth across the Wishbone bridge for about 24 hours now.02:31
*** xkapastel has joined #tomu02:52
*** AmosSam has left #tomu05:05
*** AmosSam has joined #tomu05:06
*** rohitksingh_work has joined #tomu05:30
*** xkapastel has quit IRC05:32
*** AmosSam has left #tomu05:52
*** AmosSam has joined #tomu05:56
tntxobs: you should also try large packet transfers for a while (512b bulk or isoc).07:00
*** AmosSam has left #tomu07:10
*** AmosSam has joined #tomu07:11
*** AmosSam has left #tomu07:17
*** AmosSam has joined #tomu07:18
*** rohitksingh_wor1 has joined #tomu07:37
*** rohitksingh_work has quit IRC07:38
xobstnt: I'm actually testing both right now.07:39
xobsI'm sending random data over the debug channel (writing to memory, then reading back to make sure it matches), while sending random data over the serial channel.07:40
xobsWindows doesn't like it when I do that.  But Linux will happily mux it for me.07:40
xobsConvincing Linux that write() means write() is the hardest bit.  Undoing decades of terminal cruft.07:40
MadHackerIsn't that just stty raw < /dev/ttywhatever ?07:42
MadHackerBasically taking it out of cooked mode.07:42
tntxobs: and that uses large packet sizes btw ?07:43
*** AmosSam has left #tomu07:43
xobstnt: I'm not sure what you mean by "large" packet sizes.  I was limiting it to 64-byte packets, because that's the transfer length.07:45
xobsBut let me increase it to 512 bytes.  I'll have to worry about fragmentation now, though.07:45
tntxobs: I meant packet of 512 bytes on the USB bus.07:46
tnt(in a single packet, to check tolerance to progressive drifts etc ...)07:46
xobsAh, I see.  Okay, my C program to test appears to be broken.  Time to fix it.07:46
*** AmosSam has joined #tomu07:51
alexhw[m]xobs: nice to see it running :-)07:51
keesj2019.. serial remain usefull and funky.08:22
*** AmosSam has left #tomu08:34
xobstnt: naturally, it falls over and dies when I write 129-byte packets.08:35
xobsWhich is to be expected, given that my packet depth is 128.08:35
tntxobs: fifo depth you mean ?08:38
*** AmosSam has joined #tomu08:39
xobsYeah, and I've otherwise run out of BRAM.08:39
xobsThe real solution is to move away from this FIFO implementation and towards the epmem approach.  But this works well enough for now.08:40
keesjxobs: it this using fdti/serial or the usb serial running on the fpga?08:43
xobskeesj: it's a usb serial program running on the FPGA.08:44
xobstnt: isn't the maximum packet size for FS 64 bytes?  Anything larger than that isn't valid for FS...08:51
tntOf course it is.08:51
tntyou can go up to 1023 bytes.08:51
xobsHm, you're right.08:52
tntOnly control transfers on ep0 are limited to 6408:52
xobsWhy is Linux behaving strangely when I give it a 64-byte packet?08:52
tntOh, wait, bulk transfer might actually be limited to 64.08:53
tntIsoc can be longer for sure.08:56
keesjI also have been buffering happening on the recieve side (e.g. the use serial stack running on the tinyFPGA sending data and when I connect to the serial I the buffers are then already full of stuff08:58
tntok, til that FS bulk is limited to 64 bytes, I had no idea.08:59
tnt(I only used isoc and control so far)08:59
xobstnt: your ability to grep the USB spec is far higher than mine.  Where do you see that?09:01
xobs"The USB defines the allowable maximum bulk data payload sizes to be only 8, 16, 32, or 64 bytes for full-speed endpoints and 512 bytes for high-speed endpoints. "09:02
xobsAnd yeah, 1024 bytes is fine for FS ISO endpoints.09:03
*** AmosSam has left #tomu09:03
keesjsome data here http://www.usbmadesimple.co.uk/ums_3.htm09:04
tpbTitle: USB Made Simple - Part 3 (at www.usbmadesimple.co.uk)09:04
tntoh ... a bit late :p09:05
xobsWell, with this test running, I'm reasonably happy that the core is now stable.  That makes me happy, because mass production happens in two weeks.09:06
*** AmosSam has joined #tomu09:07
keesjnice nice09:07
alexhw[m]tnt/xobs: for fs iso transfers its 1023 (5.6.3), for other fs transfers its 64 max.09:12
tntyup. But bulk can get multiple packets in s ingle frame I think, iso can't.09:13
alexhw[m]Ah, ok09:14
MadHackerxobs: Is https://www.crowdsupply.com/solokeys/somu from you, or from the Solo folks, or some cooperation between?10:37
tpbTitle: Somu | Crowd Supply (at www.crowdsupply.com)10:37
*** futarisIRCcloud has quit IRC10:40
MadHacker(my NFC question yesterday was unrelated, just happened to spot them on crowdsupply's front page)10:45
*** xkapastel has joined #tomu10:52
xobsMadHacker: it's from the Solo folks11:06
MadHackerBit cheeky with the naming, then! Oh well. :)11:09
*** Dan_au has joined #tomu11:27
xobsthey did ask!11:33
keesjperhaps the casing can be shared. it looks nice13:06
*** xkapastel has quit IRC13:21
*** rohitksingh_wor1 has quit IRC13:28
*** AmosSam has left #tomu13:41
*** AmosSam has joined #tomu13:46
*** AmosSam has left #tomu14:51
*** AmosSam has joined #tomu14:51
*** rohitksingh has joined #tomu14:56
*** alexhw has joined #tomu15:07
*** emeb has joined #tomu15:46
pollokeesj: it wouldn't diffuse the tomu led though15:49
keesjthe tomu is quite a beast in terms of distracting me with blinking stuff16:12
*** andi- has quit IRC17:06
*** andi- has joined #tomu17:15
*** rohitksingh has quit IRC18:32
*** rohitksingh has joined #tomu18:42
*** cdmatter has joined #tomu18:44
tntDespite 64b packets, still manage to do 1.15 Mbytes/s, that's not too bad.19:39
*** rohitksingh has quit IRC20:21

Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!