| + xet7 (~xet7@user/xet7) | 02:16 | |
| - norly (QUIT: Quit: Leaving.) (~norly@zhong.enpas.org) | 05:13 | |
| + norly (~norly@zhong.enpas.org) | 05:14 | |
| + mnemotron (mnemotron@194.135.47.239) | 07:53 | |
| - xet7 (QUIT: Ping timeout: 272 seconds) (~xet7@user/xet7) | 09:17 | |
| + xet7 (~xet7@user/xet7) | 13:51 | |
| - xet7 (QUIT: Remote host closed the connection) (~xet7@user/xet7) | 14:00 | |
| minute | ok so i found one general problem | 14:38 |
|---|---|---|
| minute | if a write to ZZ takes too long for 68060, at least on CS-MK1, the next one is lost | 14:38 |
| forcie | minute: possibly on 040 also? btw, did you receive my .zip with testcase above? | 14:42 |
| minute | yes, thanks | 14:45 |
| minute | i made a debug register that just sleeps for a set amount of msecs. if i sleep for more than a second, i can already do additional pokes in `mon` on the amiga | 14:50 |
| minute | will now test if this happens for reads, too, and will compare with 030 | 15:05 |
| minute | ah, this generates illegal instruction | 15:08 |
| minute | (read taking too long) | 15:08 |
| minute | so, how do Z3 harddisk controllers work... do they do everything asynchronously and you have to poll busy regs etc? | 15:09 |
| minute | in any case, the timeout appears to be around ~280ms | 15:21 |
| minute | which is a lot | 15:21 |
| minute | ah, i'm doing a test now where i'm writing to the same register thousands of times without delay and getting a lot of loss | 15:47 |
| minute | [zzdebug] zdata: 65533 expected: 61298 | 15:48 |
| minute | [zzdebug] zdata: 65534 expected: 61299 | 15:48 |
| minute | so, around 6% of writes lost :3 | 15:49 |
| minute | well, that wasn't entirely true: there was a delay, induced by printf | 16:01 |
| minute | without that, i get around 3 lost writes in 65535 | 16:01 |
| minute | (sometimes a few more) | 16:02 |
| minute | https://twitter.com/minut_e/status/1518954584861024262 | 16:06 |
| minute | 68030 loses around 1 write in 600000 | 16:22 |
| minute | strangely, setpatch makes a big difference here: without setpatch, there are a lot more lost writes on 68030 | 16:29 |
| - mnemotron (QUIT: Quit: Quit) (mnemotron@194.135.47.239) | 17:25 | |
| apolkosnik[m] | Does the mmu.library matter? | 18:07 |
| apolkosnik[m] | ACTION uploaded an image: (341KiB) < https://libera.ems.host/_matrix/media/r0/download/matrix.org/cLFCkRaeWEwQmKqDfXrmSlZH/PXL_20220426_160359377.MP.jpg > | 18:08 |
| minute | apolkosnik[m]: good question! | 18:08 |
| apolkosnik[m] | history log for mmu.library feels like a fascinating lecture | 18:18 |
| minute | communication from amiga with the fpga side works perfectly though, even for many 10s of millions of writes | 19:35 |
| minute | (implemented a resettable counter register in the mntzorro side and a test program on the amiga, no writes are missed) | 19:35 |
| apolkosnik[m] | http://amigadev.elowar.com/read/ADCD_2.1/AmigaMail_Vol2_guide/node0161.html | 19:46 |
| apolkosnik[m] | One of the folks on Acill's discord mentioned that this could be applicable | 19:48 |
| minute | > The new library prevents the 68040 from doing full | 20:09 |
| minute | line bursts to non-cachable devices | 20:09 |
| minute | the bug i'm seeing is internal to the fpga->ARM mechanism. | 20:12 |
| minute | all the writes actually arrive in mntzorro | 20:12 |
| minute | but they don't all arrive at the ARM | 20:13 |
| minute | ok, problem fixed | 20:54 |
| _Bnu | Are these ARM registers? | 20:57 |
| _Bnu | Oh, problem fixed already. | 20:57 |
| _Bnu | Because there has always* been an issue where some reads/writes never made it through to the ARM part. | 20:58 |
| _Bnu | The faster you tried to read or write, the more failures you'd get, that's why I made the RTG driver use that FPGA register for checking vblank. | 20:59 |
| minute | yeah, i only got to the bottom of it now | 21:00 |
