| Rev |
Log message |
Author |
Age |
Path |
| 28 |
switched to work on the v1 production GECKO3main. the timing is a big issue
with the prototype system. the cabeling and the possibly bad influence of the
prototype board layout makes it impossible to run this core in sync with the
host transfers.
out transfer is fully working. always in sync with the transfers and reads all
tmc headers correctly.
abort handling in the core works as well. there were some fixes needed in the
firmware, not all IN FIFO buffer where flushed.
a lot of work was done for the IN transfer (fpga to pc). the in transfer
handling in GECKO3COM_simple_test.vhd is finished.
the GECKO3COM_simple_datapath/fsm is finished.
there is still an issue left but I think the problem is in the gpif_com
module. for long IN transfers, we still not receive the correct number of bytes.
it works great for all sort of short transfers.
Here I present a time measurement to show the achieved message throughput:
(Response message was 1 byte, total with header and align bytes was 16 byte)
time for i in {0..100000}; do cat /dev/usbtmc1 > /dev/null; done
real 5m45.706s
user 0m27.498s
sys 4m52.676s
This shows that we can read data from the fpga with a rate up to 290 Hz |
nussgipfel |
5205d 10h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com_fsm.vhd |
| 27 |
some of the problems are fixed now, some still remain.
all transfersize counter where wrong. they have to count the correct number of bytes transfered.
handling of a "short word" added, to be shure that all data (also when it is shorter than a word)
is transfered.
corrected some errors with the fifo write enable signals. |
nussgipfel |
5207d 15h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com_fsm.vhd |
| 21 |
system is fully working and tested with big files over 4 Gbyte. there were some smaller bugs and glitches which caused
sometimes a deadlock. they are fixed now.
one was because the fx2 reads the done pin sometimes wrong during a long data
transfer. I don't now if this is because of crosstalk on my prototyping system or fpga internal weakness.
a simple debouncing (read again after 40 ms) did the trick.
changed firmware version to 0.5rc0, basically all features targeted for 0.5 are done. |
nussgipfel |
5234d 10h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com_fsm.vhd |
| 20 |
basic synchronous IN (fpga to host) transfer works.
detail changes:
-the scpi command "fpga:data" checks now, if the fpga is configured before it changes the context to the fpga. if the fpga is
not configured, it returns an "EXECUTE ERROR".
in the same way, the main_loop checks if the fpga looses his configuration. it disables the GPIF, switches the context back
to the fx2 if so. this is mainly to avoid undeterministic behaviour if you reconfigure the fpga via jtag.
-introduced the new signal "EOM" end of message from the "usb tmc protokoll interpreter" to the gpif_com module
-changed the GPIF waveform for the FIFO IN transfer to the new scheme.
-implemented the same waveform into the gpif_com_fsm.vhd. works well together.
-bugfixed the gpif_com_test.vhd. sends the new EOM signal, the response message generator works now as it should.
-added the missing AUTHORS README and COPYING (license) files to the core directory. |
nussgipfel |
5235d 08h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com_fsm.vhd |
| 19 |
found a pinning error in the series production documentation. fixed in the wiki and the core.
really stable host to fpga data transfer achieved. several time over 4Gbyte transfered. including switching to fpga to
host GPIF waveform to listen to possible fpga transfer tries. but fpga to host transfers still not finished.
the correct length of data is read from the gpif_fsm now. still untested data consitency when data flow throtling occours.
need bigger buffers to test this. will be done on a higher level. |
nussgipfel |
5243d 10h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com_fsm.vhd |
| 18 |
I achieved now stable OUT transfers (from the PC to the FPGA) with working throtling (when the FPGA consumes data slower than the host delivers).
The basics needed for this are implemented in the FPGA like handshaking with the FX2 and clock domain transistion from the Interface clock to the user defined system clock.
in the gpif_com_test.vhd is a message rom, containing a prepared answer message to generate an IN transfer. this is needed for the next step. |
nussgipfel |
5246d 18h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com_fsm.vhd |
| 14 |
reorganising and renaming the stuff in these project.
the core will get the name "GECKO3COM_" followed by the type "simple", "plb" or "opb"
to follow the naming in the GECKO3 wiki and to show the IP core interface.
duplicated fifo corecenerator files are merged together including a wrapper to easily supress synthesizer warnings
from unavailable, unused and unconnected pins.
the project is now organised in a way how the IP core and it's parts are beeing used. this means that the
low-level gpif access module is instantiated by the higher level modules and not the other way around.
this will make more sense when more parts of this IP core are finished (planning is finished, they have
to be implemented and tested now). |
nussgipfel |
5274d 15h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com_fsm.vhd |
| 12 |
this is the version 0.4 of the GECKO3com IP core.
This is the last version from Andreas Habegger, the maintainer for this core is now Christoph Zimmermann |
nussgipfel |
5289d 18h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com.vhd |
| 11 |
initial add of the version 0.3 of the GECKO3com IP core. originaly developped by Andreas Habegger.
this commit is just for backup and to make the project progress visible because version 0.4 is also ready and will be commited in the next step. |
nussgipfel |
5289d 18h |
/gecko3/trunk/GECKO3COM/gecko3com-ip/core/gpif_com.vhd |
