| Rev |
Age |
Author |
Path |
Log message |
Diff |
| 52864 |
5807 d 3 h |
urbaneks |
/branches/R-exp-R5/ |
Experinental branch implementing "R5" ideas - extending S language to
create a more intuitive object system |
|
| 50559 |
6088 d 9 h |
ripley |
/trunk/ |
reword one message, put up all for translation |
|
| 46808 |
6477 d 0 h |
maechler |
/trunk/ |
pf(x, ncp) more accurate for large x, using new pnbeta2() with (x,1-x) arg |
|
| 43710 |
6797 d 8 h |
ripley |
/trunk/src/nmath/ |
clean up |
|
| 42724 |
6904 d 2 h |
ripley |
/trunk/ |
gettext has defined _GNU_SOURCE in config.h on systems which need it,
so we now only need declaration where config.h is not included. |
|
| 42682 |
6907 d 3 h |
ripley |
/trunk/src/nmath/ |
need to set glibc flags even for nmath standlaone builds
make LDOUBLE available for future use (e.g. in rmultinom) |
|
| 42549 |
6918 d 11 h |
ripley |
/trunk/src/nmath/ |
replacing underflow by incorrect denormalized value is not helpful
(and not what was done for overflow). |
|
| 42427 |
6931 d 8 h |
ripley |
/trunk/src/ |
tweaks for glibc systems under -ansi and -std=c99 |
|
| 42422 |
6931 d 11 h |
ripley |
/trunk/src/ |
eliminate MATHLIB_PRIVATE
use float.c constants rather than d1mach
port.c does not actually use i1mach |
|
| 42419 |
6932 d 9 h |
ripley |
/trunk/src/ |
more tidyig up of Rmath.h |
|
| 42417 |
6932 d 9 h |
ripley |
/trunk/src/ |
tidy up headers for private nmath |
|
| 42405 |
6933 d 11 h |
ripley |
/trunk/src/nmath/ |
add header |
|
| 42396 |
6934 d 2 h |
ripley |
/trunk/src/ |
start tidying up headers.
port.c no longer used d1mach |
|
| 42303 |
6942 d 10 h |
ripley |
/trunk/src/nmath/ |
fix up mess of r42302 |
|
| 42302 |
6942 d 11 h |
ripley |
/trunk/src/nmath/ |
update reference to GPL |
|
| 37528 |
7441 d 2 h |
ripley |
/trunk/src/ |
start moving bratio to have log_p argument.
remove private functions (_raw etc) from public header files |
|
| 37461 |
7452 d 6 h |
ripley |
/trunk/src/nmath/ |
report where ML_ERROR is occurring |
|
| 37287 |
7474 d 21 h |
ripley |
/trunk/ |
improvements to [pq]nchisq
- use direct formula which allows for lower_tail = FALSE if ncp < 80
(this is often a lot more accurate).
- use starting point and lower_tail in qnchisq
- can be slower, so make interruptible |
|
| 37281 |
7475 d 8 h |
ripley |
/trunk/ |
Get ML_ERROR to do something useful, and ensure MATHLIB messages end in \n.
Warn on cancellation errors in non-central CDFs for t, f, beta, chisq |
|
| 36986 |
7508 d 5 h |
ripley |
/trunk/ |
port.c was using d1mach and i1mach even though they were declared private
more use of visibility attributes |
|
The R Project SVN
R