The memory mapped AXI interfaces for the AXI-DMAC are uni-directional.
Which means they are either write-only or read-only. Unfortunately the
Altera tools can't handle this, so we had to add dummy signals for the
other direction.
The Xilinx tools on the other hand handle uni-directional AXI interfaces
and in fact IPI can do a better job and use less resources when creating
the AXI interconnects when it knows that the interface is uni-directional.
So always disable the dummy ports for the IPI package.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
The memory mapped AXI interfaces for the AXI-DMAC are uni-directional.
Which means they are either write-only or read-only. Unfortunately the
Altera tools can't handle this, so we had to add dummy signals for the
other direction.
The Xilinx tools on the other hand handle uni-directional AXI interfaces
and in fact IPI can do a better job and use less resources when creating
the AXI interconnects when it knows that the interface is uni-directional.
So always disable the dummy ports for the IPI package.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Add .gitattributes file which sets up the eol encoding handling. This will
make sure that we get a uniform eol encoding across different operating
systems.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Add .gitattributes file which sets up the eol encoding handling. This will
make sure that we get a uniform eol encoding across different operating
systems.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
This control signal can be overwritten by the up_axis_xlast/up_axis_xlast_en bits, in order to create a single stream, which is contains multiple streams.
This can be use to fill up the DACFIFO module.
While the reset for the memory mapped AXI master is synchronous to some
clock it is not necessarily synchronous to the clock used for that
interface. So always generate a local reset signal to avoid problems that
could result from this.
While we are at it also update the code to only generate a local reset if
the interface is asynchronous to the register map, otherwise use the
register map reset.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Ignore the timing path from the current DMA address to the register map,
this is just a debug signal at the moment.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
If the internal FIFO is larger than one block ram there will be multiple
BRAMs called ram_reg[0], ram_reg[1]. Modify the BRAM constraint rule so that
it matches these as well.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
The src_response_fifo has been removed from the design, but we still need to
assert the ready and empty control signals for things to work properly.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
We really only want to apply the CDC constraints if the clocks are actually
asynchronous. Unfortunately we can't use if ... inside a xdc script. But we
can use expr which has support for a ? b : c if-like expression. We can use
that to create helper variables that contains valid clock when the clock
domains are asynchronous or {} if they are not. Passing {} as
set_false_path/set_max_delay as either the source or destination will cause
it to abort and no constraints will be added.
Also add -quiet parameters to avoid generating warning if the constraints
could not be added.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
All the FPGA internal control signals are active high, using a active low
reset inserts a extra invert LUT. By using a active high reset we can avoid
that.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Clean ran for a project will clean only the project files.
Clean-all ran for a project will clean also the library files on which the project depends.
The clean commands will only remove specific files and directories.
The top Makefile allows several options(per suggestion from jameyhicks):
make fmcomms1.zed will run "make all" in projects/fmcomms1/zed/
make clean will run "make clean" for all the projects
make clean-all will run "make clean" for all the projects and libraries
make lib will run "make all" for the library files
Instead of just marking all clock domains as asynchronous set the
appropriate constraints for each CDC path.
For single-bit synchronizers use set_false_path to not constraint the path
at at all.
For multi-bit synchronizers as used for gray counters use set_max_delay with
the source clock period domain to make sure that the signal skew will not
exceed one clock period. Otherwise one bit might overtake another and the
synchronizer no longer works correctly.
For multi-bit synchronizers implemented with hold registers use
set_max_delay with the target clock period to make sure that the skew does
not get to large, otherwise we might violate setup and hold time.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Bring back the AXIS FIFO as a separate module instead of embedding it into
the DMAC module. This makes it possible to use it in other modules outside
of the DMAC.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Move the axi_repack block to its own module. This allows it to use it
outside of the DMA controller.
Also rename it to util_axis_resize to better reflect its function.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Modify the adi_add_bus() function to take the full vlnv strings instead of just the bus type.
This makes the function more flexible and e.g. allows to handle buses from other vendors.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
When the DMA controller gets disabled in the middle of a transfer it is
possible that the resize block contains a partial sample. Starting the next
transfer the partial sample will appear the begining of the new stream and
also cause a channel shift.
