On 31 January 2013 15:55, Eero Tamminen
<oak@xxxxxxxxxxxxxx> wrote:
Hi,
For DSP I could store (in addition of sum of instructions & cycles)
also min and max cycle values for individual addresses. Disassembly
could them contain those, or just indicator that they differed.
Would that be enough?
That would be great yes. Min/max provide more detail, but take more space per line... (I suppose a full export would be best served with min/max since it can be post-processed into anything beyond that point).
One more note: having an identifiable prefix/delimiter before each profiler numeric column e.g. "@:" (or whatever works!) makes it easier to extract those columns into Excel or other tools. A minor thing - I may just apply my own patch for that since it can again be post-processed.
For CPU profiling, storing extra information per memory address would
take too much memory because there can be 14MB of addresses (so just
the 32-bit instruction count takes at max 4*14MB of RAM), whereas
DSP memory size is much smaller...
Yes I think it's less useful anyway on the CPU since it's likely happening all over the place. Cache misses are much more interesting on the CPU.
D.