*** tpb has joined #tomu | 00:00 | |
*** TheJJ has quit IRC | 01:06 | |
*** tannewt has quit IRC | 01:06 | |
*** ssb has quit IRC | 01:06 | |
*** slurdge has quit IRC | 01:06 | |
*** jerith has quit IRC | 01:06 | |
*** MadHacker has quit IRC | 01:06 | |
*** TheJJ has joined #tomu | 01:09 | |
*** tannewt has joined #tomu | 01:09 | |
*** ssb has joined #tomu | 01:09 | |
*** slurdge has joined #tomu | 01:09 | |
*** jerith has joined #tomu | 01:09 | |
*** MadHacker has joined #tomu | 01:09 | |
*** jerith has quit IRC | 01:11 | |
*** jerith_ has joined #tomu | 01:11 | |
*** Jay_jayjay has quit IRC | 02:08 | |
*** slurdge8 has joined #tomu | 02:35 | |
*** tannewt_ has joined #tomu | 02:35 | |
*** Vercas_ has joined #tomu | 02:37 | |
*** TheJJ_ has joined #tomu | 02:37 | |
*** ssb_ has joined #tomu | 02:38 | |
*** craigo has joined #tomu | 02:38 | |
*** MadHacke1 has joined #tomu | 02:38 | |
*** gio_ has joined #tomu | 02:39 | |
*** eaterof has joined #tomu | 02:43 | |
*** gio has quit IRC | 02:43 | |
*** eater has quit IRC | 02:43 | |
*** gerard- has quit IRC | 02:43 | |
*** Vercas has quit IRC | 02:43 | |
*** TheJJ has quit IRC | 02:43 | |
*** tannewt has quit IRC | 02:43 | |
*** ssb has quit IRC | 02:43 | |
*** slurdge has quit IRC | 02:43 | |
*** MadHacker has quit IRC | 02:43 | |
*** tannewt_ is now known as tannewt | 02:43 | |
*** slurdge8 is now known as slurdge | 02:44 | |
*** asante10111_ has joined #tomu | 02:49 | |
*** gerard- has joined #tomu | 02:54 | |
*** asante10111__ has joined #tomu | 03:07 | |
*** asante10111_ has quit IRC | 03:11 | |
*** whatnick has quit IRC | 04:53 | |
*** craigo has quit IRC | 08:18 | |
*** ssb_ is now known as ssb | 08:35 | |
*** whatnick has joined #tomu | 10:05 | |
*** whatnick has quit IRC | 10:52 | |
*** marcmerlin has joined #tomu | 11:45 | |
marcmerlin | howdy. I've been reading up on fomus and FPGAs. I see that fomu has 5000 LUTs, which doesn't feel like a lot, but that's somehow still enough to upload a RISC-V softcore + code to run on top? If so, impressive | 11:46 |
---|---|---|
marcmerlin | I understand that with an FPGA, I could upload any softcore that fits, but realistically people mostly use RISC-V since it's simple. | 11:47 |
marcmerlin | So, for a small FPGA, why would I want to use one over a microcontroller that already has a CPU, and where all the resources are dedicated to user code and RAM? | 11:49 |
marcmerlin | I can see that it would be cool to be able to patch your CPU and add instructions, which you can't do on a microcontroller, but I'm assuming few people do/use that. Same thing with micropython, how would it differ to run it on an FPGA over let's say an ESP32? | 11:52 |
xobs | marcmerlin: yep, enough to upload a RISC-V softcore + code. Plus USB core. The bootloader is 8 kB, which is loaded into the softcore's "bootrom". | 11:52 |
xobs | The advantages are mostly educational, and being able to understand how a CPU works. | 11:53 |
xobs | People have lamented the lack of a multiply unit, so just yesterday someone compiled a new version of the CPU to include a single-cycle multiplier. | 11:54 |
marcmerlin | Right, I understood that from the docs (workshop is great, much better than tomu that was hard to learn :) ). Got it, so on that small scale, it's not better in any way than a microcontroller | 11:54 |
tnt | marcmerlin: it's going to be much slower that fpga, that's one difference :p | 11:54 |
tnt | marcmerlin: but the huge advantage of a fpga is you get the peripherals you want ... exactly like you want them. | 11:55 |
marcmerlin | @tnt: yes, I figured an FGPA softcore will be slower than a hardware CPU (although you can add instructions to the latter) | 11:55 |
xobs | You're correct. It's slower and less full-featured than an MCU (i.e. no multiplier or fancy peripherals). | 11:55 |
xobs | But there are a few tricks. For example, because we control the USB hardware, there's a debug bridge that's exposed over the USB pins. | 11:56 |
xobs | That means you can poke at hardware registers without needing a third-party debug cable. | 11:56 |
