There is no CGI trick to count your download in this page.
This code is released on November 17th in 2009, so it may contain some bugs. Any feedback is welcome (send an email to Mutsuo Saito, saito "at sign" math.sci.hiroshima-u.ac.jp)
| version | archive | changes |
|---|---|---|
| 1.1.2 |
MTGP-src-1.1.2.zip MTGP-src-1.1.2.tar.gz |
CUDA samples are fixed to run on CUDA 5.0. (2012/10/24) |
| 1.1.1 |
MTGP-src-1.1.1.zip MTGP-src-1.1.1.tar.gz |
add jump function.
add OpenCL sample with jump.
Sorry, CUDA environment of me (Saito) seems to be broken now. (2012/10/16) |
| 1.1 |
MTGP-src-1.1.zip MTGP-src-1.1.tar.gz |
Parameter files of mtgp32 are changed for better dimension
of equidistributions. This change cause changes of outputs, but this
not cause effect on floating point outputs. The changes affect only on
9 LSB bits of unsigned integers and the 9 LSB bits are discarded when
converting to floating points. You can also use old paramter files
instead of new ones. This change reflects the improvement of MTGPDC
ver. 0.3. Sample files for cuda now can be compiled in CUDA 4.0 environment. (2011/6/6) |
| 1.0.2 |
MTGP-src-1.0.2.zip MTGP-src-1.0.2.tar.gz |
change MEXPs of CUDA-sample from 23209 and 44497 to 11213. This is because CUDA occupancy of the latter is larger than those of the formers. (2010/3/26) |
| 1.0.1 |
MTGP-src-1.0.1.zip MTGP-src-1.0.1.tar.gz |
64-bit initialization bug fixed. (2010/1/5) |
| 1.0 |
MTGP-src-1.0.zip MTGP-src-1.0.tar.gz |
First version. (2009/11/17) |
This code is released on March 26th in 2010, so it may contain some bugs. Any feedback is welcome (send an email to Mutsuo Saito, saito "at sign" math.sci.hiroshima-u.ac.jp)
| version | archive | changes |
|---|---|---|
| new! 0.3.1 |
MTGPDC-src-0.3.1.zip MTGPDC-src-0.3.1.tar.gz |
Fix the bug in mtgp64dc mentioned below. (2011/3/9) |
| 0.3 |
MTGPDC-src-0.3.zip MTGPDC-src-0.3.tar.gz |
A small bug in search_temper.hpp is fixed. The bug effects only on the least significant 5 bits, but it makes the dimension of equidistribution of the LSBs smaller than they should be. In addition, we modified the algorithm of searching tempering parameters, so that the algorithm tempers the dimensions of equidistribution of LSBs too, in addition to those of MSBs. This is to address a report on some statistical tests of these LSBs due to Jonathan Passerat-Palmbach, David R.C. Hill and Claude Mazel. We thank them.(2011/3/4) The 64 bit version, mtgp64dc in this version has a serious bug, and should not be used. BUG:All tempering parameters searched by mtgp64dc are zeros. This bug exists only in this version 0.3. It does not exist in the earlier versions, and it is fixed in the version 0.3.1 above. (2011/3/9) |
| 0.2 |
MTGPDC-src-0.2.zip MTGPDC-src-0.2.tar.gz |
some bugs are fixed. (2010/6/8) |
| 0.1 |
MTGPDC-src-0.1.zip MTGPDC-src-0.1.tar.gz |
First version. (2010/3/26) |