search

CNMAT / CNMAT-MMJ-Depot

CNMAT's expanding library of Max/MSP/Jitter patches
89 stars 17 forks source link

m29 standalone crashes in lab #61

Closed equilet closed 9 years ago

equilet commented 9 years ago

From Matt:

I copied you this morning on an email to Willa about an error I have witnessed over the last two labs. The error manifests as "spinning wheel"/crashed program. This effects c. 5-7% of lab computers for any given patch that is opened, so c.2 machines of 35 per patch. The problem is erratic in the sense that it doesn't seem to repeatedly target specific lab machines and in the sense that the software works for the majority of the users at any one time, suggesting to me that the problem might not be in the given standalone applications. But maybe I'm wrong about that. At one point, Willa suggested that it might have something to do with memory allocation on the lab machines. The only temporary fix we have adopted so far is to have GSIs pass new copies of the standalone from their share folder to affected students. This seems to work most but not all of the time.

equilet commented 9 years ago

From Ilya:

"The problem with lab machines is more than the % provided, I have around 10 machines spinning wheels. And they sometimes start spinning wheels after not spinning originally, which may be a hint...

I find it suspect that (a) I had NO ISSUES on the instructor machine, and (b) that the problem is as erratic as it is. I would recommend looking at "top" in the terminal and look at memory. I will do this next time I'm at the lab - the weird delayed nature of the problem hints at an issue that is extremely challenging to track.

Thanks a lot for the remarkable work with standalones. These issues come up and things are hard, but I think we have the best ones I've seen yet and I'm very grateful for your work. Instructor machine, at least, is rocking them solid without a single (!!!) crash or issue, across three hours of lab time, with everything turned on. "

equilet commented 9 years ago

@ilzxc Thanks for the feedback. It wouldn't kill me to rebuild all of them, but if you can narrow the issue down a bit, it'd be greatly appreciated - LMK about your findings w/ top and if you have specific steps to reproduce, that'd also be helpful. I need details like the following:

  1. what OS
  2. what version of runtime or Max
  3. what version of the standalones (or where you got them)
  4. what created date of the files you're loading
  5. which apps this is an issue on (or is it all of them?)
  6. specs of the Mac computers running these
equilet commented 9 years ago

From Matt:

up to about 3 computers could be having spinning wheel issues at any given time. So, 10+ computers total could be affected over the course of a lab but not at the same time. This is what I observed in two sessions of the previous lab on different days.

Willa told me this morning that, within the next couple of days, their specialist will test with our software in the Wheeler lab to try and determine whether or not it is a lab issue. They suspect that it is not a memory issue in the lab because larger programs, such as Illustrator, seem to be working fine on their machines. I offered to be there to demonstrate the use of the programs, so hopefully I will be able to witness this testing. I'm not sure yet if they plan to let me be there.

Information taken from a student lab machine.

OS: 10.9.5

Max version: The standalones are self-contained and don't use runtime or Max, is that correct? However, the application Max6 is available on the student machines.

version of the standalones (or where you got them): The standalones were originally taken from Ed's large hard drive. It is generally kept in his mailbox at CNMAT. Where do I find the version number for the standalones? Do you want the version numbers of all the applications or is there a version number for the whole package? Each applications seems to have a different one. For instance, Soundfile Presenter is ver. 1, Temporal Beat Slicer is 1.1, and Soundfile Interp Space is version 0.1a.

what created date of the files you're loading: These dates are taken from the lab folders: Labs 1-2, 7-12 is Jan 22, 2015 Lab 3+6, is Feb 5, 2015 Labs 4+5, 2015

which apps this is an issue on (or is it all of them?): Lab 1 with Soundfile Presenter seemed to go smoothly. Labs 2-3 have experienced significant spinning wheel, hanging program issues. Lab 4 has not happened yet but we expect similar issues since it involves the same software as Lab 3.

specs of the Mac computers running these: iMac, late 2012 Processor: 2.9 GHz intel Core i5 Memory: 8 gb 1600 MHz DDR3 Graphics card: NVIDIA Gforce GXT 660M 512 MB Software: Os x 10.9.5

Lab standalone applications seem to have vastly improved performance if you make a significant change. Whenever students open any lab standalone, we now instruct them to:

1) Open Audio status window and in this window: 2) Change i/o vector size to 512 3) Also, change signal vector size to 512

Most standalones seem to start with setting of 64 for both i/o vector size and signal vector size.

I was with Rafael in lab this morning and we saw that with these new settings we went from 10 malfunctioning computers to none malfunctioning with the Soundfile Interpolation patch. Audition and Decide I standalone had similar results.
If it is the case, will it be possible to program the future patches to start automatically with higher i/o and signal vector sizes?

equilet commented 9 years ago

"Willa suggested that it might have something to do with memory allocation on the lab machines"

I have a hard time believing that this could be even remotely related to memory allocation. I'm crossing this off my list of ideas to check.

equilet commented 9 years ago

@mattschumaker - Thanks for these details, but I still need specs.

Please review this info and post here.

A couple of notes:

equilet commented 9 years ago

loadbang -> dsp init will be present in all future builds.

mattschumaker commented 9 years ago

Version of Max: 6.1.8 (9fafea67)(32-bit)

Standalone Creation Dates: Soundfile Presenter: April 7, 2014 (modified 1-22-15) Additive Synthesis Introduction: March 6, 2013 (modified 1-22-15) Frequency Bands and Filtering: March 6, 2013 (modified 1-22-15) Keymouse Synthesizer: April 11, 2013 (modified 1-22-15) Synthesized Human Voice: March 6, 2013 (modified 1-22-15) Waveform Builder: April 11, 2013 (modified 1-22-15) Waveforms and Hearing I: March 6, 2103 (modified 1-22-15) Waveforms and Hearing II: April 11, 2103 (modified 1-22-15) Soundfile Interpolation Space: January 28, 2014 (modified 2-22-2015) Audition and Decide II: March 14, 2013 (modified 2-11-2015) Polyrhythm Explorer: March 27, 2013 (modified 2-11-2015) Temporary Beat Slicer: March 27, 2013 (modified 1-22-2015) Plucked String Playspace: April 30, 2013 (modified 1-22-2015) Plucked String Simulation: March 27, 2013 (modified 1-22-2015)

equilet commented 9 years ago

crashes have been resolved, and were happening based on DSP initialization issues. closing in favor of https://github.com/CNMAT/CNMAT-MMJ-Depot/issues/40