??? 08/01/05 11:12 Read: times |
#98571 - nohohoho Responding to: ???'s previous message |
Andy Neil said:
Jan Waclawek said:
... the major drawback is, that you need to perform the simulation/cycle counting each time whan you recompile That is why you'd use Assembler - then you'd know that it would stay the same unless you specifically changed it. No - if I'd C, and had critical routine, I'd use assembler only if unable to make it fast enough in C (and able to do so in asm). I'm sure I will find my way how to enforce staying it the same, if I've counted the cycles once (e.g. there must be something like .obj for '51, or a way to switch off further optimisation using those smart linkers you mentioned). Maybe I didn't make my point clear enough - one needs to recount the cycles each time when modifying the source - _that_ would be a pain. But, if you want, let's agree on that automatic cycle counting is rather complicated so maybe not worth doing it - I don't care at all as I don't C - it was of pure academic interest for me. Jan Waclawek |
Topic | Author | Date |
Counting cycles... possible in C? | 01/01/70 00:00 | |
You need a Profiler | 01/01/70 00:00 | |
The compiler knows | 01/01/70 00:00 | |
Source code useless | 01/01/70 00:00 | |
That's why you need Assembler! | 01/01/70 00:00 | |
nohohoho | 01/01/70 00:00 | |
Speed is not everything! | 01/01/70 00:00 | |
Cycles vary? | 01/01/70 00:00 | |
I know | 01/01/70 00:00 | |
absolutely, but what good does it do | 01/01/70 00:00 | |
other way round | 01/01/70 00:00 | |
Yes, the things posted apply in this cas | 01/01/70 00:00 | |
exactly for this reason | 01/01/70 00:00 | |
how can you automate that it is small en | 01/01/70 00:00 | |
Use Microsoft Excel | 01/01/70 00:00 | |
excel and conditional branches ? | 01/01/70 00:00 | |
That's why | 01/01/70 00:00 | |
That makes absolutotally no difference,![]() | 01/01/70 00:00 |