It also makes corporations a lot of money, e.g. some versions of IBMs S360 were pure logic gates, but the cheaper ones implemented the same instruction set in microcode.
It also allows some corporations to dig themselves out of holes. How do you think intel "mitigated" the effects of the recent timing attacks on their processors? They didn't require people to recompile their applications or operating systems, but they did change load different microcode into the processor as part of the boot routine.
Other computers have allowed users to define the processor's visible instruction set to make it easier for compilers for high level languages.
Fortunately I've never been in the position where it might have been applicable. Typically the large scale systems I've seen have gross inefficiencies at much higher levels. One case was simulating RPC over TCP/IP in unix system, by plonking a value in an RDBMS and having the other end poll for new values.
Yes. None of the money is "real" in the sense of creating any worth - it's just moving numbers around and sucking out the value from other people's work.
It's not compilers lying - it's users who don't understand what "optimisation" means. It is nothing more or less than saying how much effort the compiler should put into generating more efficient object code with the same functionality (according to the rules of the language and other requirements, such as the ABI or the additional definitions given in the compiler documentation).
Programmers misunderstanding the job of the compiler and the meaning of optimisation? Who'd have thunk /that/? :-)
I was concentrating on the "... even when they think...". I presumed that equated to either "optimisations turned off" or "optimisations not enabled".
Apart from that, yes. Except that some languages/implementations give so many "obscure" options that it makes it more tricky to understand what might happen.
Yes, I know. The point is, the programmer is wrong to think optimisations can be turned off (or not enabled) - the concept simply doesn't make sense. A compiler might let you enable or disable various passes, but that's all.
That's a matter of understanding the language. No one said programming was easy!
Umm. I have decades of direct personal experience here.
A lot of standard RHEL is used, but of course options are chosen to suit the matter at hand.
VxWorks was also much used, but is fading, being replaced by RHEL. Partly because the raw performance and memory capacity of X86 hardware now vastly exceeds that of the traditional SBCs used with RTOSs.
The true hard realtime parts are now most often done in FPGAs, which are designed like logic circuitry, not traditional software.
Be careful with that word "embedded". It does not mean or imply hard realtime or even really fast.
My classic example is the computer system implementing realtime control of a rotary cement kiln. The sample rate is a blistering ten points per hour. As long as that computer doesn't fall behind too often, we get perfect control of the kiln, and the cement plant is profitable.
Yep. As mentioned above, the peformance dominance of X86 is a large driver.
Repeats the definition of hard realtime, but the point was that hard realtime was too fragile, so people had to soften the edges. More later.
Well, it is adaptive for sure. This is a good thing.
The problem is that traditional hard realtime always proved unreliable in the field, and some elasticity was required to avoid dramatic failures under stress.
War story: Circa 1990, I was the software architect for the NATO Sea Sparrow ship self-defense system (intended to knock incoming cruise missiles down).
One fine day, a mob of software engineers appeared at my door, arging with each other. The dispute was about what to do if the engagability calculation had a divide by zero error (this hazard occurs for some geometric configurations - it's not a programming error).
Their question was if they should print out an error message and related data and stop, or press on, foregoing the ability to debug the problem.
I was stunned by the very question. If one is doing engagability analysis, there is an incoming anti-ship cruise missile coming at twice the speed of sound, with ten or twenty seconds remaining before impact. Somehow, I doubt that the operator will appreciate debugging information at that moment, in his remaining few seconds.
In this world, if your self-defense missile isn't moving on the launch rails in about five seconds, it will be impossible to defeat the incoming anti-ship missile.
So, just write diag data to the war diary, and provide a very large number as the division result, and press on, praying.
More generally, in embedded systems of any complexity, thare are many essential computations whose run time is inherently statistical, depending on randomly changing details of the outside world, and one therefore must plan for and handle the inevitable overruns.
It is nothing of the sort. They are buying and selling equities which prov ides liquidity to those who wish to buy and sell. It's a market. That's t he purpose of a market, buying and selling. Your complaint is like critici zing an auction as being "parasitism". If you don't like it, stay out of i t.
--
Rick C.
-+-+ Get 1,000 miles of free Supercharging
-+-+ Tesla referral code - https://ts.la/richard11209
Pure BS. The value is in the market providing funds for companies to grow and prosper. If there were no market there would be many fewer companies a nd a much smaller economy.
Why do people twist things around to make others sound evil when their only crime is success?
--
Rick C.
-++- Get 1,000 miles of free Supercharging
-++- Tesla referral code - https://ts.la/richard11209
There was a time that speculation was frowned upon. These days it's the rule. Don't pretend that manipulating stocks at ms intervals can in any way be advantageous for a productive economy. They are parasites, leeching the productive work of others.
You are not making any sense. You don't like computer trading but you are fine with everyone else trading?
There's nothing wrong with computer trading. I don't trade on millisecond levels so they don't impact me. They don't trade on information so I don't impact them.
The fact that you don't actually make an argument against them speaks volumes. Calling them names just makes you appear to be strange.
Do you have the same opinion of the floor traders or the many day traders? Is there anyone trading stocks you approve of?
--
Rick C.
-+++ Get 1,000 miles of free Supercharging
-+++ Tesla referral code - https://ts.la/richard11209
Is this in a particular niche or application area? Your comments don't match my understanding, but perhaps there is a particular field where RHEL is used in this way.
Again, I'd wonder what the use-case is. To me, VxWorks and RHEL cover very different kinds of system.
That has always been the case (except in the sense that some embedded boards can be more powerful for specific functionality, such as DSP work).
Agreed - that is where you put the most timing sensitive stuff (or on standard peripherals in a microcontroller, or ASICs, or whatever - dedicated hardware rather than multi-purpose software.)
Yes, I know. But I might not have expressed myself very clearly.
Embedded systems covers everything from a tiny microcontroller to massive processing systems - it roughly means things that are built into a bigger system, and don't have normal screens, keyboards, or that kind of interaction, and don't count as server systems.
RHEL is primarily for workstations and servers, not embedded systems. Linux is often used in embedded devices, but it is generally either a small Linux system such as a minimal Debian, or a custom build Linux using something like Yocto or Buildroot. OpenWRT is a popular choice if the system is network oriented, and of course Android is used. These systems are usually going to be "soft" realtime, or not realtime at all. (And the huge majority run on ARM processors, not x86, with a scattering of MIPS, RISC-V and other cpus.)
That may count as realtime (and definitely /hard/ realtime!) - realtime does not imply "fast".
Sure.
Agreed. Hard realtime is often overrated, IMHO, and people often think they need it when they don't. They also often think they've made a hard realtime system when they have not.
It sounds like a design error, rather than a programming error, but that is beside the point.
"-O0" does not turn them all off or "disable optimisations". It turns off many, but some are on by default. And some things that people might consider "optimisations" are always active.
On Monday, June 8, 2020 at 4:25:07 PM UTC-4, Lasse Langwadt Christensen wro te:
ond levels so they don't impact me. They don't trade on information so I d on't impact them.
You don't understand trading? I could explain it to you, but it would be s lightly too long for the margin of this page.
Let's just summarize it by saying no one is forced to buy or sell any equit ies. Do you somehow have the understanding they are improperly taking away from others? That would be the same as saying your competitors are improp erly taking away from you. Is that what this is about? Are you feeling li ke a victim of unfair competition?
--
Rick C.
+--- Get 1,000 miles of free Supercharging
+--- Tesla referral code - https://ts.la/richard11209
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.