hasvk: Skip zero-sized compute dispatches on gen7.

GPGPU_WALKER has no encoding for a zero group count, so passing one
through walks the full dimension and the command never retires, hanging
rcs0 (GPU HANG ecode 7:1:8ff8ffff, EIR and PGTBL_ER both clear).  A zero
dispatch is legal and defined as a no-op, so drop it at record time.
through walks the full dimension and the command never retires, hanging
rcs0 (GPU HANG ecode 7:1:8ff8ffff, EIR and PGTBL_ER both clear).  A zero
dispatch is legal and defined as a no-op, so drop it at record time.

Gen9+ uses COMPUTE_WALKER and handles this in hardware, which is why
upstream anv carries no equivalent guard.

Fixes dEQP-VK.compute.pipeline.basic.empty_workgroup_{all,x,y,z} and
empty_shader, which previously aborted the run and could take the whole
machine down.  compute.pipeline.* now 20285 cases, 0 fail.

src/intel/vulkan_hasvk/genX_cmd_buffer.c

@@ -4811,6 +4811,18 @@ void genX(CmdDispatchBase)(
    if (cmd_buffer->state.conditional_render_enabled)
       genX(cmd_emit_conditional_render_predicate)(cmd_buffer);
 
+   /* A dispatch with a zero group count is legal and defined to do nothing.
+    * GPGPU_WALKER has no encoding for "no groups at all" though, so passing
+    * zero through walks the full dimension and the command never retires:
+    * dEQP-VK.compute.pipeline.basic.empty_workgroup_* hangs rcs0 every time
+    * (GPU HANG ecode 7:1:8ff8ffff).  Gen9+ uses COMPUTE_WALKER and handles
+    * this in hardware, which is why upstream anv carries no such guard.
+    */
+   if (groupCountX == 0 || groupCountY == 0 || groupCountZ == 0) {
+      trace_intel_end_compute(&cmd_buffer->trace, 0, 0, 0, 0);
+      return;
+   }
+
    emit_cs_walker(cmd_buffer, pipeline, false, prog_data, groupCountX,
                   groupCountY, groupCountZ);