| minute | needed to add 1 more state on the fpga side that would check for zorro_ram_write_flag go from 1->0 before ending the cycle | 21:01 |
| minute | or in some cases ARM could skip a write and fpga thinks the write was done | 21:01 |
| _Bnu | Huh, no change needed for reads? | 21:01 |
| minute | probably also applies for reads yeah but i can add the same mechanism there | 21:01 |
| _Bnu | Ah. | 21:02 |
| minute | didn't debug reads yet yeah | 21:02 |
| minute | a good test is the blanker screensaver thing | 21:02 |
| minute | that comes with amigaos | 21:02 |
| minute | because it was always leaving some "dirt" behind | 21:02 |
| minute | i.e. some lines not cleared | 21:02 |
| _Bnu | Because I had immense issues with fast repeated reads, haha. | 21:02 |
| minute | ah, could be the same then | 21:02 |
| minute | so the blanker thing is now 100% clean even with nodma driver on z3 (doing all line draws through the arm mechanism) | 21:03 |
| _Bnu | That'd probably fix the problem people had running the ZZ9000 Frontier intro on some CPU cards as well. | 21:03 |
| minute | hm! | 21:03 |
| minute | oh, the read side had this fix already :0 | 21:05 |
| minute | (not sure when i added that though) | 21:05 |
| minute | ermahgerd | 21:05 |
| _Bnu | Smol sneks. | 21:19 |
| minute | schnecks | 21:34 |
| minute | how to add .mp3 files to amigaamp? | 21:45 |
| minute | it shows my directory with .mp3 and .wav stuff as empty | 21:45 |
| minute | question marks floating above my head | 21:45 |
| minute | what's the secret trick to using this program? | 21:45 |
| forcie | i guess LIBS:mpega.library | 21:50 |
| minute | yeah, i have that one | 21:50 |
| forcie | and asyncio.library | 21:51 |
| minute | the UI is just buggy or something (of this 2.x version) | 21:51 |
| minute | adding a whole folder works | 21:51 |
| minute | just not adding individual files | 21:51 |
| minute | so i can totally play back mp3s decoded by 060! | 21:51 |
| minute | (through zz9000ax) | 21:51 |
| minute | with cs mk1 | 21:51 |
| forcie | oh nice | 21:51 |
| minute | i don't get any glitches... but probably other people will? ;3 | 21:52 |
| minute | oh, it fucked up in the middle | 21:52 |
| minute | i mean, amigaamp or the OS did, something locked up | 21:53 |
| minute | that little ball is still spinning and the scroll text is still scrolling though | 21:53 |
| minute | wei0red | 21:53 |
| minute | maybe some interrupt fuckery | 21:54 |
| minute | in any case, this is a good firmware for further testing | 21:54 |
| _Bnu | Remove unused MP3s!!! | 21:56 |
| minute | dump.mntmn.com/zz9000-fwpack-110-20220426.zip | 21:56 |
| minute | well, http://dump.mntmn.com/zz9000-fwpack-110-20220426.zip | 21:56 |
| minute | or not | 21:56 |
| minute | now. | 21:56 |
| + Nowee (~NoweeEewo@lfbn-ncy-1-4-192.w83-194.abo.wanadoo.fr) | 21:58 | |
| + mnemotron (mnemotron@194.135.47.239) | 22:02 | |
| mnemotron | gerade Windows 10 frisch installiert | 22:02 |
| mnemotron | whee | 22:02 |
| mnemotron | ups, wrong channel ;) | 22:07 |
| - mnemotron (QUIT: Ping timeout: 276 seconds) (mnemotron@194.135.47.239) | 22:15 | |
| minute | on my a3000+csmk1, audio playback is rock solid and there are no more gfx glitches to be seen... can someone confirm? | 22:18 |
| minute | Chain-Q: does the csmk2 offer still stand btw? i would like to see if the audio stuff works with it | 22:19 |
| Nowee | Hey there | 22:21 |
| Nowee | if you need to test over a CSPPC, feel free to send that over :) | 22:21 |
| apolkosnik[m] | loading stuff here | 22:24 |
| apolkosnik[m] | on 3k with 16mb fast and 030, I get rando spikes | 22:26 |
| apolkosnik[m] | pops | 22:26 |
| apolkosnik[m] | probably because I set the 48k as mixing rate | 22:28 |
| Chain-Q | minute: yes it does. how and when can I get it to you? | 22:28 |
