Re: [mpir-devel] Re: FFT wrapper done
On Friday 20 January 2012 02:43:11 jason wrote:
On Jan 20, 1:07 am, Jason ja...@njkfrudils.plus.com wrote:
On Thursday 19 January 2012 09:42:35 Jason wrote:
On Thursday 19 January 2012 09:32:02 Cactus wrote:
Thanks for the testing. You are right - the t-locale.c test failure is
not
a major concern.
Has anyone tested a Linux build for the mpir-exp branch?
This will need some configure work done first , I was just waiting for it
to settle done a bit , I should have the time tonight
Brian
done it except for the new fft tuning . I had to change a minor bit about
ULLONG_MAX in tests/t-constants.c as I dont think unix has these or if
configure supports them , I'll check
Tried running the stand alone fft-tune with FLINT_BITS=64 and even
with 3GB and swap space I cant tune the second param
/* fft_tuning.h -- autogenerated by tune-fft */
#ifndef FFT_TUNING_H
#define FFT_TUNING_H
#include mpir.h
#define FFT_TAB \
{ { 4, 3 }, { 4, 3 }, { 3, 2 }, { 2, 1 }, { 2, 1 } }
#define MULMOD_TAB \
{ Segmentation fault
It clearly uses too much memory
Jason
compiling mpir with --enable-assert we get this straight away
/* fft_tuning.h -- autogenerated by tune-fft */
#ifndef FFT_TUNING_H
#define FFT_TUNING_H
#include mpir.h
#define FFT_TAB \
{ { mulmod_2expp1_basecase.c:51: GNU MP assertion failed: !((xp) + (n)
(yp) (yp) + (n) (xp))
Aborted
which is a partial overlap of src and dst in the function
--
You received this message because you are subscribed to the Google Groups
mpir-devel group.
To post to this group, send email to mpir-devel@googlegroups.com.
To unsubscribe from this group, send email to
mpir-devel+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
On Saturday 21 January 2012 15:33:12 Cactus wrote: The FFT tuning appears to work in Windows but if I turn on asserts I see the same problem that Jason reported earlier. But I don't see the out of memory problem. Brian with asserts on it fails make check anyway , so it's nothing to do with the tuning program , could be default parameters for the fft or something else -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
On Saturday 21 January 2012 17:37:27 Jason wrote: On Saturday 21 January 2012 15:33:12 Cactus wrote: The FFT tuning appears to work in Windows but if I turn on asserts I see the same problem that Jason reported earlier. But I don't see the out of memory problem. Brian with asserts on it fails make check anyway , so it's nothing to do with the tuning program , could be default parameters for the fft or something else like the other parameters or cutoff points , this was on a sandybridge -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
OK, then we do need a trace of the function calls and the parameters that are passed for the failing test. Brian is probably right. It's probably an n=0 or trunc=2*n or trunc=0 passed somewhere, or something like that. I'm glad I left the code so that it doesn't deal with these cases, as it may shake some bugs out. It is very odd that it fails on such a small example though. Bill. On 21 January 2012 17:42, Jason ja...@njkfrudils.plus.com wrote: On Saturday 21 January 2012 17:37:27 Jason wrote: On Saturday 21 January 2012 15:33:12 Cactus wrote: The FFT tuning appears to work in Windows but if I turn on asserts I see the same problem that Jason reported earlier. But I don't see the out of memory problem. Brian with asserts on it fails make check anyway , so it's nothing to do with the tuning program , could be default parameters for the fft or something else like the other parameters or cutoff points , this was on a sandybridge -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
Thanks for the testing. You are right - the t-locale.c test failure is not a major concern. Has anyone tested a Linux build for the mpir-exp branch? Brian -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/OQtHeg4FY4gJ. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
On Thursday 19 January 2012 09:32:02 Cactus wrote: Thanks for the testing. You are right - the t-locale.c test failure is not a major concern. Has anyone tested a Linux build for the mpir-exp branch? This will need some configure work done first , I was just waiting for it to settle done a bit , I should have the time tonight Brian -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
Thanks Jason, Right now the source code files for the mpn_* routines provide the interface to the new FFT are in the fft subdirectory. Does this matter - should I move them to the mpn\generic subdirectory? I don't expect these to be subject to assembler code overrides but there may be other reasons for putting them in mpn\generic that I am unaware of. Brian -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/AcVc2rA-kYYJ. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
On Thursday 19 January 2012 10:36:49 Cactus wrote: Thanks Jason, Right now the source code files for the mpn_* routines provide the interface to the new FFT are in the fft subdirectory. Does this matter - should I move them to the mpn\generic subdirectory? I don't expect these to be subject to assembler code overrides but there may be other reasons for putting them in mpn\generic that I am unaware of. Brian yes , only files which may have asm overrides need to be in the mpn/generic , I think! -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
