<pepijndevos>
Seems all build targets just yeet them at the end of the constraints
scientes has joined #litex
scientes has quit [Changing host]
<pepijndevos>
uhoh ERROR (RP0002) : The number(31) of BSRAM in the design exceeds the resource limit(26) of current device. And RAM_STYLE maybe the useful user assignment to change the inference result
<pepijndevos>
that's simple.py
<pepijndevos>
It apparently has 468K bits of bram, while SoCCore has 0x2000 bytes? So 65536 bits? Yea I can see how that wouldn't fit...
<pepijndevos>
wait no... confusion...
<zyp>
0x2000 is 8kB or 64 kbits
<pepijndevos>
yea so 468K is like 10 times more right... if i'm not completely mad
<zyp>
sounds right, yes
<pepijndevos>
Maybe it's using a whole BRAM for every tiny array in there... but hard to tell
<pepijndevos>
I also can't find out what I may do with RAM_STYLE
<pepijndevos>
I get that may cause some things to synth to lutram which sounds great... but how
<pepijndevos>
hurray, I made a bunch of random brams in vexrisc distributed and we'll see... it's veeeery tight
<pepijndevos>
Like, it uses 26 out of 26 brams and if i use one more lutram it runs out of those haha
<_florent_>
pepijndevos: but it's easier to start with the ROM in the FPGA when possible
<pepijndevos>
oooh okay :)
<pepijndevos>
and uh... on the ULX3S I have to press the reset button before it goes into lxterm, but on simple.py I don't see anything like a reset button configured, which bits take care of that?
<pepijndevos>
CHASER!!
<pepijndevos>
also with openfpgaloader... but still gives fail
<pepijndevos>
I'm just stuck at [LXTERM] Starting.... now
<_florent_>
you can use something like: CRG(platform.request(platform.default_clk_name), platform.request("rst))
<pepijndevos>
getting so cloooose to working soc...
<pepijndevos>
I should probably try the spi flash trick... how do you get the rom into the spi though... I assume that requires some trickery with the programmer.
<pepijndevos>
oops, rst is inverted...
<pepijndevos>
works!!! now some hello world...
<pepijndevos>
bleg... the fpga_101 expects a common makefile which does not seem to be there
<pepijndevos>
okay I got a binary from icerbreaker examples... but, still no cigar.
<pepijndevos>
When I press reset it does the LiteX header, of which the last line is "BIOS built on Sep 28 2020 17:00:16"
<pepijndevos>
Then nothing.
<pepijndevos>
I've passed the binary with --kernel
<_florent_>
pepijndevos: sorry i'm not sure to understand, do you have the ROM on BRAM? are you able to get the LiteX BIOS working?
<pepijndevos>
_florent_, I now have it in bram with a few things moved to lutram
<pepijndevos>
I'm not sure when the bios is considered "working", it displays an ascii art and some copyright and version info, but does not download the kernel or offer a prompt
<_florent_>
it's not fully working if you are not able to get the prompt
<pepijndevos>
alright, so just "lxterm /dev/ttyUSB0" should get me a prompt right?
<_florent_>
which CPU are you using? VexRiscv? if so could you try with SERV (this will reduce resource usage and use polling mode for UART)
<_florent_>
yes
<pepijndevos>
Yea, I was about to try a different CPU.
<pepijndevos>
Maybe VexRisc doesn't like some of its bits moved to lutram hehe
<_florent_>
ah also, if you change the CPU, be sure to remove your build directory first :) (IIRC you already had this issue when playing with the ULX3S & Rust)
<pepijndevos>
haha yea i had exactly this problem again
<pepijndevos>
heyyyy! now the bios works
<pepijndevos>
but my blinky does not... :((((
<lkcl>
_florent_: one thing that's missing is the SDRAM dqm 0/1 bits
<tpb>
Title: GitHub - litex-hub/linux at litex-rocket-rebase (at github.com)
<somlo>
started with the Antmicro driver, made it portable for 32-bit csr-data-width, fixed sdcard clocking, am generally able to mount and read/write a sdcard
<somlo>
major remaining wart is that the sdcard must be plugged in during kernel boot, and ejecting and re-inserting it is not currently handled
<somlo>
(although I can see the gateware CD register holding the correct value in the bios, so most likely additional scrubbing of the linux driver is needed)
<somlo>
right now the litex_mmc commits are "raw" (reflecting my haphazard hacking of the antmicro starting point). Once I get to the bottom of the card-detect problem, I'll do a clean-up rebase to "hide the sausage-making process" :)