Friday, 2017-03-31

*** tpb has joined #timvideos00:00
cr1901_modernmithro: Bitbang flash write works. Bad news though... I kinda lost the scrollback to screencap tho (code not ready) :(00:53
*** arcticShadow has quit IRC01:10
*** arcticShadow has joined #timvideos01:13
*** arcticShadow has quit IRC01:16
*** arcticShadow has joined #timvideos01:25
*** arcticShadow has joined #timvideos01:27
mithrocr1901_modern: Great, clean it up and send it upstream?01:33
cr1901_modernmithro: Sure. But before I send a pull request, I want to discuss formatting/file changes. Right now, I'm examining some unexpected behavior. >>01:57
cr1901_modern(I *strongly* suspect it's due to not clearing the cache properly in the currently loaded firmware, but just in case)01:57
cr1901_modernmithro: If you wish to play along, here's a text file showing what I did02:03
cr1901_modernLines 40 and 41 in particular are unexpected, taking the lm32's cache into account02:03
mithroYou didn't link a text file....02:04
cr1901_modernThat would help, wouldn't it: https://gist.github.com/cr1901/4b4e5849ede91fd48d06fbebdfc39e9902:04
cr1901_modernmithro: (Don't mind me, btw. I'm just thinking out loud)02:04
tpbTitle: spiflash.txt · GitHub (at gist.github.com)02:04
cr1901_modernmithro: I was just going to make you guess the gist ID :)02:04
mithrocr1901_modern: That also doesn't show me the code for that02:05
cr1901_modernmithro: Two seconds please02:07
cr1901_modernGah! firmware/flash.c was deleted in the tip of the tree02:08
cr1901_modernmithro: Git woes, I apologize. I am somehow on a commit that no longer exists upstream02:16
mithrocr1901_modern: Umm, firmware/flash.c never existed upstream02:17
cr1901_modernmithro: I said upstream. I meant mithro/HDMI2USB-litex-firmware*02:17
mithrocr1901_modern: Never existed there either...02:17
cr1901_modernmithro: The tip of my master points to e656e22ea079d041095b1d44738ae7bf8edb2f5002:18
mithrocr1901_modern: I've never written a firmware/flash.c ...02:18
cr1901_modern(what the hell did I do to this poor tree T_T?!)02:18
cr1901_modernOkay I found where I am in the commit graph. I managed to accidentally fast-forward to the tip of merge-arty-soc02:24
mithrocr1901_modern: arty-soc has a firmware/flash.c?02:24
cr1901_modernmithro: Yes02:25
mithroInteresting, I didn't notice that02:25
cr1901_modernAs of this commit: https://github.com/mithro/HDMI2USB-litex-firmware/commit/b70c03bfbb5d7c37a1f5f801f6b40e937798139e02:25
tpbTitle: execute firmware from sdram, add flash and flash_test · mithro/[email protected] · GitHub (at github.com)02:25
cr1901_modernmithro: Should I cherry pick back on master, or is it okay to merge back to master from arty-soc (while acknowledging it was a mistake)?02:25
mithrocr1901_modern: I'll merge the arty support in the next couple of days02:26
cr1901_modernmithro: So do cherry pick, and you'll handle the merge conflict that *might* occur from "both modified"?02:27
cr1901_modern(firmware/flash.c firmware/flash.h will be "both modified" during the merge)02:27
mithrocr1901_modern: ?02:30
cr1901_modernmithro: Actually, let me just commit, then I'll show you the problem02:30
cr1901_modernmithro: Sweet Jesus, finally got it: https://github.com/cr1901/HDMI2USB-litex-firmware03:00
tpbTitle: GitHub - cr1901/HDMI2USB-litex-firmware: A version of the HDMI2USB firmware based around the misoc+migen tools produced by m-labs. (at github.com)03:00
cr1901_modernhttps://github.com/cr1901/litex03:01
tpbTitle: GitHub - cr1901/litex: Build your hardware, easily! (at github.com)03:01
mithrocr1901_modern: What is line 91 doing?03:02
cr1901_modernmithro: Which file03:02
mithrohttps://github.com/cr1901/HDMI2USB-litex-firmware/blob/master/firmware/flash.c#L9103:03
