[Bug 93681] GLSL compilation can be very slow

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=93681

--- Comment #18 from Tapani Pälli <lemody at gmail.com> ---
(In reply to Eero Tamminen from comment #17)
> It's clear off-by-one access past the end of array:
> 
> ==14419== Invalid read of size 4
> ==14419==    at 0x6454A39: fs_reg_alloc::set_spill_costs()
> (brw_fs_reg_allocate.cpp:925)
> ==14419==    by 0x6454B26: fs_reg_alloc::choose_spill_reg()
> (brw_fs_reg_allocate.cpp:948)
> ==14419==    by 0x64556FD: fs_reg_alloc::assign_regs(bool, bool)
> (brw_fs_reg_allocate.cpp:1163)
> 
> ==14419==  Address 0x83bdcb4 is 0 bytes after a block of size 10,244 alloc'd
> ==14419==    at 0x483880B: malloc (vg_replace_malloc.c:309)
> ==14419==    by 0x5CAF04C: ralloc_size (ralloc.c:119)
> ==14419==    by 0x5CAF3C1: ralloc_array_size (ralloc.c:221)
> ==14419==    by 0x64317A9: fs_visitor::calculate_live_intervals()
> (brw_fs_live_variables.cpp:335)

Yeah. However I'm sorry, I've confused with my shader_test files (argh!), this
trace was for shader from somewhat similar bug #70613. But fortunately it seems
I hit the very same issue with the slow.shader_test. I've bisected this to
start happening with following commit, maybe should have a separate bug for
that:

b2d274c6775 intel/fs/ra: Choose a spill reg before throwing away the graph

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20190603/445bf8fb/attachment.html>