Help wanted: porting/testing the Math Toolkit to MacOS X
Any MacOS X experts willing to run some tests on the Math Toolkit, and/or help track down any porting issues? The latest code is in the sandbox under "math_toolkit", last time the tests were run on that platform there were still some unresolved issues with the Mac's strange long double type to be resolved... I'd like to get these nailed down if possible before we do a first post-review release. Many thanks, John Maddock.
Any MacOS X experts willing to run some tests on the Math Toolkit, and/or help track down any porting issues?
The latest code is in the sandbox under "math_toolkit", last time the tests were run on that platform there were still some unresolved issues with the Mac's strange long double type to be resolved... I'd like to get these nailed down if possible before we do a first post-review release.
Did you get any takers? If not, I can run some tests.... -- -- Marshall Marshall Clow Idio Software <mailto:marshall@idio.com> It is by caffeine alone I set my mind in motion. It is by the beans of Java that thoughts acquire speed, the hands acquire shaking, the shaking becomes a warning. It is by caffeine alone I set my mind in motion.
On 8/28/07, John Maddock <john@johnmaddock.co.uk> wrote:
Any MacOS X experts willing to run some tests on the Math Toolkit, and/or help track down any porting issues?
Ran the tests from the sandbox on an Intel mac, GCC 4.0.1/darwin. Errors: http://pastebin.com/m3b56938 Test results: http://dancinghacker.com/test.results HTH, let me know if you need them re-run. If you point me in a direction, I can _try_ to help with any porting issues. Stjepan
I ran the tests, sent the output to John, and am trying to look into the problems. Feel free to look into possible fixes too though, if you get the chance. Jeremy Pack On 8/28/07, Stjepan Rajko <stipe@asu.edu> wrote:
On 8/28/07, John Maddock <john@johnmaddock.co.uk> wrote:
Any MacOS X experts willing to run some tests on the Math Toolkit, and/or help track down any porting issues?
Ran the tests from the sandbox on an Intel mac, GCC 4.0.1/darwin.
Errors: http://pastebin.com/m3b56938
Test results: http://dancinghacker.com/test.results
HTH, let me know if you need them re-run. If you point me in a direction, I can _try_ to help with any porting issues.
Stjepan _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
Jeremy Pack wrote:
I ran the tests, sent the output to John, and am trying to look into the problems. Feel free to look into possible fixes too though, if you get the chance.
Many thanks to both Stjepan and Jeremy for running these, I have a couple of observations/questions: The error rates all look to what I would expect from an 80-bit (64-bit mantissa) real, but I thought the Mac used a 128-bit "double double" type? So... which OS version are you using, is this one of the new "MacIntels" or one of the older Motorola-based machines? Also can you build and run libs/config/test/math_info.cpp and let me have the program output: this will let me see a lot of useful limits-related info. If we can nail the low hanging fruit first, I can look into anything else left later. Many thanks for doing this, John.
On 8/28/07, Stjepan Rajko <stipe@asu.edu> wrote:
On 8/28/07, John Maddock <john@johnmaddock.co.uk> wrote:
Any MacOS X experts willing to run some tests on the Math Toolkit, and/or help track down any porting issues?
Ran the tests from the sandbox on an Intel mac, GCC 4.0.1/darwin.
Errors: http://pastebin.com/m3b56938
Test results: http://dancinghacker.com/test.results
HTH, let me know if you need them re-run. If you point me in a direction, I can _try_ to help with any porting issues.
Stjepan _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
John Maddock wrote:
Jeremy Pack wrote: Also can you build and run libs/config/test/math_info.cpp and let me have the program output: this will let me see a lot of useful limits-related info. If we can nail the low hanging fruit first, I can look into anything else left later.
Never mind, I found the results online - and there are two MacOS "flavors" one with an 80-bit long double and one with a 128-bit "double double". I've committed some changes that should fix most of the failures - or at least produce more sensible results. If one of you can update and re-run the tests that would be excellent! Many thanks, John.
On 8/29/07, John Maddock <john@johnmaddock.co.uk> wrote:
John Maddock wrote:
I've committed some changes that should fix most of the failures - or at least produce more sensible results. If one of you can update and re-run the tests that would be excellent!
Errors: test_triangular.cpp:23:55: error: boost/math/special_functions\fpclassify.hpp: No such file or directory Results: http://dancinghacker.com/test.results2 Stjepan
Stjepan Rajko wrote:
On 8/29/07, John Maddock <john@johnmaddock.co.uk> wrote:
John Maddock wrote:
I've committed some changes that should fix most of the failures - or at least produce more sensible results. If one of you can update and re-run the tests that would be excellent!
Errors:
test_triangular.cpp:23:55: error: boost/math/special_functions\fpclassify.hpp: No such file or directory
Results:
Thanks again! That's looking better already: I'll look into these in detail tomorrow now, but it looks like there are only a few that may signal hidden bugs/problems. John.
-----Original Message----- From: boost-bounces@lists.boost.org [mailto:boost-bounces@lists.boost.org] On Behalf Of Stjepan Rajko Sent: 29 August 2007 19:14 To: boost@lists.boost.org Subject: Re: [boost] Help wanted: porting/testing the Math Toolkit to MacOS X
On 8/29/07, John Maddock <john@johnmaddock.co.uk> wrote:
John Maddock wrote:
I've committed some changes that should fix most of the failures - or at least produce more sensible results. If one of you can update and re-run the tests that would be excellent!
