[Mesa-dev] RFC: ARB_arrays_of_arrays work in progress

[Chris Forbes]

That assert means that you've not told the backend the size of some
object correctly, and the register allocator has placed it so its true
extent hangs off the end of the available registers.

To fix it you'll need to adjust the code generation for array accesses
in i965 -- probably quite a bit.

On Thu, Jan 16, 2014 at 1:14 AM, Timothy Arceri <t_arceri at yahoo.com.au> wrote:
> First of all sorry for my first post being a reply to Ians 10.1 release
> plan.
>
> Secondly I've rebased my work on current master and reimplemented my
> work using the approach I mention in my last email. The changes are half
> the size and now some basic execution tests are passing. The new branch
> is here: https://github.com/tarceri/Mesa_arrays_of_arrays/tree/wip8
>
> However I'm having an issue where if I'm running the tests using a
> command like
> ./piglit-run.py -t ARB_arrays_of_arrays/compiler* tests/all.tests
> results/array_of_array
>
> Some of the tests crash with this error:
> glslparsertest: brw_eu_emit.c:305: brw_set_src0: Assertion `reg.nr <
> 128' failed.
>
> But when I run them directly using glslparsertest they don't crash.
> Anyone know why this would be happening? I looked at the assert but it
> didn't really make any sense to me.
>
> In case it helps I'm using the build options ./autogen.sh
> --prefix=/opt/xorg --enable-texture-float --with-dri-drivers="i965"
> --with-llvm-shared-libs --disable-dri3 --with-gallium-drivers=""
> on Fedora 19
>
> Thanks,
> Tim
>
>
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev