[Mesa-dev] freedreno: 'Unhandled NIR tex src type: 11' on A3XX

[Rob Clark]

On Sun, Jun 9, 2019 at 7:43 AM Jonathan Marek <jonathan at marek.ca> wrote:
>
> On 6/9/19 8:41 AM, Brian Masney wrote:
> > On Sat, Jun 08, 2019 at 10:58:11PM -0400, Jonathan Marek wrote:
> >> Hi,
> >>
> >> It's possible 19.1 has another issue, I only tested the master branch with
> >> my fix. I would suggest trying 19.0 or the master branch.
> >
> > The mesa master branch and 19.0.6 both give the following error when
> > glxgears starts up:
> >
> > ../src/gallium/drivers/freedreno/freedreno_batch.c:424:fd_batch_add_dep: Assertion `!batch_depends_on(dep, batch)' failed.
> >
>
> No one is testing freedreno+X11 AFAIK. This would affect all adrenos
> too, not just a3xx. I can look into it at some point, if no one else does.

This one could be timing related, I suppose.. that would be my best
guess about why I haven't seen it..

BR,
-R

>
> To test if the GPU works at all you should use kmscube. If that works
> then you can try wayland/weston, or if you really need X11 IIRC 18.1 was
> working with X11.
>
> >> FYI, I haven't pushed it anywhere but I recently rebased my Nexus 5 patches
> >> from last year (and been looking at getting call audio working).
> >
> > Fantastic!
> >
> > Brian
> >
> >
> >
> >> On 6/8/19 9:08 PM, Brian Masney wrote:
> >>> Hi,
> >>>
> >>> I'm trying to get the GPU working using the Freedreno driver (A330) on
> >>> the Nexus 5 phone. I'm using kernel 5.2rc3 with some out of tree patches
> >>> related to the GPU [1] and mesa 19.1.0-rc5 on postmarketOS. When I run
> >>> glxgears, I see the gears show up for a fraction of a second and then
> >>> it terminates due to the following error:
> >>>
> >>> -----
> >>> shader: MESA_SHADER_FRAGMENT
> >>> inputs: 1
> >>> outputs: 1
> >>> uniforms: 0
> >>> shared: 0
> >>> decl_var uniform INTERP_MODE_NONE sampler2D sampler (0, 0, 0)
> >>> decl_var shader_in INTERP_MODE_SMOOTH vec4 in_0 (VARYING_SLOT_VAR0, 0, 0)
> >>> decl_var shader_out INTERP_MODE_FLAT vec4 out_0 (FRAG_RESULT_DATA0, 0, 0)
> >>> decl_function main (0 params)
> >>>
> >>> impl main {
> >>>           block block_0:
> >>>           /* preds: */
> >>>           vec1 32 ssa_0 = load_const (0x00000000 /* 0.000000 */)
> >>>           vec2 32 ssa_1 = intrinsic load_barycentric_pixel () (1) /* interp_mode=1 */
> >>>           vec4 32 ssa_2 = intrinsic load_interpolated_input (ssa_1, ssa_0) (0, 0) /* base=0 */ /* component=0 */  /* in_0 */
> >>>           vec1 32 ssa_3 = deref_var &sampler (uniform sampler2D)
> >>>           vec2 32 ssa_4 = vec2 ssa_2.x, ssa_2.y
> >>>           vec4 32 ssa_5 = tex ssa_3 (texture_deref), ssa_3 (sampler_deref), ssa_4 (coord)
> >>> Unhandled NIR tex src type: 11
> >>>
> >>>
> >>>           intrinsic store_output (ssa_5, ssa_0) (0, 15, 0) /* base=0 */ /* wrmask=xyzw */ /* component=0 */       /* out_0 */
> >>>           /* succs: block_1 */
> >>>           block block_1:
> >>> }
> >>>
> >>> Assertion failed: !"" (../src/freedreno/ir3/ir3_context.c: ir3_context_error: 407)
> >>> -----
> >>>
> >>> I verified that the mesa 19.1.0-rc5 release contains this recent a3xx
> >>> fix from Jonathan:
> >>> https://gitlab.freedesktop.org/mesa/mesa/commit/1db86d8b62860380c34af77ae62b019ed2376443
> >>>
> >>> Any suggestions?
> >>>
> >>> [1] https://github.com/masneyb/linux/commits/v5.2-rc3-nexus5-gpu-wip
> >>>       The GPU specific patches start at Rob's patch 'qcom-scm: add support
> >>>       to restore secure config' on that list. I submitted the patches
> >>>       below that a few weeks ago to the upstream kernel and I expect
> >>>       they'll be merged. Once I have a working GPU, I plan to start
> >>>       working on the interconnect support in the kernel for msm8974 so
> >>>       that the clock hacks can be dropped.
> >>>
> >>> Brian
> >>>