linsolve bug

[Albert Chan]

FYI, this is result of time(linsolve(a,b)). Timing 2nd digit is just a rough guess.

Code:

XCas-1.5.0:         TIME
    53 bits double  6.0E-5
    16 digits MPFR  2.8E-4
    32 digits MPFR  3.2E-4
    
XCas-1.9.0:
    48 bits trunc   6.2E-6
    16 digits MPFR  1.2E-4
    32 digits MPFR  1.3E-4

MPFR timings are basically the same, regardless of setting digits to 16 or 32, or even 100.

Order of magnitude difference between 48 bits and IEEE double is odd.
Perhaps comparing wildly different XCas versions is not a good idea.
But, that's the only 2 versions I had to test.

Anyone has XCas 1.90 (win32), compiled with the -DDOUBLEVAL flag ?