I fixed an interesting bug recently. By way of background, I work on industrial robotics control software, which is a mix of time-critical software which has to do its job by a tight deadline (variously between 1 and 10 ms) or a safety watchdog will trigger, and other software that does tasks that aren’t time sensitive. We keep the timing-insensitive software from interfering with the real-time software’s performance by having the scheduler strictly prioritize the deadline-critical tasks over the “best effort” tasks, and further by massively overprovisioning the system with far more CPU and memory than the tasks could ever require.

The bug (as you may have guessed) is that the deadline was sometimes getting missed, triggering the watchdog. What makes the bug interesting is that it was caused by us giving the system too much excess CPU and memory!

The deadline overruns happen when the computer runs best-effort tasks. It only runs those tasks a small fraction of the time, but no matter, they shouldn’t be interfering with the deadline completion at all (and having our software fail its deadlines is unacceptable, of course). The real-time tasks’ peak usage is only about a quarter of the computer’s CPU capacity, and the system gives them that always: 100% of the time they want CPU, they get CPU. They are never delayed, and never have to wait on a resource held by a best-effort task. When they miss the deadlines, it’s always because they’ve gotten jobs that take the usual amount of work to complete, and had their usual amount of time to do it, yet the jobs somehow just take more time to finish. There’s no resource contention, nor are the caches an issue.

Yet when the best-effort tasks execute, the calculations done by the real-time tasks run slower on the same CPU for the same job with the same cache hit rate and with no memory or I/O contention. What’s going on? After hitting some false leads, I discovered that they’re going slower because the CPUs are running at a lower frequency. It turns out the CPUs’ clocks have been stepped down because they’re getting too hot. (Yes, the surrounding air is much warmer than it “should” be, but we don’t have a choice about that.)

The computer is hot because all the CPUs are going at full blast, because all of the best-effort tasks are executing because the computer can fit them all in memory. Half a minute later, the tasks are done, the computer cools off, the CPU frequency gets stepped back up, and the deadline overruns cease. An hour later, the hourly background tasks all go again, the fans spin up to full, but they’re not enough and the CPU frequency steps down, hence the watchdog alerts about missed deadlines.

Okay, maybe we should change the background tasks so they execute in a staggered fashion. But before thinking about how I might do that, I tried disabling ⅔ of the computer’s CPUs instead. The hourly processing now takes 200s instead of 30, but it’s still far below 3600, and that’s what matters. Now the computer stays nice and cool, so the CPUs stay nice and quick, and the deadlines all get met. We were using too many CPUs to get the calculations we needed to get done fast, done fast. Who’d’a thunk?

There’s this video going around “An Open Source CPU!?” about SiFive’s implementation of RISC-V being “an open source CPU”, but it’s so duplicitous that I wanted to tear out my hair as I watched it. RISC-V is an interesting and worthy project, but the video is misleading in so many ways it’s hard to know where to begin.

For one thing, the x86-16 ISA wasn’t just hacked up in a few weeks from nothing. Intel’s architects spent months on it, and it was based on Intel’s 8-bit 8080 ISA, which was in turn informed by the experience of their earlier 8008 ISA. It had also already been on the market for a few years by the time IBM got around to the PC. For another thing, the x86-32 ISA was hardly lacking competition during its era of desktop dominance, with Motorola’s 68000, IBM’s Power Architecture, and DEC’s Alpha all big-name-backed ISAs trying to displace it. When the time came for 64-bit PCs, even Intel itself failed in its attempt to push a new kind of ISA for the PC market with its Itanium architecture.

Regarding the RISC/CISC divide, it is true that CISC processors don’t execute their ISA instructions directly in hardware now, but instead run RISC microcode (because RISC instruction-handling circuitry executes faster and is easier to design) for a CISC ISA interpreter. But that doesn’t mean RISC is best for application or system code, because microcode has no instruction cache! In the embedded systems and mobile-device markets, ARM had to abandon the “elegant” simplicity of their RISC ISA and supplement it with non-RISC “Thumb” and then “Thumb-2” extensions to raise its code density (important on small processors with small instruction caches) to keep x86 ISA microcontrollers, with their dense CISC code and ever-decreasing prices, at bay. As is often the case when competing technologies butt heads, the best solution is to find a way to use the advantages of both!

The video’s title is “An Open Source CPU!?” — and this is its biggest lie, because SiFive’s CPU being open is presented as something new, which is very much not true. Even if you dismiss all CISC designs as unworthy, there are open RISC specs with open source implementations that predate RISC-V, both grassroots and of commercial origin. Furthermore, SiFive’s CPU not only isn’t the first open source RISC processor, it’s not even the first open source RISC-V processor! You can get open source implementations of RISC-V here (VexRiscv) or here (ORCA) or here (Sodor).

This video is just an advertisement of SiFive’s pretending to be an independent informational video, spreading misinformation about the history and present state of microprocessors. (For your amusement, go and pause the video around 6m12s — the presenter is stating the inferiority of other open ISAs compared to RISC-V while showing a changelog on GitHub. But the changelog shown is not that of anything to do with another ISA, but the opposite: it’s actually the changelog for the port of gdb to RISC-V! The presenter could have grabbed a screenshot from the GitHub pages of the ZPU project with insignificant additional effort and shown something that was actually applicable to his claim, but once you’ve stepped off the straight-and-narrow path of honesty onto the wide highway of dishonesty, why bother?)

