https://bugs.freedesktop.org/show_bug.cgi?id=50317
--- Comment #4 from Tom Stellard <tstel...@gmail.com> 2012-06-18 06:39:24 PDT
---
(In reply to comment #3)
> Just hit this bug with HoN.
> Apparently there is a pattern for AMDIL.sin, but not for AMDGPU.sin:
>
> class SIN_Common <bits<32> inst> : R600_1OP <
> inst, "SIN",
> [(set R600_Reg32:$dst, (int_AMDIL_sin R600_Reg32:$src))]>{
> let Trig = 1;
> }
That pattern in the SIN_Common class is stale and should be removed.
For evergreen and newer, there is a pattern for int_AMGPU_sin:
def : TRIG_eg <SIN_eg, int_AMDGPU_sin>;
R700 chips normalize trigonometric inputs the same was as Evergreen+, so the
correct fix would be to create a SIN_r700 instruction and use the TRIG_eg class
to create a pattern for it. This will also require adding a predicate to
identify r700 chips, because r600 normalizes the inputs differently.
--
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
