Non correctly rounded division

[m8pple]

This is not a big issue, but I noticed that floating-point division is not
correctly rounded. It seems to be within an ULP, so maybe the Altera IP
is not correctly rounded, though I would have hoped it was.

Or is it configured as faithful to save area?

Not a massive deal, but when doing bit-exact verification it might
trip people up, so mentioning in the limitations would be useful.

HOST [00001000] 10.000000 : Fail : for 11.31642818450928 (41351017) / 19.53116416931152 (419c3fd3), got 0.5794036388397217 (3f1453cc), expected 0.5794036984443665 (3f1453cd), absErr=5.96046e-08, relErr=1.02872e-07
HOST [00001000] 10.000000 : Fail : for 9.290762901306152 (4114a6f7) / 57.35811996459961 (42656eb7), got 0.161978155374527 (3e25dd9a), expected 0.1619781702756882 (3e25dd9b), absErr=1.49012e-08, relErr=9.19949e-08
HOST [00001000] 10.000000 : Fail : for 0.3579070866107941 (3eb73f99) / 35 (420c0000), got 0.01022591628134251 (3c278a9a), expected 0.01022591721266508 (3c278a9b), absErr=9.31323e-10, relErr=9.10747e-08
HOST [00001000] 10.000000 : Fail : for 16.59977340698242 (4184cc56) / 25.88802337646484 (41cf1aac), got 0.6412143707275391 (3f2426a0), expected 0.6412144303321838 (3f2426a1), absErr=5.96046e-08, relErr=9.29559e-08
HOST [00001000] 10.000000 : Fail : for 1.119377136230469 (3f8f47c0) / 27.29496002197266 (41da5c14), got 0.04101039469242096 (3d27fa84), expected 0.04101039841771126 (3d27fa85), absErr=3.72529e-09, relErr=9.08377e-08
HOST [00001000] 10.000000 : Fail : for 43.38915634155273 (422d8e7f) / 23.39516830444336 (41bb294e), got 1.854620337486267 (3fed6433), expected 1.854620456695557 (3fed6434), absErr=1.19209e-07, relErr=6.42769e-08
HOST [00001000] 10.000000 : Fail : for 31.42159271240234 (41fb5f6c) / 11.96130084991455 (413f617d), got 2.626937627792358 (40281fbf), expected 2.626937866210938 (40281fc0), absErr=2.38419e-07, relErr=9.07591e-08
HOST [00001000] 10.000000 : Fail : for 40 (42200000) / 2.136669158935547 (4008bf30), got 18.72072601318359 (4195c40c), expected 18.72072792053223 (4195c40d), absErr=1.90735e-06, relErr=1.01884e-07
HOST [00001000] 10.000000 : Fail : for 29.69193649291992 (41ed8916) / 21.7627124786377 (41ae1a09), got 1.364349007606506 (3faea2fd), expected 1.364349126815796 (3faea2fe), absErr=1.19209e-07, relErr=8.73745e-08
HOST [00001000] 10.000000 : Fail : for 7.052730560302734 (40e1aff8) / 19 (41980000), got 0.3711963295936584 (3ebe0d72), expected 0.3711963593959808 (3ebe0d73), absErr=2.98023e-08, relErr=8.02872e-08
HOST [00001000] 10.000000 : Fail : for 3.747602224349976 (406fd8b7) / 9.308246612548828 (4114ee94), got 0.4026109576225281 (3ece2306), expected 0.4026109874248505 (3ece2307), absErr=2.98023e-08, relErr=7.40226e-08
HOST [00001000] 10.000000 : Fail : for 2.358560562133789 (4016f2a8) / 53 (42540000), got 0.04450114071369171 (3d3646d4), expected 0.04450114443898201 (3d3646d5), absErr=3.72529e-09, relErr=8.37122e-08
HOST [00001000] 10.000000 : Fail : for 61.79955673217773 (427732bf) / 51.87379455566406 (424f7ec4), got 1.191344380378723 (3f987df9), expected 1.191344499588013 (3f987dfa), absErr=1.19209e-07, relErr=1.00063e-07

[mn416]

Hi David,

Thanks for the report, I'll add something to the limitations section.

FYI, the docs for Altera's FP cores are here:

https://www.altera.com/en_US/pdfs/literature/ug/ug_altfp_mfug.pdf