broadcom/compiler: Set up passthrough Z when doing FS discards.

In order to keep early-Z from writing early in a discard shader, you need
to set the "modifies Z" bit in the shader state (which the new
prog_data.discards will indicate).  Then, in the shader we do a TLB write
to make Z passthrough happen (the QPU result is ignored, so we use a NULL
source).

src/broadcom/compiler/nir_to_vir.c

@@ -1038,6 +1038,17 @@ emit_frag_end(struct v3d_compile *c)
                                        TLB_TYPE_DEPTH |
                                        TLB_DEPTH_TYPE_PER_PIXEL |
                                        0xffffff00);
+        } else if (c->s->info.fs.uses_discard) {
+                struct qinst *inst = vir_MOV_dest(c,
+                                                  vir_reg(QFILE_TLBU, 0),
+                                                  vir_reg(QFILE_NULL, 0));
+                vir_set_cond(inst, discard_cond);
+
+                inst->src[vir_get_implicit_uniform_src(inst)] =
+                        vir_uniform_ui(c,
+                                       TLB_TYPE_DEPTH |
+                                       TLB_DEPTH_TYPE_INVARIANT |
+                                       0xffffff00);
         }
 
         /* XXX: Performance improvement: Merge Z write and color writes TLB

src/broadcom/compiler/v3d_compiler.h

@@ -579,6 +579,7 @@ struct v3d_fs_prog_data {
         BITSET_WORD flat_shade_flags[BITSET_WORDS(V3D_MAX_FS_INPUTS)];
 
         bool writes_z;
+        bool discard;
 };
 
 /* Special nir_load_input intrinsic index for loading the current TLB

src/broadcom/compiler/vir.c

@@ -762,8 +762,9 @@ uint64_t *v3d_compile_fs(const struct v3d_compiler *compiler,
 
         v3d_set_prog_data(c, &prog_data->base);
         v3d_set_fs_prog_data_inputs(c, prog_data);
-        if (c->s->info.outputs_written & (1 << FRAG_RESULT_DEPTH))
-                prog_data->writes_z = true;
+        prog_data->writes_z = (c->s->info.outputs_written &
+                               (1 << FRAG_RESULT_DEPTH));
+        prog_data->discard = c->s->info.fs.uses_discard;
 
         return v3d_return_qpu_insts(c, final_assembly_size);
 }