On Sunday 13 December 2020 14:22:08 John Dammeyer wrote: > > From: Chris Albertson [mailto:albertson.ch...@gmail.com] > > > > I don't call it a problem. With the three board designs, you can > > make a replacement for one of the "wings". splitting it up makes > > the design of each really easy. And as said PCBs this size cost > > under $1 each even in tiny quantity. So what I see you have here is > > the start of a family of boards. > > > > My problem with this is that I think the Beagle board is grossly > > underpowered. It is OK if you are making a battery-powered, > > portable milling machine but if you have access to AC mains power > > why use a beagle? > > Chris, I think you just made my case for me. It's not so much that > the Beagle is underpowered with the two co-processors. It's that the > OS and system has exploded in size and inefficiency because the > developers are always running the latest and greatest development > systems. Way back a Pentium-33 had no trouble running EMACs and a > Pentium-33 was way slower than a BBB. > > And we've allowed this inefficiency to happen without complaint. I > have a Panasonic Blu-ray player. Always bugged me that it took so > long after the power switch was pressed before it responded to the > button to open the drawer. It had no trouble immediately scrolling a > message on the display telling me it was powering up. Turns out it's > running Linux. So a 20 second start up time is considered permissible > even though it's rude to the user. > > The BBB with Replicape and OctoPrint can sit in the connecting state > for quite some time while it explores all the serial ports at two > different baud rates before it finds the BBB. Or, you can select > which serial port you want to use and it's connected right away. > > How much of MachineKit on a BBB is spent searching for and doing > general stuff that really isn't needed for an embedded system. How > much of the software is written with the idea that there is 4GB to 8GB > of 64 bit wide memory on a system that has 512MB of RAM. That the end > user wants to watch movies and surf the web. > > I used to make fun of IBM PCs with their 8 bit external bus 8088, 640K > RAM and hard drive DOS systems. Running DBASE-II took 5 seconds to > get to the DBASE-II prompt. My 8 bit Z-80 system with 56K bytes RAM > and 8" floppy disks did it in under 2 seconds. > > Yes. Apples and Oranges. Needless to say the PC was more powerful. > But my point is that if the MachineKit port was designed for 64 bit > PCs with even just 1GB RAM and fast hard drives the likelihood of it > being efficient without a total rewrite on a smaller 32 bit processor > with 512MB won't happen. > > So I'll throw up the question. Is as you said, "the Beagle board is > grossly underpowered", or has LinuxCNC/MachineKit suffered now the > same Code Bloat that Microsoft Windows and Apple have, making the need > for bigger processors with more memory mandatory? > > John > And I will say not to the degree you might think. Sure, LinuxCNC has grown, a lot, but its largely in the modules that have been added to expand its capabilities in different environments. What actually gets loaded and used in the hal files on any given machine has not grown that much. Maybe 25% in the last decade as the transition to 64 bit commodity hardware has caused some of that "bloat". But I've not edited my hal files to call in any that added plasmac or qtvcp stuff that has been added to master in the last year.
Where the machine I'm sitting here using has has grown in power considerably. The 2.1 GHz 4 core phenom 8 gig machine I built for this chair in 2008 started a fire at one of its mobo usb ports a year ago was replaced by another Asus board, carrying a 6 core I5 and 32Gig of dram. And its running the same hard drive and stretch install it ran on the phenom. The latency of the old phenom was horrible, several milliseconds. No way in hell I could have run a machine with it even with rt-preempt kernels. So I installed enough LinuxCNC of it to at least run a sim which gave me latency-test. And this machine now is by far the quickest of the 5 here, I'd have no trouble at all running my 4 axis GO704 with it. And that with a stock, stretch kernel. 4.19.0-0.bpo.9-rt-amd64, which has not changed by the bigger, more powerfull cpu and 64 bit mobo. Since its the same hd its booting from as was booting the phenom, the 50x better latency figures can only have come from the improved hardware. But even on the 6040 or GO704, the hal files have not grown by much and then only because I built new interfaces for 2 of them due to a need for more i/o. latency-test -period on this machine is 3344 ns, the old Dell, running the GO704 is 13968, the Intel D525MW boards that folks raved about years ago are 31862, and the rpi4 is 16944, 2x faster than the Intel's. So this machine is by far the quickest of the lot. Yet no single core in this machine is running faster than 900Mhz with no detectable heat, the hottest core of the 6 in this i5 is running at 30C right now. That's how much the hardware has improved in that same decade. That, I'm impressed with. How much bloat in the actually running linuxcnc does not enter into it in comparison. Thanks for reading this far. Cheers, Gene Heskett -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) If we desire respect for the law, we must first make the law respectable. - Louis D. Brandeis Genes Web page <http://geneslinuxbox.net:6309/gene> _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
