> Hi, > > I have an alpha 21164-based 164LX machine and when I use Lame 3.13 or > greater, it causes a floating exception error when using VBR with > encoding. Lame 3.11 did not have this problem and it only occurs when I > use VBR. This is independent of the linked math library too. > > Thanks for any help, > Mike Do you think this is not the same as the problem reported earlier? The stuff related to noise[ch] (see below) is new in 3.13. If that is causing the problem, you can just delete it since it is only used for the experimental -Z option. It looks like the problem is just an underflow. Linux by default will round to 0 and not complain, but you can set it to terminate on underflow. Any idea if the alpha does this by default? Mark ---- From: Nils Faerber <[EMAIL PROTECTED]> Subject: Lame 3.13 and 3.14 problem on Alpha... To: [EMAIL PROTECTED] Date: Sun, 27 Jun 1999 23:08:56 +0200 (MESZ) In-Reply-To: <199906200341.VAA28648@mrwilson> from "Mark Taylor" at Jun 19, 99 09:41:14 pm Content-Type: text/plain; charset=US-ASCII X-UIDL: 28baf30f39bf1dc389481769e0d5084d X-UID: 105 Hello Mark, me again ;) Another Alpha user and me have a problem with lame 3.13 and 3.14beta. On some samples we get a SIGFPE in loop.c in line 1453 where noise[ch]+=xfsf[ch][i+1][sfb]-l3_xmin->s[gr][ch][sfb][i]; At the point of the exception noise[ch] is about 1.0E-311 which must be something near to zero. So the following addition does not complete and causes a SIGFPE. I tried to fix it but failed :( My current fix is to set signal(SIGFPE,SIG_IGN) at the beginning of void calc_noise2() but that can only be a temporary fix. Do you have any idea how this could happen? It must have been something you changed from lame 3.12pre9 to 3.13... If you have any idea please let me know... I also checked this on an Intel box and it did not happen there. CU nils -- Nils Faerber (Linux Nils) eMail: [EMAIL PROTECTED] Student of computer science http://www.si.unix-ag.org/~nils/ Unix user group, University of Siegen, Germany Siegen ... the arctic rain forest! -- -- MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
