| --- Log opened Thu Nov 06 00:00:16 2014 | ||
| poke53282 | libxul.so: 93MB. Not really better | 01:24 |
|---|---|---|
| dalias | :) | 01:31 |
| stekern | poke53282: yeah, I know, there's that limit. shouldn't be insanely hard to fix. | 05:36 |
| stekern | isn't that's what the difference between -fpic and -fPIC is for btw? | 05:36 |
| stekern | -fpic is short range and -fPIC is long range | 05:39 |
| poke53282 | I didn't know, that there is a difference. | 05:46 |
| poke53282 | Never heard of a difference between pic and PIC. | 05:46 |
| olofk | Anyone handy with PDF reformatting here? I have a PDF document with portrait pages, but the content is landscape, so when I print it there are huge borders | 08:39 |
| olofk | ok, crappy solution of the month. Import pdf into Libreoffice Draw. Change page size from portrait a4 to landscape a5. Manually move the contents back into the visible area, export as PDF and upscale to A4 in PDF viewer print dialog | 09:26 |
| olofk | I feel like a windows user :( | 09:26 |
| olofk | maxpaln: Read about your bfm improvements. Good news. Got anything to share? | 09:34 |
| stekern | olofk: but a power windows user | 10:29 |
| stekern | my wife would have took a photo of the screen and then printed that | 10:30 |
| maxpaln | olofk: yes, I think so - my testing has been limited so I would say there is a fair chance of there being bugs and/or cases that I haven't thought of. I have also canibalised several of the existing functions for burst write and read from wb_bfm_master so at the moment you can't have both sets of files compiled at the same time. But it appears to be behaving correctly. I'll package something | 10:45 |
| maxpaln | up shortly and share (probably via dropbox or something unless there is a preferred route) | 10:45 |
| olofk | maxpaln: Dropbox is fine. Just interested in trying it out for now. We can clean it up later | 11:03 |
| olofk | stekern: She would probably have been done faster and spent less time crying than I did | 11:03 |
| -!- aburgess_ is now known as aburgess | 11:14 | |
| maxpaln | olofk: here is a zip of the backend directory (you should recognise stuff in here - it's the same directory structure as you pointed me at ages ago for your original code). I've also included my testbench as it shows how i've instantiated the testbench. There are some top level parameters for controlling transaction and subtransaction numbers. I'll be interested to see if it works for you | 11:34 |
| maxpaln | too!! :-) | 11:34 |
| maxpaln | I need to pop out for a few hours - I'll be back online to answer questions then | 11:34 |
| maxpaln | Oooops, forgot the link - here it is: https://www.dropbox.com/s/xvwvo6va9cvpudr/backend.rar?dl=0 | 11:34 |
| olofk | That was my first question :) | 11:35 |
| olofk | I'll tell you when I have taken it for a spin | 11:35 |
| maxpaln | I've added several levels of verbosity to the output debug - setting VERBOSE to 3 gives you the most, Its probably needed for debugging problems | 11:38 |
| olofk | maxpaln: Great. I like multiple verbosity levels | 11:55 |
| olofk | maxpaln: I see that you are using 2d arrays in the task headers. I'm a bit worried that the tools we use don't support enough systemverilog yet | 12:35 |
| olofk | Supporting 2d arrays as wires and regs, but not in module, function or task headers is on my top 5 list of retarded design decisions in verilog | 12:37 |
| LoneTech | is the implicit declaration up there too? | 12:43 |
| olofk | LoneTech: Absolutely | 12:54 |
| maxpaln | olofk: those tasks are proving to be a PITA - they contain a mixture of blocking and non-blocking assignments, it was a pain to unpick the sequence of events | 13:24 |
| maxpaln | I didn't realise 2d arrays in task headers was a systemverilog thing - | 13:25 |
| maxpaln | one of my favourite additions to Verilog 2000 was the none qualifier for the `default_nettype directive | 13:28 |
| maxpaln | it eliminates impicit declarations! :-) | 13:28 |
| LoneTech | that is actually in verilog 2005, though I suspect implementation is spotty, as with `unconnected_drive | 13:29 |
| maxpaln | I've found it works well in Modelsim, Active-HDL etc. I am not sure about broader support. | 13:36 |
| maxpaln | olofk: are you able to use the code with 2d arrays? It may be possible to move them to the module declarations if not. | 13:36 |
| maxpaln | FYI, I am finding this new randomised test very enlightening - 3 corner case bugs found now!!! | 15:08 |
| juliusb | constrained random FTW | 15:45 |
| maxpaln | it's almost like those verification guys know what they're talking about :-) | 15:51 |
| poke53282 | olofk: Try inkscape | 17:14 |
| poke53282 | stekern: Can you try this code please: http://pastie.org/9700857 | 18:43 |
| poke53282 | Might be a problem with my filesystem. | 18:43 |
| poke53282 | you might have to increase the name field "char name[64]" | 18:48 |
| olofk | maxpaln: Yes. The BFM code is extremely unreadable, and I'm finding it very hard to debug. Not my proudest moment :) | 20:19 |
| olofk | maxpaln: Both icarus and my Modelsim Altera edition chokes on the syntax :( | 20:44 |
| olofk | mafm: Adding the -sv switch to modelsim works better | 21:13 |
| olofk | I mean maxpaln | 21:14 |
| olofk | I was hoping that Icarus would handle enough sv by now :/ | 21:14 |
| olofk | Using Icarus from git doesn't help | 21:55 |
| --- Log closed Fri Nov 07 00:00:17 2014 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!