To avoid this make sure to reset and flush the resize blocks when the DMA
controller is disabled.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
We need to make sure to not prematurely de-assert the s_valid signal for the
request splitter when disabling the DMAC. Otherwise it is possible that
under certain conditions the DMAC is disabled with a partially accepted
request and when it is enabled again it will continue in an inconsistent
state which can lead to transfer corruption or pipeline stalls.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
All components should use the internal 'do_enable' signal instead of the
external 'enable' signal. The former correctly incorporates the shutdown
sequence and does not get asserted again until the shutdown has been
completed. Using the external signal can cause problems when it is disabled
and enabled again in close proximity.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Currently when the DMAC gets disabled the request_generator will still
generate all remaining burst requests for the currently active transfer.
While these requests will be ignored by the source and destination component
this can still take a fair amount of time for long transfers.
So just stop generating burst requests once the DMAC is being disabled.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Refactor the fifo_inf modules to always correctly generate the underflow and
overflow status signals. Before it was possible that in some cases they
were not generated when they should have been.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Tcl command "if" is not supported by Vivado XDC, therefore the tool throw some critical warnings, and does not
apply the constraints, which can cause timing violations at case of some carriers.
The following solution is much more compact and is supported by the XDC, and more importantly prevents
unwanted critical errors and timing violations.
Vivado 2014.4 is too greedy, when it needs to optimize. See more about the issue here: https://ez.analog.com/thread/48214
The response_dest_resp is unused, so not save to concatenate with a valid signal like the eot.
This silences warnings from the tools about having no clock assigned to the bus.
Also fix the name of the TVALID signal.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
The dummy a{r,w}len fields should have the same width as the real a{w,r}len
fields in order to not break auto AXI bus version detection.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Clear the pipeline when no transfers are active to make sure that we do not
get residual data on the first sample for the next transfer.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
We can remove the Altera toplevel wrapper if we switch the axi4 control bus
to axi4lite and add the few missing signals that are required by the Altera
interconnect to both the control and the data buses.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
This is unused and unneeded. The AXI interconnect will make sure that a
peripheral only gets requests that are meant for it, there is no need to
check the address in the peripheral itself.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
We must make sure that the response ID is the same as the request ID when we
accepted the request. Otherwise we might respond with the wrong ID and the
system will lockup.
Also set rlast to 1 instead of 0.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Make sure that the address generator behaves correctly when the buswidth is not
64-bit. Also since the source and destination can have different widths add
separate parameters for source and destination address alignment.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
We do not know which 'last' condition to use before hand, but we can pre-compute
the result for both conditions and then use them. This removes the comparison
from the already pretty long combinatorial path.
Also simplify a few expressions.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
The read data also becomes available only with a delay of one clock cycle,
sending the ack too early will result in bogus register reads.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
During an ID sync the request_id might increment, we should not generate a
response in this case. Since the ID sync only happens when the core is disabled
check that the core is enabled before generating a response.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
The pending_burst signal and the expression id != request_id are almost
identical. pending_burst goes high with a delay of one clock cycle, but the
important thing is that it goes low on the same clock cycle as the expression.
By using pending_burst here instead of 'id != request_id' we can reduce the
fanout of the 'id' register and improve the timing of the core.
Vivado doesn't handle the case where we use symbolic constants for the default
value properly, so update this to use plain integers.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
There were a few place in the core where it assumed a 64-bit wide bus. Make this
configurable using parameters. The patch also adds support for having different
DMA bus widths on the source and destination side.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
If the address generator is disabled the very same cycle as it tries to put a
new address on the bus, it will keep sending this address forever and the core
will lock up
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Right now there is always a period of one clock cycle where we can not transfer
any data when switching between two transfers. This patch modifies the data
mover to allow for zero latency. This fixes problems on the FMCOMMS1 platform
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Istvan Csomortani
Lars-Peter Clausen
Rejeesh Kutty
Paul Cercueil
Adrian Costina