| *** tpb <[email protected]> has joined #f4pga | 00:00 | |
| *** marzoul <[email protected]> has quit IRC (Ping timeout: 240 seconds) | 00:03 | |
| *** cr1901_ <cr1901_!~cr1901@2601:8d:8600:911:cf1:8507:720a:c17> has quit IRC (Remote host closed the connection) | 15:58 | |
| *** cr1901_ <cr1901_!~cr1901@2601:8d:8600:911:cf1:8507:720a:c17> has joined #f4pga | 15:58 | |
| *** cr1901_ is now known as cr1901 | 15:59 | |
| *** marzoul <[email protected]> has joined #f4pga | 19:02 | |
| marzoul | Hi | 20:38 |
|---|---|---|
| marzoul | I have started a very casual attempt at adding Virtex-7 devices to prjxray. | 20:38 |
| marzoul | Where : branch virtex7 in fork https://github.com/marzoul/prjxray | 20:38 |
| marzoul | There is no intention of building a clean and structured patch, it is more unorganized changes that I did in order to make existing fuzzers go as far as possible before a fail is reached. | 20:38 |
| marzoul | The goal is more to identify the significant differences that support of Virtex-7 would bring compared to current prjxray. | 20:38 |
| marzoul | Obviously some of the changes share same goals of independent efforts to support Kintex-7. | 20:39 |
| marzoul | Summary of changes needed : | 20:40 |
| marzoul | - some parts have unbonded IOs / PADs even in largest known package, so iterating over pad names in Vivado will not enumerate all IOBs | 20:40 |
| marzoul | - some IOBs might be bonded to different pad names depending on package ? why not, would need a check | 20:40 |
| marzoul | - some parts do NOT have low-perf pins, only high-perf, so default design generation must be adapted on many fuzzers, along with other device parameters | 20:40 |
| marzoul | - PCIe fuzzers are not designed to handle multiple PCIe sites | 20:40 |
| marzoul | - transceivers are GTXE2, not GTP, need tweaking | 20:40 |
| marzoul | Doing these hacks, I realized that the following directions could be of interest for prjxray : | 20:44 |
| marzoul | - GTXE2 vs GTP could be detected from database itselt (there will be GTH also at some point) - alternatively, define an env var in settings/xx.sh to enable the right fuzzers | 20:44 |
| marzoul | - number of specimens in fuzzers could/should be based on number or primitives of interest in device | 20:44 |
| marzoul | Final note : This is very casual hacks on spare computer time. Also I have no broad of fundamental understanding of current prjxray project as a whole, so these attempts are also a means for me to progress. So will be no fast progress on this - and anybody is welcome to take and improve on what I did (or throw it all away and start over xD). | 20:48 |
| marzoul | Note : Using make with options -j10 MAX_VIVADO_PROCESS=4 | 20:52 |
| marzoul | Fuzzer000 succeeded | 20:52 |
| marzoul | Fuzzer075 succeeded after 23 sec | 20:52 |
| marzoul | Fuzzer001 succeeded after 2 min | 20:52 |
| marzoul | Fuzzer073 succeeded after 4 min 20 sec | 20:52 |
| marzoul | Fuzzer005 fails after 1h50 - 2h | 20:52 |
| marzoul | Fuzzer072 succeeded after 24-25h | 20:52 |
| marzoul | Fuzzer074 not started (started after 005) | 20:52 |
| marzoul | max mem 30.1 GB, reached during fuzzers 005 + 07 | 20:52 |
| marzoul | Note : fuzzer005 fails at the end, when making database, with following error: | 20:54 |
| marzoul | Traceback (most recent call last): | 20:54 |
| marzoul | main() | 20:54 |
| marzoul | File "/home/prostboa/Projets/projet-prjxray/prjxray/fuzzers/005-tilegrid/add_tdb.py", line 162, in main | 20:54 |
| marzoul | run(args.fn_in, args.fn_out, verbose=args.verbose) | 20:54 |
| marzoul | File "/home/prostboa/Projets/projet-prjxray/prjxray/fuzzers/005-tilegrid/add_tdb.py", line 145, in run | 20:54 |
| marzoul | localutil.add_tile_bits( | 20:54 |
| marzoul | File "/home/prostboa/Projets/projet-prjxray/prjxray/fuzzers/005-tilegrid/util.py", line 111, in add_tile_bits | 20:54 |
| marzoul | assert offset + words <= 101, ( | 20:54 |
| marzoul | AssertionError: ('CMT_TOP_L_UPPER_T_X250Y148', 125, 98, 27, 'CLB_IO_CLK') | 20:54 |
| marzoul | ( I'm finished for these flood messages now ;-) ) | 20:54 |
| *** marzoul2 <[email protected]> has joined #f4pga | 22:12 | |
| *** marzoul <[email protected]> has quit IRC (Ping timeout: 272 seconds) | 23:19 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!