tpbTitle: HDMI2USB-litex-firmware/flash.c at master · cr1901/HDMI2USB-litex-firmware · GitHub (at github.com)03:03
cr1901_modernmithro: For the sake of testing the flash, I'm resetting the free area back to 0xFF03:03
mithrocr1901_modern: So... what is the problem?03:06
cr1901_modernmithro: Okay, so the gist I generated with commit 5b167704d55db44dd279ef867543a408454b56e9, before I added the flush cache command. The problem is these two lines: https://gist.github.com/cr1901/4b4e5849ede91fd48d06fbebdfc39e99#file-spiflash-txt-L40-L4103:10
tpbTitle: spiflash.txt · GitHub (at gist.github.com)03:10
cr1901_modernBefore I ran this firmware, I ensured that the SPI flash region to be tested was already zeroed out. This means that a read from that area should return all zeros; it doesn't.03:10
mithrocr1901_modern: But of course you need to flush the cache after you modify the spiflash?03:10
cr1901_modernmithro: Yes, but at power on for the FPGA the flash region that is tested by bitbang_test is already zeroed out. Thus it won't be in the cache until I run bitbang_test03:12
cr1901_modernmithro: I am *fairly* certain flushing the cache solves the problem, I just created this gist as a "starting point" just in case.03:12
cr1901_modernOnly thing I can think of is that the flash region wasn't fully zeroed out for some reason during the previous test03:14
mithrocr1901_modern: I don't know how the dcache actually works - but I wouldn't be surprised if it works on sizes more than individual bytes, so just reading until you get the area is going to cause the cache to contain part of it?03:14
cr1901_modernmithro: That's plausible. I'm not going to dwell on it right now. The problem as far as I can tell disappears if I flush the dcache03:17
cr1901_modern(which is expected, I just don't want the problem to be something more sinister like "writes aren't honored" :P)03:17
*** tvCommitBot has joined #timvideos03:18
tvCommitBot[website] mithro pushed 1 new commit to master: https://git.io/vSc4303:18
tvCommitBotwebsite/master 1fdeb48 Tim 'mithro' Ansell: Adding new lines to fix lists.03:18
*** tvCommitBot has left #timvideos03:18
cr1901_modernmithro: Do you think we should use XMODEM? Not sure if this is a concern for you, but XMODEM in and of itself has no knowledge of file size03:30
*** rohitksingh_work has joined #timvideos03:51
*** arcticShadow has quit IRC03:54
*** arcticShadow has joined #timvideos04:00
cr1901_modernmithro: I figured it out. It's not a bug. I didn't take into account that putting in/deleting code will ultimately move the "free area" I used for testing. So when I add new code in, it will push the area I use for testing up to higher addresses that were never touched by the bitbang test05:45
*** sb0 has quit IRC06:02
*** sb0 has joined #timvideos06:52
paddatrappermithro: GSoC application submitted08:59
shenkipaddatrapper: what did you submit a propsal on?10:28
paddatrappershenki: USB audio10:29
paddatrapperThough not the HDMI capture side10:29
shenkiah cool. ive done some work with that before10:29
shenkiyou'r egoing to make it act like a USB sound card?10:29
*** danielki has joined #timvideos10:55
*** sb0 has quit IRC11:12
paddatrappershenki: yup. Stereo PCM capture12:39
*** rohitksingh_work has quit IRC12:55
*** SamSagaZ has quit IRC13:23
*** danielki has quit IRC15:52
*** sb0 has joined #timvideos15:57
*** danielki has joined #timvideos17:20
*** hyadez has quit IRC17:45
*** sam_ has quit IRC18:09
*** hyadez has joined #timvideos18:44
*** Sewar has quit IRC19:56
*** userxyz has joined #timvideos20:22
*** userxyz has quit IRC21:13
*** danielki has quit IRC21:15
*** SamSagaZ has joined #timvideos21:58

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