[Bug 78679] Gen4-5 code lost: runtime_check_aads_emit

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=78679

--- Comment #3 from Iago Toral <itoral at igalia.com> ---
This is weird, but for some reason I cannot reproduce the GPU hang any more on
IronLake... I guess the hang was being caused for some other reason.

Anyway, I have spent some time implementing the removed code paths used when
runtime_check_aads_emit was set to TRUE and that are activated with the
attached example on IronLake and I only get GPU hangs.

I have checked the generated assembly code with INTEL_DEBUG="wm"  and I don't
see anything wrong with the generated program so my guess is that IronLake is
not intended to be used like that (*). On the other hand, with the current code
the AA cases all seem to work fine in IronLake: whether I render antialiased
lines or wire-frame polygons it seems to be doing  antialiasing properly so I
suspect runtime_check_aads_emit was only relevant in Gen <= 4 but Gen 5 is okay
without it.

To support this hypothesis even further there is the fact that the
runtime_check_aads_emit flag was only checked in brw_wm_emit.c, which as far as
I can see handles Gen4 code generation only.

Unfortunately, I don't have any Gen4 hardware here so I cannot verify this.

(*) Actually, I found that SIMD16 code generated was incorrect due to wrong
assignment of GRF registers. I could fix this for my tests by altering
fs_visitor::run() to use assign_regs_trivial() instead of assign_regs() though,
but I wonder if there is a bug lurking in there anyway. I will send an e-mail
to mesa-dev describing this situation in more detail.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20140521/9d163e87/attachment.html>