| GentlemanEnginee | Hello | 03:43 |
|---|---|---|
| juliusb | GentlemanEnginee: hi | 11:18 |
| stekern | juliusb: I'm maybe missing something, but what would be the big advantage over just setting breakpoints in the exception vector? | 11:42 |
| juliusb | well you could do that, I guess | 11:48 |
| juliusb | but this then relies on not even having to do that | 11:48 |
| juliusb | but you're right, you could | 11:48 |
| juliusb | but for some reason I think it is even better, though, let me think about it :) | 11:49 |
| stekern | yeah, I'm not dismissing it completely, just couldn't come up with an use case ;) | 12:23 |
| juliusb | well, say you don't trust your system to gracefully recover from a bus error or something | 13:57 |
| juliusb | you may not ever properly fetch that l.trap | 13:57 |
| juliusb | then you can probe the system via the debug bus in its broken state | 13:58 |
| juliusb | but I like the idea that by default the thing will halt on any condition which will mean the software can't properly execute | 14:00 |
| juliusb | and you can set that up in hardware | 14:01 |
| juliusb | also, if you're executing from ROM you may not have enough HW breakpoints to set up one for each exception vector | 14:01 |
| juliusb | (we have none in mor1kx atm :P) | 14:02 |
| stekern | I value watchpoints over hw breakpoints any time | 14:09 |
| juliusb | we have no hardware support for those either :-/ | 14:23 |
| juliusb | atm, anyway :) | 14:23 |
| stekern | I know | 14:26 |
| stekern | point being, watchpoints should get attention before hw breakpoints if at all =D | 14:27 |
| stekern | in the meantime, blink a led | 14:27 |
| stekern | or just debug things with a waveform and objdump | 14:28 |
| LoneTech | there's a bit of similarity between a breakpoint and a watchpoint unit placed on the instruction bus | 14:30 |
| stekern | true true | 14:30 |
| LoneTech | btw, we do use the stall on exception function - it's how the hardware debug unit uses software breakpoints. not terribly nice, as the break instruction replaced something in the pipeline | 14:33 |
| juliusb | LoneTech: sure, but the point is in the DSR we can make it stop on any exception, we just make it stop on l.trap exception to achieve the software breakpoint goal. | 15:15 |
| juliusb | well, we only use one of them. | 15:15 |
| LoneTech | true | 15:22 |
| -!- Netsplit *.net <-> *.split quits: zhai365, Amadiro | 16:31 | |
| -!- Netsplit over, joins: zhai365 | 16:31 | |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!