lundi 2 juin 2014

A bug again [eng]

Again a strong bug which I had found since a long time before...

It is located in the calculation time of the trigonometric functions emulated by the 68060.library !
 
And the Warp3D programs use it massively. Indeed, with my little TrigoSpeedTest program, the results are alarming because they are very slow :

I get these snails with a Blizzard 1260 and also with a Apollo 1260 under Kickstart 3.1 and 3.9. And yes, I also tested older versions of the 68060.library...

For cons, the results are correct with a MK2 060  and a MK3 060 under a 4000D :

So, software or hardware bug ? Good question... I included in my Kickstart 3.9 the little IntAckFix patch (on Aminet) which I discovered some days ago, but still the same results for this problem of the day !

Yet, amusing fact : this bug disappeared when the Kickstart is mapped in fastram. With Blitzkick or remapollo, I get the waited results !

Nevertheless this annoying bug should be fixed... Do you have a idea from where it comes ? Go to my email of course...
(translated by Squaley)
     

1 commentaire:

  1. I checked on my Amiga A1200 Blizzard 1260 / 64MHz.
    together with a library 68060.library 46.20, the results:
    TrigoSpeedTest
    2586
    2550
    2201


    68060.library 46.20 + MuFastZero MOVESSP FASTEXEC ON FORCENATIVE
    TrigoSpeedTest
    0178
    0182
    0192
    0870
    0899


    With a library of 68060.library ver. 43.5 (Thor) + MuRedox (replacement Oxypatcher) +
    + MuFastZero MOVESSP FASTEXEC ON:
    TrigoSpeedTest
    0149
    0172
    0157
    0548
    0559


    68060.library ver. 43.5 (Thor) + MuFastZero MOVESSP FASTEXEC ON (without MuRedox):
    TrigoSpeedTest
    0207
    0210
    0216
    1132
    1145


    the same without MoveSSP
    68060.library ver. 43.5 (Thor) + MuFastZero FASTEXEC ON (without MuRedox):
    TrigoSpeedTest
    2651
    2616
    2266


    Conclusions. To get good results you need to move the SSP to Fast RAM.

    RépondreSupprimer

Laissez vos commentaires ici :