>>>>> "Florent" == Florent Kermarrec <[email protected]> writes:
Florent> Hi, as tested by Sebastien, programmation of the bitstream on
Florent> the KC705 take a long time with XC3SPROG default configuration:
Florent> Using Libftdi, Using JTAG frequency 1.500 MHz from undivided
Florent> clock JTAG chainpos: 0 Device IDCODE = 0x33651093 Desc:
Florent> XC7K325T Created from NCD file: top;UserID=0XFFFFFFFF Target
Florent> device: 7k325tffg900 Created: 2014/08/20 12:38:28 Bitstream
Florent> length: 91548896 bits done. Programming time 61919.7 ms
Florent> This is due the size of the bistream on the KC705 and the
Florent> 1.500MHz clock used.
Florent> Frequency can be changed with the -J option, but if it's
Florent> greater than max frequency of the cable it will use that one.
Florent> Strangely in cablelist.txt we see: # Use 1500000 for all cable
Florent> connected cables and max for all on board cables [...] jtaghs1
Florent> ftdi 1500000 0x0403:0x6010:Digilent Adept USB
Florent> Device:0:0x80:0x80:0x00:0x0
Florent> while digilent say it can support 30MHz:
Florent> http://www.digilentinc.com/Data/Products/JTAG-HS1/JTAG-HS1_rm.pdf
Florent> with: jtaghs1 ftdi 3000000 0x0403:0x6010:Digilent
Florent> Adept USB Device:0:0x80:0x80:0x00:0x0
Florent> we are then able to use higher frequencies, but 7.5MHz seems to
Florent> be already a good compromise:
Florent> Using Libftdi, Using JTAG frequency 7.500 MHz from undivided
Florent> clock JTAG chainpos: 0 Device IDCODE = 0x33651093 Desc:
Florent> XC7K325T Created from NCD file: top;UserID=0XFFFFFFFF Target
Florent> device: 7k325tffg900 Created: 2014/08/20 12:38:28 Bitstream
Florent> length: 91548896 bits done. Programming time 13395.4 ms
Florent> It seems that the cable database can be defined in a
Florent> environment variable ($CABLEDB). What do you suggest? Include
Florent> our cablelist.txt in mibuild (with only the cable we used) and
Florent> setting it as an environment variable in XC3SProg __init__?
The xc3sprog manpage tells the way the cablelist is found:
- if $CABLEDB exists and the file $CABLEDB points to, use that file
- otherwise if cablelist.txt exists in current directory, use it
- else fall back to built in cablelist.txt.
If the FT2232H is on board and there are no signal integrity issues you can
probably also go to even higher rates.
I suggest to test some configuration and then fixing it in
cablelist.txt. Please feed the list than also to me to include in
sourceforge xc3sprog upstream repo.
Florent> Now we can also compress the generated bitstream with vivado,
Florent> which give us:
Florent> Using Libftdi, Using JTAG frequency 7.500 MHz from undivided
Florent> clock JTAG chainpos: 0 Device IDCODE = 0x33651093 Desc:
Florent> XC7K325T Created from NCD file: top;UserID=0XFFFFFFFF Target
Florent> device: 7k325tffg900 Created: 2014/08/20 16:56:31 Bitstream
Florent> length: 18139744 bits done. Programming time 2490.7 ms
Florent> The patch in attachment allows bitstream compression.
The patch doesn't apply to https://svn.code.sf.net/p/xc3sprog/code/trunk.
What else was changed by M-LABS? I didn't see any patches sent for inclusion
in upstream.
Bye
--
Uwe Bonnes [email protected]
Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt
--------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------
_______________________________________________
M-Labs devel mailing list
[email protected]
https://ssl.serverraum.org/lists/listinfo/devel