In short, that video is complete and utter BS, the presenter should be downvoted into YouTube oblivion, you should never pay one red cent for anything from SiFive, and RISC-V deserves better. It’s strong enough that it should be able to fight honestly, and having its proponents descend to dishonest shilling for one corporation is disgraceful.

Today’s xkcd riffs a common frustration about busy files: programs that can’t do their intended operations on them rarely (if ever) specify the file that’s busy or tell the user which programs are using the file — and without that knowledge, the user can’t fix the problem.

There are tools that can help you discover those things, as I describe using ten years ago, but I’d already been running into EBUSY problems for ten years by then, and the old guard before me for another ten years beyond that. Why haven’t things improved?

Sometimes file_operation(filename) fails — hey, this happens, busy files are a thing, ain’t nothing your program can do about that. But report to the user that file_operation on filename failed for reason strerror(errno)! Not reporting the filename is just sloppy programming.

It’s not like the tools have gotten any better. What, you think it’s acceptable to make the user trace your program, looking through a haystack of every system call to find the needle which is the program’s critical failure? Is that work you’re putting on the user’s shoulders less than sticking the filename in the error message? This has sucked for two generations now. Please help making it suck less.

It doesn’t help to zip already compressed file formats, like jpeg or mpeg, it only wastes CPU. Uncompressed video, audio, or still images will compress better with algorithms which are dedicated for video, audio, or images than they will with general compressors. JPEG is for photos — for images which are line drawings use PNG.

No, it won’t make you type faster (or probably not by much… top speed typists all use Dvorak, but I didn’t notice any improvement) or more “cheaply”, but it is easier on your hands, and reducing finger fatigue makes it a clear winner in my book.

How to switch:
Windows NT:
    Start ➙ Settings ➙ Control Panel ➙ Keyboard ➙ Input Locales ➙ Properties ➙ Keyboard Layout ➙ US-Dvorak
    Hit “OK”, then “Apply”.

Windows 9x:
    Same, but you’ll probably need to insert the install CD when it says, as the layout isn’t part of the default install, so have it ready.

Windows XP:
    Start ➙ Control Panel ➙ Regional and Language Options ➙ Languages ➙ Details ➙ Settings ➙ Default Input Language ➙ United States-Dvorak
    If it’s not availabe, under “Settings” go to Installed Services ➙ Keyboard ➙ Add and select it, then you’ll be able to set it as the default.

Linux console:
    “loadkeys dvorak” will change the current keymap to Dvorak. (Or sometimes “loadkeys dvorak/dvorak”, look around /lib/kbd/keymaps.) Making it the default usually involves setting the “KEYTABLE” attribute in /etc/sysconfig/keyboard to whatever you used with loadkeys.

X11 on PCs:
    xmodmap pentdvor.xmod

X11 on Sun 5x keyboards:
    xmodmap sun5dvor.xmod

Other X11:
    First, run xmodmap -pke > qwerty.xmod so you have the current keymap safely saved in case things get messed up. You’ll want to copy “xmodmap qwerty.xmod” into a cut buffer so you can paste it with only the mouse.
    Run xmodmap q2d.xmod
    This maps the Qwerty keysyms to Dvorak, and can fail sometimes, like if two keycodes generate one of the moved keysyms. Because it operates on keysyms, not keycodes, this file will rearrange your keyboard again if you run it twice, producing garbage. Good thing you have that recovery command in the cut buffer, right?
    You can save the keymap you produce with q2d with “xmodmap -pke > dvorak.xmod” and return to it with “xmodmap dvorak.xmod” after that. If you’re not quite happy with it, you can tweak the xmod file yourself, or get the “xkeycaps” program which will do it for you.

I promised you a more upbeast post, so here it is…

Work recently gave me two very nice LCD monitors to replace those I’d been using before. 24″ TFT from Dell.

“Unfortunately”, the graphics card in the desktop I had (which I’d been using since getting hired) couldn’t drive these things, so they gave me a HP xw4200 to replace my xw4100, double the RAM and faster processor.

Now my computer can display on my monitors, but side by side things are very wide. But, these monitors can rotate on their stands! (Couldn’t do that with my old ones.) So, I want them vertical and side by side. Well, someone else had done the hard work of figuring out how to make that work, so I went and followed his lead. The distillation of that came down to:

• Install the x.org X server (this deserves a post in itself)
• Edit xorg.conf, turning off Xinerama and turning on xrandr (for now they’re incompatible):
  • In the “Device” Section, add Option "NoTwinViewXineramaInfo" "on"
    and Option "RandRRotation" "on"
  • Change TwinViewOrientation to “Above” or “Below”
  • In the “ServerLayout” Section, add Option "Xinerama" "off"
• Put xrandr -o left at the start of your .xsession

Before and after screen shots at 1/8 scale:

I finally have enough space to have all the xterms I want! I’ve been trying to reach this state since forever, first with screen then with just piling the xterms onto all the real estate I could get. I don’t have to drop context and have space left over now. It’s surprising how good this feels– It’s this slight pressure that I never really noticed, the small frustation of having to look for somewhere to type– gone.