Re:There Are Only Four Billion Floats, So Test Them All (2014)
by gajjanag on Thursday September 14, 2017 @07:27PM
(#15253347)
Nice!
I have noticed that there is a lot of high quality mathematical/numerical software coming out of INRIA; and have often wondered of late why we do not have an equivalent in the US. Do you have any thoughts on why this is the case; and what the closest equivalent in the USA may be?
Re:There Are Only Four Billion Floats, So Test Them All (2014)
by hinkley on Thursday September 14, 2017 @08:29PM
(#15253631)
It's more a matter of what have they done for us lately. We had a number of supercomputing centers in the 90's (at least one has shuttered), and among other things NCSA brought most people their first telnet client and their first web browser via NSF funding that Al Gore pushed for (see also: Al Gore invented the Internet)
After that they were working with INRIA on collaboration software fifteen years before WebEx was a thing. They also had large divisions looking at VR and data visualization.
Re:There Are Only Four Billion Floats, So Test Them All (2014)
by hinkley on Thursday September 14, 2017 @08:34PM
(#15253661)
Oh, and also Apache web server was a fork of NCSA's httpd.
Re:There Are Only Four Billion Floats, So Test Them All (2014)
by jzwinck on Thursday September 14, 2017 @08:22PM
(#15253597)
Intel has MKL/TBB/IPP/etc which are now under one umbrella: https://software.intel.com/en-us/performance-libraries
Re:There Are Only Four Billion Floats, So Test Them All (2014)
by andrewbinstock on Thursday September 14, 2017 @08:04PM
(#15253511)
This is largely done by private companies in the US. In the past, ATT Bell Labs and Xerox PARC were shining examples. Today, every major technology company has labs and other internal organizations contributing research to new technologies: Netflix, Google, Microsoft, etc.
Re:There Are Only Four Billion Floats, So Test Them All (2014)
by Datsundere on Thursday September 14, 2017 @07:37PM
(#15253397)
When I took a testing course, I was taught that Correlated Active Clause Coverage was good enough for most test cases. This term might not be as widely popular
[For finding worst-case inputs] the approach taken was to perform an exhaustive search, taking several hours on the second author’s laptop, of IEEE single-precision inputs, using only a few arguments from Theorem 1 to prune the search.
Even though (2^32)^4 was far too large for it to be tractable to test all 4-tuples of single-precision inputs, a "pruned" search was able to easily cover and/or exclude the entire search space.