maculata

Followup: 1 Channel Scope with Persistence has been running for 4 days without memory growth. Persistence has added 3,523,590 acquisitions. Average ethernet data rate in this configuration (ADP3450 to Mac M1) has been 3.5MBytes/second. Works great! Thanks.

Followup: 1 Channel Scope with Persistence has been running for 4 days without memory growth. Persistence has added 3,523,590 acquisitions. Average ethernet data rate in this configuration (ADP3450 to Mac M1) has been 3.5MBytes/second.

I confirmed that the build link above fixes the growing real memory and beachball waits for the M1 Mac

Thanks for the info. Yes, I was ignoring the 36GByte virtual memory and only checking the iStat "memory" or "real" reported numbers - those changes as I resized the window and no longer increases with run time. After a continuous 20 hours run, there was no crashes, no beachball and those two istat reported memory numbers were basically unchanged and the ethernet adp3450 to mac data rate did not drop. Interestingly, Hiding the Scope tab with the help tab reliably slowed the ethernet adp3450 to mac data rate by about 10% - it recovers the speed when the scope tab is brought back up. It's reliable! The Mac install now appears completely useable - thanks. I have no real complaints now. From above: (Device manager confirms Ethernet; Ethernet speed test showed 57.5+ MiBps both ways) This appears to somewhat underestimate the actual achievable network data rate during run.

Installed on Mac M1 Max 32GByte MacOS 12.1 (21C52) without incident. Does not seem to understand dark window mode (about says) 3.18.15 64-bit Qt5.9.9 macOS 10.16 Qt5.9.9 Device manager confirms Ethernet; Ethernet speed test showed 57.5+ MiBps both ways it says 32768 samples at 33.333MHz holdoff 0 seconds; it seems to run faster - hard to tell. On open to existing configuration (single scope and single trace) Memory 299.5 MB real 335.7 MB virtual 36.94 GB On run Jumps to Memory 359.1 MB real 395.2 MB virtual 36.79 GB and stays steady (wanders around a few 10 MB at most After longer runs and some stop / run cycles memory still steady and no beachballs! Is there a way to verify it is running the m1 arm build? Is there a way to tell how many traces per second for updates? Moving from second monitor to laptop monitor caused a increase in memory (memory 720MB real 750MB) and the trace seems really clean :( every window resize causes memory to jump by ~350GBytes. That tolerable since memory is not growing continually. 2 CPUs (of 10) pegged, neither laptop fan has started (both are o rpm); graphics memory? 33% Is there a bug list or release note for this branch? It's already much better, thanks. (No running memory growth and no slowdown in trace refresh after 20 minutes) More speed tests: external display 4k fairly big window, ethernet settles to 5.5 MB/sec (ADP3450 to Mac) mac laptop display 5k (or whatever) window about 1/4 the screen, now pulling about 14.8 MB/sec (ADP3450 to Mac) Still running, during some recent resize for above speed tests, memory went back to ~250MB!

I have setup an UBUNTU 20.04.2 ARM64 environment on the M1 Mac described above via current version of Parallels. I downloaded diglent.adapt.runtime_2.26.1-amd64.deb sudo dpkg -i diglent.adapt.runtime_2.26.1-amd64.deb and get this error: package architecture (amd64) does not match system (arm64) Is there a way around that? I run into a a similar problem under a windows11 environment on the M1 Mac via parallels Also adding amd64 as a foreign architecture to dpkg leads to dependency errors.

Waveforms 3.17.1, 3.18.14, 3.18.15 Mac M1 Max 32GByte MacOS 12.11 ADP 3450 operating over USB or Ethernet While running the scope only, even for 1 trace, with buffers set to 1. Using both Mac and 3rd party tools I can watch memory grow and never shrink while Run. For Ethernet, Growth is maybe 3 MBytes per second. If you leave it run for very long times (hours) it crashes. If you hit Stop after it is running for minutes a beachball appears. If you wait sufficiently long, the beachball goes away and the GUI works again. It appears that the beachball is up longer when you have let it run longer. Quick Run / Stops (a few seconds) there is no beachball. In the beta versions, sometimes the beachball appears in response to other GUI actions during run, not just stop. Unless left on for hours, there is never a crash during the beachball, just several minutes of waiting on beachball. In watching the interaction it seems to me like it is allocating memory for every retrace and that when you stop or trigger certain GUI functions, it has to wind thru all the new memory allocations and do something - if run just a few seconds then no beachball, if a long run then long waits with the beachball. No variation of settings on scope trace run escapes this memory leak. The memory leak is only solved by quit/restart of waveforms. Ethernet beachball longer (and leaks memory about 10x faster) per minute of run. I tried a number of random internet solutions relative to mac / m1 mac / waveforms / beachball / configurations of the cursor, ... with no solution. They all seemed to be voodoo. <><><> Is there an archive all traces in memory while running option that I cannot find? Or needs to be unchecked? Adjusting number of buffers seems to have no effect. I don't understand why a single scope should take 2.29GBytes of memory while I wrote this post. Thanks