This changes the SoC interconnect such that the main 64-bit wishbone out
of the processor is first split between only 3 slaves (BRAM, DRAM and a
general "IO" bus) instead of all the slaves in the SoC.
The IO bus leg is then latched and down-converted to 32 bits data width,
before going through a second address decoder for the various IO devices.
This significantly reduces routing and timing pressure on the main bus,
allowing to get rid of frequent timing violations when synthetizing on
small'ish FPGAs such as the Artix-7 35T found on the original Arty board.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
When using litedram, request a much longer PLL reset. This seems to
help get rid of all the grabled output after config.
Also use the clean system_rst out of litedram as our source of reset
for the rest of the SoC (it is synchronized with system_clk and takes
pll_locked into account already)
In some cases we need to keep the reset held for much longer,
so use counters rather than shift registers.
Additionally, some signals such as ext_rst and pll_locked
or signals going from the ext_clk domain to the pll_clk
domain need to be treated as async, and testing them without
synchronizers is asking for trouble.
Finally, make the external reset also reset the PLL.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Things have changed a bit in upstream LiteX. LiteDRAM now exposes a
wishbone for the CSRs for example.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The rx synchronizers were ... non existent. Someone forgot to add
a if rising_edge(clk) to the process.
For tx, ensure that we have a default value so that TX stays high
from TPGA configuration to the reset being sampled on the first clock
cycle.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The pp_fifo decides whether top = bottom means empty or full based
on whether the previous operation was a push or a pop.
If the fifo performs both in one cycle, it sets the previous op to
pop. That means that a full fifo being added a character and removed
one at the same time becomes empty.
Instead, just leave the previous op alone. If the fifo was empty, it
remains so, if it was full ditto.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This uses the new header files for register definitions and
extracts the core frequency from syscon rather than hard coding it.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
These provides some info about the SoC (though it's still somewhat
incomplete and needs more work, see comments).
There's also a control register for selecting DRAM vs. BRAM at 0
(and for soft-resetting the SoC but that isn't wired up yet).
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The old toplevel.vhdl becomes top-generic.vhdl, which is to be used
by platforms that do not have a litedram option.
Arty has its own top-arty.vhdl which supports litedram and is now
hooked up
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Micropython has been able to fit into 384kB for ages, so lets reduce our
simulated RAM. This is useful for testing if micropython will run on an
ECP5 85k, which has enough BRAM for 384kB but not enough for 512kB.
Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
This replaces the simple_ram_behavioural and mw_soc_memory modules
with a common wishbone_bram_wrapper.vhdl that interfaces the
pipelined WB with a lower-level RAM module, along with an FPGA
and a sim variants of the latter.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This adds an output buffer to help with timing and allows the BRAMs
to actually pipeline.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Vivado by default tries to flatten the module hierarchy to improve
placement and timing. However this makes debugging timing issues
really hard as the net names in the timing report can be pretty
bogus.
This adds a generic that can be used to control attributes to stop
vivado from flattening the main core components. The resulting design
will have worst timing overall but it will be easier to understand
what the worst timing path are and address them.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This doesn't yet pipeline the block RAM, just generate a valid stall
signal so it's compatible with a pipelined master
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The shared variable used for FIFO memory is not VHDL 2008 compliant.
I can't see why it needs to be a shared variable since reads and writes
update top and bottom synchronously, meaning they don't need same cycle
access to the FIFO memory.
Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
We can now pass both the input clock and target clock frequency
via generics. Add support for both 50Mhz and 100Mhz target freqs
for both cases.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The memory wishbone doesn't clear ACK and move the state machine on
until STB is de-asserted. This seems like it isn't compliant with
the spec and results in a maximum throughput of 1 transfer every
3 cycles.
Fixing this improves the situation to one transfer every 2 cycles.
Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
(And rename it to mw_soc_memory).
This makes soc.vhdl simpler and provides the same interface as
the simulated memory, which will help when sharing soc.vhdl
with sim later
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This will be useful when we start needing different toplevels for
different boards.
We keep the reset and clock generators in the toplevel as they will
eventually be taken over by litedram when we integrate it, and they
are more likely to change on different system types.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This adds support for the Digilane Cmod A7-35.
I had to use the MMCM because the clock (12 MHz) is below the PLL
minimum of 19 MHz.
Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
The old reset code was overly complicated and never worked properly.
Replace it with a simpler sequence that uses a couple of shift registers
to assert resets:
- Wait a number of external clock cycles before removing reset from
the PLL.
- After the PLL locks and the external reset button isn't pressed,
wait a number of PLL clock cycles before removing reset from the SOC.
Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
Mikey points out that our stack grows down from 512kB and our
heap is below that too, so we can reduce our BRAM requirements,
which allowing some smaller FPGA boards to work. Not sure why
I thought we were using memory between 512kB and 1MB.
Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
Benjamin Herrenschmidt