From my experience, the default iteration time of the while loop in micro-controller or PC programming is 1 ms, whatever the language is C, C++, Python, LabVIEW or Arduino. However, the CPU frequency can usually reach several GHz, so the loop time technically could achieve micro-second or even nano-second level. Is it possible to achieve faster loop time in PC-based programming?
However, I note FPGA can achieve faster loop. For example, when I use NI crio-FPGA, I can easily get 20us. Though the computation of each loop is complex, the speed is still very fast. But the clock frequency is only 40 MHz. I know FPGA uses parallel process, but some calculation must be done step by step.
I just guess PC-based programming can go faster someway? Who know how, please?
Minimum loop time for a useful loop can be as low as some 5-10 instructions - deoending on CPU/MCU architecture, the code already loaded in an eventual cache etc.
Now it depends on the CPU/MCU clock rate how much time these instructions equal to.
Just as some hint: these days, motor inverters run @ 20 - 50 kHz - equalling 20 - 50 microseconds for a full sequence through the corresponding control code. And this code is much more than a simple loop...
It mostly depend on your data cache locality, at least for languages like C or C++. For example, if you look over an array or std::vector of integers, you will get MUCH more performance than 1ms per iteration!
Of course, a prerequisite for a program to be executed as fast as technically possible is that there is just this single task running (especially: no operation system!). This is easily achieved and the usual case on simple microcontroller systems. But on more complex hardware like PCs, with multi-tasking operating systems, not intended for real time applications, the application is one of many tasks.
If the PC hardware includes something like a parallel interface, and the purpose of the loop is to output a clock signal, for example, under a multi-tasking OS the frequency would not only be low but the signal would not even be periodic!
I guess the fastest loop execution can be achieved with limited effort by running good old DOS (Version 5 or 6 should still be present in well-furnished households ;-) on a 32-bit-PC, clock frequency about 2.4 GHz. In the initialization part of your C program (or assembly language but not C++ or other deviations from the straight path), switch off the timer interrupt of DOS, and then the whole machine is exclusively yours! I feel that after filling the cache, an empty loop (e. g. just counting, using a 32 bit integer) should take less than 10 ns per iteration.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python