<lekernel>
well, don't go there if you expect top-notch hardware hacking, but from a social/political point of view, it's very interesting =]
<lekernel>
it's a busy 11-story squatted former bank
<lekernel>
artists, researchers (yeah, there are even some academic people there), musicians, squatters, hackers ...
<Fallenou>
humm ok
<Fallenou>
has the seat 42
<lekernel>
I definitely recommend it
<Fallenou>
I try not to stand out of the crowd in London
<Fallenou>
I suppose you understand why :p
<lekernel>
jackgassett: I think you should be able to rewrite the proprietary bram2bit fairly easily in your project
<lekernel>
in fact many of the bitstream elements required for this purpose are documented by xilinx
<lekernel>
BRAM data doesn't seem to be intertwined with other configuration data (interconnect, slices, etc.)
<lekernel>
instead you have commands at the end of the bitstream that write configuration data with special values of the frame address register (FAR) - this configuration data simply gets pushed into the block RAM
<lekernel>
and the FAR value tells which BRAM element to write to
<lekernel>
you can simply get the FAR values for your placed/routed design by filling the BRAM with random data and examining what gets written where
<lekernel>
you'll also need to generate a frame CRC, but fortunately this is documented for spartan 3 (but not 6 :/)
<Fallenou>
from France
<kristianpaul>
Fallenou: salut :-)
<kristianpaul>
Okay,i had tested with the scope the memory card using TP, i can confirm CLK (159Khz) also visualized square signals coming from CMD when bios run, later with rtems
<kristianpaul>
but DAT0-3 _always_ are up (3.2V)..
<kristianpaul>
i guess at least if any response from card is coming it should appear in DAT0, but nothing hapened with both toshiba 512B and 1Gb Nokia cards..
<Fallenou>
yop kristianpaul :)
<kristianpaul>
Fallenou: btw are you moving to he US finally to work with intel?
<Fallenou>
no no I'm not
<Fallenou>
I went to London instead of US
<Fallenou>
maybe later :) dunno !
<Fallenou>
I didn't studied yet the list of interesting companies I'd like to work with/at
<Fallenou>
I will do that when I will have my diploma :p
<CIA-43>
antares: Sebastien Bourdeauducq master * rfdd2b9a / (6 files): place: new placer beginning to work (wip, buggy, no flip flops) - http://bit.ly/e7ykyH
<Fallenou>
:)
<kristianpaul>
oh
<lekernel>
#{~! this placer thing is so messy to write
<lekernel>
I think we need a tool akin to lemon/re2c/flex/bison but which processes fpga architecture descriptions and writes the messy parts of placer/packer systems
<Fallenou>
hehe flex bison is quite awesom :)
<lekernel>
now we also need to make it fast...
<Fallenou>
to write big parsers for you
<lekernel>
you should try lemon+re2c
<lekernel>
cleaner than the GNU stuff
<Fallenou>
oh ok
<lekernel>
the GNU tools use many GNUisms which make non-portable parsers (do you see a pattern here?), which also are non-reentrant by default (and making them reentrant is messy)
<lekernel>
also, lemon+re2c are simple things which do not require autocrap to build
<lekernel>
you could run a lemon+re2c parser on a microcontroller... now good luck with the GNU things
<kristianpaul>
oh, really?
<lekernel>
sure, I did it on AEMB/LM32 even without a C library
<kristianpaul>
:o
<kristianpaul>
without a C library, looks even more portable it seems then
<Fallenou>
maybe they reinvented the wheel
<Fallenou>
you are always more likely to be portale reinventing the wheel
<lekernel>
well, in fact, they do use a few C library functions. what I meant is without newlib/glibc/etc.
<kristianpaul>
ah ok :-)
<lekernel>
at the beginning I tried with the GNU tools but there were endless problems
<Fallenou>
ok
<lekernel>
lemon+re2c went without a hitch
<kristianpaul>
lekernel: still interested on a gcc replacement? ;)
<lekernel>
it's not something i'm going to spend time on, but yes