marcmerlin | @tnt: I see what you mean. If you want 4 hardware SPI busses, you only have 2 on an ESP32 and then you do software SPI. So the FPGA would let you add extra SPI outputs that would be faster than software SPI on ESP32 | 11:57 |
marcmerlin | xobs: yes, I saw the USB debug bridge, that's very cool indeed. You need a special cable to do that on other chips | 11:58 |
marcmerlin | Now, I know that at bigger scale (driving 8 RGBPanel strings in parallel), FPGAs do better, because they can do so much I/O | 11:59 |
tnt | marcmerlin: yeah .. or stuff that just plain doesn't exist. I have a UP5k interfacing to E1 (telecom interfaces), or to SPDIF ... I would have needed external chips to do that with any MCU at all. | 11:59 |
marcmerlin | can't you use digital pins and PWM to do that on any MCU? | 12:00 |
marcmerlin | I've never seen it, but I'm guessing an arduino style chip could output composite video on a pin for instance HDMI would be harder due to lack of required speed | 12:01 |
tnt | E1, no. SPDIF I guess you could possibly bitbang it maybe although at like 3 MHz that's a pretty high rate to bitbang and you gotta be _spot_on_ timing. | 12:01 |
tnt | And as soo as you bitbang anything on a MCU ... you lose any CPU speed advantage your hw CPU has because you're spending all your time bitbanging stuff. | 12:02 |
marcmerlin | ok, that's where I may need to learn more. Yes, bigbanging ties up the CPU and requires a spot on ISR. But isn't that true for an FPGA too? I'm assuming you would generate the output not in C code that runs on the softcore, or you're no better off, correct? | 12:04 |
xobs | You could generate a hardware block that does it. | 12:28 |
xobs | For example, a ws2812b "neopixels" driver could be native, where you'd have a framebuffer and hardware would take care of the timing. | 12:29 |
marcmerlin | gotcha, thanks. I need to do more reading and learning | 13:00 |
*** TheJJ_ has quit IRC | 14:33 | |
*** TheJJ has joined #tomu | 14:35 | |
*** Jay_jayjay has joined #tomu | 15:19 | |
*** Jay_jayjay has quit IRC | 15:54 | |
*** Jay_jayjay has joined #tomu | 15:55 | |
*** Jay_jayjay has quit IRC | 16:00 | |
*** srk has quit IRC | 16:02 | |
*** Jay_jayjay has joined #tomu | 16:04 | |
*** srk has joined #tomu | 16:07 | |
*** Jay_jayjay has quit IRC | 16:08 | |
*** Jay_jayjay has joined #tomu | 16:21 | |
*** Jay_jayjay has quit IRC | 17:17 | |
*** Jay_jayjay has joined #tomu | 17:24 | |
*** NoGodDamnIdea has joined #tomu | 17:39 | |
*** Jay_jayjay has quit IRC | 18:07 | |
*** Jay_jayjay has joined #tomu | 18:18 | |
*** Jay_jayjay has quit IRC | 18:23 | |
*** AnimaInvicta has joined #tomu | 19:08 | |
*** Jay_jayjay has joined #tomu | 19:34 | |
*** Jay_jayjay has quit IRC | 20:21 | |
*** Jay_jayjay has joined #tomu | 20:38 | |
*** NoGodDamnIdea has quit IRC | 20:45 | |
*** NoGodDamnIdea has joined #tomu | 20:52 | |
*** Jay_jayjay has quit IRC | 21:34 | |
*** Jay_jayjay has joined #tomu | 21:37 | |
*** Jay_jayjay has quit IRC | 21:42 | |
marcmerlin | l | 22:00 |
*** AnimaInvicta has left #tomu | 22:04 | |
acathla | Are you the marcmerlin on marmerlins.org ? | 22:12 |
marcmerlin | oops, sorry for sending that stray character. Yes, that's me ( http://marc.merlins.org/ I assume you meant) | 22:14 |
acathla | Yep. I used your SA-exim patch and conf files a lot :) | 22:17 |
*** Jay_jayjay has joined #tomu | 22:17 | |
marcmerlin | oh my, that's like a totally previous life, 15 years ago :) | 22:20 |
acathla | You should update it :) | 22:21 |
marcmerlin | I'll admit that I've kind of lost interest. I still run my own Email, but I'm done writing exim configs (the exim4 config file in debian is partially mine), and exim code :) | 22:25 |
marcmerlin | but the cool part is that someone else in debian has been maintaining it for me | 22:26 |
*** Jay_jayjay has quit IRC | 22:43 | |
*** asante10111__ has quit IRC | 23:09 | |
*** Jay_jayjay has joined #tomu | 23:51 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!