Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Axilian]

Did audio analysing get multi-threaded in 25.027? as is doesn't seem to be working properly for me

I just tried to add an SACD iso image to my library ( first time adding a SACD image since going to 25.027) and on running audio analysis MC completely took over the CPU (all 4 cores and 4 HTs at 100% ), ground the whole the whole PC to a  halt and crashed it.

That's the first time my PC has crashed in a very long time  and the CPU had been 'utilised' so well that it didn't even have enough time to write to the error log before going down to a BSOD

For Info I've rebooted the PC and all seems to be working ok , the SACD file that caused the issue seems to have been imported into my media library correctly and is able to play


UPDATE : tried to audio analyse the same SCAD again and once again the the CPU gets taken over (this time I cancelled the process before everything went pear shaped)
Tried another SCAD image (that was already in my library) and the same thing happened so it doesn't look it's an error in the original SACD image causing the issue

For info I run 25.027 x64 in windows 10 Pro (latest update)
PC is  i7 920 , 12Gb RAM

[RD James]

It's not only CPU usage. It causes disk thrashing too, which can significantly slow down analysis.
https://yabb.jriver.com/interact/index.php/topic,120421.msg832671.html#msg832671

This change really needs to be rolled back or changed so that the amount of threading used depends on the content being analyzed.

[Hendrik]

I agree that its not ideal to always thread audio analysis, and I send the references to both threads to Matt to re-consider this. However, if your system crashes "just" because someone uses the CPU to 100%, your system is not stable to begin with - it should never do that. Slow down and become unusable until its done, maybe, but not crash.

[Axilian]

The crash appears to be  because I have some other software running on the PC with High priority and had forgotten it was scheduled to make an access call around the time I was playing around with the SACD iso so not totally MC's fault but it's certainly a factor 

What you described is actually what happened the usage just started jumping as soon as started analysing the sacd audio. From what I can tell the  other software then tried to make a call and that's what threw it over the edge. Never the less, there is no way the CPU should be clogging like that - I have run some pretty major data analysis software on this particular PC (Shimadzu Lab solution for mass spectrometry & Agilent OpenLab CDS) and never see anything like  the system was  acting (normally it maxs on 1 or 2 cores and leaves the others for stability)

As an experiment I've also had a quick go at doing the analysis of the SACD file on my laptop using an install of JMRC on that (it's a more humble Core i7-820QM CPU) and the the laptop doesn't crash but slows down to a near crawl with disk thrashing)

[Axilian]

as reported previously - I am getting crashes on trying to import SACD ISO images . Previously I though it was due to running other software while using MC but I've now tried importing on several computers and I get the same BSOD reproducibly and every time on any SACD iso even with no other programs running (no AV . no firewall etc)

I get the following error on the BSOD  -
Bug Check 0x162: KERNEL_AUTO_BOOST_INVALID_LOCK_RELEASE

which on checking  Microsoft error messages gives

reason:

This is typically caused when some thread releases a lock on behalf of another thread (which is not legal with AutoBoost tracking enabled) or when some thread tries to release a lock it no longer owns.

As no other software on any of my systems was changed except installing / 'upgrading' to 25.027 I can only point to the multi-threading as an issue

[Hendrik]

BSODs cannot be caused by software. If a software has a bug it would just crash. BSODs are only caused by malfunctioning hardware or bad drivers.
The kind of threads the documentation talks about are kernel/driver threads, not anything Media Center does.

[Axilian]

As a test I just uninstalled 25.027 x64 and reinstalled 25.024 x64  - I just tried importing exactly the same SACD iso  as I had tried to import (and failed ) under 25.027 and  - no problem at all - imported fine - time of import 5.08 minutes - so that still leads me to believe that there's something in MC 25.027 that's causing the issue whether it's triggering an kernel / driver issue or not


Update - just had a random thought  - the BSOD report refers to auto boost - if that's referrring to the CPU auto boost (i.e most modern processor can run one core faster when others aren't needed) than that may be the issue as I don't run intels turbo boost (it's disabled in my BIOS) as i run all my Cores  at the same frequency all the time due to some of the software I use needing to be core balanced)

[JimH]

Please don't start new topics on this problem.

[Axilian]

Sorry - I hadn't realised I could edit the title back to not solved.

Anyway i appear to have found a way around it -

I ended up going into my BIOS and  downclocking the CPU by 10% . I then rebooted and JMRC auto-updated to 25.027 . I then re-imported the same ISO as before. This time the iso did import , so maybe the CPU was struggling with the multi-threading (it is quite a old CPU now though it had no issues with a multi-thread 3ds max render).

Interestingly compared to the time I posted below for 25.024   (5.08 minutes) this time it took 9:50 minutes rather longer than expected even with a 10% down clock taken into account

[Axilian]

Well I've now had a chance to play if quite a few more SACD images and it does look like 25.027 is consistent slower at importing and analysing SACD iso files (and possibly large flac files as well)

in general 25.024 was taking 5-6 minutes for a sacd iso but it appears 25.027 is taking approximately 10-12 minute approximately

smaller flac files do appear to analysing faster than before but it's marginal at best on my system (we are counting 5-6 seconds faster per track)

For me at least  the trade off of the faster processing of the smaller flac file is not worth it with the increased processing time of the sacd iso (and    not to mention the other effect of slowing the PC down so it's difficult to alt tab to another program and work while it's processing - which was easy before .

Is it possible to give a option to switch off the multi-threading - as currently for me at least it's just not worth it