[libre-riscv-dev] [Bug 44] IEEE754 FPU inverse-sqrt

bugzilla-daemon at libre-riscv.org bugzilla-daemon at libre-riscv.org
Thu May 2 09:34:02 BST 2019


--- Comment #8 from Jacob Lifshay <programmerjake at gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #7)
> What is the sort of performance actually needed? Is this critical for
> certain circumstances?
Some common lighting algorithms use per-pixel interpolated surface normal
vectors, those need to be normalized before they can be used in lighting
calculations, requiring a 3x3 dot product, an inverse-sqrt, and a 1x3 scalar
vector multiplication.

So, we should be able to support at least 1 inverse-sqrt per pixel or so (with
possibly somewhat reduced framerate). Note that this is in addition to the
depth division that is required per-pixel for basic 3D, so budget accordingly.

You are receiving this mail because:
You are on the CC list for the bug.

More information about the libre-riscv-dev mailing list