Stats in GBMill (GReductor)
There is already some implementation for gathering stats in G-basis computation.
Perhaps separate stats from progress logging?
#1 Updated by John Abbott over 2 years ago
Mario asked about gathering stats for his involutive basis code.
I suggested using a
struct to keep all the stats stuff together.
struct rather than a
class? Not entirely sure; a
struct seems simpler: by having all data members public any code can increment counters etc.. With a
class it would probably be necessary to have lots of one line functions (at least one to increment a counter, and another to read the counter), yet without any obvious gain.
Anyway, we can try first with a
struct; if that proves not to be as easy/clean as hoped then it should be easy to switch to a
#2 Updated by John Abbott over 2 years ago
I suppose Mario ought to write about this. He has followed my design suggestion, and seems to be happy with it. If it continues to work well, perhaps we should consider adapting the GBMill stats gatherer to work in a similar way.
After speaking to Mario we felt that it was appropriate to have separate mechanisms for gathering stats and for producing progress/logging info. The GBMill implementation seems to have both combined (see
TmpGRStats.H). I also note that the counters in the GBMill impl are
unsigned int; it would probably be better to replace them with
long (or possibly
unsigned long) since
unsigned int may be limited to
#4 Updated by Anna Maria Bigatti 11 months ago
- Subject changed from Stats in GBMill to Stats in GBMill (GReductor)
- % Done changed from 0 to 20
- Estimated time set to 6.00 h
This issue is about "Stats", i.e. gathering info about number of pairs, timings, etc.
This is strictly related with "verbosity", which prints them out, but without recording data.
I'm now resurrecting the GRStats code, making it compatible/complementary with the new verbosity mechanism.