Rev 3 | Details | Compare with Previous | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 2 | pj | 1 | <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN"> |
| 2 | <html> |
||
| 3 | <head><title> |
||
| 4 | FFTW FAQ - Section 5 |
||
| 5 | </title> |
||
| 6 | <link rev="made" href="mailto:fftw@theory.lcs.mit.edu"> |
||
| 7 | <link rel="Contents" href="index.html"> |
||
| 8 | <link rel="Start" href="index.html"> |
||
| 9 | <link rel="Previous" href="section4.html"><link rel="Bookmark" title="FFTW FAQ" href="index.html"> |
||
| 10 | </head><body text="#000000" bgcolor="#FFFFFF"><h1> |
||
| 11 | FFTW FAQ - Section 5 <br> |
||
| 12 | Known bugs |
||
| 13 | </h1> |
||
| 14 | |||
| 15 | <ul> |
||
| 16 | <li><a href="#rfftwndbug" rel=subdocument>Q5.1. FFTW 1.1 crashes in rfftwnd on Linux.</a> |
||
| 17 | <li><a href="#fftwmpibug" rel=subdocument>Q5.2. The MPI transforms in FFTW 1.2 give incorrect results/leak |
||
| 18 | memory.</a> |
||
| 19 | <li><a href="#testsingbug" rel=subdocument>Q5.3. The test programs in FFTW 1.2.1 fail when I change FFTW to use single |
||
| 20 | precision.</a> |
||
| 21 | <li><a href="#teststoobig" rel=subdocument>Q5.4. The test program in FFTW 1.2.1 fails for n > |
||
| 22 | 46340.</a> |
||
| 23 | <li><a href="#linuxthreads" rel=subdocument>Q5.5. The threaded code fails on Linux Redhat 5.0</a> |
||
| 24 | <li><a href="#bigrfftwnd" rel=subdocument>Q5.6. FFTW 2.0's rfftwnd fails for rank > 1 transforms with a final |
||
| 25 | dimension >= 65536.</a> |
||
| 26 | <li><a href="#primebug" rel=subdocument>Q5.7. FFTW 2.0's complex transforms give the wrong results with prime |
||
| 27 | factors 17 to 97.</a> |
||
| 28 | <li><a href="#mpichbug" rel=subdocument>Q5.8. FFTW 2.1.1's MPI test programs crash with |
||
| 29 | MPICH.</a> |
||
| 30 | </ul><hr> |
||
| 31 | |||
| 32 | <h2><A name="rfftwndbug"> |
||
| 33 | Question 5.1. FFTW 1.1 crashes in rfftwnd on |
||
| 34 | Linux. |
||
| 35 | </A></h2> |
||
| 36 | |||
| 37 | This bug was fixed in FFTW 1.2. There was a bug in |
||
| 38 | <code>rfftwnd</code> causing an incorrect amount of memory to be allocated. The bug showed |
||
| 39 | up in Linux with libc-5.3.12 (and nowhere else that we know of). |
||
| 40 | |||
| 41 | <h2><A name="fftwmpibug"> |
||
| 42 | Question 5.2. The MPI transforms in FFTW 1.2 give incorrect |
||
| 43 | results/leak memory. |
||
| 44 | </A></h2> |
||
| 45 | |||
| 46 | These bugs were corrected in FFTW 1.2.1. The MPI transforms (really, |
||
| 47 | just the transpose routines) in FFTW 1.2 had bugs that could cause |
||
| 48 | errors in some situations. |
||
| 49 | <h2><A name="testsingbug"> |
||
| 50 | Question 5.3. The test programs in FFTW 1.2.1 fail when I change FFTW |
||
| 51 | to use single precision. |
||
| 52 | </A></h2> |
||
| 53 | |||
| 54 | This bug was fixed in FFTW 1.3. (Older versions of FFTW did |
||
| 55 | work in single precision, but the test programs didn't--the error |
||
| 56 | tolerances in the tests were set for double precision.) |
||
| 57 | |||
| 58 | <h2><A name="teststoobig"> |
||
| 59 | Question 5.4. The test program in FFTW 1.2.1 fails for n > |
||
| 60 | 46340. |
||
| 61 | </A></h2> |
||
| 62 | |||
| 63 | This bug was fixed in FFTW 1.3. FFTW 1.2.1 produced the right answer, |
||
| 64 | but the test program was wrong. For large n, n*n in the naive |
||
| 65 | transform that we used for comparison overflows 32 bit integer |
||
| 66 | precision, breaking the test. |
||
| 67 | <h2><A name="linuxthreads"> |
||
| 68 | Question 5.5. The threaded code fails on Linux Redhat |
||
| 69 | 5.0 |
||
| 70 | </A></h2> |
||
| 71 | |||
| 72 | We had problems with glibc-2.0.5. The code should work with |
||
| 73 | glibc-2.0.7. |
||
| 74 | <h2><A name="bigrfftwnd"> |
||
| 75 | Question 5.6. FFTW 2.0's rfftwnd fails for rank > 1 transforms |
||
| 76 | with a final dimension >= 65536. |
||
| 77 | </A></h2> |
||
| 78 | |||
| 79 | This bug was fixed in FFTW 2.0.1. (There was a 32-bit integer |
||
| 80 | overflow due to a poorly-parenthesized expression.) |
||
| 81 | <h2><A name="primebug"> |
||
| 82 | Question 5.7. FFTW 2.0's complex transforms give the wrong results |
||
| 83 | with prime factors 17 to 97. |
||
| 84 | </A></h2> |
||
| 85 | |||
| 86 | There was a bug in the complex transforms that could cause incorrect |
||
| 87 | results under (hopefully rare) circumstances for lengths with |
||
| 88 | intermediate-size prime factors (17-97). This bug was fixed in FFTW |
||
| 89 | 2.1.1. |
||
| 90 | <h2><A name="mpichbug"> |
||
| 91 | Question 5.8. FFTW 2.1.1's MPI test programs crash with |
||
| 92 | MPICH. |
||
| 93 | </A></h2> |
||
| 94 | |||
| 95 | This was fixed in FFTW 2.1.2. The 2.1/2.1.1 MPI test programs crashed |
||
| 96 | when using the MPICH implementation of MPI with the |
||
| 97 | <code>ch_p4</code> device (TCP/IP); the transforms themselves worked fine. (The source |
||
| 98 | of the bug was some strange constraints that MPICH imposes on access |
||
| 99 | to the program argument list.) <hr> |
||
| 100 | Back: <a href="section4.html" rev=precedes>Internals of FFTW</a>.<br> |
||
| 101 | <a href="index.html" rev=subdocument>Return to contents</a>.<p> |
||
| 102 | <address> |
||
| 103 | <A href="http://theory.lcs.mit.edu/~fftw/">Matteo Frigo and Steven G. Johnson</A> / <A href="mailto:fftw@theory.lcs.mit.edu">fftw@theory.lcs.mit.edu</A> |
||
| 104 | - 18 May 1999 |
||
| 105 | </address><br> |
||
| 106 | Extracted from FFTW Frequently Asked Questions with Answers, |
||
| 107 | Copyright © 1999 Massachusetts Institute of Technology. |
||
| 108 | </body></html> |