Aqqafa7+@$'_)#*#aq*q* Sent via BlackBerry from T-Mobile -Original Message- From: Case Van Horsen cas...@gmail.com Sender: mpir-devel@googlegroups.com Date: Sun, 15 Jan 2012 15:38:35 To: mpir-devel@googlegroups.com Reply-To: mpir-devel@googlegroups.com Subject: Re: [mpir-devel] Re: FFT wrapper done On Sun, Jan 15, 2012 at 3:33 PM, Cactus rieman...@gmail.com wrote: Thanks Case - I can reproduce the problem and I will do some work on it with Bill and get a fix as soon as possible. Is this the only test that fails? Brian I also get a failure on t-locale.c t-locale.c LIBCMT.lib(lconv.obj) : error LNK2005: localeconv already defined in t-locale.obj t-locale.exe : fatal error LNK1169: one or more multiply defined symbols found ERROR IIRC, that test fails on on quite a few systems so I wasn't too worried about. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/wKc01kTkr2gJ. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
Bill found the problem with the FFT code and I have now uploaded the fix to the mpir-exp branch in the MPIR repository. This fixes the problem for me but I would be grateful if you could check it out. Thanks for helping to test this branch. Do you have any applications that will exercise the MPIR-EXP 64-bit integer support on Windows x64? Brian -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/NO9B_zsFDt8J. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
On Mon, Jan 16, 2012 at 1:41 AM, Cactus rieman...@gmail.com wrote: Bill found the problem with the FFT code and I have now uploaded the fix to the mpir-exp branch in the MPIR repository. This fixes the problem for me but I would be grateful if you could check it out. Thanks for the quick fix. Tests pass on my machine. Thanks for helping to test this branch. Do you have any applications that will exercise the MPIR-EXP 64-bit integer support on Windows x64? I will be using it with GMPY2. Case Brian -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/NO9B_zsFDt8J. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
On Sun, Jan 15, 2012 at 10:34 AM, Cactus rieman...@gmail.com wrote: Thanks for trying this Case - I appreciate your efforts. Unfortunately I don't use Jason's command line build but since it all works in Visual Studio, I would guess that it must be something missing in this build. If it is possible to check all the files that are compiled into the build, it would be worth checking that the file mulmod_2expp1_basecase.c is being included. It has a name that has a long prefix that is the same as another file but I think the days when this would have been a problem are long passed. I can't find mulmod_2expp1_basecase.c on either a Windows or Linux checkout. Is the file possibly missing from SVN? Case Another clue may be that the missing symbol is internal to the library and is hence defined in gmp-impl.h rather than mpir.h - is the right version of gmp-impl.h being picked up when this file is compiled? Brian -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/_nSqACWqfocJ. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
You are right, the file is not in the SVN. My apologies for this - I assumed that a file in the SVN could be renamed and would stay in the SVN but it doesn't seem so. I've now uploaded it and corrected the other problem you found as well. Brian -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/VxXg-tuIUhEJ. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
Thanks Case - I can reproduce the problem and I will do some work on it with Bill and get a fix as soon as possible. Is this the only test that fails? Brian -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/wKc01kTkr2gJ. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
On Sun, Jan 15, 2012 at 3:33 PM, Cactus rieman...@gmail.com wrote: Thanks Case - I can reproduce the problem and I will do some work on it with Bill and get a fix as soon as possible. Is this the only test that fails? Brian I also get a failure on t-locale.c t-locale.c LIBCMT.lib(lconv.obj) : error LNK2005: localeconv already defined in t-locale.obj t-locale.exe : fatal error LNK1169: one or more multiply defined symbols found ERROR IIRC, that test fails on on quite a few systems so I wasn't too worried about. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/wKc01kTkr2gJ. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
Re: [mpir-devel] Re: FFT wrapper done
On Sun, Jan 15, 2012 at 3:38 PM, Case Van Horsen cas...@gmail.com wrote: On Sun, Jan 15, 2012 at 3:33 PM, Cactus rieman...@gmail.com wrote: Thanks Case - I can reproduce the problem and I will do some work on it with Bill and get a fix as soon as possible. Is this the only test that fails? Brian I also get a failure on t-locale.c t-locale.c LIBCMT.lib(lconv.obj) : error LNK2005: localeconv already defined in t-locale.obj t-locale.exe : fatal error LNK1169: one or more multiply defined symbols found ERROR It does not fail on a VS2008 SDK build. Does the MSVC version check in t-locale.c need to be updated? IIRC, that test fails on on quite a few systems so I wasn't too worried about. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To view this discussion on the web visit https://groups.google.com/d/msg/mpir-devel/-/wKc01kTkr2gJ. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en. -- You received this message because you are subscribed to the Google Groups mpir-devel group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
