03:43 <Dar1us> is it possible to get migen to just generate a verilog file for manual integration? If so, does someone have a cookbook for it? Thanks

07:06 <rjo> cr1901_modern: same as matlab/numpy. the indices vary c-contiguously (last fastest) and the product of all dimensions is the shape (needs to match). ravel() is a special case of reshape(), transpose() is a special case of swapaxes().

07:06 <rjo> Dar1us: migen.fhdl.verilog.convert() ?

07:08 <rjo> s/is the shape/is the size/

07:09 <Dar1us> rjo: yeah.. I am struggling to work out the goo around it to make it produce something sensible

07:09 <rjo> cr1901_modern: but i don't expect to see much use of swapaces() and reshape() for more than 2 dimensions.

09:47 <sb0cn> Dar1us, https://github.com/m-labs/migen/blob/master/examples/basic/namer.py#L44

09:48 <sb0cn> or https://github.com/m-labs/migen/blob/master/examples/basic/memory.py

09:48 <sb0cn> the "goo" as you call it is (I suppose) for handling memory initialization files ($readmemh)

10:44 <cr1901_modern> rjo: Ack. Hopefully I memorize the semantics at some point. It's not like I got reshape right without looking it up in Matlab either :P.

11:58 <Dar1us> ah well they left, but that is good goo

12:28 <tpw_rules> rjo: matlab and numpy are very much not the same!! matlab is fortran-contiguous so the first index varies fastest

13:58 <rjo> tpw_rules: yes. i meant that once the layout is defined, then the definition is of reshape() is.

16:17 <ZirconiumX> wq: Is there a way of measuring "code coverage" by simulator tests?

16:18 <whitequark> ZirconiumX: unfortunately not

16:18 <whitequark> if you come up with a metric i could probably implement it fairly easily in 0.2

16:18 <ZirconiumX> I think edge coverage would probably be my choice here

16:19 <ZirconiumX> For example, if you have an m.If, it'd be nice to measure "this 'branch' has been both taken and not taken"

16:20 <ZirconiumX> (I'm well aware they're not actually branches, thus the scare quotes)

16:20 <whitequark> mhmm

16:21 <whitequark> that's very easy to do.

16:21 <whitequark> so you basically want SC/CC>

16:21 <whitequark> *?

16:21 <ZirconiumX> Can you expand the acronyms? I'm not too smart at these things

16:22 <ZirconiumX> CC = cyclomatic complexity?

16:24 <ZirconiumX> My theory behind picking something like that, is that if you have, say, an enum of input types, then edge coverage would require you to specify each one for example

16:34 <whitequark> statement coverage/condition coverage

16:34 <ZirconiumX> Sure