_florent_ changed the topic of #litex to: LiteX FPGA SoC builder and Cores / Github : https://github.com/enjoy-digital, https://github.com/litex-hub / Logs: https://libera.irclog.whitequark.org/litex
tpb has quit [Remote host closed the connection]
tpb has joined #litex
Emantor has quit [Quit: ZNC - http://znc.in]
Emantor has joined #litex
Degi has quit [Ping timeout: 246 seconds]
Degi has joined #litex
Stary has quit [*.net *.split]
mobius has quit [*.net *.split]
jtf has quit [*.net *.split]
mobius has joined #litex
jtf has joined #litex
Stary has joined #litex
_whitelogger has joined #litex
tucanae47 has joined #litex
yootis has joined #litex
_alice has joined #litex
mithro has joined #litex
x56_ has joined #litex
<_florent_> somlo: The aim of these changes is to allow fixed mapping of the register offsets and one use case are the linux drivers.
<_florent_> somlo: LiteEth is collecting the CSR differently and this case does not seems to be covered correctly, I'll fix this, sorry for that.
shoragan has quit [Ping timeout: 264 seconds]
mtretter has quit [Ping timeout: 268 seconds]
shoragan has joined #litex
mtretter has joined #litex
<_florent_> somlo: This is good now (compared the generated mappings with your example)
<DerekKozel[m]> Thanks for this feature
cr1901 has quit [Remote host closed the connection]
<_florent_> DerekKozel[m]: This was required for quite some time :) (and isn't finished), next I'll prepare the design we were discussing for you GnuRadio tests.
<_florent_> you/your
_alice is now known as lili
hrberg has quit [Ping timeout: 246 seconds]
toshywoshy has quit [Ping timeout: 248 seconds]
<somlo> _florent_: thanks, it looks fine now (getting the same offsets as before)
toshywoshy has joined #litex
cr1901 has joined #litex
<_florent_> DerekKozel[m]: An example of minimal ComputeEngine for a LitePCIe design: https://github.com/enjoy-digital/acorn_pcie_compute_test/commit/b226ccad7049f71ef4485a256ea78703f82500c7
<_florent_> DerekKozel[m]: With this, you can use litepcie_test play to inject data and litepcie_test record to get back the modified data
<_florent_> DerekKozel[m]: So you can pipeline different engines on the same DMA
<_florent_> DerekKozel[m]: but can also have several independant DMAs, each with its own processing pipeline
<_florent_> DerekKozel[m]: To configure the number of DMAs in the design: you can use the ndmas of add_pcie
<_florent_> DerekKozel[m]: and you can select the DMA with litepcie_test with -c X
<_florent_> DerekKozel[m]: In the ComputeEngine, the sink/source endpoint are LiteX streams which are very similar to AXI stream
<_florent_> Derivating from stream.PipelinedActor avoid having to handle the control signals manually (valid/ready), you can just focus on the data processing.
<DerekKozel[m]> Excellent
<DerekKozel[m]> The Pipeline class/concept I had missed before, that is ideal and so simple. I'll get that running and building dynamically.
<DerekKozel[m]> Thank you
<DerekKozel[m]> I can easily stay busy for a week or two with this example. Registers and Verilog will be the two most useful things after that.
toshywoshy has quit [Ping timeout: 276 seconds]
toshywoshy has joined #litex