Porting GMP 5.0.1 to IBM BlueGene/P
Andrew Mattingly
andrew_mattingly at au1.ibm.com
Thu Apr 28 02:46:29 CEST 2011
I am trying to port GMP to run on BG/P (the chip is powerpc-like, but
different enough to cause issues sometimes), but even when I specify
"--host=none-unknown-linux-gnu" to get a generic build, many of the
verification tests fail (details below). Does anyone recognise these
problems and can you suggest a way around them? Any assistance is greatly
appreciated.
t-bdiv reports:
*******************************************************************************
mpn_mu_bdiv_qr inconsistent in test 5
N= 000fffff ffffffff ffffffff ffffffff ... 00000000 00000000 00000000
00000000
D= 000003ff ffffffff ffffffff ffffffff ... ffffffff ffffffff ffffffff
ffffffff
Q= d222dc33 0031ff94 e41b919f 87e1ec00 ... 00000000 00000000 00000000
00000000
R= 000ffcb7 748f33ff 3801ac6f 925c7974 ... 4e800083 7ed8300a a7c0bf83
ff5afe31
Rb= 0, Cy=0
T= 00100000 00000000 00000000 00a2f793 ... 00000000 00000000 00000000
00000000
nn = 2713, dn = 727, qn = 1986
*******************************************************************************
t-div reports:
*******************************************************************************
mpn_dcpi1_div_qr failed test 1: q too small
N= 000fffff ffffffff ffffffff ffffffff ... 00000000 00000000 00000000
00000000
D= 80000fff ffffffff ffffffff ffffffff ... ffffffff ffffffff ffffffff
ffffffff
Q= 00000000 001ffffc 00007fff f00001ff ... 82352d1a 35a7a2c8 59b1b958
80fe43b4
R= 7a95c83d e2094b0c 571b4b9b 9eb17452 ... 82352d1a 35a7a2c8 59b1b958
80fe43b4
N-Q*D=00000000 00000000 00000000 00000000 ... 82352d1a 35a7a2c8 59b1b958
80fe43b4
nn = 5045, dn = 939, qn = 4107
t-divis just hangs and doesn't report anything
t-fdiv reports:
ERROR
dividend =
-7FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF...
divisor =
-7FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF!
FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF800000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000...
t-gcd reports:
ERROR in test 0
mpz_gcdext returned invalid result
op1=110545CB3F76B25B353B668A9B13477A440FCB5713B52617A8F161F8C02183534B2B1FC3CFF63A4585F1D66A466FED1EB2D3775746ACFAE663546EF8E7...
op2=4415172CFDDAC96CD4ED9A2A6C4D1DE9103F2D5C4ED4985EA3C587E300860D4D2CA83DBE111E22967ED4D43C4C05797EFC55FFE82F7AE078907806...
mpz_gcdext returns:
1FFFFFFFFFFFFFFC01FFFC000000000000000FFFFFFFFFFFFFFFFFFFFF00000000000000000000001FFFFFFFFC000001
s=-3B38BB5C3333C2CC262A0643CCF27B735C16FB1898D6A186F1344D35958DD31797A7CCCC84EFB3C557DA4E94CD71CDE5A0A5E38CFA413BEC26C967B1...
The following tests also fail:
t-hgcd, t-invert, t-mullo, t-perfpow, t-pow, t-powm, t-powm_ui, t-root,
t-sqrtrem, t-tdiv, t-toom8h
Kind regards
Andrew Mattingly
IBM Australia
More information about the gmp-discuss
mailing list