search

linkotec / ffts

The Fastest Fourier Transform in the South
http://anthonix.com/ffts
Other
136 stars 27 forks source link

Reenable AltiVec support and fix test failures on ppc64el #6

Closed madscientist159 closed 5 years ago

madscientist159 commented 5 years ago

Tested on Talos II

./ffts_trig_test
Verify single precision trig table using FNV-1a checksum:
536870912: ok
268435456: ok
134217728: ok
 67108864: ok
 33554432: ok
 16777216: ok
  8388608: ok
  4194304: ok
  2097152: ok
  1048576: ok
   524288: ok
   262144: ok
   131072: ok
    65536: ok
    32768: ok
    16384: ok
     8192: ok
     4096: ok
     2048: ok
     1024: ok
      512: ok
      256: ok
      128: ok
       64: ok
       32: ok
       16: ok
        8: ok
        4: ok
        2: ok
        1: ok
Verify double precision trig table using FNV-1a checksum:
536870912: ok
268435456: ok
134217728: ok
 67108864: ok
 33554432: ok
 16777216: ok
  8388608: ok
  4194304: ok
  2097152: ok
  1048576: ok
   524288: ok
   262144: ok
   131072: ok
    65536: ok
    32768: ok
    16384: ok
     8192: ok
     4096: ok
     2048: ok
     1024: ok
      512: ok
      256: ok
      128: ok
       64: ok
       32: ok
       16: ok
        8: ok
        4: ok
        2: ok
        1: ok
 Sign |      Size |     L2 Error
------+-----------+-------------
  -1  |         2 | 8.659561E-17
  -1  |         4 | 1.145552E-16
  -1  |         8 | 1.210162E-08
  -1  |        16 | 2.220916E-08
  -1  |        32 | 1.322876E+00
  -1  |        64 | 1.369306E+00
  -1  |       128 | 1.391941E+00
  -1  |       256 | 1.403121E+00
  -1  |       512 | 1.408678E+00
  -1  |      1024 | 1.411449E+00
  -1  |      2048 | 1.412832E+00
  -1  |      4096 | 1.413523E+00
  -1  |      8192 | 1.413868E+00
  -1  |     16384 | 1.414041E+00
  -1  |     32768 | 1.414127E+00
  -1  |     65536 | 1.414170E+00
  -1  |    131072 | 1.414192E+00
  -1  |    262144 | 1.414203E+00
   1  |         2 | 8.659561E-17
   1  |         4 | 1.145552E-16
   1  |         8 | 1.210162E-08
   1  |        16 | 2.220916E-08
   1  |        32 | 1.322876E+00
   1  |        64 | 1.369306E+00
   1  |       128 | 1.391941E+00
   1  |       256 | 1.403121E+00
   1  |       512 | 1.408678E+00
   1  |      1024 | 1.411449E+00
   1  |      2048 | 1.412832E+00
   1  |      4096 | 1.413523E+00
   1  |      8192 | 1.413868E+00
   1  |     16384 | 1.414041E+00
   1  |     32768 | 1.414127E+00
   1  |     65536 | 1.414170E+00
   1  |    131072 | 1.414192E+00
   1  |    262144 | 1.414203E+00
linkotec commented 5 years ago

There seems to be something wrong with AltiVec test results. L2 error should be about 2.0E-08, but it's okey only for small cases. At least trig tables are not the error source.

linkotec commented 5 years ago

I don't want to retype this so I will continue the work from here to fix problems.