| *** tpb has joined #vtr-dev | 00:00 | |
| elms | mithro: I realized that the something is still not 100% right. I essentially have a iceblink in parallel (counter->LEDs)with the ram test. But only the RAM LED was blinking. Replacing the LED with a constant instead of the RAM value and then the counter->LED part routes and works on the icestick. My attempt so far today to debug the routing has not been fruitful. | 17:00 |
|---|---|---|
| mithro | kem_: Ping? Any chance you are around and have time to chat? | 17:00 |
| mithro | elms: So the reading / writing block ram doesn't seem to be working? | 17:01 |
| elms | no I think the reading from RAM is working with the data section. Haven't tested writing (ie checker) | 17:02 |
| daveshah | elms: have you looked at the icebox_vlog output? | 17:02 |
| mithro | daveshah: That is what we were looking at last night | 17:03 |
| elms | Looking at it again right now. I'm still working on my reading verilog skills | 17:06 |
| daveshah | I found symbol output helps a lot. Did you get that working? | 17:07 |
| elms | daveshah: what do you mean by symbol output? | 17:09 |
| daveshah | elms: if you use sym statements in the asc file, you can map an icebox net number to an input net name | 17:09 |
| daveshah | Have a look at the arachne-pnr asc output for an example | 17:10 |
| mithro | daveshah: Yeah, we would need to figure out how to get symbol data through HLC to the ASC | 17:31 |
| mithro | elms: Okay, I'm going to try and get the ice40 stuff merged into symbiflow-arch-defs without the ram for the moment | 17:32 |
| daveshah | Yeah, you'd probably need a tag in HLC or something | 17:32 |
| daveshah | If you have the icebox net numbers as metadata in VPR though, then the VPR side shouldn't be too bad | 17:33 |
| mithro | daveshah: We have all the data we need in the HLC comments we already egnerate | 17:33 |
| daveshah | mithro: Given this is debug info I would argue hot comments are actually OK | 17:33 |
| mithro | daveshah: hot comments? | 17:34 |
| elms | mithro: ok. Have we tested routing with any other projects with more complex logic? | 17:34 |
| daveshah | mithro: comments with a meaning beyond comments | 17:34 |
| mithro | elms: Beyond blinky? | 17:34 |
| mithro | elms: Not really | 17:34 |
| mithro | elms: Have you tried doing equivalence check? | 17:35 |
| daveshah | mithro/elms: rs232demo is nice as something a bit more than blinky but no RAM, and has a testbench | 17:37 |
| daveshah | https://github.com/cliffordwolf/icestorm/blob/master/examples/icestick/rs232demo.v | 17:37 |
| tpb | Title: icestorm/rs232demo.v at master · cliffordwolf/icestorm · GitHub (at github.com) | 17:37 |
| elms | yeah and was running into an issue that I think is related to modeling the ram behavior | 17:37 |
| elms | daveshah: thanks | 17:37 |
| mithro | Yeah we should get that in | 17:37 |
| mithro | elms: While daveshah is around you should ask him about that | 17:37 |
| daveshah | I am not exactly sure how the equivalence check works with RAM | 17:37 |
| mithro | elms: And he can tell you that he doesn't know how equivalence checking works either ;-) | 17:38 |
| daveshah | Beware it does also test for a limited number of cycles without temporal induction | 17:38 |
| daveshah | IMO simulation would probably be better | 17:38 |
| daveshah | I suspect the problem is the SB_RAM40_4K | 17:38 |
| elms | That's what I was going to try next. | 17:38 |
| elms | simulation I mean. | 17:39 |
| daveshah | Another nasty gotcha, that all of the demo stuff deals with but you must beware of, is in hardware initialised BRAM will be invalid until about 30 cycles after startup | 17:40 |
| mithro | elms: Any luck with that simulation? | 20:20 |
| elms | mithro: some. I have it on my laptop and will lookmore during my flight. | 21:29 |
| mithro | elms: Okay - can you send me an email with the status when you stop working for the holidays? | 21:29 |
| elms | Will do | 21:31 |
Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!