| apolkosnik[m] | Frontier gl-demo gets stuck when the extra memory is showing up | 22:36 |
| apolkosnik[m] | on my a4k with csmk1 | 22:36 |
| apolkosnik[m] | though no gfx artifacts so far | 22:36 |
| minute | apolkosnik[m]: did you have gfx artifacts before? | 22:37 |
| minute | Chain-Q: cool, maybe i can pick it up on friday or sometime next week? | 22:38 |
| minute | apolkosnik[m]: ah yes 48k is a bit much for 030 | 22:38 |
| minute | apolkosnik[m]: how about 24khz? | 22:38 |
| apolkosnik[m] | yes, though not very many of them | 22:39 |
| minute | apolkosnik[m]: which player and what are you playing? | 22:40 |
| minute | apolkosnik[m]: ah, that was the answer about the gfx artifacts. | 22:40 |
| apolkosnik[m] | Delitracker, and Hippo | 22:41 |
| minute | ah, delitracker was super slow for me through AHI | 22:46 |
| minute | hippo much better | 22:46 |
| minute | i even get dropouts on deli on 060 with the ahi noteplayer, dunno why | 22:46 |
| minute | LOL "The Maestrix" works with Octamed Soundstudio and ZZ9000AX | 22:46 |
| minute | i'm playing the hired guns title MED with 16 bit OSS mixing | 22:47 |
| minute | that sounds very interesting | 22:47 |
| apolkosnik[m] | got gfx artifacts | 22:52 |
| apolkosnik[m] | ACTION uploaded an image: (350KiB) < https://libera.ems.host/_matrix/media/r0/download/matrix.org/sBREnOLYFthONqrplySvPqPs/PXL_20220426_205218849.jpg > | 22:52 |
| minute | interesting | 22:53 |
| minute | apolkosnik[m]: can you run the amiga blanker commodity? (set it to 5 seconds or so) | 22:53 |
| minute | apolkosnik[m]: is it clean or does it leave lines? | 22:54 |
| minute | wow, mod playback with OSS through the maestrix @ 48khz is really something | 22:54 |
| apolkosnik[m] | ACTION uploaded an image: (104KiB) < https://libera.ems.host/_matrix/media/r0/download/matrix.org/eUhImKDXzTqfJaqJZAjYZCPO/PXL_20220426_205654293.jpg > | 22:57 |
| minute | apolkosnik[m]: oh wow ok. but i have another idea... you're probably using the dma driver | 22:58 |
| minute | apolkosnik[m]: can you try http://dump.mntmn.com/ZZ9000.card.nodma ? (copy to Libs:Picasso96/ZZ9000.card ) | 22:59 |
| Chain-Q | minute: a few of us meet in the IN-Berlin (Lehterstr, near the HbF) this Saturday for an Amiga Meetup... if that works for you, that would be the best | 23:02 |
| Chain-Q | but otherwise yeah, you can pick it up on Friday or next week too | 23:02 |
| Chain-Q | I just can't go anywhere on Friday, and I'm the other side of town (near the Rathaus Lichtenberg) | 23:03 |
| Chain-Q | +on | 23:03 |
| minute | Chain-Q: saturday might work, i'll get back to you on this later! | 23:03 |
| Chain-Q | okie | 23:03 |
| minute | thanks! | 23:04 |
| Chain-Q | meetup starts at 15:00, btw, on Saturday, open-end, kinda. we're usually there until midnight | 23:04 |
| apolkosnik[m] | ok, copied over | 23:07 |
| apolkosnik[m] | no artifacts in bla ker | 23:09 |
| apolkosnik[m] | lol, I ran the real frontier, blanker kicked in, and I can hear the wind sound, and the blanker took over the display and broke | 23:13 |
| minute | well | 23:22 |
| minute | ok, so then there's also a problem with DMA writes or a cache coherency problem | 23:22 |
| _Bnu | Any MMU libraries are always going to break DMA RTG forever. | 23:27 |
| _Bnu | I mean anything that activates the MMU and misconfigures it. | 23:28 |
| minute | so that it allows caching in that area? | 23:28 |
| _Bnu | Yeah, because then it'll be all "no that data is already there ahhaha lol :)" | 23:32 |
| minute | lol | 23:35 |
| - Nowee (QUIT: Quit: bye) (~NoweeEewo@lfbn-ncy-1-4-192.w83-194.abo.wanadoo.fr) | 23:39 | |
| apolkosnik[m] | Which frontier intro was supposed to be accelerated? | 23:41 |
| apolkosnik[m] | It disappeared from my desktop, now if I run fintro it's under 20 fps | 23:43 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
