| ??? 09/25/06 08:30 Read: times |
#124973 - False assumption Responding to: ???'s previous message |
Jez Smith said:
...the idea is that the optimisation routines dont actualy break your code,they just mangle loop counters and change the way things are accessed Unfortunately, especially in a real-time embedded system, this can affect the actual operation in the actual system. In particular, this can show up timing problems; the timings of the "real" (optimised) and "debug" (non-optimised) code will be different - so bugs that manifest in the "real" (optimised) system may well not manifest in the "debug" (non-optimised) version! Another one is memory usage: this will differ between the "real" (optimised) and "debug" (non-optimised) builds; so, again, it is possible that bugs will manifest in one but not the other. Then, of course, there's the just plain bad 'C' code that makes unwarranted assumptions and unsafe reliance on the nature of the generated code... so it still works 'in the large'.Thats the plan anyway. As well all know, things don't always go to plan! |
| Topic | Author | Date |
| debuging C code | 01/01/70 00:00 | |
| Management solution | 01/01/70 00:00 | |
| I kind of hope | 01/01/70 00:00 | |
| False assumption | 01/01/70 00:00 | |
| to HLL or not to HLL | 01/01/70 00:00 | |
| Things for which an HLL cannot be trusted... | 01/01/70 00:00 | |
| sometimes there's no other way | 01/01/70 00:00 | |
| that was not me, but Andy | 01/01/70 00:00 | |
| Timing - end of wrong stick | 01/01/70 00:00 | |
| Thats not the point | 01/01/70 00:00 | |
| better, don't use C at all.. :-) | 01/01/70 00:00 | |
| experience and knowledge | 01/01/70 00:00 | |
| the key word is "manifest" | 01/01/70 00:00 | |
| NASA | 01/01/70 00:00 | |
| against NASA rules, sorry Andy posted before | 01/01/70 00:00 | |
| Testing | 01/01/70 00:00 | |
| Pointers | 01/01/70 00:00 | |
| In the real world | 01/01/70 00:00 | |
| worse than that | 01/01/70 00:00 | |
customers complaining or not ... | 01/01/70 00:00 |