> The C55 and C64 builds diverge in exactly the same place as before (byte > 0x1000). The output from C55 build 11387 (svn head) diverges from C55 build > 11343 slightly later (byte 0x1116). Similarly, the output from C64 build > 11387 (svn head) diverges from C64 build 11343 slightly later (byte 0x1126). > So, your change clearly had an effect, not just on the 16-bit machine, but > there is something else going on.I made other minor changes that are responsible for the difference between 11343 and 11387. Can you check if the tables from kiss-fft are the same on C55 and C64. That could explain the problem. Jean-Marc
>> The C55 and C64 builds diverge in exactly the same place as before (byte >> 0x1000). The output from C55 build 11387 (svn head) diverges from C55 >> build >> 11343 slightly later (byte 0x1116). Similarly, the output from C64 build >> 11387 (svn head) diverges from C64 build 11343 slightly later (byte >> 0x1126). >> So, your change clearly had an effect, not just on the 16-bit machine, >> but >> there is something else going on. > > I made other minor changes that are responsible for the difference > between 11343 and 11387. Can you check if the tables from kiss-fft are > the same on C55 and C64. That could explain the problem.I found one location in the two super_twiddle tables which differed in the LSB. I patched this in the C55 build, and it did not change the canceled output. - Jim
> I found one location in the two super_twiddle tables which differed in the > LSB. I patched this in the C55 build, and it did not change the canceled > output.Can you investigate where exactly the divergence occurs? Maybe when st->adapted gets set to 1? Jean-Marc