HGCAL in phase-2 premix workflow .999: missing HEB rechits, low response in HEF and EE

[slava77]

followed by recent (e.g. https://github.com/cms-sw/cmssw/pull/31906#issuecomment-722578846 or https://github.com/cms-sw/cmssw/pull/32022#issuecomment-722025986) it looks like some of the premixing is not working in phase 2

here are a few plots comparing in CMSSW_11_2_X_2020-11-04-1100 23234.0 in black with 23434.999 in red (10 events; presumably the same sim), using the matrix outputs from the baseline jenkins job

HGCHEBRecHits are empty in 23434.999

[kpedro88]

@adas1994 when I change the preceding line like this:

if (std::get<2>(*(findPos)) == tof and findPos != PhitRefs_bx0[id].begin()) {

I no longer see invalid writes in valgrind.

[adas1994]

The correct lines should actually be like :

std::get<0>((findPos)) += charge; std::get<1>((findPos)) += charge;

I am running tests with this modification, but I am yet to check it with Valgrind. If you can check it with Valgrind, that would be great. Otherwise I would do it by tomorrow anyway. Thanks a lot @kpedro88 for identifying the bug ; as I said earlier, due to my misunderstanding of Valgrind, I had probably completely ignored it's diagnosis.

On Fri, Feb 12, 2021 at 6:45 PM Kevin Pedro notifications@github.com wrote:

@adas1994 https://github.com/adas1994 when I change the preceding line like this:

if (std::get<2>(*(findPos)) == tof and findPos != PhitRefs_bx0[id].begin()) {

I no longer see invalid writes in valgrind.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/cms-sw/cmssw/issues/32050#issuecomment-778516196, or unsubscribe https://github.com/notifications/unsubscribe-auth/AG64M4MMGEIHC7U4KJOI2YLS6W4TDANCNFSM4TM2PLZA .

-- Sincerely,

Abhishek Das.

[kpedro88]

The change you suggested also succeeds in valgrind without invalid writes.

[adas1994]

I have been able to complete my first batch of tests after my initial modification. The results are partially correct and partially wrong, so I would need some more time to do further modifications. But I am glad that the ball is finally rolling after months of stagnation. I could not be more grateful to @kpedro88 for his help.

On Sat, Feb 13, 2021 at 2:18 AM Kevin Pedro notifications@github.com wrote:

The change you suggested also succeeds in valgrind without invalid writes.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/cms-sw/cmssw/issues/32050#issuecomment-778575716, or unsubscribe https://github.com/notifications/unsubscribe-auth/AG64M4LJWNIUBT677TRU2V3S6YRTLANCNFSM4TM2PLZA .

-- Sincerely,

Abhishek Das.

[srimanob]

Hi @adas1994 It has been a while since the last update on the issue, could you please update if the issue is handled somehow? Thanks.

[adas1994]

I had an exam from the grad school, so did not spend time on this problem. I am starting to work on it from today.

On Sun, Apr 11, 2021 at 7:48 AM Phat Srimanobhas @.***> wrote:

Hi @adas1994 https://github.com/adas1994 It has been a while since the last update on the issue, could you please update if the issue is handled somehow? Thanks.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/cms-sw/cmssw/issues/32050#issuecomment-817294264, or unsubscribe https://github.com/notifications/unsubscribe-auth/AG64M4L76JJGD3KK7EFZSYLTIGD7XANCNFSM4TM2PLZA .

-- Sincerely,

Abhishek Das.

[adas1994]

@kpedro88 I have reviewed my code for this PR purpose and found no algorithmic glitch thus far, but the results keep being somewhat discrepant for StdMix and PreMix. I was wondering if it is possible to order a slew of tests with various samples here on github to run on a commit I had made. I am hoping that it will help me pinpoint any lurking bug.

[kpedro88]

@adas1994 you can test additional workflows in a PR, but this currently does not generate comparisons to the baseline, because the additional workflows were not tested in the IB. It's probably better to continue testing privately.

[srimanob]

@adas1994 What kind of private samples you would like to try?

[adas1994]

@srimanob I was hoping to get my commit tested through the regular test suite one gets after a PR submission. As Kevin says, if that is not possible, privately on my computing node, I would reuse Single Photon, Single K0_L and ttbar samples.

[srimanob]

@adas1994 Do you have an update on this? Thanks.

[srimanob]

assign hgcal-dpg

[cmsbuild]

New categories assigned: hgcal-dpg

@cseez,@rovere you have been requested to review this Pull request/Issue and eventually sign? Thanks

[adas1994]

I don't have any updates on it. I tried to solve it for a couple of weeks and could not. Since then I have been extremely busy and I will remain so for at least next 2 months. I suggest asking somebody else to solve it with a fresh open issue or wait for 2 more months for me to have another look at it. I am sorry.

On Wed, Jul 21, 2021 at 9:09 AM cmsbuild @.***> wrote:

New categories assigned: hgcal-dpg

@cseez @.*** https://github.com/rovere you have been requested to review this Pull request/Issue and eventually sign? Thanks

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/cms-sw/cmssw/issues/32050#issuecomment-883948037, or unsubscribe https://github.com/notifications/unsubscribe-auth/AG64M4P5XXNLVBYAZXBM4MTTYZXDNANCNFSM4TM2PLZA .

-- Sincerely,

Abhishek Das.