Errors:
test_triangular.cpp:23:55: error: boost/math/special_functions\fpclassify.hpp: No such file or directory
Is this a mixed /\ problem? (Whoever created this \/\/ muddle should be shot, however rich!) My/the current version has all /// #include <boost/math/special_functions/fpclassify.hpp> This file fpclassify.hpp exists in the current sandbox version. Does this help? Paul --- Paul A Bristow Prizet Farmhouse, Kendal, Cumbria UK LA8 8AB +44 1539561830 & SMS, Mobile +44 7714 330204 & SMS pbristow@hetp.u-net.com
Stjepan Rajko wrote:
On 8/29/07, John Maddock <john@johnmaddock.co.uk> wrote:
John Maddock wrote:
I've committed some changes that should fix most of the failures - or at least produce more sensible results. If one of you can update and re-run the tests that would be excellent!
Errors:
test_triangular.cpp:23:55: error: boost/math/special_functions\fpclassify.hpp: No such file or directory
Results:
Thanks again Stjepan. I think I've fixed most of the "trivial" failures (limits too tight) in SVN, but there are a couple of functions where I'm not sure what's going on. After an SVN update can you try running the program below: hopefully comparing the results to Linux will give me some clue what's going on, and whether there's anything to worry about or fix! #define BOOST_MATH_INSTRUMENT #include <boost/math/special_functions.hpp> int main() { std::cout << std::setprecision(20); long double d = boost::math::erfc(89.71L); std::cout << "erfc(89.71L) = " << d << std::endl; d = boost::math::cyl_bessel_k(0.0L, 8072.0L); std::cout << "cyl_bessel_k(0.0L, 8072.0L) = " << d << std::endl; return 0; } Many thanks, John.
On 8/30/07, John Maddock <john@johnmaddock.co.uk> wrote:
Thanks again Stjepan.
No problem, glad to be of help.
I think I've fixed most of the "trivial" failures (limits too tight) in SVN, but there are a couple of functions where I'm not sure what's going on.
I reran the tests. No compile errors this time :-) Test output: http://dancinghacker.com/test.results3
After an SVN update can you try running the program below: hopefully comparing the results to Linux will give me some clue what's going on, and whether there's anything to worry about or fix!
http://pastebin.com/m123ef625 Cheers, Stjepan
Stjepan Rajko wrote:
I reran the tests. No compile errors this time :-)
Test output:
http://dancinghacker.com/test.results3
After an SVN update can you try running the program below: hopefully comparing the results to Linux will give me some clue what's going on, and whether there's anything to worry about or fix!
Looks like I quashed most of the problems, but still missed a few, also looks like expl is inaccurate on that platform? Can you check what std::exp(-8072.0L) evaluates to: should be 2.371057450122979885196291188186517859e-3506 according to functions.wolfram.com. Thanks again, John.
On 8/31/07, John Maddock <john@johnmaddock.co.uk> wrote:
Looks like I quashed most of the problems, but still missed a few, also looks like expl is inaccurate on that platform? Can you check what std::exp(-8072.0L) evaluates to: should be 2.371057450122979885196291188186517859e-3506 according to functions.wolfram.com.
It's: 2.371057450122979876482639519674937e-3506 Best regards, Stjepan
Stjepan Rajko wrote:
On 8/31/07, John Maddock <john@johnmaddock.co.uk> wrote:
Looks like I quashed most of the problems, but still missed a few, also looks like expl is inaccurate on that platform? Can you check what std::exp(-8072.0L) evaluates to: should be 2.371057450122979885196291188186517859e-3506 according to functions.wolfram.com.
It's:
2.371057450122979876482639519674937e-3506
OK that's about 33 eps, which seems to account for the bulk of the error in this case: so that would appear to be why erfc and bessel_k are less accurate on that platform. Shame :-( John. P.S. Are you able to test with Apple's 128-bit "double double" at all? Or is than only on Mach systems? Thanks.
On 8/31/07, John Maddock <john@johnmaddock.co.uk> wrote:
P.S. Are you able to test with Apple's 128-bit "double double" at all? Or is than only on Mach systems? Thanks.
I am really not sure about that. If you have an idea of what compiler switch should tell GCC to do that (or whatever else is needed to make that test happen), I'll give it a try. Stjepan
Stjepan Rajko wrote:
On 8/31/07, John Maddock <john@johnmaddock.co.uk> wrote:
P.S. Are you able to test with Apple's 128-bit "double double" at all? Or is than only on Mach systems? Thanks.
I am really not sure about that. If you have an idea of what compiler switch should tell GCC to do that (or whatever else is needed to make that test happen), I'll give it a try.
I've no idea, it's the difference between: http://engineering.meta-comm.com/boost-regression/trunk/developer/output/San... and http://engineering.meta-comm.com/boost-regression/trunk/developer/output/OSL... It appears one is on PowerPC, the other on Intel. Regards, John.
participants (5)
-
Jeremy Pack -
John Maddock -
Marshall Clow -
Paul A Bristow -
Stjepan Rajko