20180610, 16:11  #2861 
Jun 2003
2×5^{2}×97 Posts 

20180610, 16:17  #2862 
Einyen
Dec 2003
Denmark
7·431 Posts 
It's must be stopped somewhere above 99.90% ? or maybe above 99.95%?

20180610, 17:51  #2863 
"Yves"
Jul 2017
Belgium
3×17 Posts 
One week ago , this exponent was already displayed as 100.00 % done

20180610, 17:56  #2864 
Jun 2015
Vallejo, CA/.
2×5×97 Posts 

20180610, 18:11  #2865  
Sep 2003
3^{2}×7×41 Posts 
Quote:
So anything above 99.95% is rounded to 100.0%, in according with the usual rules for printf rounding to one decimal place. In fact I verified this just now: 51637528 / 51663607 = 99.94952% = displays as 99.9% after Manual Communication 51638156 / 51663607 = 99.95074% = displays as 100.0% after Manual Communication 

20180610, 18:14  #2866  
Sep 2003
5027_{8} Posts 
Quote:
The threshold for displaying 100.0%, as I mentioned, is 99.95% iterations completed. 

20180610, 20:26  #2867  
Jun 2015
Vallejo, CA/.
2×5×97 Posts 
Quote:


20180610, 22:50  #2868 
Sep 2003
A17_{16} Posts 

20180611, 08:28  #2869 
Romulan Interpreter
Jun 2011
Thailand
10001111010000_{2} Posts 
Beating that horse more, if one lets the default numbers be, for the frequency of reports, or if he uses "reasonable" values (like every 1 million iterations, which is quite ok for a test of 80M+ even if it takes 40 days on a shitty computer, it still makes two reports per day), then every "step" of the report is about 100/80=1.25%. Therefore the number of exponents for which "shit happens" (i.e. last report falls in between 99.95% and 100%) is so small, and they are so rare, they can be counted on the fingers of one hand...
(Well, not exactly, but still, we don't see any problem here...) Last fiddled with by LaurV on 20180611 at 08:40 
20180611, 10:29  #2870 
Undefined
"The unspeakable one"
Jun 2006
My evil lair
1011110000000_{2} Posts 
Or, ya 'no, just fix it for real by turning off rounding. Or if that proves to be too difficult in C for some reason then just subtract 0.05 before printing the value.
Also, while you're at it, remove the trailing '0'. It is unneeded and gives a false impression about the precision being shown. 
20180613, 04:06  #2871 
Serpentine Vermin Jar
Jul 2014
2·5·7·47 Posts 

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Newer X64 build needed  Googulator  Msieve  73  20200830 07:47 
Performance of cudaecm on newer hardware?  fivemack  GMPECM  14  20150212 20:10 
Cause this don't belong in the milestone thread  bcp19  Data  30  20120908 15:09 
Newer msieves are slow on Core i7  mklasson  Msieve  9  20090218 12:58 
Use of large memory pages possible with newer linux kernels  Dresdenboy  Software  3  20031208 14:47 