add support for the kintex/virtex high performance banks

[hansfbaier]

Signed-off-by: Hans Baier hansfbaier@gmail.com

[hansfbaier]

@kgugala @mkurc-ant Any ideas why the kintex job was aborted?

[pgielda]

I am guessing 6h timeout?

[hansfbaier]

@pgielda @mithro @kgugala Unfortunately it seems that kintex7 really needs that much time. I tried reducing the ROI, but then 025-bram-config fails complaining it has not enough BRAM instances in the ROI. What should I do now?

[pgielda]

FYI @tmichalak

[tmichalak]

@hansfbaier the default timeout for a job, before GitHub cancels it, is 360 minutes, which is exactly what seems to happen here.
You can change that using timeout-minutes per job or step - in your case using that in the run tests step should solve the problem.

[kgugala]

hmm, looks like the timeout setting was ignored?

[tmichalak]

Looks to me that the timeout-minutes was added to the 'Build' step while the step that times out is the run test step. @hansfbaier

[hansfbaier]

@kgugala @tmichalak Yes I added it to the wrong step. Thanks!

[hansfbaier]

@tmichalak Now I have it on 'Run Tests' but unfortunately it did not work.

[hansfbaier]

Looks like we have reached a hard limit on GitHub here

To have a job running more than 6 hours, we would need to have a self hosted agent, see
https://stackoverflow.com/questions/68187987/github-actions-job-timeout

[kgugala]

@hansfbaier the CI here is run in custom runners

[kgugala]

timeout seems to be working. Kintex job is now running for 7h

[hansfbaier]

@kgugala Great! Moving the timeout configuration into the parent element did the trick!

[hansfbaier]

@kgugala It looks like some of the other series tests are flickering. Race condition?

[kgugala]

there might be some instabilities - in general fuzzers, by nature, random so sometimes we hit a case where the flow fails unexpectedly

[hansfbaier]

Ah, the 037-iob18-pips fuzzer does not terminate, the todo's do not vanish:

RIOI.IOI_ODELAY0_CLKIN.IOI_OCLK_0
RIOI.IOI_ODELAY1_CLKIN.IOI_OCLK_1
RIOI.RIOI_O0.RIOI_ODELAY0_DATAOUT
RIOI.RIOI_O0.RIOI_OLOGIC0_OQ
RIOI.RIOI_O1.RIOI_ODELAY1_DATAOUT
RIOI.RIOI_O1.RIOI_OLOGIC1_OQ

These are the unsolved TODOs so far.
I probably need to understand if they are necessary.
Any thoughts?

[hansfbaier]

Great! All the fuzzers ran successfully now, only checkdb still reports collisions. Got to figure those out.

[hansfbaier]

@tmichalak Kintex is green!! Artix and Spartan seem to have their usual flickers. The changes should have no effect whatsoever on those, since I only changed the bitfilters in the iob18 fuzzers, which only run for Kintex.

I also tested the IO banks with this blinky: https://github.com/kintex-chatter/demo-projects/tree/main/blinky-stlv7325

VID_20221122_110518.mp4

The blinky uses a differential clock input, a single ended button and LED and two differential LEDs, all connected to the high performance banks.

[hansfbaier]

My newly commited conditional in a Makefile might fail for non kintex. Will look into it tomorrow. Will be a trivial fix. Kintex is redy for review nevertheless.

[hansfbaier]

Another Makefile needs the same adjustment. Also, fuzzers/036-iob18-ologic/bits.dbf and fuzzers/035a-iob18-idelay/bits.dbf I have just copy and pasted. Should this be empty by default? Is the fileformat documented somewhere?

[tmichalak]

@hansfbaier the .dbf is for zero groups filtering and it can be empty by default. You can find a short documentation for this format here

[tmichalak]

Please update the date in the license headers in the new files you're adding e.g. 2017-2022

[tmichalak]

@hansfbaier CI is indeed green. Just had a two small comments. Other than that LGTM.

[hansfbaier]

@tmichalak Thank you, will tidy up tomorrow!

[hansfbaier]

@tmichalak I applied your review notes and rebased to master. Concerning the .dbf files I removed/adjusted the parts which are obviously not applicaple, but kept the rest of the bits, because in the fuzzer results it looks like the same frame range is used and thus I suspect by default that the changes are valid, especially as some of the io logic fuzzers gave identical results to the high range bank fuzzers so that I actually deleted them and just copied their bits over.

[hansfbaier]

@tmichalak It is green again. Except Artix, but that seems to be that flickering timeout in the fuzzer 065b-gtp-common-pips again

2022-11-23T22:21:04.8795880Z 22:21:04 | 2022-11-23T22:21:04 - xc7a200tffg1156-1/065b-gtp-common-pips      - 1h56m: + '[' -f /root/prjxray/prjxray/fuzzers/065b-gtp-common-pips/build_xc7a200tffg1156-1/todo/timeout ']'
2022-11-23T22:21:04.8796540Z 22:21:04 | 2022-11-23T22:21:04 - xc7a200tffg1156-1/065b-gtp-common-pips      - 1h56m: + echo 'ERROR: timeout'
2022-11-23T22:21:04.8797097Z 22:21:04 | 2022-11-23T22:21:04 - xc7a200tffg1156-1/065b-gtp-common-pips      - 1h56m: + exit 1
2022-11-23T22:21:04.8797705Z 22:21:04 | 2022-11-23T22:21:04 - xc7a200tffg1156-1/065b-gtp-common-pips      - 1h56m: make[2]: *** [../pip_loop.mk:68: run] Error 1
2022-11-23T22:21:04.8798403Z 22:21:04 | 2022-11-23T22:21:04 - xc7a200tffg1156-1/065b-gtp-common-pips      - 1h56m: 
2022-11-23T22:21:04.8799154Z 22:21:04 | 2022-11-23T22:21:04 - xc7a200tffg1156-1/065b-gtp-common-pips      - 1h56m: --------------------------------------------------------------------------
2022-11-23T22:21:04.8799847Z 22:21:04 | 2022-11-23T22:21:04 - xc7a200tffg1156-1/065b-gtp-common-pips      - 1h56m: !Failed! @ 2022-11-23T22:21:04.581598 with exit code: 2

Is there an issue for that fuzzer?

[tmichalak]

@hansfbaier nice work. Thanks a lot !